Ovarian Cancers Harbor Defects in Nonhomologous End Joining Resulting in Resistance to Rucaparib

Double-strand breaks (DSB; ref. 1), the most lethal forms of DNA damage, are repaired by 2 main pathways: nonhomologous end joining (NHEJ) and homologous recombination (HR). These pathways are distinct in that HR copies identical DNA sequences from sister chromatids resulting in error-free repair (2), whereas NHEJ joins the broken DNA ends with limited processing (3). In vitro studies have demonstrated that complementary DNA ends are joined in an efficient and accurate manner by NHEJ (4, 5). However, the modification required for partially or completely incompatible DNA ends results in losses of sequence at the resultant junctions, such that NHEJ is potentially a mutagenic process (3, 6). More recent studies have demonstrated an alternative end joining mechanism (A-EJ), which uses regions of microhomology at internal sites on the DNA substrate. Unlike HR, A-EJ is inherently error-prone, as the use of microhomology leads to deletions of sequences from the strand being repaired and to chromosomal translocations (7, 8). This mechanism has been suggested to function in the absence of NHEJ (9–14) and more recently in absence of HR (7, 8).

NHEJ has been demonstrated to function throughout the cell cycle (1, 15). The NHEJ pathway is initiated by the binding of the Ku heterodimer (Ku70 and Ku80) to DSBs, and the subsequent association and autophosphorylation of the DNA-dependent protein kinase catalytic subunit (DNA-PKcs; ref. 16). This DNA-PK complex facilitates ligation by recruitment of the XRCC4/LIG4 complex. Mutations in NHEJ components are associated with immunodeficiency and developmental abnormalities (17, 18) as well as cancers (6, 19–22), underscoring the importance of the NHEJ pathway in maintaining genome integrity.

DNA damage repair (DDR) is increasingly recognized as an important determinant of response to cancer therapeutics. This interest was initially provoked by the paradigm shifting discovery that inhibition of base excision repair with PARP1 inhibitors (PARPi) was synthetically lethal in HR-defective (HRD) tumors. PARPi were therefore selectively targeting the defect arising in the tumors, but not in normal tissues (23–27). In epithelial ovarian cancer (EOC), HRD is reported in 50% of cases (28) and evidence is building for the efficacy of PARPi. This has been assumed to be as a result of synthetic lethality, with PARPi preventing effective base excision repair leading to stalled replication forks which in turn could not be repaired by HR. However, a number of studies also indicate a connection between components of the NHEJ pathway and PARP1 (29–35), culminating in the suggestion by Patel and colleagues that dysfunctional NHEJ is important in generating genomic instability in PARPi-treated, HRD cells (36). Moreover, they also demonstrated that inhibition of DNA-PK results in HR function recovery and PARPi resistance in vitro (37).

The suggestion that NHEJ status is important in determining sensitivity to PARP inhibitors is in keeping with evidence that NHEJ is a fast pathway which is the pathway of choice for the repair of DSBs with HR only being employed for unrepaired DSBs (38).

The incidence of NHEJ dysfunction has not been explored in primary EOC to date. Here we demonstrate that more than 40% of primary ovarian cancer (PCO) cultures are NHEJ-defective (NHEJD), which is associated with resistance to rucaparib ex vivo.

Ethical approval was granted (12/NW/0202) for the collection of ascites from consented patients undergoing surgery for EOC at the Queen Elizabeth Hospital, Gateshead, UK. Clinical details were recorded and specimens registered and handled in accordance with the Human Tissue Act. Samples were assigned a reference number to retain anonymity.

PCO cultures were generated and maintained as previously described (39, 40). Briefly, 20 mL of ascites was added to 20 mL of warmed Sigma RPMI-1460 HEPES modified culture medium supplemented with 20% v/v fetal calf serum and 100 μL/mL penicillin and streptomycin in T75 flasks and incubated at 37°C, 5% CO₂ humidified air.

