MRE11 and EXO1 nucleases degrade reversed forks and elicit MUS81-dependent fork rescue in BRCA2-deficient cells

The breast cancer susceptibility proteins BRCA1 and BRCA2 have emerged as key stabilizing factors for the maintenance of replication fork integrity following replication stress. In their absence, stalled replication forks are extensively degraded by the MRE11 nuclease, leading to chemotherapeutic sensitivity. Here we report that BRCA proteins prevent nucleolytic degradation by protecting replication forks that have undergone fork reversal upon drug treatment. The unprotected regressed arms of reversed forks are the entry point for MRE11 in BRCA-deficient cells. The CtIP protein initiates MRE11-dependent degradation, which is extended by the EXO1 nuclease. Next, we show that the initial limited resection of the regressed arms establishes the substrate for MUS81 in BRCA2-deficient cells. In turn, MUS81 cleavage of regressed forks with a ssDNA tail promotes POLD3-dependent fork rescue. We propose that targeting this pathway may represent a new strategy to modulate BRCA2-deficient cancer cell response to chemotherapeutics that cause fork degradation.

G ermline mutations in the Breast Cancer Susceptibility genes BRCA1/BRCA2 account for the vast majority of familial breast cancer cases [1][2][3][4] . Aside from their wellestablished roles in homologous recombination (HR), BRCA proteins are emerging as key factors required for the maintenance of replication fork stability following replication stress induction [5][6][7][8] . In their absence, replication forks are extensively degraded by the MRE11 nuclease. MRE11-dependent degradation of replication forks observed in the absence of BRCA proteins leads to long stretches of ssDNA (>4-5 kb) and is emerging as one of the leading causes of the sensitivity to therapies that target DNA or that inhibit specific repair pathways such as PARP inhibitors 5 . The mechanism leading to this extensive fork degradation phenotype in the absence of BRCA1 or BRCA2 remains unclear. For example, the exact structure(s) of the replication intermediates targeted by nucleases in BRCA-deficient cells is unknown. Moreover, MRE11 has limited nucleolytic activity 9 and is unlikely to be the only nuclease responsible for degrading several kb of DNA in BRCA-deficient cells. Finally, the fate of the extensively resected forks upon drug removal has never been investigated in detail, even though it is tightly linked to the increased chromosomal aberrations and DNA damage sensitivity of BRCA-deficient cells.
Replication fork reversal is a key protective mechanism that allows replication forks to reverse their course when they encounter DNA lesions [10][11][12][13][14] . Interestingly, the same HR factors controlling MRE11 nuclease activity and ssDNA accumulation are also emerging as crucial players involved in fork remodeling [14][15][16] . In particular, the central recombinase RAD51 is essential for fork reversal upon chemotherapeutic treatment 14 . By analogy with its bacterial homologue RecA, RAD51 may be recruited to ssDNA stretches formed at replication fork junctions and promote the initial step of fork reversal by invading the complementary strand. In this context, HR proteins may also be required to stabilize forks in their reversed state by protecting the double-stranded end of the regressed arm from nucleolytic degradation.
In this study, we combine electron microscopy (EM) with genome-wide single-molecule DNA fiber approaches to define the mechanism by which the BRCA proteins protect replication forks from nucleolytic degradation following replication stress induction. We show that the main function of BRCA proteins in this context is to protect the regressed arms of replication forks that have reversed upon drug treatment from nucleolytic degradation. In their absence, CtIP initiates the MRE11-dependent degradation of the unprotected regressed arms and EXO1 contributes to extend fork degradation. Next, we investigate how cells cope with these extensively resected forks upon drug removal. In particular, we find that MUS81 cleavage rescues the resected forks in BRCA2-, but not BRCA1-deficient cells through a break-induced replication (BIR)-like mechanism mediated by POLD3dependent DNA synthesis. Our findings revisit the functions of central HR factors in DNA replication and are crucial to understanding how targeting BIR-dependent pathways can modulate current chemotherapeutic modalities.
