Sensitization to Radiation and Alkylating Agents by Inhibitors of Poly(ADP-ribose) Polymerase Is Enhanced in Cells Deficient in DNA Double-Strand Break Repair

The cytotoxic effects of ionizing radiation and alkylating chemotherapeutic drugs are mediated through DNA damage. Ionizing radiation induces single-stranded and double-stranded DNA breaks in an approximate ratio of 25:1 (1), whereas alkylating agents induce base damage that gives rise to single-strand breaks (SSB) and stalled replication forks. SSB induced by ionizing radiation or alkylating agents are predominantly repaired by base excision repair, whereas double-strand breaks (DSB) are repaired mainly by the nonhomologous end-joining (NHEJ) pathway (2). DSB associated with DNA replication or occurring during G₂ phase are also repaired by homologous recombination (2). Radiation-induced DSB are frequently associated with base or sugar damage, repair of which requires contributions from base excision repair proteins such as polynucleotide kinase (3). Furthermore, DSB can arise from SSB occurring in close proximity or as a consequence of DNA replication.

Because the cytotoxicity of DNA damaging agents correlates with the number of unrepaired DSB (4), inhibition of DNA repair represents a mechanism by which the therapeutic effects of these agents might by enhanced. Such enhancement must be tumor specific if outcomes are to be improved. Whereas cancer cells are typically characterized by aberrant cell cycle checkpoint control, defective DNA repair pathways, and accelerated proliferation rates, normal tissue cells have intact cell cycle checkpoints and DNA repair pathways (5). In addition, some critical normal tissues are composed almost entirely of nonreplicating cells. A sensitizing agent that was effective only in cells with high replication rates and/or DNA repair defects would therefore have great clinical potential.

Poly(ADP-ribose) polymerase 1 (PARP-1) is a DNA damage sensing protein that binds to SSB (reviewed in ref. 6), a process that activates its catalytic function and facilitates DNA repair. Inhibition of PARP activity reduces the rate of SSB repair and increases cellular sensitivity to ionizing radiation and alkylating agents such as methylmethane sulfonate (MMS; refs. 7–10). We have previously shown that the radiosensitizing effect of PARP inhibition requires DNA replication and that enhanced conversion of unrepaired SSB to DSB during S phase is the likely mechanism (11). Another study showed that PARP inhibition exacerbates the cytotoxicity of the alkylating agent temozolomide by enhancing conversion of SSB to DSB during S phase (12).

The cellular consequences of PARP inhibition are dictated by DNA repair proficiency. This is illustrated by the extreme sensitivity to PARP inhibitors of tumor cells deficient in BRCA1 or BRCA2 (13, 14). This phenomenon is caused by accumulation of endogenously arising DNA damage that would normally be repaired by PARP–dependent processes. In replicating cells, this damage triggers replication fork collapse, repair of which requires homologous recombination and is therefore dependent on BRCA1 and BRCA2. The impact of DNA repair defects on the sensitizing effects of PARP inhibitors in combination with DNA damaging agents remains unclear, however. Various studies have indicated that PARP inhibition exacerbates radiosensitivity (15) or chromosomal instability associated with NHEJ deficiency (16), but the evidence is inconsistent (17). Although it has been reported that PARP-1 binds to DSB and interacts with NHEJ proteins (18), there is no evidence that PARP activity is required for resolution of DSB in NHEJ proficient cells. However, an alternative repair pathway has been postulated to execute end joining of DSB in cells that are deficient in NHEJ (19). This process has been termed "backup" or "alternative" end joining, and there is evidence that it is compromised by inhibition of PARP (20). The aim of this study was to investigate the relationship between NHEJ integrity and the sensitizing effects of PARP inhibition.

NHEJ comprises core components (Ku70/80, DNA-protein kinase catalytic subunits, DNA Ligase IV, XRCC4, and XLF) that are responsible and sufficient for repair most radiation-induced DSB (reviewed in ref. 21). Loss of function of these proteins is associated with marked radiosensitivity and immunodeficiency, the latter as a consequence of impaired variable (diversity) joining recombination. A subset of radiation-induced DSB (∼15%) requires more extensive processing. Repair of these breaks occurs with slow kinetics and requires additional factors, including Artemis and ataxia telangiectasia mutated (ATM). Despite the relatively small size of this subset, defects in Artemis function are associated with significant radiosensitivity (22, 23).