All cell lines, unless stated otherwise, were grown in RPMI-1640 media supplemented with 10% FBS and 100 units/mL penicillin/streptomycin incubated at 37°C in 5% CO_2. V3 (DNA-PK_CS defective) and V3YAC cells (V3 cells complemented with human DNA-PK_CS) were a kind gift from Professor Jeggo. V3YAC cells were grown in full medium with G418 (400 μg/mL). A2780, a human ovarian carcinoma cell line, and CP70, MMR-deficient variant of A2780, 5-fold resistant to cisplatin relative to the parental A2780, were a kind gift from Prof. R. Brown (Cancer Research UK Beatson Laboratories, Glasgow, Scotland). SKOV-3, OVCAR-3, IGROV-1, and MDAH are all human ovarian adenocarcinoma cell lines and were purchased from ATCC. PEO1 cell line was derived from a poorly differentiated serous adenocarcinoma and PEO4 cell line derived from the same patient after clinical resistance developed to chemotherapy. Both were purchased from the European Collection of Cell Cultures.

OSEC2 and OSEC4 cell lines developed at Newcastle University (Newcastle Upon Tyne, UK) from normal ovarian surface epithelium using a temperature sensitive SV-40 large T antigen construct were incubated at 33°C (41).

UWB1-289, a BRCA1-null human EOC cell line derived from papillary serous ovarian carcinoma, was cultured in 50% RPMI-1640 media supplemented with 10% FBS and 100 units/mL penicillin/streptomycin and 50% (v/v) MEBM BulletKit media (Lonza) supplemented with 10% FBS. UWB1-289-BRCA1 is derived from UWB1-289 cells in which BRCA1 was restored were cultured in full media with 400 μg/mL G418. Both were obtained from ATCC.

Cell extracts were prepared as previously described (6). Briefly, three T175 flasks at 80% confluence were trypsinized, lysed in 500 μL of hypotonic buffer, and homogenized. After the addition of 0.5 vol of high salt buffer, the extracts were centrifuged for 56 minutes at 70,000 rpm (213,000 × g) at 4°C in a Beckman TLA120.2 rotor. Protein concentration was determined using the BSA protein assay according to manufacturer's instructions (ThermoScientific). Samples were snap-frozen and stored at −80°C.

Vectors which on digestion with BstXI yielded a 3-2 kb plasmid and 1.2-kb λ fragment with either compatible (Co) (CCACTAAG_GTGG and GGTG_ATTCCACC) or 2 base pair (2I; CCACTAAG_GTGG and GGTG_AAACCACC) and 4 base pair (4I; CCACTAAG_GTGG and GGTG_TAAGCACC) incompatible ends. Vectors were kindly donated by Dr. Ann Kiltie (Oxford). DNA fragments were gel-purified using spin columns (Qiagen). End-joining reactions were carried out as previously described (6) with 45 μg protein extract and 100 ng DNA substrate for 2.5 hours. DNA was extracted with Tris-buffered phenol/chloroform/isoamyl alcohol. Analysis was performed by agarose (0.7%) gel electrophoresis and GelRed (VWR) staining. Image capture was carried out using G:Box and GeneSnap system and analyzed using GeneTools (SynGene).

For the analysis of joined products, end-joining reactions were ethanol-precipitated and amplified using ThermoPrimeTaq with ReadyMix PCR buffer (Thermo Scientific) in the presence of internal plasmid primers pFOR (5′-CCGGCGAACGTGGCGAGAAAG) and pREV (5′-GACTGGAAAGCGGGCAGTGAG) for 40 cycles (30 seconds at 94°C, 30 seconds at 55°C, 30 seconds at 72°C, full-length product size, 551 bp). Analysis was performed by agarose (1%) gel electrophoresis and GelRed staining.

Plasmid pGL2 (Promega) was linearized using either HindIII or EcoRI and linearization was confirmed by agarose gel electrophoresis. The linearized DNA was purified using the QIAquick PCR Purification Kit (Qiagen), dissolved in sterilized water, and transfected into cells using Lipofectamine LTX (Invitrogen) as per manufacturer's instructions. The transfectants were harvested 48 hours after transfection and assayed for luciferase activity as described previously (42).

