Evaluation of candidate biomarkers to predict cancer cell sensitivity or resistance to PARP-1 inhibitor treatment

Deficiency in the MRN complex and sensitivity to PARP-1 inhibition

To examine whether PARP-1 inhibition is selectively lethal to various cellular models deficient in components of the DSB-sensing and -processing complex of Mre11, Rad50 and Nbs1, we tested sensitivity of a panel of human cancer cell lines with differential status of this important tumor suppressor complex.^3,21 Given that only a subset of carcinoma-derived cell lines can perform robustly in a clonogenic assay, we first established a more universally applicable, shorter-term viability assay to determine sensitivity to PARP-1i. We chose as a positive control the pancreatic cancer cell line CAPAN-1 (harboring the naturally occurring BRCA2 617delT frameshift mutation accompanied by the loss of the second allele)²² and, in parallel, examined an isogenic pair of human SV40-immortalized fibroblasts either deficient in Nbs1, NBS-1LBI or complemented with wt-Nbs1, NBS-1LBI+Nbs1. The CAPAN-1 and NBS-1LBI cell lines have been reported to show profound sensitivity to PARP-1i due to their defect in BRCA2 and Nbs1, respectively.^16,23 After 4 d of exponential growth (see Methods for details), viable cells were counted and the viability expressed as a cell number normalized to untreated control (Fig. 1). These initial experiments confirmed differences in responses of Nbs1-deficient vs. Nbs1-proficient human fibroblasts and pronounced sensitivity of the BRCA2-deficient CAPAN-1 cells to PARP-1i treatment, at the same time providing support for our assay as an informative approach to monitor the impact of PARP-1 inhibition on cellular viability.

Deficiency in the MRN complex sensitizes breast cancer cells to PARP-1i

Exposure of BRCA1- or BRCA2-depleted cells to PARP-1i reportedly leads to cell cycle arrest, predominantly in G₂ phase, followed by an increase of apoptotic cell death (Farmer et al., 2005). To test whether exposure of Cal51 breast cancer cells to PARP-1i induces cell death, rather than cell cycle arrest only, we performed a well- established XTT-based cell proliferation assay coupled with lactate dehydrogenase (LDH) activity measurement. LDH is an intracellular enzyme, which, when released into the media by dying/dead cells, catalyzes the conversion of lactate to pyruvate while reducing NAD⁺ to NADH/H⁺. In the second step of this assay, NADH/H⁺ is used to reduce tetrazolium salt into colorimetrically detectable formazan. As shown in Figure 2A, reduced cell proliferation caused by PARP-1i treatment is accompanied by increased cell death as indicated by the higher activity of LDH.

To further validate the hypothesis that MRN deficiency increases sensitivity to PARP-1i, we compared the Cal51 cells that harbor a functional MRN complex (wt-MRN) with stable Cal51-derived transfectants expressing shRNA against either Mre11 or Nbs1 to knockdown the respective MRN-complex proteins (Fig. 2B and C). The 48-h treatment with PARPi was sufficient to abolish PARsylation spontaneously occurring in these cells (Fig. S1), and the 4 d viability assay showed enhanced toxicity of PARPi toward Cal51 cells upon partial knockdown of Mre11 and Nbs1 proteins, respectively, compared with the parental wt-MRN cells (Fig. 2B). These results extend the previous analyses to defects of different MRN complex components and suggest that even partial depletion of Mre11 or NBS1 translates into enhanced sensitivity of human breast cancer cells to PARP-1 inhibition.