To elucidate the relative impact of defects in core NHEJ and the Artemis/ATM pathway on the sensitizing properties of PARP inhibition, we evaluated the effects of olaparib (AZD2281, previously KU-005946) on radiation and MMS responses in repair-defective cell lines. Olaparib is a potent and specific inhibitor of PARP-1 and PARP-2 and is currently undergoing phase II clinical trials as a single agent (24). Our data indicate that the sensitizing effects of olaparib are enhanced in cells with defective DSB repair and support the hypothesis that the sensitizing effects of PARP inhibitors will be more pronounced in tumors than in normal tissues.

CJ176 primary and human telomerase reverse transcriptase (hTERT; immortalized derivative) are Artemis-deficient human fibroblast cell lines. AT5-BIVA are ATM-deficient, SV40 immortalized human fibroblasts. HSC62 (kindly provided by Dr. M. Digweed) are primary fibroblasts from a patient with a homozygous mutation (IVS19-1 G to A) in BRCA2. 48BR primary and 1BR hTERT cells are wild-type (WT) controls. WT, Artemis^−/−, ATM^−/−, and LigaseIV^−/−p53^−/− mouse embryo fibroblasts (MEF) were kind gifts from Dr. F. Alt (Artemis^−/−), Dr. T. Mak (WT, ATM^−/−), and Dr. P. McKinnon (LigaseIV^−/−p53^−/−). Human hTERT cell lines and MEFs were cultured in minimal essential medium (MEM) or α-MEM respectively, supplemented with 10% fetal calf serum, 2 mmol/L l-glutamate, penicillin (100 units/mL), and streptomycin (0.1 mg/mL). Primary cell lines and HeLa cells (obtained from ECACC) were cultured in MEM with 15% fetal calf serum. All cell culture reagents were from Gibco; other reagents were from Sigma unless otherwise stated.

Cells were irradiated in medium using either a ¹³⁷Cs γ source (Gammacell 1000; dose rate, 7.5 Gy/min) or 250 kVp X-rays delivered at 12 mA (dose rate, 0.5 Gy/min). Olaparib (gift of KuDOS Pharmaceuticals/AstraZeneca) was used at an end concentration of 500 nmol/L, a noncytotoxic dose that abolished PARP activity in the cell lines indicated (Fig. 1A). PJ34 (Calbiochem) and KU55933 (ATM inhibitor; gift of KuDOS Pharmaceuticals/AstraZeneca) were used at end concentrations of 10 μmol/L. To inhibit DNA replication during PARP inhibition, aphidicolin (5 μmol/L) was administered for three hours, commencing 1 hour preradiation. For clonogenic survival assays using MMS, aphidicolin and olaparib were administered 1 hour before, 1 hour during, and 1 or 22 hours after treatment.

Cells were trypsinized, seeded into 4-cm plates, and allowed to adhere at 37°C before drug treatment or irradiation. For radiation or MMS assays, cells were exposed to olaparib for 24 hours, commencing 2 hours pretreatment. For continuous olaparib exposure, medium was replaced every 48 hours to ensure constant levels of the drug. Plates were incubated at 37°C for 7 to 8 days then stained with methylene blue. Colonies of 50 cells or more were counted manually, and survival curves were derived from a minimum of three independent experiments, each done in triplicate. Surviving fraction was corrected for independent cytotoxic effects of olaparib or aphidicolin, except in Supplementary Fig. S2A, wherein significant toxicity was observed. Linear quadratic (for radiation) or exponential (for MMS) models were fitted to the data sets to generate survival curves. Radiation or MMS doses associated with surviving fractions of 10%, 37%, or 50% were calculated from the fitted curves. Sensitizer enhancement ratios (SER; ref. 25) for olaparib were calculated as in Eq. 1:

wherein Dx is the dose of ionizing radiation or MMS associated with a surviving fraction of x%. Surviving fraction after 2 Gy (SF₂) values were obtained from fitted survival data, and SF₂ ratios with and without olaparib were calculated.