Cells were seeded onto glass coverslips and treated with 2-Gy ionizing radiation and rucaparib at 10 μmol/L concentration for 24 hours to induce DSB. All experiments were performed alongside untreated controls with equivalent 0.1% DMSO. Cells were then fixed and rehydrated before staining with 1:100 mouse monoclonal anti-γH2AX (Upstate, Millipore Corp.) and 1:100 goat polyclonal anti-Rad51 (Calbiochem, EMD Biosciences, Inc.) antibodies with appropriate secondary fluorochrome-conjugated antibodies, as previously described (43).

ImageJ counting software (44, 45) was used to count γH2AX and Rad51 nucleic foci. Cells were classed as HR-competent if there was more than a 2-fold increase in Rad51 foci after DNA damage, confirmed by a 2-fold increase in γH2AX.

Extraction of RNA was performed using an RNeasy Mini kit (Qiagen) as per manufacturer's instructions. RNA was eluted in 30 μL RNase-free water and quantified on the Nanodrop ND-1000 Spectrophotometer (Lab Tech International). Total RNA (1.6 μg) was incubated at 65°C for 5 minutes followed by 37°C for 5 minutes prior to addition of Promega MMLV-reverse transcriptase master mix (4 μL 5× Moloney Murine Leukaemia Virus RT buffer, 2 μL 4mmol/L dNTPs, 1 μL 50 μmol/L Oligo dT15, and 0.3 μL MMLV reverse transcriptase) and incubation at 37°C for 1 hour followed by 95°C for 5 minutes. cDNA (2 μL) was loaded on to a 386-well plate in triplicate with Invitrogen SYBR green Master Mix (dNTPs, optimized buffer, UDG, ROX reference dye, AmpliTaq DNA polymerase UP and SYBR green ER dye) and the 2.5 mmol/L of the appropriate forward and reverse primers. Primers used were purchased from Sigma-Aldrich. Primers sequences were: DNAPK-1, 5′–CTAACTCGCCAGTTTATCAATC–3′; 5′–TTTTTCCAATCAAAGGAGGG–3′; DNA-PK-2, 5′–GATCTGAAGAGATATGCTGTG–3′; 5′–GTTTCAGAAAGGATTCCAGG–3′; XRCC5, 5′-TTCATTCAGTGAGAGTCTGAG-3′; 5′-CGATTTATAGGCTGCAATCC-3′; XRCC6, 5′-AAGAAGAGTTGGATGACCAG-3′; 5′-GTCACTTCTGTATGTGAAGC-3′; LIG4, 5′-ATTTCTCCCGTTTTTGACTC-3′; 5′-GAATCTTCTCGTTTAACTGGC-3′; XRCC4-1, 5′-AGCTGCTGTAAGTAAAGATG-3′; 5′-CCAAGATTTCTTTGCATTCG-3′; XRCC4-2, 5′-CCAAGTAGAAAAAGGAGACAG-3′; 5′-GCTTTTCCTTTTCTTGAAGC-3′; XRCC4-3, 5′-CTAGAGAAAGTTGAAAACCCAG-3′; 5′-ATCGTCCTTGAACATCATTC–3′; GAPDH, 5′–CGACCACTTTGTCAAGCTCA–3′; 5′–GGGTCTTACTCCTTGGAGGC–3′. Samples were run on an AbiPrism Applied Biosystems real-time PCR machine for 10 minutes at 95°C, 40 cycles (15 seconds at 95°C, 60 seconds at 60°C), 15 seconds at 95°C, 15 seconds at 60°C, 15 seconds at 95°C. Data were analyzed using SDS2.3 software.

