Inhibition of Parp1 by BMN673 Effectively Sensitizes Cells to Radiotherapy by Upsetting the Balance of Repair Pathways Processing DNA Double-Strand Breaks

Cell culture

All cell lines were maintained at 37°C in an atmosphere with 5% CO₂ and 95% air. Hamster cells CHO10B4 wild-type (WT), V3 (DNA-PKcsm), a gift from Dr. D. Chen (UT Southwestern Medical Center, Dallas, TX), human cells HCT116 wt, A549, U2OS, and U2OS DR-GFP (a gift from Dr. J. Stark [Beckman Research Institute of the City of Hope]; ref. 25) were grown in McCoy's 5A medium supplemented with 10% fetal bovine serum (FBS) and antibiotics. The CHO mutant irs1SF (Xrcc3m), a gift from Dr. L. Thompson (Lawrence Livermore National Labora-tory), and 82-6 hTert (human fibroblast cells), a gift from Dr. M. Loebrich (Darmstadt University of Technology, Germany), were grown in minimum essential medium (MEM) supplemen-ted with 10% FBS and antibiotics. Mouse embryonic fibroblasts (MEF), WT, Parp1^−/− (26), a gift from Dr. Z.Q. Wang (Leibniz Institute on Aging, Jena) and human RPE-1 cells, a gift from Dr. K.W. Caldecott (University of Sussex, UK), were grown in Dulbecco's modified MEM (DMEM) supplemented with 10% FBS and antibiotics. BT-12 (human, atypical teratoid/rhabdoid tumor) cells, a gift from Dr. P. Houghton, were grown in RPMI medium supplemented with 15% FBS and antibiotics. CHLA-9 (human, Ewing sarcoma) cells (a gift from Dr. P. Houghton [Nationwide Children's Hospital, Columbus, Ohio]) were grown in Iscove's Modified Dulbecco's Medium supplemented with 20%FBS, 1% insulin–transferrin–selenium (Invitrogen) and antibio-tics. Cell lines were passaged thrice a week. A549, HCT116, 82-6 hTert, and U2OS-282C were authenticated using Multiplex Cell Authentication by Multiplexion as described (27). Human hTert RPE-1 cells were directly traced to ATCC (28). BT12 and CHLA-9 cells were obtained directly from the original source; the cell lines are not widely distributed. Cells were tested for mycoplasma before freezing using MycoAlert Plus Mycoplasma detection kit from Lonza (LT07-705).

Inhibitors

BMN673 was obtained from Medivation, a collaborator of NCI. The Parp1/2 inhibitor PJ34 (ref. 29; Calbiochem) was used at 5 μmol/L final concentration. 8-(4-Dibenzothienyl)-2-(4-morpholinyl)-4H-1-benzopyran-4-one (NU7441, Tocris), a highly specific DNA-PKcs inhibitor (30), was used at 5 μmol/L final concentration. Rad51 inhibitor B-02 (ref. 31; Merck-Millipore) was used at 25 μmol/L final concentration. AG14361 (specific Parp1 inhibitor; ref. 32), olaparib (Parp1/2 inhibitor), ME0328 (Parp3 inhibitor; ref. 33), and UPF1069 (Parp2 inhibitor; ref. 34) were purchased from Selleckchem and used at 400 nmol/L (AG14361), 1 μmol/L (UPF1069) or 3 μmol/L (all others).

Radiation exposure

Irradiations were carried out with an X-ray machine (GE Healthcare) operated at 320 kV, 10 mA with a 1.65 mm Al filter (effective photon energy approximately 90 kV), at a distance of 50 cm and a dose rate of approximately 1.3 Gy/minute. Cells were returned to the incubator immediately after exposure to IR.

Clonogenic survival assay

Analysis of chromosomal translocations

Cytogenetic analysis was done in irradiated G₂ phase cells as described before (14, 35). Exponentially growing cells were exposed to 1 Gy X-rays and were incubated at 37°C for 4 hours before adding 0.1 μg/mL Colcemid (L-6221, Biochrom AG) for 1 hour. Metaphase-enriched cultures were harvested and processed. Bright field microscopy (Olympus, Vanox-T) and a Meta-Systems station (Altlussheim) with a microscope (AxioImager.Z2, Zeiss) and automated image capture and analysis capabilities were used for scoring chromosome aberrations. Standard criteria were used for scoring chromatid translocations.

