Progression through mitosis promotes PARP inhibitor-induced cytotoxicity in homologous recombination-deficient cancer cells

Mutations in homologous recombination (HR) genes BRCA1 and BRCA2 predispose to tumorigenesis. HR-deficient cancers are hypersensitive to Poly (ADP ribose)-polymerase (PARP) inhibitors, but can acquire resistance and relapse. Mechanistic understanding how PARP inhibition induces cytotoxicity in HR-deficient cancer cells is incomplete. Here we find PARP inhibition to compromise replication fork stability in HR-deficient cancer cells, leading to mitotic DNA damage and consequent chromatin bridges and lagging chromosomes in anaphase, frequently leading to cytokinesis failure, multinucleation and cell death. PARP-inhibitor-induced multinucleated cells fail clonogenic outgrowth, and high percentages of multinucleated cells are found in vivo in remnants of PARP inhibitor-treated Brca2−/−;p53−/− and Brca1−/−;p53−/− mammary mouse tumours, suggesting that mitotic progression promotes PARP-inhibitor-induced cell death. Indeed, enforced mitotic bypass through EMI1 depletion abrogates PARP-inhibitor-induced cytotoxicity. These findings provide insight into the cytotoxic effects of PARP inhibition, and point at combination therapies to potentiate PARP inhibitor treatment of HR-deficient tumours.

B RCA1 and BRCA2 function in the repair of DNA doublestrand breaks (DSBs) through homologous recombination (HR), and ensure the protection of stalled replication forks 1 . Defective HR is thought to underlie the progressive accumulation of genomic aberrations, leading to malignant transformation. Indeed, mutations in BRCA1 and BRCA2 predispose to tumorigenesis, most frequently involving breast and ovarian cancer [2][3][4] . Due to their DNA repair defect, BRCA1/2 mutant cancer cells are more sensitive to platinum-based chemotherapeutics, as observed in preclinical models and in clinical studies [5][6][7] . In addition, BRCA1/2 mutant cancers were found to be selectively sensitive to inhibition of the poly-(ADP)ribose polymerase PARP1 (refs 7-9). Unfortunately, however, BRCA1/2 mutant cancers can acquire resistance and relapse 10 .
Mechanistically, PARP1 promotes the repair of non-toxic single-strand DNA breaks 11 , which are converted into potentially toxic DSBs during S-phase 8,9 . These DSBs depend on HR for repair, and hence were suggested to cause cell death in HR-defective cancer cells. However, the number of singlestrand DNA breaks were not found to be increased after PARP1 depletion or PARP inhibition [11][12][13] , and the synthetic lethal interaction between PARP inhibition and HR deficiency may therefore involve other mechanisms 14,15 . Indeed, PARP1 and BRCA1/2 were shown to orchestrate the protection and restart of stalled replication forks [16][17][18][19][20] . Analogously, PARP1 activity increases during replication 21 , and sensitivity to PARP inhibition in BRCA2 mutant cancer cells can be rescued by mutations that prevent replication fork degradation 22 .
Here, we study the mechanisms by which PARP-inhibitorinduced DNA lesions affect mitotic progression. We describe that PARP inhibition compromises replication fork stability and leads to DNA lesions that are transmitted into mitosis. During mitosis, these DNA lesions cause chromatin bridges and lead to cytokinesis failure, multinucleation and cell death. Importantly, our data show that progression through mitosis promotes PARPinhibitor-induced cell death, since forced mitotic bypass. abrogates PARP-inhibitor-induced cytotoxicity.

PARP-inhibitor-induced lesions are transmitted into mitosis.
To explore the consequences of PARP inhibition on mitotic progression in HR-defective cancer cells, we depleted BRCA2 in HeLa cells (Fig. 1a). As expected, treatment with the PARP inhibitor olaparib resulted in selective killing of BRCA2-depleted cells (Fig. 1b). In line with roles for BRCA2 and PARP in facilitating replication fork stability 22 , we observed compromised replication fork protection using single DNA fibre analysis upon BRCA2 depletion, which was aggravated upon PARP inhibition (Fig. 1c,d). These findings show that PARP inhibition in BRCA2deficient cancer cells incrementally interferes with replication fork stability. In line with previous studies showing involvement of Mre11 and PTIP in degradation of stalled replication fork in BRCA2-deficient cells, Mre11 inhibition using mirin or PTIP depletion alleviated the fork protection defects ( Supplementary  Fig. 1A,B) 20,22 .