Western blotting was assessed as previously described (46). Briefly, 40 μg of total protein from each samples was loaded and resolved by electrophoresis in 3% to 8% SDS-PAGE gradient gels (Bio-Rad) and transferred to nitrocellulose membrane (Hybond C Membrane; GE Healthcare). Blots were then incubated using appropriate antibodies: DNA-PK_cs, [1:500, at 4°C, overnight (ON); SantaCruz Biotechnology]; Ku70 [1:800, at 4°C, ON (Abcam)]; Ku80 [1:800, at 4°C, ON (Abcam)]; XRCC4 [1:1000, at 4°C, ON (AbDSerotec)]; ligase IV [1:800, at 4°C, ON (Abcam)]; and GAPDH [1:3,000, at room temperature (RT), for 1 hour (Santa Cruz)] followed by HRP-conjugated, goat anti-rabbit, or goat anti-mouse IgG-HRP secondary antibody [1:1000 at RT, for 1 hour (Dako)]. Image capture and analysis was carried out using the Fuji LAS-300 Image Analyzer System (FujiFilm).

Sulforhodamine B (SRB) assay was used to assess cytotoxicity and cell growth as previously described (47). Briefly, cells were seeded at a concentration of 1,000 cells per well and after adherence were treated with different concentrations of rucaparib or cisplatin ± 1 μmol/L of DNA-PK inhibitor NU7441 for 10 days before fixation, stained, and assessed by spectrophotometer.

Immunofluorescent experiments were carried out as previously described (43). Briefly, cells were fixed after 24 hours with 10 μmol/L rucaparib ± 1 μmol/L NU7441 and 2-Gy X-ray irradiation for HR assay or 1 hour after 2-Gy irradiation for pDNA-PKcs. The γH2AX, RAD51, or pDNA-PKcs foci were detected by immunofluorescence using appropriate antibodies: anti-phospho-histone γH2AX [Ser139; 1:100 dilution, at RT, for 1 hour (Upstate, Millipore Corporation)]; rabbit polyclonal anti-Rad51 [1:100 dilution, at 4°C, ON (Calbiochem, EMD Biosciences, Inc.)]; or DNA-PKcs phospho S2056 [1:500 dilution, at 4°C, ON (Abcam)] followed by Alexa Fluor 546 goat anti-mouse or 488 goat anti-rabbit IgG secondary antibody [1:1,000 dilution at RT, for 1 hour, protected from light (Invitrogen)]. Images were captured using a Leica DMR microscope and RT SE6 Slider camera Advanced Spot software version 3.408 (Diagnostic Instruments Inc.). Automated analysis using ImageJ software and a custom macro of foci in more than 50 cells per field of view was carried out.

PARP1 activity was measured using a validated assay as previously described (48). Briefly, PARP activity in 1,000 permeabilized cells was maximally stimulated with a double-stranded oligonucleotide in the presence of excess NAD (350 μmol/L) and the amount of ADP-ribose polymer formed quantified by immunoblot using anti-PAR antibody (clone 10H, from Professor Dr. Alex Burkle, University of Konstanz, Konstanz, Germany) by reference to a PAR standard curve (Enzo Life Sciences). Data are expressed as percentage of PAR of L1210 control.

Statistical analysis was performed using GraphPad Prism version 6.00 (GraphPad Software) Unpaired Student t tests or Mann–Whitney tests were used depending on a D'Agostino and Pearson omnibus normality test. Multiple comparisons were performed using one-way ANOVA with Tukey multiple comparisons correction. All statistical tests were 2-sided and considered statistically significant if the P value was less than 0.05.