Pulsed-field gel electrophoresis

Pulsed-field gel electrophoresis (PFGE) was used to assess induction and repair of DSB. The methodology has been described previously (36). PFGE gels were scanned in a fluorescence scanner and the fraction of DNA released (FDR) from the plug into the lane quantified by ImageQuant 5.2 software (GE Healthcare). Dose response curves are plotted as FDR versus radiation dose. These curves are used to calculate the equivalent Gy dose values (DEQ) for each FDR measured at a given repair time point; repair kinetics are given as plots of DEQ versus time.

Indirect immunofluorescence

Cells were grown on coverslips and treated with the indicated inhibitors for 1 hour prior to IR. Coverslips were rinsed in phosphate-buffered saline (PBS) and fixed with 2% paraformaldehyde (PFA) solution followed by permeabilization with P-solution (100 mmol/L Tris pH 7.4, 50 mmol/L EDTA, 0.5% Triton X-100) at different times after IR. Fixed cells were blocked in PBG solution (0.2% gelatin, 0.5% BSA fraction V in PBS) overnight at 4°C. Cells were incubated with appropriately diluted antibodies as indicated for 1.5 hour at room temperature. After washing with PBS, cells were incubated with AlexaFluor488 (AF488)-, AF568-, or AF647-conjugated secondary antibodies, as required, for 1 hour at room temperature. Cells were finally counterstained with DAPI (50 ng/mL in distilled water) for 10 minutes at room temperature and mounted in Prolong-Antifade mounting media. Samples were scanned on a Leica TCS-SP5 confocal microscope and foci were counted using Imaris software (Bitplane). Sources and dilutions of antibodies used are listed in Supplementary Table S2.

Poly (ADP-ribose) PAR staining

Cells were grown on coverslips for ~48 hours and treated with the indicated inhibitors or DMSO for 1 hour at 37°C. H₂O₂ was diluted in PBS and added at a final concentration of 10 mmol/L. Cells were then processed for immunofluorescence as described above.

DSB repair analysis using genomically integrated reporter constructs

U2OS cells (2 × 10⁶) containing an HRR reporter construct (25) were transfected with 1 μg I-SceI expression plasmid and allowed to attach for 90 minutes. Then, cells were treated with indicated inhibitors. Seventy-two hours later, cells were analyzed for GFP expression by flow cytometry.

Statistical analysis

Graphs were created in SigmaPlot 11.0. Statistical significance was determined using Student t test available in SigmaPlot 11.0. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Among Parpi, BMN673 exerts the strongest radiosensitization with a short, time and sequence flexible exposure

We began our investigations with CHO and mouse cells to take advantage of the large repertoire of available DSB repair mutants that enable analysis of radiosensitization mechanisms. CHO cells preexposed to a variety of Parpi for 1 hour, irradiated and plated immediately thereafter in growth medium, also supplemented with inhibitors, show surprisingly variable degrees of radiosensitization (Fig. 1A). Strikingly, BMN673 is by far the strongest radiosensitizer leading to radiosensitivities only known from cNHEJ or HRR mutants (Fig. 1B). All inhibitors are used at concentrations sufficiently high to reduce H₂O₂ induced parylation below detection (Supplementary Fig. S1A). We conclude that differences in radiosensitization derive from mechanisms operating beyond simple Parp inhibition.

BMN673 is an effective radiosensitizer with 10 nmol/L generating radiosensitization comparable to that of 3 μmol/L olaparib (compare Figs. 1A and C). Notably, the radiosensitizing effect of BMN673 plateaus at about 50 nmol/L, a concentration that is clinically achievable (37). The cytotoxicity data for CHO cells after treatment with various concentrations of BMN673 are given in Supplementary Table S3, where it can be seen that the IC₅₀ is about 200 nmol/L. BMN673-induced radiosensitization derives from inhibition of Parp1, as specific inhibition of Parp2 with UPF1069 or of Parp3 with ME0328 fails to generate statistically significant levels of radiosensitization (Fig. 1D, see Supplementary Table S1 for DMF₁₀ values). In addition, Parp1^−/− MEFs show no radiosensitization by BMN673 (Supplementary Fig. S1B).