Alkaline comet assays were done as previously described (26). Cells suspended in medium (2.8 × 10⁵ cells/mL) were exposed to 30-Gy γ-irradiation or to 200 μmol/L H₂O₂ for 20 minutes on ice and incubated for the indicated repair periods at 37°C in medium. Where indicated, cells were incubated with 500 nmol/L olaparib for 2 hours before DNA damage and during repair incubations. DNA strand breakage was expressed as "comet tail moment" (27). Tail moment was measured for 100 cells per sample using Comet Assay IV software (Perceptive Instruments). Average damage remaining of at least three independent experiments was calculated.

γ-H2AX foci assays were done specifically in G₁ cells as previously described (11). Cells were stained with centromere protein F (1:100; Santa Cruz) for human cells or phospho-histone H3 (1:300; Upstate Biotechnology) for mouse cells to identify G₂ phase cells. Mitotic cells were identified by morphology, and S-phase cells were identified by diffuse low-level γ-H2AX staining. Human fibroblasts (Fig. 5A) were treated with 3 μmol/L aphidicolin after irradiation to inhibit DNA replication and improve identification of S-phase cells by intensifying γ-H2AX staining (28). Thus, DSB repair in G₁ cells was monitored by enumerating γ-H2AX foci in cells that were deemed not to be in G₂, S, or mitosis.

Slides stained as above were visualized using a Nikon Eclipse 50i microscope or a Zeiss Axioplan microscope at ×40 magnification, and image processing was done using Simple PCI software. Signal intensity within selected regions of interest was analyzed using NIH Image-J. MMS-induced γ-H2AX staining intensity was calculated by subtracting mean signal in untreated nuclei from that in MMS-treated nuclei. H₂O₂-induced poly(ADP-ribose) synthesis was measured as mean fluorescence intensity of poly(ADP-ribose) signal per nucleus.

Cells were harvested by scraping and fixed in ice-cold 70% ethanol before staining with propidium iodide (0.45 μg/mL), RNase (0.45 mg/mL), and 0.045% Tween. Resuspended cells were analyzed for DNA content on a FACS Canto flow cytometer; data was processed with FACS Diva software (Becton Dickinson).

All data were derived from at least three independent experiments. For repair foci, at least 30 nuclei were counted for each experiment, except where stated, and statistical significance was determined using Student's two-tailed t test. For clonogenic survival experiments, mean surviving fraction ± SE was plotted. SER and SF₂ values were derived from individual experiments to enable the calculation of mean values and SEM. Mean SER values were assessed for significance by Mann-Whitney U test.

Continuous exposure to PARP inhibition is toxic to homologous recombination–deficient cells, and shorter exposures increase the sensitivity of repair proficient cells to MMS or ionizing radiation. To investigate the influence of NHEJ on these outcomes, we tested the effect of the PARP inhibitor olaparib on survival responses of WT, Artemis, ATM, and Ligase IV–deficient MEFs. We first showed that all four cell lines were deficient in p53 protein (data not shown) and that asynchronous undamaged populations showed no significant differences in cell cycle distribution (Fig. 1A). Quantitative immunofluorescent detection of poly(ADP-ribose) synthesis after treatment with hydrogen peroxide was done (Supplementary Fig. S1), showing no significant difference in PARP activity between cells lines, and that 0.5 μmol/L olaparib inhibited PARP activity to baseline levels (Fig. 1B). Twenty-four-hour exposure to a range of olaparib doses had minimal effect on clonogenic survival (Fig. 1C), but ATM^−/− cells were significantly more sensitive to prolonged PARP inhibition than the other cell lines (Fig. 1D; P < 0.01; all doses). This is consistent with previous findings (29, 30) and can be explained by the involvement of ATM in the homologous recombination pathway (29) and by defects in S-phase checkpoint signaling that have been clearly defined in ATM-deficient cells. In contrast, Artemis^−/− and Ligase IV^−/− cells were no more sensitive to continuous PARP inhibition than WT MEFs. Hence, NHEJ does not play an important role in repair of endogenously arising damage in replicating cells, even in the presence of PARP inhibition.