A number of assays are described in the literature to assess NHEJ function (49). Most of these assays only assess the rejoining of compatible ends, which does not represent the complexity of DNA DSBs that occur in cells. We therefore assessed rejoining of compatible (Co) and incompatible (2I, containing mismatches of 2 bases and 4I, containing mismatches of 4 bases) vector ends following the addition of cell extracts. T4 ligase ligated Co substrates, but incompatible substrates (2I and 4I) could not be joined without the addition of the appropriate λDNA fragment which formed compatible ends with each 2I and 4I substrates (Fig. 1A). OSEC2 cells rejoined 34.8% of Co, 15.9% of 2I, and 13.7% of 4I substrates (Fig. 1A and B). Addition of the λDNA fragment increased the rejoining rate of incompatible (P < 0.001 of 2I and P = 0.0004 of 4I) but had no effect on rejoining of the compatible substrates. As both 2I and 4I had similar rejoining rates, assessment in cell lines and PCO panels was performed using Co and 2I substrates only. Comparison of rejoining in paired DNA-PK–deficient and -proficient cell lines demonstrated that while compatible ends are largely rejoined correctly, DNA-PK–deficient V3 and M059J cells were unable to rejoin 2I substrates (Fig. 1C). Furthermore, addition of DNA-PK inhibitor NU7441 inhibited rejoining in DNA-PKcs–proficient V3YAC cells but had no effect on DNA-PKcs–deficient V3 cells (Fig. 1D).

To ensure the cell-free extract assay represented the cellular end-joining accurately, NHEJ function was assessed in a panel of established cell lines using the cell extract and a cellular luciferase assay (Fig. 2). While the OSEC cell lines derived from normal ovarian epithelium were able to rejoin 2I ends accurately, 4 of the 6 EOC cell lines were unable to rejoin 2I substrates, thus indicating NHEJ deficiency. This correlated with the cellular end-joining assay. Mean accurate cellular rejoining rate was 30.17% [95% confidence interval (CI), 25%–37.6%] by cell lines able to rejoin 2I substrates compared with 9.9% (95% CI 4.39–14.0, P = 0.03) by cell lines unable to rejoin 2I substrates, when assessed using the luciferase cellular assay (Pearson correlation: r = 0.79, P = 0.007). We have previously demonstrated that vector transfection into PCO cultures is not possible (39); therefore, NHEJ was assessed in PCO cultures with the validated extract assay only.

We next assessed end joining in a panel of primary ovarian cancer cultures. PCO cultures had a reduced end-joining rate compared with NHEJ-competent (NHEJC) control cell lines. There was significant intersample variability (range, 5%–39% of loaded DNA, Fig. 3A and B). PCR analysis of the junctions formed demonstrated that the rejoining of the Co substrate was accurate (Fig. 3C).

We found that 20 of the 47 PCO cultures were NHEJD, as demonstrated by incubation with 2I substrates producing either no products or forming products of significantly smaller size (example PCR product bands is shown in Fig. 3C). Furthermore, some cultures formed multiple bands of different sizes indicating loss of differing numbers of nucleotides. Extensive resection has been demonstrated to be due to use of microhomologies in this vector in the absence of a functional NHEJ pathway (6). NHEJ competence was seen to be independent of culture growth rate, with a mean doubling time of 117 hours for NHEJC and 115 hours for NHEJD cultures. Patient characteristics detailed in Supplementary Table ST1 show that there was no significant difference between the NHEJC and NHEJD cultures in any of the clinical parameters assessed.

Sensitivity of rucaparib and cisplatin was assessed in the cell line panel and all primary cultures. In contrast to HRD association with increased rucaparib sensitivity, NHEJD cultures were resistant to rucaparib (P = 0.0022, Fig. 4A), as well as established cell lines (P < 0.0001; Fig. 4B). Furthermore, NU7441 induced resistance to rucaparib in the sensitive PCO cultures (P = 0.014). When HR and NHEJ functions were taken into account, only NHEJC/HRD cultures were sensitive to rucaparib (compared with NHEJC/HRC P = 0.034, NHEJD/HRC P = 0.0002, and NHEJD/HRD P = 0.0045).

No correlation of cisplatin sensitivity was found with NHEJ function or inhibition (Fig. 4C). Cisplatin was found to inhibit NHEJ significantly even at 4 nmol/L concentration (Fig. 4D). This was consistent with the finding of no association of NHEJ function with progression-free survival (PFS) or overall survival in our cohort of patients who were treated with a standard platinum-based therapy after a median follow-up of 20 months (Supplementary Table ST1).