Defective replication fork stability upon PARP inhibition was further underscored by the increase in FANCD2 foci in interphase cells upon BRCA2 depletion. A significant further increase was observed when BRCA2-depleted cells were treated with PARP inhibitor (Fig. 1e). Surprisingly, the increase in FANCD2 foci was only accompanied by minor increases in the numbers of g-H2AX foci in interphase, suggesting that replication lesions do not per se result in DNA breaks ( Supplementary  Fig. 1C).
The observed replication lesions were not resolved before mitotic entry, as increased numbers of FANCD2 foci were observed in BRCA2-depleted mitotic cells (Fig. 1f). Again, the numbers of FANCD2 foci increased further upon PARP inhibitor treatment (Fig. 1f). Of note, in PARP inhibitor-treated, BRCA2-depleted mitotic cells, numbers of g-H2AX foci were increased similarly to FANCD2 foci ( Supplementary Fig. 1D). Combined, these data show that PARP inhibition in BRCA2defective cells leads to replication intermediates that are transmitted into mitosis.
PARP inhibition causes mitotic chromatin bridges. Since persistence of unresolved replication intermediates into mitosis may interfere with proper chromosome segregation, we tested whether PARP inhibition-induced mitotic aberrancies. Whereas PARP inhibition did not affect the percentages of anaphase or telophase cells containing chromatin bridges in control cells, depletion of BRCA2 in HeLa cells led to an increased percentage of cells showing chromatin bridge formation, which remained unresolved up until telophase (14 and 17% in BRCA2-depleted cells versus 2% in control-depleted cells) (Fig. 2a,b), in line with previous observations 31 . Strikingly, the number of BRCA2depleted cells containing anaphase chromatin bridges markedly increased upon olaparib treatment (59 and 65% in BRCA2depleted, olaparib-treated cells versus 20% in control-depleted cells treated with olaparib) (Fig. 2a,b). Interestingly, whereas most of the olaparib-induced chromatin bridges were resolved before telophase in control cells, chromatin bridges persisted throughout mitosis in BRCA2-depleted cells (49 and 57% in BRCA2-depleted cells versus 6% in control-depleted cells) (Fig. 2b). Furthermore, PARP inhibition also markedly increased the numbers of lagging chromosomes in BRCA2-depleted cells (58 and 53% in BRCA2depleted cells versus 8% in control-depleted cells) (Fig. 2b, right panel).
These observations were not specific for BRCA2, as depletion of BRCA1 (Supplementary Fig. 2A,B) or RAD51 ( Supplementary  Fig. 2C,D) also showed a clear induction of PARP-inhibitorinduced chromatin bridges persisting throughout mitosis. Notably, BRCA1 or RAD51 depletion in HeLa cells also increased the amount of lagging chromosomes upon PARP inhibition, although to a lesser extent in BRCA1-depleted cells when compared to RAD51 or BRCA2-depleted cells (Fig. 2b, right panel and Supplementary Fig. 2B,D, right panels). Similar results were obtained upon depletion of BRCA1 or BRCA2 in BT-549 breast cancer cells treated with olaparib. Specifically, PARP inhibitor sensitivity greatly increased upon doxycycline-induced shRNAs depletion of BRCA1 or BRCA2 ( Supplementary Fig. 2E,F). Importantly, the numbers of unresolved chromosome bridges (Fig. 2c), and lagging chromosomes ( Supplementary Fig. 2G) increased significantly in BRCA1/2-depleted BT-549 cells upon olaparib treatment.