When repair inhibitors are combined with RT, it is important to ensure that their concentrations in the blood will be sufficiently high at the time of patient irradiation and that they will be maintained high for several hours after irradiation—in order to efficiently interfere with DSB processing. The precise administration schedule will depend on their behavior as repair inhibitors and their pharmacokinetics. To begin understanding the properties of BMN673 as radiosensitizer, we tested the drug exposure times required for maximum effect in CHO cells. Strikingly, we discovered that treatment with 50 nmol/L for only 1 hour prior to IR is sufficient to generate nearly maximum radiosensitization (Fig. 2A; see Supplementary Table S1 for DMF₁₀ values). An experiment in which cells were first plated for colony formation, treated with BMN673 for 1 hour, irradiated and incubated for up to 72 hours before transferring to BMN673 free growth medium, allows similar conclusions (Supplementary Fig. S1C).

We inquired whether radiosensitization established in CHO cells also holds for human tumor cell lines. Because our team participates in the NRG sarcoma group, which focuses on the development of novel treatments for tumors of mesenchymal origin, we tested BMN673 radiosensitization in tumor cell lines of such origin. Preexposure for 1 hour with BMN673 of human rhabdoid BT12 cells fails to generate detectable radiosensitization, but continuous treatment generates marked radiosensitization (Fig 2B; see Supplementary Table S1 for DMF₁₀ values). We conclude that exposure time for maximum radiosensitization will be cell line dependent and that biomarkers of response are required for optimal administration in the clinical setting (see below). Notably, BMN673 is a much stronger radiosensitizer than PJ34, which is completely ineffective in BT12 cells (Fig. 2C; Supplementary Table S1 for DMF₁₀ values).

The human Ewing sarcoma CHLA9 cells also show considerable radiosensitization after 1 hour, and even more after continuous, treatment with low concentrations (10 nmol/L) of BMN673 (higher concentrations were toxic), although PJ34 is relatively ineffective (Fig. 2D; Supplementary Table S1 for DMF₁₀ values). Considering that BMN673, as single agent, shows limited activity against xenografts grown from these cells (38), the results suggest benefits from combining BMN673 with IR in these childhood cancers.

We also examined BMN673 radiosensitization in human tumor cell lines widely used in mechanistic studies of the DNA damage response (Fig. 3). We observed marked and practically maximal radiosensitization in U2OS, HCT116, and A549 cells treated with 25 nmol/L BMN673 for 1 hour before irradiation and then immediately plated for colony formation in medium without BMN673 (Fig. 3A-C; Supplementary Fig. S2 for results on inhibition of parylation and Supplementary Table S1 for DMF₁₀ values). These results extend the validity of the results presented above with CHO cells.

An important consideration for the clinical application of BMN673 is the radiosensitization of normal cells that will lead to normal tissue toxicity. We therefore tested BMN673 radiosensitization in two normal human cell lines. Normal human fibroblasts 82-6 hTert and normal retinal epithelial cells RPE-1 show no signs of radiosensitization by BMN673 (Fig. 3D and E). In these cells, BMN673 strongly inhibits H₂O₂-induced parylation, confirming inhibition of Parp1 (Supplementary Fig. S2). We conclude that BMN673 may act specifically on tumor cells, fulfilling a key requirement for a clinically relevant, tumor-specific radiosensitizer.

IR is applied in daily fractions in the clinical setting. We examined, therefore, the effect of IR dose fractionation on BMN673-mediated radiosensitization. The results summarized in Fig. 3F, G, and H again show cell line-dependent reduction in BMN673-mediated radiosensitization following the application of 6 Gy in 3 fractions of 2 Gy separated by 4 or 24 hours.

Collectively, the low and clinically achievable concentrations of BMN673 for maximum radiosensitization, the specificity for tumor cells, the mere 1 hour or so of required drug exposure, and the embedded flexibility in the timing of IR and drug administration uncover properties never before reported for a radiosensitizer. Together, they define the "perfect" scenario for the combined application of BMN673 with IR in cancer treatment and make elucidation of the underpinning molecular mechanisms a high priority task. Work along these lines is described next.