Analysis of NHEJ pathway components showed that protein expression of Ku70, Ku80, and DNA-PKcs, normalized to GAPDH, was significantly lower in NHEJD cultures (Ku70, P = 0.0013; Ku80, P = 0.002; and DNA-PKcs, P < 0.0001; Fig. 5A). These were found to be good predictors [area under the curve (AUC), 0.798, 0.762, and 0.852 respectively; Fig. 5B] for NHEJ function and may therefore be suitable candidates for biomarkers. Discordance between protein and mRNA expression was noted, as previously reported in other studies (50, 51). DNA-PK autophosphorylation correlated with NHEJ function in cell lines but not in the PCO cultures (Supplementary Fig. S1).

We have previously demonstrated that 50% of ovarian cultures are functionally HRD (28), therefore upon the finding that 40% of primary ovarian cancer cultures are also NHEJD, we assessed the interaction of NHEJ and HR. The addition of the DNA-PK inhibitor, NU7441, resulted in a significant upregulation of RAD51 foci after 2-Gy irradiation in OSEC2 cells (Fig. 6A), demonstrating an increase in HR repair. NU7441 also recovered HR competence in the BRCA1-deficient cell line, but it had no effect in the HR-competent or BRCA2-defective cell lines (Fig. 6B). Furthermore, the mean fold increase in RAD51 foci in DNA-PK–deficient M059J cells was significantly higher than isogenic DNA-PK–proficient M059FUS-1 cells (P < 0.0001, Fig. 7C).

In our cohort of PCO cultures, NHEJ function was independent of HR competence. 15 cultures were functional for both pathways, 7 cultures were defective for both pathways, whereas 11 and 14 cultures showed defects in NHEJ and HR, respectively. RAD51 foci increase was higher in NHEJD than in NHEJC cultures (P < 0.0024, Fig. 7A). DNA-PKcs expression was higher in HRC cultures than in HRD cultures (P < 0.0001, Fig. 7A). When taking both HR and NHEJ function into account, while both NHEJC/HRC and NHEJD/HRC were found to have RAD51 foci fold increase >2, the mean RAD51 foci fold increase for NHEJC/HRC group was lower than for NHEJD/HRC group (Fig. 7B). The differences in RAD51 foci increase between all 4 groups were independent of the amount of DNA DSBs, as determined by γH2AX foci formation. Importantly, no correlation between either PARP1 activity or mRNA expression and HR or NHEJ competence was found (Supplementary Fig. S1C).

Here we have described our findings that NHEJ is defective in more than 40% of ex vivo EOCs. We found NHEJ to be independent of HR function and PARP1 activity. In contrast to HR (where cells with the HRD phenotype are sensitive to PARPi), we have demonstrated that cells defective in NHEJ are resistant to PARPi. By considering the function of both pathways, we have shown that only the NHEJC/HRD cultures are sensitive to rucaparib. This finding potentially explains the resistance observed in some HRD tumors. Finally, we suggest that expression of the NHEJ-related proteins Ku70, Ku80, and DNA-PKcs may be useful as biomarkers to determine NHEJ status in cancer samples.

The sensitivity of HRD cancers to PARPi was initially attributed to the concept of synthetic lethality, on the basis of the theory that HR-defective cells are unable to repair DNA DSBs (25, 52). However, the majority of DNA DSBs are repaired by NHEJ (1, 15). Furthermore, cell line studies demonstrate interaction between the NHEJ pathway, PARP1, and subsequent resistance to PARPi (37). The suggested role for NHEJ in PARPi sensitivity was through upregulation of error-prone NHEJ in HRD cells (37). Recent studies have demonstrated that the error-prone A-EJ functions in the absence of NHEJ and competes with HR (7, 8, 53). Clearly, the interaction between the various DSB repair pathways is complex and understanding is compounded by the commonality of the early part of the process. In this study, we were not able to assess the cell-cycle–specific function of both pathways but this may provide further insight into the interaction (54).