We next investigated whether permanent genetic inactivation of BRCA1 shows a similar increase in the amount of mitotic chromatin bridges, when compared to acute siRNA-mediated BRCA2/BRCA1 inactivation. Indeed, the human HCC1937 breast cancer cell line, harbouring a BRCA1 deletion as well as a hypomorphic BRCA1 allele with a 5382insC frameshift mutation, showed increased chromatin bridges in anaphase and telophase upon olaparib treatment (Fig. 2d), as well as lagging chromosomes ( Supplementary Fig. 2H). To validate these results using isogenic models, we next used a tumour cell line derived from a K14cre;Brca2 del/del ;p53 del/del mouse mammary tumour (denoted as Brca2 À / À ) 32 . As a control, we used an isogenic cell line in which BRCA2 was reconstituted using an infectious bacterial artificial chromosome (iBAC), harbouring the mouse Brca2 gene (denoted as Brca2 iBAC ) 33     In accordance with our observations after transient BRCA2 depletion, PARP inhibition in Brca2 À / À cells resulted in an increased percentage of cells harbouring chromatin bridges in anaphase (71% in Brca2 À / À cells versus 17% in Brca2 iBAC cells), which to a large degree remained unresolved until telophase (42% in Brca2 À / À cells versus 5% in Brca2 iBAC cells) (Fig. 2e, left panel). Again, also the percentage of cells with lagging chromosomes was increased (39% versus 8% in Brca2 À / À and Brca2 iBAC cells, respectively) ( Fig. 2e, right panel). These observations again likely reflected generic consequences of defective HR, as very similar defects were observed in a tumour cell line derived from K14cre;Brca1 F5-13/F5-13 ;p53 F2-10/F2-10 mice ( Supplementary Fig. 3C). These observations were further validated in DLD-1 human colorectal adenocarcinoma cells, in which BRCA2 was inactivated using CRISPR-Cas9. In complete accordance to what was observed in mouse Brca2-null cells, BRCA2 À / À DLD-1 cells were sensitive to PARP inhibition ( Supplementary Fig. 3D), and PARP inhibition greatly enhanced the formation of chromatin bridges as well as lagging chromosomes in BRCA2 À / À , but not in BRCA2 þ / þ DLD-1 cells ( Supplementary Fig. 3E). Combined, these findings indicate that PARP inhibition induces mitotic defects when HR is inactivated acutely or permanently, in a species-independent fashion.
Since the formation of mitotic chromatin bridges and lagging chromosomes in HR-deficient cancer cells are a likely consequence of disturbed replication fork integrity, we tested the involvement of Mre11 and PTIP. Mre11 inhibition or PTIP depletion, which alleviated the PARP-inhibitor-induced replication fork instability ( Supplementary Fig. 1A,B), also reduced the amounts of mitotic chromatin bridges (Fig. 2f,g). These findings further corroborate that aberrant control of replication fork stability underlies PARP-inhibitor-induced chromosome bridge formation in mitosis. Notably, the amount of lagging chromosomes was not reduced upon Mre11 or PTIP inactivation, suggesting different biological origins of these lesion (Supplementary Fig. 3F,G).
PARP trapping is required for chromatin bridge formation. Cytotoxicity of PARP inhibitors was previously associated with the ability of PARP inhibitor to trap PARP onto DNA, rather than its effects on PARP catalytic activity 15 . To test whether PARP trapping is required for the observed mitotic defects, we depleted PARP1 using siRNA (Fig. 3a). In contrast to treatment with olaparib, a PARP inhibitor which has trapping activity, depletion of PARP1 did not significantly induce mitotic chromatin bridges, nor lagging chromosomes in BRCA2-depleted cells ( Fig. 3b and Supplementary Fig. 4A). Also, very similar levels of mitotic chromatin bridges and lagging chromosomes were observed in response to the structurally unrelated PARP inhibitor AZD2461 (Fig. 3c,d and Supplementary Fig. 4B). Of note, the observed increase in chromosome bridges in PARP inhibitortreated cells was much more pronounced when compared to cisplatin treatment, at a dose that efficiently caused DNA breaks as judged by g-H2AX (Fig. 3c,d).
Since PARP is involved in multiple cellular processes, we next tested whether the PARP-inhibitor-induced mitotic defects required treatment during S-phase. To this end, cells were synchronized at the G1/S-phase of the cell cycle using a doublethymidine block (Fig. 3e). Cells treated during S-phase displayed significantly increased numbers of mitotic cells with FANCD2 foci (Fig. 3f), chromosome bridges as well as lagging chromosomes (Fig. 3g). In line with expectations, BRCA2-depleted cells with chromosome bridges also contained higher numbers of mitotic FANCD2 foci, when compared to cells without chromosome bridges ( Supplementary Fig. 4C). Importantly, when cells were treated with olaparib past S-phase (at 7 h after release from thymidine block), the number of FANCD2 foci in mitotic cells (Fig. 3f), chromosome bridges as well as lagging chromosomes ( Fig. 3g) was significantly reduced. Taken together, these data show that PARP trapping during S-phase is required for induction of mitotic chromosome bridges.