EXO1 contributes to fork resection in BRCA-deficient cells. Two distinct pathways act redundantly to mediate processive double-strand break (DSB) resection downstream from the MRE11-RAD50-NBS1 (MRN) and CtIP factors in eukaryotic cells: one requires DNA2 and the other EXO1 [17][18][19][20][21] . We sought to investigate whether DNA2 and EXO1 also contribute to the extended fork degradation phenotype of BRCA1-or BRCA2deficient cells following genotoxic stress induction. We knocked down EXO1 or DNA2 in different cell lines, including the BRCA2-mutant ovarian cancer cells PEO1 (and the isogenic PEO4 cells, which contain a second point mutation restoring BRCA2 function), the Fanconi anemia BRCA2-mutant line EUFA423 (and its derivative expressing wild-type BRCA2), the BRCA1 mutant ovarian cancer cell line UWB1.289 (and its complemented derivative expressing wild-type BRCA1), plus the human osteoscarcoma U2OS cells, which were siRNA-depleted for BRCA1 or BRCA2. Nucleolytic resection following replication fork stalling was monitored by pulse-labeling cells with the first thymidine analog IdU (red label) followed by treatment with hydroxyurea (HU) and concomitant labeling with the second thymidine analog, CldU (green label) (Fig. 1a). Shortening of the first tract was measured as a readout of degradation only on forks characterized by contiguous IdU-CldU signals (and not on forks that have only the IdU label) to ensure that the shortening phenotype is indeed due to nucleolytic resection of stalled replication forks and not to premature termination events 22 . Upon HU treatment, BRCA1-and BRCA2-deficient cells showed a marked reduction (30-50% corresponding to > 5 kb of DNA) in the IdU tract length (Fig. 1b, c and Supplementary Fig. 1a). Moreover, MRE11 inhibition or knockdown partially rescued fork degradation consistent with the previous data [5][6][7] (Fig. 1b, c and Supplementary Fig. 1). Analogous results were obtained using an alternative labeling scheme, where HU was added after thymidine labeling, suggesting that the results were not affected by the particular labeling scheme used in our work ( Supplementary  Fig. 1b). EXO1 knockdown by two different siRNAs yielded the same fork protection phenotype observed with MRE11 inhibition, indicating that EXO1 also contributes to fork degradation in BRCA1-and BRCA2-deficient cells upon HU treatment ( Fig. 1 and Supplementary Figs. 1a-e and 2a, b). Interestingly, combined ablation of MRE11 and EXO1 activities further rescued the fork degradation phenotype of BRCA-deficient cells, suggesting that the two nucleases may be able to act independently on the stalled replication intermediates. The same results were obtained by treating BRCA2-deficient cells with DNA damaging agents such as cisplatin or UV-C, supporting the notion that different genotoxic agents trigger a similar fork resection mechanism whereby MRE11 and EXO1 extensively degrade replication forks in the absence of key HR factors ( Supplementary Fig. 3a, b). Conversely, DNA2 knockdown did not restore fork protection (Supplementary Fig. 2c-g), in agreement with previous findings 23 . The same results were recapitulated by treating BRCA2-deficient cells with the NSC-105808 DNA2 inhibitor 24 (Supplementary Fig. 2e). Taken together, these results suggest that human EXO1, but not DNA2, contributes to replication fork degradation in BRCA1and BRCA2-deficient cells.