Our previous data supported the hypothesis that PARP inhibition promotes replication-dependent conversion of unrepaired SSB to potentially toxic DSB (11) To investigate whether NHEJ influences the response to these DSB, we first measured the effect of PARP inhibition and NHEJ status on sensitivity to the alkylating agent MMS, which induces predominantly SSB (Fig. 2A). WT MEFs were sensitized by PARP inhibition as expected (SER₃₇ = 2.36), but the magnitude of this effect was markedly increased in Artemis^−/− (SER₃₇ = 7.90; P < 0.05) and ATM^−/− (SER₃₇ = 4.48; P < 0.05) cells (Fig. 2A; Table 1). These findings are consistent with a model whereby PARP inhibition promotes the generation of DSB from MMS-induced SSB and that these lesions are more toxic in the absence of ATM or Artemis. Ligase IV^−/− cells were markedly sensitized by olaparib (SER₃₇ = 14.29; P < 0.05), indicating that core NHEJ is required for repair of lesions arising under these conditions. In the absence of PARP inhibition, Ligase IV–deficient MEFs showed mild sensitivity to high doses of MMS, indicating that MMS-induced lesions, when present at high density, can generate DSB that are repaired by core NHEJ. Artemis- and ATM-deficient cells were no more sensitive to MMS than controls, indicating that MMS-induced lesions do not require processing by these proteins. The mechanisms underlying these findings are explored in more detail later.

Ionizing radiation induces DNA damage comprising SSB and DSB in a ratio of ∼25:1, so the cytotoxic effects of DSB arising from SSB are obscured by those of directly induced DSB. In addition, radiation induces far fewer SSB than MMS at the doses used in these experiments. Taking this into account, the clonogenic survival data presented in Fig. 2B are consistent with the MMS observations. In the absence of olaparib, Ligase IV^−/− cells were highly sensitive to radiation whereas Artemis^−/− and ATM^−/− cells showed intermediate sensitivity as expected. Whereas PARP inhibition had a modest radiosensitizing effect on WT MEFs (SER₅₀ = 1.33), consistent with previous reports (31), higher levels of radiosensitization were observed in Artemis (SER₅₀ = 1.61; P < 0.05) and Ligase IV–deficient cells (SER₅₀ = 1.63; P < 0.05; Fig. 2B; Table 1). Sensitization of ATM-deficient cells was also greater than in WT cells, but this did not reach statistical significance (SER₅₀ = 1.56).

To validate the relevance of these findings to the treatment of cancer, the impact of the ATM inhibitor KU-55933 on the radiosensitizing effects of olaparib was measured in HeLa cells. The magnitude of the sensitizing effect of olaparib was enhanced on a background of ATM inhibition (SER₅₀ = 1.87 compared with 1.50; Fig. 2C; Table 1). For additional clinical relevance, surviving fractions at 2 Gy (SF₂) were calculated for all survival experiments and SF₂ ratios in the presence and absence of olaparib derived (Table 1). In all cases, the ratio was markedly increased in DNA repair–defective backgrounds, with the greatest effect observed in Ligase IV–deficient MEFs.

To investigate whether the radiosensitizing effects of olaparib in NHEJ-deficient cells are mediated by replication-dependent conversion of SSB to DSB, the DNA polymerase inhibitor aphidicolin was used to inhibit DNA replication during the period of PARP inhibition. Radiosensitization by olaparib was completely rescued by aphidicolin in WT, Artemis^−/−, and ATM^−/− cells (Fig. 3A-C) but was unaffected in Ligase IV–deficient cells (Fig. 3D). This indicates that DNA replication is not necessary for radiosensitization of Ligase IV–defective cells.