Chromatin bridges cause multinucleation and cell death. Unresolved chromatin bridges can cause genomic aberrations, multinucleation and cell death. To investigate the consequences of PARP-inhibitor-induced chromatin bridges for HR-deficient cells, live-cell imaging was used in combination with stable expression of fluorescently tagged Histone-H2B to visualize chromosome dynamics (Fig. 4a). Although some chromatin bridge events were observed in DMSO-treated control cells, the majority of mitoses proceeded either without any visible chromatin bridges (cells with chromatin bridge: 24%) or with chromatin bridges that were resolved during the course of mitosis (6%) (Fig. 4b). Very comparable results were found for controldepleted cells treated with PARP inhibitor (chromatin bridge: 32%; resolved bridge: 14%), or BRCA2-depleted cells treated with DMSO (chromatin bridge: 26%; resolved bridge: 12%) (Fig. 4b).
In stark contrast, the amount of aberrant mitoses was strongly increased in BRCA2-depleted cells treated with PARP inhibitor (chromatin bridge: 64%; resolved bridge: 13%) ( Fig. 4b and Supplementary Fig. 5A). Furthermore, in a large fraction of BRCA2-depleted cells, unresolved mitotic chromatin bridges resulted in failed cytokinesis leading to multinucleation (29% of total mitoses), or were followed by cell death, (22% of total mitoses), indicating that mitotic failure following PARP inhibitor treatment is often detrimental for BRCA2-deficient cells ( Fig. 4b and Supplementary Fig. 5A).
Similar phenotypes were observed in Brca2-null mouse mammary tumour cells, expressing GFP-tagged Histone-H2B and mCherry-tagged a-Tubulin (Fig. 4c). Again, PARP inhibitor treatment of Brca2-null mouse cells greatly enhanced the number of cells with chromatin bridges (82% versus 22% in Brca2 À / À and Brca2 iBAC , respectively, Fig. 4c). Furthermore, most of the chromatin bridges in Brca2 À / À cells were not resolved during mitosis and lead to cytokinesis failure, accompanied with multinucleation (32% versus 5% in Brca2 À / À and Brca2 iBAC , respectively), or cell death (16% versus 1% in Brca2 À / À and Brca2 iBAC , respectively) (Fig. 4c). These effects were not caused by expression of GFP-HistoneH2B or mCherry-a-Tubulin, as similar amounts of cells with 44n DNA content were observed using flow cytometry in BRCA2-depleted HeLa cells or Brca1/2null cells lacking these reporters ( Fig. 4d-f, Supplementary  Fig. 5B-D). In conclusion, PARP inhibition in cells with inactivated BRCA2 leads to chromatin bridges, which frequently remain unresolved and are associated with multinucleation and cell death.