MRE11 and EXO1 target reversed forks in BRCA-deficient cells. Next, we sought to investigate the actual structure(s) of the replication intermediates targeted by MRE11 and EXO1 in BRCA1-or BRCA2-deficient cells. We visualized the fine architecture of the replication intermediates using a combination of in vivo psoralen cross-linking and EM 25 (Fig. 2a). Our analysis showed a substantial fraction of reversed replication forks (~25% of molecules analyzed) in control U2OS cells treated with 4 mM HU, consistent with previous data 14,26 (Fig. 2b). BRCA1-or BRCA2-knockdown resulted in a significantly lower frequency of fork reversal events (~10 and 11%, respectively) compared to HU-treated control cells, suggesting that BRCA proteins are required either to promote fork reversal or to prevent nucleolytic processing of forks that have already reversed following HU treatment. To distinguish between these two possibilities, we repeated the EM analysis while inhibiting MRE11 activity or knocking down EXO1 in BRCA-deficient cells. We found that ablation of either nuclease rescues the frequency of reversed forks to levels observed in control cells, suggesting that BRCA proteins protect forks that have already undergone fork reversal from nucleolytic degradation. Similar results were obtained by inhibiting MRE11 in BRCA2-mutant PEO1 cells (Fig. 2c). Timecourse experiments performed by treating BRCA2-deficient U2OS cells with HU for 0, 30, or 120 min indicated that fork reversal precedes fork degradation, as predicted by our model (Fig. 2d-f). Collectively, these results suggest that BRCA proteins protect reversed forks from nucleolytic degradation and that the unprotected reversed forks are the entry point for MRE11 and EXO1 in BRCA1-and BRCA2-deficient cells.
CtIP initiates MRE11-dependent resection of reversed forks. The regressed arm of a reversed replication fork resembles by all means a one-ended DSB. Previous biochemical studies showed that the 5′-3′ endonuclease activity of MRE11 initiates the resection process and that CtIP is required to promote MRE11 endonuclease activity at the 5′ strand 27 . Resection is then continued by the 5′-3′ exonuclease activity of EXO1 19,28,29 . As an alternative approach to confirm that MRE11-dependent resection starts from the regressed arm of reversed replication forks, we repeated the DNA fiber experiments with CtIP knockdown cells and found that CtIP loss prevents fork degradation in BRCA2deficient cells ( Supplementary Fig. 3c, d). These data suggest that the same pathway that initiates DSB resection in the context of HR is required to process the open dsDNA end of the regressed arm. To further validate this conclusion, we reasoned that degradation should not take place in a genetic background that prevents reversed fork formation-i.e., RAD51 knockdown 14 .
Indeed, loss of RAD51 suppressed fork degradation in BRCA2depleted U2OS cells, confirming that RAD51 acts upstream of BRCA2 to promote reversed fork formation (Fig. 3a, b). In addition, RAD51 depletion severely compromised fork restart, suggesting that fork remodeling is an essential requirement for efficient resumption of DNA synthesis upon HU removal (Figs. 3c, d).
Resection does not affect fork restart upon BRCA2 loss. The extended nucleolytic degradation of stalled replication intermediates in BRCA1-and BRCA2-deficient cells leads to increased chromosomal aberrations and genome instability 5, 6 . However, the molecular mechanism linking fork degradation with chromosomal instability remains elusive. Thereby, we set out to investigate how cells cope with these extensively degraded forks focusing on BRCA2-deficient cells. Fork degradation associated with BRCA2 loss did not significantly affect fork restart (Fig. 3e), in agreement with previous findings 5, 6 . However, by increasing the timing of CldU labeling from 15 to 90 min, we found that thymidine incorporation after HU removal was reduced in BRCA2-deficient compared to BRCA2-proficient cells (Fig. 3f). This reduction in the CldU tract length suggests that BRCA2 loss, though not affecting the number of restarting forks, either impairs fork progression after restart or causes a delayed restart of the resected forks. Ablation of MRE11 and/or EXO1 activity rescued CldU tract length, suggesting that the extended nascent strand degradation associated with BRCA2 loss is the leading cause of the observed defect in fork progression or timing of  (Fig. 3f). Moreover, BRCA2 loss led to increased chromosomal aberrations when cells were challenged with HU, and this effect was again partially rescued by ablation of MRE11 and/ or EXO1 activity, supporting a link between fork degradation, defects in fork progression or restart upon drug removal and increased chromosomal aberrations (Fig. 3g).