Efflux transporters confer resistance of Mre11-deficient colon cancer cells to single-agent treatment with PARP-1i

Next, we focused on human colon cancer, a tumor type with known MRN defects. Among the five colorectal cancer cell lines we examined for expression of proteins of the MRN complex, the HT29 cell line lacked the Nbs1 protein and showed an aberrantly reduced level of Mre11 complex (Fig. 3B). The cell line HCT116, from which p53-proficient and p53-deleted variants are available, harbors a mixture of wild-type Mre11 and a splice variant that leads to a partial deletion of sequence in the N-terminal nuclease domain. This mutant allele of Mre11 preserves some functions of the wild-type protein and suppresses other functions of the MRN complex in a dominant-negative manner.^24,25 Our immunoblotting analysis of total lysates from HCT116 cells revealed detectable but significantly reduced levels of components of the MRN complex (Fig. 3B), suggesting destabilization of the MRN complex by the Mre11 mutation in this cell line. Exposure of colon cancer cells to PARP-1i revealed mild sensitivity of the p53-proficient version of HCT116 cells, while the p53-deleted HCT116 cells were more resistant, similar to the MRN-proficient colon cancer cell line SW620 (Fig. 3A). Ectopic expression of Mre11 in HCT116 cells resulted in increased levels of endogenous Nbs1 and Rad50 proteins (Fig. 3D), and such reconstitution led to a shift in the PARP-1i response curve (Fig. 3C).

We speculated that the inefficient response of MRN-deficient colon cancer cells to PARP-1i could be related to multidrug resistance, a frequent feature of colorectal cancer, as P-glycoprotein (P-gp) overexpression was associated with PARP-1i resistance.^26-28 To address this possibility, we treated the P-gp efflux pump expressing HCT116 cells with PARP-1i and co-treated them or not with the P-gp inhibitor Verapamil. Monitoring the intracellular levels of PARP-1i directly by mass spectrometry showed (Fig. S3) that administration of a non-toxic concentration of Verapamil more than doubled the intracellular concentration of PARP-1i, and this correlated with sensitization of the HCT116 cells in our 4-d proliferation test (Fig. 4). We conclude that one mechanism of resistance to PARP-1i treatment in human MRN-deficient colorectal cancer is the enhanced P-gp-mediated drug efflux, and that this type of resistance can potentially be overcome, at least in part, by suitable inhibitors of the multidrug resistance pumps.

p53 status and response of MRN-deficient cancer cells to PARP-1i

Mutation or loss of p53 can confer resistance of diverse types of cancer cells to multiple classes of drugs.²⁹ Impaired apoptotic and/or checkpoint mechanisms in p53-mutant tumors allow such cancer cells to proliferate even under exposure to genotoxic stress. BRCA2-depleted cells, however, are sensitive to PARP-1i regardless of p53 status.⁹ Our findings, on the other hand, suggested there might be differential responses of HCT116 colon cancer cells with wild-type vs. deleted p53 to PARP-1i treatment (Fig. 3A). Intrigued by this observation, we further investigated the possibility that p53 aberrations could contribute to the observed resistance of the MRN-deficient colon cancer cells to PARP-1i. First, we noted the p53 status in the 20 cell lines of our panel, tested for sensitivity to PARP-1i (Table S1). To further examine the potential effect of p53 status on response to PARP-1i using another MRN-deficient cancer cell type, we silenced the wt-p53 in the breast cancer cell line Cal51, in which the MRN complex was disabled by a knockdown of Nbs1 through stable expression of shRNA (Nbs1 kd) (Fig. 5A). Consistent with the data obtained for colon cancer cells in the 4-d XTT assay, we observed an increased resistance to PARP-1i in the p53-depleted Cal51-(Nbs1 kd) cells, compared with their Cal51-(Nbs1 kd)/wt-p53 counterparts (Fig. 5B). However, in contrast to the differences observed in the short-term proliferation assay, p53 status did not appear to have an impact on sensitivity of MRN-deficient cells to PARP-1i assessed by the longer-term colony formation assay, at least in the HCT116 colon cancer model (Fig. 5C).