Nevertheless, we have found that NHEJ function is independent of HR competence and that inhibition of NHEJ resulted in up-regulation of HR function in HRC and BRCA1-deficient cells. In our cohort, the cultures which were NHEJD were resistant to rucaparib, irrespective of HR function. This is supported by the observation that NU7441 caused rucaparib resistance in all sensitive cultures, independent of HR function. When taking both pathways into account, only NHEJC/HRD cultures were found to be sensitive to rucaparib. Here, we demonstrate the role of NHEJ function in ex vivo primary cultures. Therefore, we propose that in EOC, in the absence of HR, error-prone NHEJ results in sensitivity to PARPi. Conversely, absence of NHEJ function results in PARPi resistance in HRD cells. This may be through A-EJ; however, assessment of this pathway in primary EOC is still needed. Because of the inhibitory effect of cisplatin on NHEJ (55), this model is limited to rucaparib sensitivity only.

The hypothesis put forward for the role of NHEJ in PARPi resistance is based on the error proneness of NHEJ. The errors in repair are suggested to cause lethal defects in DNA, which, in the absence of HR, results in apoptosis. Therefore, NHEJC/HRD cells are sensitive to PARPi. Cells with competent NHEJ and HR pathways are able to repair DNA damage and are, therefore, resistant to PARPi. In the absence of NHEJ, the slower error-free HR takes over repair. This notion is supported by findings of greater HR function, demonstrated by greater RAD51 foci formation in the DNA-PK–deficient cell lines in this study as well as the existing literature (56). Therefore, in the absence of NHEJ function, the lack of error-prone repair results in resistance to PARPi (37).

Our finding of selectivity for a DNA-PKcs inhibitor to selectively revert the BRCA1-mutant cell line but not the BRCA2-mutant line is interesting but can be explained by the differing functions of BRCA1 and BRCA2 in the DDR pathways. The role of BRCA1 is to inhibit Rif1 which in turn selects for NHEJ. HR can still therefore function in the absence of BRCA1 but preferentially cells will proceed down the NHEJ pathway because there is no inhibition of Rif1 and therefore no inhibition of NHEJ. But if NHEJ is inhibited or defective, even without BRCA1, HR can go ahead.

The role of BRCA2 is within HR pathway itself. Therefore, inhibiting NHEJ in the absence of BRCA2 still does not activate HR because HR itself is broken.

The ability to select the correct patient, for the correct treatment, at the right time, is required for personalized medicine. Our findings suggest that accurate selection will be compromised if HR function alone is assessed and assessment of NHEJ may also be required. While attempts are being made to develop predictive biomarkers of HR, we suggest that biomarkers for NHEJ should also be developed to aid patient selection for PARPi therapy.

N.J. Curtin holds ownership interest (including patents) in rucaparib. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. McCormick, N.J. Curtin, R.J. Edmondson

Development of methodology: A. McCormick, N.J. Curtin, R.J. Edmondson

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. McCormick, P. Donoghue, R. O'Sullivan, R.L. O'Donnell, J. Murray, A. Kaufmann, R.J. Edmondson

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. McCormick, M. Dixon, R. O'Sullivan, R.L. O'Donnell, J. Murray, N.J. Curtin, R.J. Edmondson

Writing, review, and/or revision of the manuscript: A. McCormick, R.L. O'Donnell, N.J. Curtin, R.J. Edmondson

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. McCormick, M. Dixon, R.L. O'Donnell, J. Murray

Study supervision: A. McCormick, N.J. Curtin, R.J. Edmondson

We thank Dr. Anne Kiltie for the kind donation of vector constructs and the protocol for the end-joining assay which allowed this project to take place. We also thank the patients and the clinical staff at Queen Elizabeth Hospital for the PCO sample donation.

This work was supported by Cancer Research UK grant number C27826/A12498.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.