MUS81 cleaves partially resected reversed forks. We next examined the mechanism that rescues resected forks upon drug removal. Using neutral comet assays we found that BRCA2 loss leads to DSB accumulation upon HU treatment (Figs. 4a, b). However, the frequency of DSBs decreased to control levels 15 min after HU removal, suggesting that these DSBs are quickly repaired after fork restart. Interestingly, MUS81 depletion prevented DSB accumulation in BRCA2-deficient cells (Fig. 4c). A short-interfering RNA (siRNA)-resistant MUS81 cDNA, but not a catalytically inactive version (MUS81(D338A/D339A)), restored DSB accumulation in MUS81 siRNA-depleted cells ( Fig. 4c and Supplementary Fig. 3f). Therefore, we propose that MUS81 endonuclease activity is required to cleave partially resected forks, leading to transient DSB accumulation in the absence of BRCA2. Our observation that MUS81 loss did not significantly affect fork resection supports the idea that the extended fork degradation phenotype was not due to the resection of the DSBs created by MUS81 cleavage (Supplementary Fig. 3g). Moreover, the notion that MUS81 acts downstream of MRE11 and EXO1 was supported by the observation that the DSB accumulation of BRCA2deficent cells was significantly reduced in a genetic background that prevents fork resection-i.e., loss of MRE11, EXO1, or CtIP activity ( Fig. 4c and Supplementary Fig. 3e).
To define the exact substrate cleaved by MUS81 in BRCA2deficient cells, we inspected the structure of the replication intermediates that accumulate in BRCA2-deficient cells upon MUS81 depletion. Our EM analysis revealed that MUS81 loss causes a dramatic increase in reversed fork frequency in BRCA2deficient cells (Fig. 5a, b). Next, we evaluated the ssDNA composition of the regressed arms by detecting local differences in filament thickness. MUS81 depletion led to a significant increase in the percentage of regressed forks that are partially or entirely single-stranded, strongly suggesting that MUS81 cleaves partially resected regressed forks with a ssDNA tail (Fig. 5b). Interestingly, MUS81 depletion also led to a significant increase in the percentage of daughter strands with ssDNA at the fork junction, suggesting that in the absence of MUS81 cleavage asymmetric fork resection continues beyond the length of the regressed arm, leading to partially resected 3-way junction structures (Fig. 5c). Of note, the small fraction of HU-induced DSBs of BRCA2-proficient cells is also rescued by the loss of MUS81, but not of MRE11 or EXO1 (Supplementary Fig. 3h). This observation is consistent with the previous biochemical data showing that MUS81 can also cleave intact forks, although with much lower efficiency compared to flap-fork substrates 30 . Taken together, these results suggest that the initial resection of the regressed arms leads to the formation of reversed forks with a ssDNA flap, which are cleaved by MUS81.
To test whether MUS81-dependent cleavage of resected regressed forks is required for fork restart, we quantified the percentage of stalled forks in MUS81 and BRCA2 co-depleted cells (Fig. 5d). MUS81 loss slightly increased fork stalling in BRCA2-proficient cells, in agreement with previous findings 31, 32 . This effect was significantly more dramatic in the absence of BRCA2 and was confirmed using two different MUS81 siRNAs ( Fig. 5d and Supplementary Fig. 4a, b). Genetic knockdown-rescue experiments confirmed that complementation in MUS81-depleted U2OS cells with siRNA-resistant wild-type MUS81, but not with the catalytically inactive mutant, abrogated the effect of MUS81 depletion on replication fork stalling upon HU treatment (Fig. 5d).