PARP-1i can sensitize cancer cells to Camptothecin or ionizing radiation regardless of the p53 status

Apart from the BRCA1 and BRCA2 defects that strongly sensitize cells to PARP-1i, the other genetic defects that show synthetic sickness effects when combined with PARPi have an overall less pronounced impact. Consequently, tumors with such non-BRCA1/2 aberrations are less suitable for a single-agent treatment with PARPi. Rather, in such clinical scenarios, it could potentially be advantageous to apply PARP-1i in combination treatments to sensitize tumors to standard-of-care chemo- or radiotherapy.

To explore this possibility in a model system, we first examined the impact of PARP-1i combined with the genotoxic drug camptothecin (CPT, a topoisomerase I inhibitor commonly used in clinical oncology) in our model of MRN-deficient colon cancer cells HCT116, either with wt-p53 or deleted for p53, the latter being more resistant to PARP-1i treatment alone (see Fig. 3A). First, we noted that the p53 status affected the response of this cell line to CPT alone, in that the p53-deleted HCT116 cells were clearly more resistant compared with their p53-wild-type counterparts (Fig. 6A). Notably, concomitant treatment with PARP-1i (at a moderate dose that alone showed virtually no effect on proliferation of the p53-deleted HCT116 cells in this assay, Fig. 3A) enhanced the anti-proliferative effects of CPT in either variant of the HCT116 cells, suggesting that PARP-1i sensitizes colorectal carcinoma cells to chemotherapy independently of p53 status (Fig. 6A).

Second, we explored a conceptually analogous scenario, testing the impact of added PARP-1i to treatment of two p53-mutant human prostate carcinoma cell lines by ionizing radiation (IR), a modality commonly used to treat prostate cancer. As can be seen from the results of clonogenic assays showed in Figure 6B, both PC3 and DU145 prostate cancer cell lines responded similarly to clinically relevant doses of 2 Gy and 4 Gy of IR alone. Interestingly, addition of a moderate dose of the PARP-1i potentiated the impact of IR on the PC3 cell line, while the response of DU145 cells to IR remained unaffected regardless of PARP-1i treatment (Fig. 6B).

Together, these results suggest that PARP-1i can sensitize subsets of diverse types of common human cancers to CPT and/or ionizing radiation, and that the added PARP-1i in such combined treatment may help eliminate even p53-mutant tumors that are otherwise often more resistant to standard-of-care non-surgical therapies.

Spontaneous PARsylation and Rad51 foci formation as candidate biomarkers of response to PARP-1i

Next, we thought to explore the potential of two functional aspects of the DDR machinery as candidate predictive biomarkers relevant for PARP-1i treatment, namely endogenous PARsylation and focus formation by the HR protein Rad51.

First, as the primary function of PARP-1i is to block the enzymatic activity of PARP-1 and thereby prevent or reduce PAR formation, we argued that cells which do not spontaneously produce detectable PAR without exogenous stimuli are unlikely to be successfully targeted with PARP-1i as a single agent. To this end, we first investigated whether there are any differences in spontaneous PARsylation among the cell lines tested for sensitivity to PARP-1i. Western blot analysis of whole-cell lysates derived from the cells exponentially grown under standard conditions revealed notable variation in endogenous PAR levels among the cell lines of our panel (see examples in Fig. 7B). When comparing the ability of the cells to produce PAR, with response to PARP-1i treatment, we noted a trend for cells with undetectable endogenous PAR to show greater resistance to PARP-1i, as compared with more robust responses of those cancer cell lines producing PAR at detectable levels (Fig. 7A and B). These findings support our working hypothesis of PAR as a potentially useful predictive biomarker, a notion that is further considered in the discussion.