As PARP inhibitor treatment resulted in a large fraction of multinucleated cells upon cytokinesis failure, we wondered to what extent these cells contribute to clonogenic survival. We therefore sorted BRCA2-depleted cells based on DNA content and separately plated cells with 2n, 4n and 44n DNA content ( Fig. 5a and Supplementary Fig. 6A-C). As expected, DMSOtreated, BRCA2-depleted HeLa cells with either 2n or 4n DNA content resulted in efficient clonogenic outgrowth (Fig. 5b). Similarly, 2n or 4n Brca2 À / À cells showed comparable numbers of colonies (Fig. 5c). Notably, DMSO-treated cells with 44n DNA content showed a B50% decrease in clonogenic potential, indicating that multinucleation reduces viability, but does not per se preclude long-term survival of tumour cells, in line with previous reports 34, 35 . Indeed, multinucleated cells did not display an intrinsic inability to replicate, as judged by BrdU incorporation (Supplementary Fig. 6D,E). Importantly, upon PARP inhibition, clonogenic survival was markedly decreased in both BRCA2-depleted HeLa cells as well as Brca2 À / À cells, with 4n DNA-containing cells consistently showing a more pronounced decrease in clonogenic outgrowth (Fig. 5b,c). Notably, cells with 44n DNA content showed a near-complete loss of colony formation, showing that PARP-induced multinucleation precludes long-term viability in cells lacking functional BRCA2.  Cisplatin Olaparib --+ --- ---- To test whether these observations could be extrapolated in vivo, we analysed Brca2 À / À ;p53 À / À mammary tumours, generated in K14cre;Brca2 F/F ;p53 F/F mice, orthotopically transplanted into syngeneic wild-type (wt) mice, and treated with vehicle or olaparib for 28 days (Fig. 5d). Notably, remnants of olaparib-treated Brca2 À / À ;p53 À / À tumours showed significantly increased numbers of multinucleated cells (Fig. 5d,e). To test if this phenotype is generic for HR-deficient tumours, we also analysed Brca1 À / À ;p53 À / À mammary tumours, derived from K14cre;Brca1 F/F ;p53 F/F mice (Fig. 5f). Again, tumour remnants of olaparib-treated Brca1 À / À ;p53 À / À tumours showed significantly increased numbers of multinucleated cells (Fig. 5f,g). Taken together, these data confirm PARP-inhibitorinduced multinucleation in BRCA1-or BRCA2-deficient tumour cells, and suggest that failed mitosis may contribute to the cytotoxicity of PARP inhibition in these cancer cells.
Mitotic progression promotes PARP-inhibitor cytotoxicity. We next investigated to what extent the progression through mitosis contributes to the cytotoxic effects of PARP inhibitor treatment in these tumour cells. To address this, we aimed to prevent progression through mitosis, while still allowing for DNA replication. To this end, we inactivated the early mitotic inhibitor-1 (EMI1). During interphase, EMI1 keeps the APC/C E3 ligase inactive, and thereby allows for the accumulation of numerous mitotic regulators, including B-type cyclins 36 . Inactivation of EMI1 leads to premature APC/C activation in G2-phase, interferes with cyclin B accumulation and consequently precludes mitotic entry. As a result, EMI1-depleted cells bypass mitosis and enter cycles of endoreplication 37 . Indeed, following EMI1 depletion, HeLa cells were unable to enter mitosis, yet continued DNA replication as judged by a large population of endoreplicating cells (44n DNA content) by flow cytometry (Fig. 6a,b). Subsequently, HeLa cells were depleted for BRCA2 either alone or in combination with EMI1, and induction of apoptosis was analysed by annexin-V staining (Fig. 6c). Whereas olaparib treatment resulted in clear induction of apoptosis in BRCA2-depleted cells, co-depletion of EMI1 rescued the induction of apoptosis (Fig. 6c,d). Next, we assessed whether these effects translated into altered short-term cell survival. Interestingly, olaparib sensitivity of BRCA2-depleted cells was largely nullified by concomitant EMI1 depletion (Fig. 6e). These observations were confirmed in BRCA2 À / À DLD-1 cells, in which PARP inhibitor sensitivity was rescued when mitosis was bypassed due to EMI1 depletion (Fig. 6f,g). Importantly, EMI1 depletion did not alleviate induction of DNA lesions in response to PARP inhibition in BRCA2-depleted cells, as judged by foci analysis of FANCD2 (Supplementary Fig. 7A) and g-H2AX ( Supplementary Fig. 7B). Also, loss of PARP inhibitor sensitivity could not be attributed to decreased proliferation rates, since EdU incorporation was not impaired after BRCA2 and EMI1 co-depletion, when compared to depletion of BRCA2 alone. (Supplementary Fig. 7C). Taken together, our data show that forced bypass of mitosis results in decreased PARP inhibitioninduced cytotoxicity, and indicate that progression through mitosis promotes cell death in BRCA2-deficient cancer cells treated with PARP inhibitor.