MUS81 has two partners in human cells, EME1 and EME2, and the two proteins are thought to interact with MUS81 at different stages of the cell cycle. Genetic knockdown experiments showed that EME2, but not EME1, is involved in the same fork restart pathway, suggesting that the function of the MUS81-EME2 complex is restricted to the S-phase ( Supplementary  Fig. 4c, d), in agreement with previous findings 32 . MUS81 cleavage was previously shown to promote POLD3-dependent DNA synthesis at common fragile sites 33 and telomeric 34 loci in human cells. POLD3 is one of the accessory subunits of the replicative polymerase POL δ and was recently shown to be required for BIR, a specialized HR pathway to repair DSBs at stalled forks 35,36 . Similar to MUS81 depletion, POLD3 loss increased fork stalling in BRCA2-deficient cells ( Fig. 6a and Supplementary Fig. 4e). Moreover, it led to a severe defect in fork progression, suggesting that POLD3-dependent DNA synthesis is required to restart the MUS81-cleaved resected forks (Fig. 6b). Interestingly, POLD3/BRCA2 double depletion caused a further increase in DSB accumulation compared to depletion of BRCA2 alone and these DSBs were again rescued by MUS81 depletion confirming that POLD3 acts downstream of MUS81 in the pathway (Fig. 6c). The notion that BRCA2-deficient cells rely on the MUS81-dependent pathway to resume DNA synthesis is supported by our observation that cell viability is reduced in MUS81/BRCA2 co-depleted cells relative to BRCA2-depleted cells following HU treatment (Fig. 5e). Conversely, PARP inhibitor sensitivity was not significantly affected by MUS81 depletion consistent with the notion that PARP inhibitors, differently from HU, prevent reversed fork accumulation 10,13 and thereby do not require a MUS81-dependent pathway of fork rescue ( Supplementary Fig. 4f). Collectively, these results suggest that BRCA2-deficent cells rely on a MUS81/POLD3-dependent mechanism to rescue resected replication forks following treatment with genotoxic agents that induce replication fork reversal and degradation in a BRCA2-deficient background.
MUS81 does not rescue forks in BRCA1-deficient cells. Our data suggest that BRCA1 shares a function similar to BRCA2 in reversed fork protection (Fig. 1). However, the MUS81 pathway is not required to rescue forks in BRCA1-deficient cells (Supplementary Fig. 4g, h), suggesting that different pathways mediate the restart of the resected forks depending on the particular genetic background. These findings are in agreement with recent studies, suggesting that MUS81 depletion differentially affects chemotherapeutic sensitivity in BRCA2-versus BRCA1-deficient cells (Alan D'Andrea personal communication) and with our immunofluorescence experiments showing that MUS81 and POLD3 foci accumulate specifically in BRCA2-, but not in BRCA1-deficient cells (Fig. 6d).

Discussion
This work defines the mechanism that leads to the extensive fork degradation phenotype observed in BRCA2-deficient cells and provides novel insights into the molecular steps that rescue the resected forks upon drug removal. We propose a model whereby BRCA2 protects the regressed arms of replication forks that have reversed upon drug treatment from nucleolytic degradation. In its absence, the double-stranded ends formed by fork reversal are targeted by the CtIP, MRE11, and EXO1 nucleases to initiate the degradation of the stalled replication intermediates (Fig. 6e). Recent data suggest that RAD51 promotes reversed fork formation 14 and is enriched on nascent DNA independently of BRCA2 5 . We propose that RAD51 has two distinct functions during replication stress: a BRCA2-independent function in promoting the initial step of reversed fork formation and a BRCA2-dependent function, whereby BRCA proteins protect