Second, the inability of Rad51 protein to form subnuclear foci in response to DNA damage is regarded as an indication of a functional defect in the HR repair pathway, and therefore a potential surrogate marker that could be useful in predicting responses to PARP-1i. To assess this notion in our panel of human cancer cell lines, we examined the extent of spontaneous Rad51 foci and those formed in response to PARP-1i treatment using immunofluorescence with a well-validated antibody to Rad51.³⁰ The assay was also validated by the fact that neither spontaneous nor PARP-1i-induced Rad51 foci were observed in SUM149 and CAPAN1 cancer cell lines defective in BRCA1 and BRCA2, respectively, and used here as controls expected to be deficient in this assay (Fig. 7C). In contrast to the BRCA1/BRCA2-defective cell lines, however, we could detect sizeable fractions of cells with spontaneous Rad51 foci,\ and an increased formation of such foci after a 24-exposure to PARP-1i, in all other cancer cell lines of our panel examined in this assay (Fig. 7C; Fig. S2 and Table S2). As the cell lines capable of forming Rad51 foci included also several MRN-deficient cell types and other models that showed enhanced sensitivity to PARP-1i (Table S1), our data raise some concerns about the general applicability of this assay in predicting responses to PARP-1i.

Loss of 53BP1 and resistance of BRCA1-deficient breast cancer cells to PARP-1i

Recent evidence suggests that there is a biological selection and enhanced viability and growth among the BRCA-defective human tumors and mouse models, for those cells with aberrantly reduced or lost 53BP1, an important DDR mediator protein that channels DNA DSB lesions preferentially for repair by NHEJ, at the expense of HR.^31,32 As the loss of 53BP1 allows at least partial rescue of the recombination repair, including the ability to form Rad51 foci, these results suggested that absence of 53BP1 could highlight a new class of resistance mechanism to PARP-1i in the otherwise exquisitely sensitive BRCA1-defective cancer cells. To test this possibility in our experimental models, we first examined the total levels of 53BP1 protein in whole-cell lysates by immunoblotting. Despite some differences in protein abundance, 53BP1 was expressed in all cancer cell lines of our panel, including the two BRCA1-defective breast cancer cell lines SUM149 and MDA-MB-436 (Fig. 8B and data not shown). Therefore, we created lentivirus vectors and knocked-down 53BP1 in the MDA- MB-436 and control Cal51 cells, by two different shRNAs (Fig. 8B and data not shown). While the knockdown of 53BP1 had no impact on response of the Cal51 cells to PARP-1i, reduction of 53BP1 resulted in a partial, yet statistically significant (p < 0.01 for both concentrations), reduction in sensitivity to PARP-1i in the BRCA1-deficient MDA-MB-436 cells (Fig. 8A). Considering the robust but nevertheless incomplete loss of 53BP1 in these knockdown experiments (Fig. 8B and C), it is possible that a complete lack of 53BP1 would cause an even more pronounced degree of resistance to PARP-1i. In an attempt to relate these results to clinical settings, we evaluated the proportion of complete vs. partial loss of 53BP1 by sensitive immunohistochemical analysis in our previously published cohort (n = 81) of human breast carcinomas.³³ Notably, while complete or near-complete lack of 53BP1 protein was very rare, the vast majority of cases among the 25% of tumors with aberrantly reduced 53BP1 featured heterogeneous expression of the protein within the tumor cell areas (Fig. 8D), reminiscent of the patterns observed in our MDA-436 cells after shRNA-mediated knockdown (Figs. 8B and C). Overall, these results indicate that our model for aberrant 53BP1 reduction in breast cancer cells mimicks the patterns seen in clinical specimens, and that such reduced levels of 53BP1 can result in enhanced resistance to treatment with PARP-1i, particularly in HR-deficient tumors, such as those with BRCA1 mutation.

Cell culture

Human fibroblasts BJ and NBS-1LBI, NBS-1LBI+Nbs1; colon cancer cell lines HCT116, SW480, SW620, HCT15, HT29; osteosarcoma cell U2OS; breast cancer cell lines CAL51 (CAL51-wt, CAL51-Mre11 kd, CAL51-Nbs1 kd), MDA-MB-436, SUM149; prostate cancer cell lines PC3, DU-145; ovarian cancer cell line OVCAR3; and pancreatic cancer cell line CAPAN1 were cultured as described.⁵⁰ Lymphoblastoma cell lines K256, SR were maintained in RPMI 1640 medium supplemented with 10% (v/v) FBS and 100 U Penicillin and 100 μl/ml Strepromycin. All cell lines were maintained at 37°C in a humidified atmosphere at 5% CO₂. Reagents for cell cultivation were obtained from Gibco Invitrogen.