Discussion
Our findings on PARP-inhibitor-induced cytotoxicity in HR-deficient cancer cells extend on recent findings that mitotic processing of DNA lesions is linked to genome stability 24,25,[38][39][40] . Further, our data challenge the dogma that accumulation of DNA DSBs due to combined loss of base excision repair and HR is the main contributor to synthetic lethality. While DSBs do occur, other aberrant replication intermediates also arise during replication in HR-deficient cells treated with PARP inhibitors [23][24][25][26][27] . Furthermore, our data show that these lesions do not immediately lead to cell death, but can be transferred into mitosis, resulting in chromatin bridging and subsequent cytotoxicity. Inactivation of HR components BRCA1/2 or RAD51 (ref. 31), or Fanconi Anemia components 30 , has been previously linked to mitotic defects. Loss of either BRCA1/2 or RAD51 was found to increase the percentage of cells with ultrafine anaphase bridges as well as bulky chromatin bridges during anaphase 31 , in agreement with our data. Previously, PARP inhibition was shown to promote mitotic aberrancies and multinucleation 41 . Similar to our data, these reports showed that PARP inhibition alone does not appear to induce severe effects on mitosis in wt cells. Rather, our data indicate that severe mitotic defects only arise when PARP inhibitors are combined with a HR defect.
Chromatin bridges were previously described to frequently arise as a consequence of unresolved replication lesions 26 . Since PARP inhibition in HR-deficient cells also leads to replication fork instability, it is conceivable that unresolved replication intermediates underlie the formation of chromatin bridges in HR-deficient cells upon PARP inhibition. Although the exact nature of these lesions remains obscure, the observed FANCD2positive foci in mitosis suggest that under-replicated regions may persist after replication fork stalling. Indeed, secondary mutations that rescue replication fork stability in BRCA2-deficient cancer cells, render them resistant to PARP inhibition 22 .
Late-stage replication intermediates that persist up until mitosis are normally cleared by DNA resolvases 40,42 . Specifically, the structure-specific endonuclease complex MUS81-EME1 operates in conjunction with SLX4 and GEN1 to resolve DNA joint molecules, and these enzymes are known to be highly active during mitosis and are required for proper chromosome segregation 39,43 . One could therefore speculate Interestingly, forced mitotic bypass through EMI1 depletion could largely rescue viability of HR-deficient cells upon PARP inhibition. This implies that progression through mitosis facilitates PARP-induced cytotoxicity, at least in short-term assays. Since EMI1 is an essential gene in vivo 44 and is also required for long-term growth in vitro 45-47 , we do not consider EMI1 downregulation as a clinically relevant means to achieve long-term PARP inhibitor resistance. Rather, EMI1 served as a tool to bypass mitosis without impairing replication. In line with this notion, RNA sequencing analysis of Brca2 mutant cancers that were either sensitive or resistant to PARP inhibition did not provide evidence that EMI1 loss is involved in PARP inhibitor resistance ( Supplementary Fig. 7D).
In addition, our results suggest that PARP-inhibitor-induced cytotoxicity requires cycles of both replication and mitosis, and that tumour cells that remain in G1-or G2-phase longer are more resistant to PARP-inhibitor-induced cytotoxicity. Conversely, drugs that promote mitotic entry, such as WEE1 or DDR kinase inhibitors, may potentiate PARP inhibitor treatment. Preliminary evidence indeed shows additive effects of combined inhibition of PARP and WEE1 in BRCA2-deficient cells, which warrants further investigation (Supplementary Fig. 8).
Alternatively, targeting the mitotic spindle using tubulin poisons could be an interesting approach to potentiate PARP inhibitor treatment in HR-deficient cells. Further characterization   of the nature of the DNA lesions that underlie mitotic chromatin bridges and the pathways that respond to these structures is required to elucidate how PARP inhibitor therapy functions at the molecular and cellular level. These insights could then aid in designing rational combination therapies to potentiate PARP inhibitor treatment.
The KP3.33 cell line (denoted in the manuscript text as p53 À / À ) was established from a mammary tumour from K14cre;p53 F2-10/F2-10 mice 32 . The KB2P1.21R1 cell line (denoted as Brca2 iBAC ) was generated through stable introduction of an iBAC containing the entire mouse Brca2 gene into the KB2P1.21 cell line 32 . All mouse cell lines were cultured in DMEM/F-12 medium, supplemented with 10% FCS, 50 units per ml penicillin, 50 mg ml À 1 streptomycin, 5 mg ml À 1 insulin (Sigma), 5 ng ml À 1 epidermal growth factor (Life Technologies) and 5 ng ml À 1 cholera toxin (Gentaur), at 37°C under hypoxic conditions (1% O 2 , 5% CO 2 ). Cell cycle synchronization was achieved using a double-thymidine block. Specifically, cells were treated with thymidine (2 mM) for 17 h, washed twice with PBS and were incubated in warm medium for 9 h. Subsequently, cells were again incubated in thymidine for 17 h, washed with PBS and released in pre-warmed medium and collected at indicated time points. For treatment of cells during S-phase, cells were treated immediately following release from thymidine. For treatment after S-phase, cells were treated at 7 h after release from thymidine.