To eliminate the possibility that the sensitizing effects of PARP inhibition in ATM- and Artemis-deficient cells might reflect direct involvement of either protein in SSB repair, alkaline comet assays were done in ATM^−/− and Artemis^−/− MEFs (Supplementary Fig. S2). Both cell lines exhibited normal repair kinetics after treatment with ionizing radiation or the SSB-inducing agent hydrogen peroxide. Addition of olaparib delayed repair as expected, but the effect was no greater in ATM- or Artemis-deficient cells than in WT controls. Together with the replication-dependent effect on survival, these observations implicate increased conversion of SSB to cytotoxic DSB during DNA replication as the primary mechanism underlying the enhanced radiosensitizing effect of PARP inhibition in ATM^−/− and Artemis^−/− cells.

We hypothesized that the increased effects of PARP inhibition in Artemis-deficient cells were a consequence of defective repair of replication-dependent DSB. To verify this, we exposed cells to MMS and quantified levels of phosphorylated histone H2AX (γ-H2AX) immunofluorescence in S-phase MEFs. In G₀ and G₁ cells γ-H2AX foci have been shown to correlate well with DSB numbers as measured by neutral comet assay, pulsed field gel electrophoresis, and survival after ionizing radiation (32). However, phosphorylation of H2AX also occurs diffusely in response to damage other than DSB and during apoptosis and may be stimulated by replication fork collapse. As a result, individual γ-H2AX foci cannot be accurately detected in S-phase cells. To quantify replication-associated DNA damage, therefore, we analyzed total γ-H2AX staining intensity in S-phase nuclei. These were identified by negative phospho-histone H3 and diffuse γ-H2AX staining (28). In combination with MMS, PARP inhibition increased total γ-H2AX signal in WT and Artemis^−/− cells, with similar increases occurring up to 4 hours after MMS treatment. At 24 hours, however, S-phase Artemis^−/− cells exhibited significantly greater γ-H2AX signal than WT (Fig. 4A; quantified in Fig. 4B; P < 0.05), indicating impaired resolution of damage. By measuring γ-H2AX staining intensity in phospho-histone H3–positive cells, we showed that the excess damage persisted as cells progressed into G₂ (Fig. 4C), with Artemis^−/− cells staining much more intensely than WT (P < 0.05 at 24 hours; Fig. 4C and D). These findings are consistent with the clonogenic survival data (Fig. 2A) and support the hypothesis that PARP inhibition promotes replication-associated conversion of SSB to DSB that require Artemis for efficient repair.

Survival data indicated that DNA replication was not required for radiosensitization of Ligase IV–deficient cells by olaparib (Fig. 3D). To explore replication-independent mechanisms, we measured the effect of PARP inhibition on induction and resolution of γ-H2AX foci after ionizing radiation, specifically in G1 phase nuclei. Cells were costained for centromere protein F (human fibroblasts) or phospho-histone H3 (MEFs) to identify G₂ (centromere protein F or phospho-histone H3 positive) and G₁ (centromere protein F or phospho-histone H3 negative) nuclei (28). S-phase nuclei were identified by intermediate centromere protein F staining (human cells) and diffuse background γH2AX staining. Mitotic or apoptotic cells were identified by their distinctive 4′,6-diamidino-2-phenylindole staining pattern. By excluding these nuclei, the specificity of the γH2AX foci assay for true DSB was increased. To further substantiate DSB specificity, we showed that the induction of γ-H2AX foci by ionizing radiation was abolished by simultaneous inhibition of the DSB-dependent kinases ATM and DNA-PK in human fibroblasts and MEFs (Supplementary Fig. S3A and B). Hence, although ionizing radiation induces high levels of both SSB and DSB, H2AX phosphorylation occurs only in response to signaling events that are specifically activated by DSB.

As reported previously (33), G₁ phase Artemis- and ATM-deficient cells exhibited a partial DSB repair defect (Fig. 5A). PARP inhibition did not affect induction or resolution of γ-H2AX foci, consistent with the model that DNA replication is required to generate excess double-stranded lesions in these cell lines.