Stable transfections

Colon cancer cell line HCT116 p53wt was transfected with pcDNA 3.1 Mre11-GFP vector (encoding the full-length wild-type human Mre11 fused with the green fluorescent protein) together with pBABEpuro (5:1 ration) using FuGene 6 (Roche) reagent according to manufacturer’s protocol. Stable clones expressing the transgene were selected in standard DMEM with added puromycine (1 μg/ml). Breast cancer cell lines CAL-51 and MDA MB 436 were transduced with lentiviral particles containing non-targeting shRNA (NT shRNA) or 53BP1 shRNA (Sigma Aldrich). Stable cell lines were selected (5 μg/ml puromycin treatment) and maintained with 0.2 μg/ml puromycin. Breast cancer cell line CAL51-wt and CAL51 shRNA knockdowns of Mre11 and Nbs1 (CAL51-Mre11, CAL51-Nbs1) were kindly provided by KuDOS Pharmaceuticals Ltd. (now part of AstraZeneca).

The cell lines Nbs1 wt reconstituted Nbs1-Tert and NBS-1LBI cell lines were prepared and described by Horejsi et al.⁵¹

RNA interference

Silencing of p53 expression was induced by respective ON-TARGET plus SMARTpool siRNA, the control siRNA sequence used was GGGAGGACAAGACGUUCUAdTdT (Dharmacon). Introduction of siRNA into the cells was performed by an electroporator using Nucleofector kit V (program P-20) (Amaxa/Lonza), according to manufacturer’s instructions. Efficacy of gene silencing was validated by western blot analysis.

Western blot analysis

Western blotting was performed as previously described.⁵² Briefly, dishes with cells were washed with ice-cold PBS, and subsequently lysed in EB lysis buffer (50 mM Tris pH7.5, 150 mM NaCl, 0.5% Igepal 360, 1.0 mM EDTA, 1 mM NaF, 1.0 mM PMSF, 10 mM β-glycerolphosphate, 1 μg/ml aprotitnin, 1 μg/ml leupeptin) on the dishes, scrapped off and lysed on ice for 20 min. After centrifugation (10 min/16,000 g/5°C) supernatant was used. Protein concentration in each supernatant was determined spectrophotometrically (wavelength 595 nm) with Bradford reagent (Bio-Rad) or Pierce BCA Protein Assay Kit (Thermo Scientific). Nuclear extracts from CAL51 and MDA MB 436 cell lines to assess BRCA1 protein level were isolated using NE-PER Nuclear and Cytoplasmic Extraction Kit, according to manufacturer’s protocols (Thermo Scientific).

Protein loading buffer (4×, 200 mM Tris-HCL, pH 7.8, 400 mM DTT, 8% SDS, 0.4% bromphenol blue, and 40% glycerol) was added to each sample and heated at 95°C for 3 min. Proteins were separated on 7.5% SDS-PAGE and blotted onto nitrocellulose membrane (Advantec). Large proteins (53BP1 and BRCA1) were separated on Novex 4% Tris-Glycine gels (Invitrogen) and transferred to nitrocellulose membrane by using I-blot system and reagents, according to manufacturer’s protocols (Invitrogen). After blocking with 5% (w/v) milk in PBS containing 0.1% Tween-20, membranes were probed with indicated primary antibodies: Nbs1 (1D7, GeneTex, Inc.), Rad50 (13B3, GeneTex, Inc.), Mre11 (NB100–142, Novus Biologicals), SMC1 (ab 9262, Abcam), goat γ-tubulin (C-20, Santa Cruz Biotech., Inc.), PARP-1 (H-250, Santa Cruz Biotech., Inc.), PAR (LP96–10, BD PharMingen, 53BP1 (Santa Cruz Biotech., Inc.), α-tubulin (Sigma Aldrich), BRCA1 (Santa Cruz Biotech., Inc.) overnight at 4°C. Finally, the membranes were rinsed with PBS + 0.1% Tween-20 for 30 min and incubated with appropriate horseradish peroxidase conjugated secondary antibodies (Pierce) for 1 h at room temperature (RT). Followed by washing in PBS + 0.1% Tween-20 for 30 min, protein bands were detected using enhanced chemiluminiscence kit, ECL solution (Amersham Biosciences).