MTT assays. HeLa, DLD-1, BT-549, HCC1937, KB2P1.21 and KB2P1.21R1 tumour cell lines were plated in 96-wells plates. BT-549 cells were pre-treated with 1 mg ml À 1 doxycycline for 48 h. HeLa were plated at 2,000 cells per well, DLD-1 cells at 5,000 cells per well, BT-549 and HCC1937 cells at 1,000 cells per well, and KB2P1.21 and KB2P1.21R1 were plated at 1,200 cells per well. Cells were allowed to attach for 3 or 24 h and were treated with indicated concentrations of olaparib, MK-1775, or MK4827 (all from Axon Medchem, Groningen, the Netherlands) for 3 or 4 days. Methyl-thiazol tetrazolium (MTT) was added to cells at a concentration of 5 mg ml À 1 for 4 h, after which culture medium was removed and formazan crystals were dissolved in DMSO. Absorbance values were determined using a Bio-Rad benchmark III Biorad microtiter spectrophotometer at a wavelength of 520 nm. Proliferation was determined as the relative decrease in signal compared to DMSO-treated cells. Unless mentioned otherwise, statistical significance was tested using Student's t-test.
Western blotting. Cells were lysed using mammalian protein extraction reagent (Thermo Scientific), supplemented with protease inhibitor and phosphatase inhibitor cocktail (Thermo Scientific). Protein content was determined with a Bradford assay after which 20 mg of protein sample was separated by sodium dodecyl sulfate (SDS)/PAGE, transferred to polyvinylidene fluoride (immobilon) membranes and blocked in 5% skimmed milk (Sigma) in Tris-buffered saline (TBS) containing 0.05% Tween20 (Sigma).
Flow cytometry. For apoptosis analysis by annexin V staining, total cell populations were collected by trypsinization and stained with annexin-V-FITC (1:20) and propidium iodide as per manufacturer's instructions (Immune Quality Products). Cells were then analysed on a LSR-II (Becton Dickinson) cytometer using FACSDiva software (Becton Dickinson). For cell cycle analysis, BrdU and phospho-HistoneH3 analysis, cells were fixed in ice-cold 70% ethanol or methanol for at least 6 h and were then immunostained with an Alexa-488-conjugated antibody targeting BrdU (MoBU1, #B35130, 1:200), or anti-phospho-histone-H3 (Ser10, Cell Signaling, #9701, 1:300) in combination with Alexa-488-conjugated secondary antibodies (1:300). DNA staining was performed using propidium iodide in the presence of RNAse. At least 10,000 events per sample were analysed on a FACScalibur (Becton Dickinson). Data was analysed using FlowJo software.
Live-cell microscopy. KB2P2.21 and KB2P2.21R1 transduced with H2B-EGFP-IRES-a-tubulin-mCherry and HeLa cells transduced with H2B-EGFP were seeded in eight-chambered cover glass plates (Lab-Tek-II, Nunc) at 10,000 cells per well. Cells were then treated with 0.5 mM olaparib at 24 h before imaging, and were followed for at least 36 h on a DeltaVision Elite microscope, equipped with a CoolSNAP HQ2 camera and a Â 40 immersion objective (U-APO 340, numerical aperture: 1.35). Images were obtained each 10 min, with 12 images being acquired in the Z-axis, at 0.5 mm interval. Image analysis was done using SoftWorX software (Applied Precision/GE Healthcare). The fate of all cells that entered mitosis and proceeded at least until anaphase were included for analysis.