BMN673 suppresses DSB repair at low IR doses

We used γ-H2AX immunofluorescence staining to assess DSB processing at low doses of IR. CHO cells exposed to 2 Gy show pronounced increase in γ-H2AX foci formation, as compared with nonirradiated controls (Supplementary Fig. S3A and S3B), which reaches a maximum at 1 hour after IR. The number of foci decreases and reaches values only slightly above background at 8 hours (Supplementary Fig. S3A), suggesting efficient processing of DSBs. A similar response is also observed in cells incubated with AG14361 for 1 hour before IR and 8 hour thereafter, indicating no detectable effect of the inhibitor on DSB processing (Supplementary Fig. S3A and S3B). Notably, treatment with BMN673 increases the number of γ-H2AX foci scored at 1 hour and strongly inhibits their resolution in the ensuing 8 hours, documenting profound global inhibition of DSB processing. Also, the intensity of the γ-H2AX foci is increased after treatment with BMN673, an effect also observed with AG14361. Similar results are also obtained with A549 cells (Supplementary Fig. S3C). Although our results clearly implicate inhibition of DSB processing in BMN673 radiosensitization, they do not provide information as to whether the effect relies on inhibition of a specific DSB repair pathway (cNHEJ, HRR, or altEJ). Pathway specificity of BMN673-induced DSB repair inhibition was therefore studied next.

BMN673 favors HRR by promoting hyper-resection and suppressing recruitment of 53BP1 at DSBs

To study BMN673 effects on HRR, we measured DNA end resection (referred to as "resection" here), the first step in HRR, by means of RPA foci formation—a widely accepted marker, detected here as RPA70 immunofluorescence.

Although a relatively small number of RPA70 foci forms in cells exposed to 2Gy, a dramatic increase occurs after incubation with BMN673, particularly at 3 and 6 hours after IR (Fig. 4A-C). Treatment of nonirradiated cells with BMN673 generates a modest increase in RPA70 foci; therefore, Fig. 4C shows the results of Fig. 4B after subtraction of this background. The dramatic increase in RPA foci formation suggests a shift in the fraction of DSB shunted for resection, i.e., a shift from cNHEJ to either HRR or altEJ.

53BP1 is thought to suppress resection at DSB (39), and this prompted us to inquire whether the increase in resection observed is accompanied by a decrease in 53BP1 foci. In the absence of IR, CHO cells display low numbers of 53BP1 foci that increase abruptly 1 hour after exposure to 2 Gy (Fig. 4D-F). 53BP1 foci numbers gradually decrease in irradiated cells at 3 and 6 hours, indicating completion in DSB processing. Notably, treatment with BMN673 causes a large increase in 53BP1 foci even in nonirradiated cells, which becomes apparent at 1 hour and reaches values over 5-fold above background after 3 and 6 hours of incubation (Fig. 4E). IR causes only a small additional increase in 53BP1 foci in this setting (Fig. 4F), an effect clearly evident after subtraction of the background in nonirradiated cells. We conclude that BMN673 suppresses 53BP1 foci formation at early times after IR.

EdU (5-ethynyl-2′-deoxyuridine) labeling of S-phase cells (Supplementary Fig. S4A) shows that 53BP1 foci developing in nonirradiated cells after treatment with BMN673 are specific for S-phase cells and reflect therefore effects of Parpi on DNA replication. IR exposure in S-phase causes only a modest further increase in 53BP1 foci formation. In EdU-negative cells (Supplementary Fig. S4B), BMN673 has no effect on 53BP1 foci formation in the absence of IR and only a modest effect after exposure to IR. Thus, suppression of IR-dependent 53BP1 foci formation by BMN673 is specific for S-phase cells.

For direct analysis of the effect of BMN673 on HRR, we used immunofluorescence to score formation and resolution of foci formed by Rad51, the central mitotic recombinase in eukaryotic cells. CHO cells exposed to 2 Gy develop Rad51 foci 3 hours later demonstrating engagement of HRR(Fig. 4G). Rad51 foci decrease at 8 hours signifying the gradual completion of this form of processing. Notably, treatment with BMN673 causes initially a 3-fold increase in Rad51 foci (Fig. 4G and H), suggesting increased engagement of HRR. However, at 8 hours, although a clear reduction in Rad51 foci is measured, residual numbers remain high, suggesting that processing by HRR is incomplete in BMN673-treated cells. Similar results are obtained at higher concentration of BMN673 (Fig. 4G). Rad51 foci colocalize with γ-H2AX foci as would be expected from DSB processing by HRR (Supplementary Fig. S5A and S5B).