Cell proliferation assays and cell death assessment

Exponentially growing cells were seeded in 6-well culture plates at appropriate densities (20 × 10⁴ cancer cells, 15 × 10⁴ fibroblasts) and attached overnight. PARP-1 inhibitors KU 58948 or AZD2281/olaparib (KuDOS Pharmaceuticals Ltd, now part of AstraZeneca) and CPT (Sigma Aldrich) were added to the medium at indicated concentrations. In case of combined treatment, CPT in various concentrations (1.23 nM, 3.7 nM, 11.11 nM) plus 1 μM PARP-1 inhibitor was added to the medium. Cells were grown for 2 d then split to 50% (to keep the optimal cell density) and grown for additional 2 d with medium only. After 4 d of growth, the cells were counted on cell counter (BD Biosciences, CellQuest Pro Software), and cell number was calculated as a percentage of mock-treated cells. All experiments were performed in triplicates and repeated at least twice. After transfection with siRNA against p53, cells were seeded into 6-well plates, attached ON and treated with PARP-1 inhibitor as described above. Stable Cal51- and MDA-436-derived cell lines expressing NT shRNA or 53BP1 shRNA were seeded in 24-well culture plates (1 × 10⁴ cells/well) in triplicate and exposed to 0.1 and 1.0 μM PARPi (or vehiculum) for a period of time corresponding to four cell cycles. Surviving fraction (SF) was determined by cell counting and compared with control (vehiculum-treated) cells. Results are means from four independent experiments ± s.d. To assess p value, two tailed Student’s t-test was used (p < 0.05).

Proliferation activity of PARP-1 inhibitor-treated CAL51 cells and their derivatives, coupled with cell death evaluation, was assessed by 96-well plate based XTT [2,3-bis(2-methoxy-4-nitro-5-sulfophenyl)-2H-tetrazolium-5-carboxyanilide inner salt] colorimetric assay and lactate dehydrogenase (LDH) activity test, respectively, according to manufacturer’s protocols (Roche). Briefly, cells were seeded at the concentration of 5,000 cells per well and attached overnight. The following day, PARP-1 inhibitor KU 58948 was added to the cells at concentrations of 10⁻⁴ – 10⁻⁸M. After 4 d growth with the drug, medium was removed to the new 96-well plate to evaluate the LDH activity released to the medium by the death cells. Remaining cells were washed twice with PBS, and absorbance was measured 4 h after addition of XTT reagents at 480 and 690 nm using VersaMax spectrophotometer.

The XTT assay was used to evaluate proliferation activity of HCT116 cells after treatment with KU 58948 alone and in combination with Pg-p inhibitor. Briefly, cells were treated with KU 58948 in concentrations 10⁻⁸ – 10⁻⁴ M or pre-treated for 1 h with 5 μg/ ml Verapamil prior to addition of KU 58948. The cells were grown for 4 d, and subsequently the cell viability was assessed using the XTT test according to manufacturer’s instructions.

Clonogenic survival assay and irradiation

Cells were seeded in triplicates into 6-well plates (500 cells per well) and left to stabilize for 24 h. After that, the cells were incubated with 0.1 μM KU 58948 for 24 h and then irradiated. Treated cultures were incubated for additional 10 d in drug-free medium. Finally, the cultures were fixed, stained with crystal violet and colonies containing more than 50 cells were counted. Irradiation sessions were performed at room temperature using Linear accelerator Siemens primus (6MV).