the already formed reversed forks from nucleolytic degradation by stabilizing the RAD51 filament on the regressed arm. In BRCA2deficient cells, this second function is lost, leading to the nascent strand degradation phenotype observed with BRCA2 mutants unable to stabilize RAD51 on ssDNA 6 or with RAD51 mutants that destabilize the RAD51 nucleofilament 37,38 . Because of the severe defect in fork restart associated with RAD51 depletion are reported here, we also suggest that fork remodeling is a central mechanism of replication stress response following prolonged drug treatment. Collectively, these findings shed light on the longdebated function of central HR factors in fork remodeling and go beyond the oversimplified concept that HR factors are simply required for DSB repair during replication stress. Our findings also reveal that MUS81-dependent cleavage of the resected forks is required for fork restart in BRCA2-deficient cells. The discovery that replication stalling induces RAD51 foci formation in a MUS81-independent manner is consistent with the idea that MUS81 is required to cleave forks that have already undergone RAD51-dependent fork remodeling 31 . We propose that MUS81 acts downstream of MRE11-and EXO1-mediated degradation. On the basis of our findings that CtIP is required to promote fork resection and MRE11 endonuclease activity at the 5′ strand, we also propose that the initial degradation of the regressed arms generates a reversed fork with a 3′-ssDNA tail that is then cleaved by MUS81 to mediate fork restart. This notion is consistent with   30,39,40 . In the absence of MUS81 cleavage, the nucleolytic degradation might quickly proceed to degrade nascent strands behind the junction finally leading to the extensively resected forks observed by DNA fiber ( Supplementary  Fig. 4i). Recent studies suggest that completion of DNA replication at common fragile sites 33 and telomeric 34 loci occurs via a specialized form of DNA repair originally characterized in yeast and termed BIR, whereby MUS81 cleavage of stalled replication forks produces a migrating bubble that drives POLD3-dependent DNA synthesis 41,42,43 . We propose that a similar mechanism is responsible to rescue partially resected regressed forks in BRCA2deficient cells. We speculate that the MUS81/POLD3-dependent pathway used to rescue resected forks in BRCA2-deficient cells might represent a novel anticancer drug target specific for BRCA2-defective tumors to be used in combination with chemotherapeutics that cause replication fork reversal and degradation.  46 . U2OS and EUFA 423 F/HAB2 45 cells were grown in DMEM media supplemented with 10% fetal bovine serum (FBS), 100 U ml −1 penicillin and 100 μg ml −1 streptomycin at 37°C in 5% CO 2 . Ovarian cancer PEO4/PEO1 44 were cultivated in RPMI media supplemented with 10% fetal bovine serum (FBS), 100 U ml −1 penicillin and 100 μg ml −1 streptomycin and UWB1/UWB1 + BRCA1 46 cells were grown in 50% RPMI media, 50% MEGM bullet kit (Lonza CC-3150) completed with 3% FBS, 100 U ml −1 penicillin and 100 μg ml −1 streptomycin at 37°C in 5% CO 2 .

Methods
Drug and reagents. The MRE11 inhibitor Mirin was from Sigma-Aldrich. The NSC-105808 DNA2 inhibitor was a gift from Gregorz Ira 24 . The DNA2 inhibitor was used at a concentration of 0.3 μM for 24 h. Hydroxyurea (Sigma-Aldrich) was dissolved in double-distilled H 2 O to obtain a 100 mM (7.6 mg ml −1 ) solution.
Cisplatin (Sigma-Aldrich) was dissolved in PBS 10× to obtain a 5 mM stock. UV-C was used at 40 mJ cm −2 as described in the labeling scheme. The MUS81 vectors used for the genetic complementation experiments was a gift from by Ian Hickson 33 . All vectors were amplified in DH5α Escherichia coli and extracted with GeneJET Plasmid Midiprep Kit (ThermoFisher Scientific).