We introduced a functional reporter assay to study the effect of Parpi on HRR. U2OS-DRGFP cells have integrated a construct in their genome that carries two nonfunctional copies of the GFP gene (ref. 25; Supplementary Fig. S6A). Introduction of a DSB within this construct by transient expression of I-Sce I, a restriction endonuclease recognizing a DNA sequence normally not present in the human genome, and processing by HRR results in expression of GFP that is detected by flow cytometry. Supplementary Fig. S6A shows that BMN673, added 1.5 hours after transfection and kept for the duration of the experiment (72 hours), has no effect on HRR when tested at 50 nmol/L and even 10 μmol/L fails to generate any effect. HRR remains unaffected also by olaparib, AG14361, or PJ34. In contrast to Parpi, a Rad51 inhibitor (B02) abrogates HRR and DNA-PK inhibitor (NU7441) exerts the expected increase in the frequency of HRR events.

How could BMN673, on the one hand, increase resection and Rad51 foci formation and, on the other hand, leave HRR as measured by functional assays unchanged? We considered the possibility that inhibition of cNHEJ by BMN673 allows resection at DSB that feeds futile HRR. Futile HRR events, in turn, may be rescued by altEJ causing translocations (40). We therefore examined the effect of BMN673 on translocations and cNHEJ.

Contrary to other Parpi, BMN673 increases IR-induced translocations

Our previous studies show that IR-induced translocations, particularly those generated under conditions of HRR or cNHEJ deficiency, require Parp1-mediated altEJ (14, 35) and are reduced after treatment with PJ34 or olaparib (14, 35, 41). We inquired whether BMN673 exerts similar effects. CHO cells irradiated in G₂ phase and treated with NU7441 to inhibit cNHEJ show a marked increase in the incidence of chromosomal translocations as they reach metaphase 4 hours after IR (Fig. 5A and B). Treatment of these cells with PJ34, olaparib, or AG14361 causes the expected decrease in translocations. Strikingly, BMN673 treatment in combination with NU7441 has opposite effects of increasing the incidence of translocations by nearly a factor of two (Fig. 5B). BMN673 also increases translocation in HCT116 cells exposed to 1 Gy (Fig. 5C). Notably, RPE-1 cells show no increase in translocation formation after BMN673 treatment (Fig. 5C), in line with the lack of radiosensitization shown in Fig. 3E. We conclude that, compared with other Parpi, BMN673 has a distinct spectrum of activities on IR-induced lesions.

BMN673 compromises cNHEJ at high doses of IR

DSB generated in CHO cells after exposure to 20 Gy are processed with fast kinetics in untreated cells (Fig. 6A) with nearly 90% being removed within 1 hour. Remaining DSBs are removed within 8 hours with slower kinetics. It is thought that DSB processing detected under these conditions by PFGE mainly reflects the function of cNHEJ (42).

Notably, although PJ34 has no detectable effect and olaparib only inhibits the slow component, BMN673 has a pronounced effect on DSB processing (Fig. 6A). Indeed, the inhibition achieved by BMN673 is indistinguishable from that of NU7441, a specific inhibitor of DNA-PKcs (Fig. 6A). Because combined treatment with NU7441 and BMN673 generates additional inhibition, we conclude that beyond cNHEJ, BMN673 also inhibits altEJ, which is also assessed in this type of experiments (36).

BMN673 inhibits DSB processing to a greater degree than PJ34 and olaparib in MEFs as well (Fig. 6B), an effect that is dependent on Parp1, as it is absent in Parp1^−/− mutants (Fig. 6C). BMN673 inhibits DSB processing stronger than olaparib or PJ34 in V3 as well, a CHO mutant defective in DNA-PKcs (Fig. 6D). Finally, BMN673 exerts marked inhibition of cNHEJ in CHLA9 cells and a modest inhibition in BT12 cells (Supplementary Fig. S6B).