Intracellular detection of KU 58948

HCT116 cells were incubated with 1μM PARP-1 inhibitor KU 58948 and in combination with 5 μg/ ml Verapamil for 1 min and 24 h, respectively. The cell pellet was resuspended in 200 μl of methanol, and an analyte (KU 58948) was extracted from the supernatant using solid-phase extraction microcolumn with a C18 phase (SPE C18). The eluates were resuspended in 30 μl of 10% methanol, vortexed and sonicated for 5 min. One-third of the solution was injected to UPLC-ESI-QqQ system (Acquity UPLC and XEVO MS, Waters). The quantitation of the analyte in the cells was performed using an external calibration with KU 58958 as calibrant.

Immunofluorescence analysis

Immunofluorescence analysis for performed as previously described.⁵³ For in situ analysis of p53, cells were grown on glass coverslips, rinsed briefly in cold PBS and fixed in a 1:1 mixture of ice cold methanol:acetone at RT for 10 min. After drying at RT, cells on coverslips were stained with primary antibody against p53 (DO-01, prepared in house) for 1 h/RT. As secondary antibodies goat anti-mouse coupled to Alexa Flour 488 nm (Invitrogen) were used. After staining procedure, 1 h/ RT (light shielded), the coverslips were dried by ethanol series and mounted on standard microscopic glass using Vecta Shield mounting medium with DAPI (Vector Laboratories). Examination was done using fluorescent microscope (Zeiss Axioplan II). Images were acquired through a PLAN-Neofluar 40×/1.3 oil immersion objective and photographed by digital camera (Cool Snap).

For Rad51 analysis, cells were seeded into 96-well plate (Costar, black, flat bottom) and attached overnight. Cells were treated with PARP-1 inhibitor KU 58948 1 μM for 24 h and subsequently fixed in 4% formaldehyde for 15 min RT. Fixation was followed by cell permeabilization with 0.5% Triton X-100 (in PBS) for 4 min RT and washing 3× in PBS. Cells were then blocked with 3% BSA in PBS for 1 h RT. Primary antibody mix Rad51 (GeneTex Clone 14B4) and cyclinA (Santa Cruz Biotech., Inc.) was added on the cells ON at 4°C. Plate was washed 3× with TBS + 0.1% Tween, and the secondary antibody mix Alexa Fluor 488 and 546 was applied for 1 h RT. Cells were washed again 3× with TBS + 0.1% Tween and incubated with Hoechst (Invitrogen) for 5 min RT. Finally, plate was washed 1× in PBS and cells were covered with 75 μL PBS. Images were acquired on ArrayScan VTI HCS Reader (Thermo Scientific) 20x objective and Rad51 foci in cyclin A-positive cells analyzed by Cellomics software.

For 53BP1 analysis, attached cells in 96-well plate (Costar) were irradiated with 2 Gy and after 30 min fixed as described for Rad51/cyclinA. Cells were incubated with antibody against 53BP1 (Santa Cruz Biotech., Inc.) ON at 4°C. After the washing step, secondary antibody Alexa Fluor 488 was applied for 1 h RT which was followed by 5 min incubation with Hoechst (Invitrogen). The plate was washed 1× in PBS and cells were covered with 75 μL PBS. Representative images were acquired on ArrayScan VTI HCS Reader (Thermo Scientific) 20× objective.

Immunohistochemistry on paraffin sections

For immunohistochemical analysis of archival formalin-fixed, paraffin-embedded human breast carcinomas, the tissue sections were deparaffinized and processed for sensitive immunoperoxidase staining with the primary mouse monoclonal antibody to 53BP1 (kindly donated by Thanos Halazonetis, University of Geneva). The primary antibody was incubated overnight, followed by detection using the Vectastain Elite kit (Vector Laboratories) and nickel sulfate enhancement without nuclear counterstaining, as described previously,^25,33 followed by evaluation of the staining patterns by an experienced oncopathologist.