RNA interference. All transient gene depletions were carried out using the Lipofectamine RNAiMax transfection reagent (Life Technologies), except for EXO1 gene silencing that was performed using TransIT-siQuest (Mirus). SMARTpool siRNA from Dharmacon were employed to deplete the following genes: BRCA2 (L-003462-00, 10 nM, 24 or 48 h) as described 47 , BRCA1 (L-003461-00, 50 nM, 48 h) as described 48 , MRE11A (L-009271-00, 40 nM, 48 h) as described 49    DNA fiber assay. Briefly, asynchronously growing cells were labeled with two thymidine analogs: 20 μM 5-iodo-2′-deoxyuridine (IdU; Sigma-Aldrich) followed by 200 μM 5-chloro-2′-deoxyuridine (CldU; Sigma-Aldrich) for the indicated times 13,51 . Cells were washed twice with PBS after the first pulse and treated with the indicated doses of the genotoxic agents. After the indicated times, cells were collected and resuspended in PBS at 100,000 cells per ml. A total of 2 μl of this cell solution was mixed with 8 μl of lysis buffer (200 mM Tris.HCl pH 7.5; 50 mM EDTA; 0.5 % SDS) on a glass slide. After 6 min, the slides were tilted at a 20-45°a ngle, and the resulting DNA spreads were air dried, fixed in 3:1 methanol/acetic acid and stored at 4°C. The DNA fibers were denatured with 2.5 M HCl for 1 h, washed with PBS, and blocked with 5% BSA in PBS/ 0.1% Tween 20 for 1 h. DNA immunostaining was performed with rat anti-BrdU antibody (1:50; AbCys SA, ABC117 7513) for CldU and mouse anti-BrdU antibody (1:50; Becton Dickson, 347580) for IdU in a humid chamber at 37°C for 1 h. The following secondary antibodies were used: anti-rat Alexa 488 (1:100; Molecular Probes, A21470) and anti-mouse Alexa 546 (1:100; Molecular Probes, A21123) at 37°C for 45 min. The slides were air dried and mounted with Prolong Gold Antifade reagent (Invitrogen). Images were sequentially acquired (for double-label) with LAS AF software using TCS SP5 confocal microscope (Leica). A 63×/1.4 oil immersion objective was used. Images were captured at room temperature. The DNA tract lengths were measured using ImageJ and the pixel length values were converted into micrometers using the scale bars created by the microscope. n ≥ 150 fiber tracts scored for each data set. The statistics for all these experiments measuring changes in the size of the IdU or CldU tracts were calculated on the total number of DNA tracts measured in each given sample (usually n ≥ 250). For the fork restart experiments, the percentage of stalled forks was calculated on the basis of at least 250 tracts counted in each independent experiment. All DNA fiber experiments were performed in duplicate or triplicate, as indicated in the figure legends. Additional information on the minimal number of tracts that should be measured for a reliable estimation of changes in fork speed within a given sample can be found in refs. 26,52 .
Electron microscopy. For the EM analysis of replication intermediates, 5-10 × 10 6 U2OS or PEO1/4 cells were collected and genomic DNA was cross-linked by two rounds of incubation in 10 μg ml −1 4,5′,8-trimethylpsoralen (Sigma-Aldrich) and 3 min of irradiation with 366 nm UV light on a precooled metal block 10,26 . Cells were lysed and genomic DNA was isolated from the nuclei by proteinase K (Roche) digestion and phenol-chloroform extraction. DNA was purified by isopropanol precipitation, digested with PvuII HF in the proper buffer for 3-5 h at 37°C and replication intermediates were enriched on a benzoylated naphthoylated DEAE-cellulose (Sigma-Aldrich) column. EM samples were prepared by spreading the DNA on carbon-coated grids in the presence of benzyl-dimethylalkylammonium chloride and visualized by platinum rotary shadowing. Images were acquired on a transmission electron microscope (JEOL 1200 EX) with sidemounted camera (AMTXR41 supported by AMT software v601) and analyzed with ImageJ (National Institutes of Health). EM analysis allows distinguishing duplex DNA-which is expected to appear as a 10 nm thick fiber, after the platimun/ carbon coating step necessary for EM visualization-from ssDNA, which has a reduced thickness of 5-7 nm. The criteria used for the unequivocal assignment of reversed forks include the presence of a rhomboid structure at the junction itself in order to provide a clear indication that the junction is opened up and that the fourway junction structure is not simply the result of the occasional crossing of two DNA molecules 25 . In addition, the length of the two arms corresponding to the newly replicated duplex should be equal (b = c), whereas the length of the parental arm and the regressed arm can vary (a ≠ b = c ≠ d). Conversely, canonical Holliday junction structures will be characterized by arms of equal length two by two (a = b, c = d).