DNA fibre analysis. To assess replication fork protection. HeLa cells were pulselabelled with CIdU (25 mM) for 60 min. Next, cells were washed with medium and incubated with hydroxyurea (HU, 5 mM) for 5 h. Cells were harvested using trypsine and lysed on microscopy slides in lysis buffer (0.5% SDS, 200 mM Tris (pH 7.4), 50 mM EDTA). DNA fibres were spread by tilting the slide and were subsequently air dried and fixed in methanol/acetic acid (3:1) for 10 min. For immunolabelling, spreads were treated with 2.5 M HCl for 1.5 h. CIdU was detected by staining with rat anti-BrdU (1:1,000, AbD Serotec) for 1 h and was further incubated with AlexaFluor 488-conjugated anti-rat IgG (1:500) for 1.5 h. Images were acquired on a Leica DM-6000RXA fluorescence microscope, equipped with Leica Application Suite software. The lengths of CIdU and IdU tracks were measured blindly using ImageJ software. Statistical analysis was performed using two-sided Mann-Whitney tests with 95% confidence intervals.
Generation of mammary tumours. Brca1 À / À ;p53 À / À and Brca2 À / À ;p53 À / À mammary tumours were generated in K14cre;Brca1 F/F ;p53 F/F and K14cre;Brca2 F/F ; p53 F/F mice, respectively, genotyped, and orthotopically transplanted into syngeneic wt mice as described 7 . Starting 2 weeks after tumour grafting, in female FVB/N mice (6-8 weeks old), the onset of tumour growth was checked at least three times per week. Mammary tumour size was determined by caliper measurements. When mammary tumours reached a size of B200 mm 3 , treatment was initiated. Olaparib was used by diluting 50 mg per ml stocks in DMSO with 10% 2-hydroxyl-propyl-b-cyclodextrine/PBS such that the final volume administered by intraperitoneally (i.p.) injection was 10 ml g À 1 of body weight. Olaparib (50mg kg À 1 ) was given i.p. daily for 21 or 28 consecutive days. Controls were dosed with vehicle only. Animals were killed with CO 2 at the end of treatment when the minimal residual disease stage was reached. At this point, olaparibtreated tumour explants had an approximate size of 1 mm 3 , whereas controltreated tumour explants had a volume of B1 cm 3 . Tumour samples were fixed in 4% formaline and processed for hematoxylin/eosin staining. S.d. represent 10 different fields, containing at least 100 cells. All experimental procedures on animals were approved by the Animal Ethics Committee of the Netherlands Cancer Institute.
Strand-specific libraries were generated using the TruSeq Stranded mRNA sample preparation kit (Illumina Inc., San Diego, RS-122-2101/2), according to the manufacturer's instructions (Illumina, Part #15031047 Rev. E). Briefly, polyadenylated RNA from intact total RNA was purified using oligo-dT beads. Following purification, the RNA was fragmented, random primed and reverse transcribed using SuperScript II Reverse Transcriptase (Invitrogen, part #18064-014) with the addition of Actinomycin D. Second strand synthesis was performed using Polymerase I and RNaseH with replacement of dTTP for dUTP. The generated cDNA fragments were 3 0 -end adenylated and ligated to Illumina Pairedend sequencing adapters and subsequently amplified by 12 cycles of polymerase chain reaction. The libraries were analysed on a 2100 Bioanalyzer using a 7500 chip (Agilent, Santa Clara, CA), diluted and pooled equimolar into a 10 nM sequencing stock solution. Illumina TruSeq mRNA libraries were sequenced with 50 base single reads on a HiSeq2000 using V3 chemistry (Illumina Inc., San Diego).
The resulting reads were trimmed using Cutadapt (version 1.12) 8 to remove any remaining adapter sequences, filtering reads shorter than 20 bp after trimming to ensure efficient mapping. The trimmed reads were aligned to the GRCm38 reference genome using STAR (version 2.5.2b) 9 . QC statistics from Fastqc (version 0.11.5) and the above-mentioned tools were collected and summarized using Multiqc (version 0.8) 51 . Gene expression counts were generated by featureCounts (version 1.5.0-post3) 52 , using gene definitions from Ensembl GRCm38 version 76. Normalized expression values were obtained by correcting for differences in sequencing depth between samples using DESeq median-of-ratios approach 53 , and subsequent log-transformation the normalized counts.
Data availability. The data that support the findings of this study are available from the corresponding author upon reasonable request. RNA sequence data has been deposited at the European Nucleotide Archive (ENA) under accession number PRJEB20535.