Metaphase spreads. Cells were treated with 4 mM HU for 6 h, washed twice with PBS, and released for 24 h in fresh medium. During the last 4 h, 10 μM nocodazole was added to the medium. Cells were collected, washed and resuspended in 10 ml of warmed hypotonic solution (10 mM KCl, 10% FBS) for 10 min at 37°C. Cells were fixed by adding 500 μl of cold fixation buffer (acetic acid 1: 3 ethanol). Cell pellets were washed four times with the cold fixation buffer and stored in this buffer at 4°C overnight. The nuclei were spread on cold slides. The slides were air dried overnight and mounted with Prolong Gold Antifade (Invitrogen) with DAPI. Images were acquired with a fluorescence microscope (LEICA DMU 4000B; 63×/ 1.40-0.60 NA oil) coupled to the LEICA DFC345FX camera. The images were analyzed with ImageJ. At least 50 metaphases per sample were scored in each experiment.
Neutral comet assay for DSB detection. A total of 700 cells were resuspended in 70 μl 0.5% low melting point agarose (Trevigen, 4250-050-02) and spread on a comet slide (Trevigen, 4250-200-03). Cells were lysed in a cold lysis solution (Trevigen, 4250-050-01) at 4°C for 30 min. DNA migration was performed in TBE buffer at 1 V cm −1 for 30 min. Slides were washed in milliQ water, fixed with ethanol 70% for 30 min and dried at room temperature. Comets were labeled with SYBR ® Gold Nucleic Acid Gel Stain (ThermoFisher) for 30 min. Images were acquired with a fluorescence microscope (LEICA DMU 4000B; 20×/0.4 CORR) coupled to the LEICA DFC345FX camera. The images were analyzed using ImageJ. At least 150 comets were scored per sample in each experiment.
Cell viability. Cell viability was determined using the Cell Proliferation Kit II (XTT, Roche) 53 . Briefly, cells were seeded at 13,000 cells per well in a 24-well plate the day prior to treatment. Cells were treated chronically with the indicated doses of HU and cell viability was assessed 3 days after. The absorbance was measured at 450 nm with a reference wavelength at 650 nm. Results were expressed as percentage of the untreated control.
Immunofluorescence microscopy. After treatment with 4 mM HU, U2OS cells were extracted with 1× PBS containing 0.2% Triton X-100 and protease inhibitors (Pierce) for 5 min on ice prior to fixation with 3.2% paraformaldehyde. The cells were then washed extensively with IF wash buffer (1× PBS, 0.5% NP-40, and 0.02% NaN 3 ), then blocked with IF Blocking Buffer (IF wash buffer plus 10% FBS) for at least 30 min. Anti-MUS81 mouse monoclonal (1:200; ab14387 Abcam) or anti-POLD3 mouse monoclonal (1:350; H00010714-M01 Abnova) antibodies were diluted in IF Blocking Buffer overnight at 4°C. After staining with secondary antibodies (1:1000, conjugated with Alexa Fluor 594; A11032 Life Technologies) and Hoechst 33342 (1:1000, Sigma-Aldrich), samples were mounted using Prolong Gold mounting medium (Invitrogen). Epifluorescent microscopy was performed on an Olympus fluorescent microscope (BX-53) using a UPlanS-Apo 100×/1.4 oil immersion lens, Hamamatsu ORCA-Flash 4.0LT digital camera, and cell-Sens Dimension software. The raw images were exported into Adobe Photoshop, and for any adjustments in image contrast or brightness, the levels function was applied as previously described 54 . For foci quantitation, at least 150 cells were analyzed in triplicate.
Statistical analysis. Statistical analysis was performed using Prism (GraphPad Software). The statistical significance in each case was calculated as indicated in each figure legend.