G₁ phase Ligase IV– and Ku80-deficient cells exhibited a more marked but not complete DSB repair defect (Fig. 5B). In these cells, PARP inhibition significantly increased the number of persistent γ-H2AX foci (Fig. 5B; P < 0.01 at 4 and 8 h for Ligase IV^−/− cells; P < 0.05 and P < 0.01 at 4 and 8 h, respectively, for Ku80^−/− cells). Indeed, residual DSB repair function seemed to be abolished by PARP inhibition in these NHEJ-deficient cells. This supports the hypothesis that the activity of an alternative end-joining pathway is promoted in the absence of either Ku80 or Ligase IV and eliminated by inhibition of PARP (20, 34).

To exclude the possibility that these observations reflected effects on H2AX phosphorylation or focus dynamics rather than DSB repair, we showed that treatment with the PARP inhibitor did not affect the number of foci induced by irradiation (Supplementary Fig. S4A), the rate at which they resolved (Fig. 5A and Supplementary Fig. S4A), or the signal intensity per focus at 30 minutes or 8 hours after irradiation (Supplementary Fig. S4B and C).

Our observation that Ligase IV^−/− cells were sensitive to high doses of MMS in the absence of PARP inhibition (Fig. 2A) and were more profoundly sensitized to MMS by PARP inhibition than other repair-defective cells raised the possibility that additional mechanisms might be operating. We hypothesized that PARP inhibition would delay repair of SSB induced by MMS and increase the probability of unrepaired lesions giving rise to DSB by interacting either with adjacent SSB or with transcription machinery (35). Either mechanism might be exacerbated by persistent binding of inhibited PARP at the damaged sites. To investigate this hypothesis, we monitored the appearance of γ-H2AX foci in WT and repair-deficient cells. Effects of DNA replication were eliminated by restricting the analysis to cells in G₁ phase, and the specificity of γ-H2AX foci for DSB was validated as described above and illustrated in Supplementary Fig. S3.

PARP inhibition was required for generation of γ-H2AX foci in G₁ phase WT fibroblasts exposed to MMS for 1 hour (Fig. 6A; Supplementary Fig. S5). PARP inhibition significantly increased DSB formation between 1 and 4 hours after MMS treatment (P < 0.001, P < 0.05, and P < 0.01 at 1, 2, and 4 h, respectively; Fig. 6B). In WT cells, γ-H2AX foci were few (6-8 per cell) and were almost completely repaired within 8 hours (Fig. 6B), but in Ligase IV–deficient MEFs, foci continued to accumulate over 8 hours (Fig. 6C). This accumulation was observed in the presence and absence of the PARP inhibitor, but the increase in foci associated with PARP inhibition was much greater than in WT cells and was highly significant (P < 0.05 at 1 h; P < 0.001 at 2, 4, and 8 h). At 24 hours, cells exhibited intense γ-H2AX staining indicative of apoptosis and foci could not be counted. In G₁-phase Artemis-deficient human fibroblasts (CJ176), the kinetics of induction and repair of γ-H2AX foci were essentially normal and the effect of PARP inhibition was no greater than in WT fibroblasts (Fig. 6D), indicating that DNA replication is required to generate DSB that are substrates for Artemis. Similarly, BRCA2-mutant homologous recombination–defective cells (HSC62) showed normal DSB repair kinetics and the same response to PARP inhibition as WT (Fig. 6E). Hence, DSB arising from MMS-induced damage in the absence of replication are repaired by Ligase IV–dependent NHEJ.

These data show that treatment with MMS yields low levels of DSB, most of which are not normally detected because they are rapidly repaired by core NHEJ. In the presence of a PARP inhibitor, DSB induction is greatly increased and these lesions are also repaired by Ligase IV–dependent NHEJ. These novel observations provide a possible explanation for the relative sensitivity of Ligase IV^−/− MEFs to high doses of MMS and the gross sensitizing effect of PARP inhibition (Fig. 2A).

PARP inhibitors have clinical potential as sensitizers to be used in combination with ionizing radiation or with alkylating agents. It is important to establish whether these effects will be tumor specific and whether therapeutic benefit can be predicted by the integrity of DNA repair pathways in the target tumor. Our previous study showed that radiosensitization of human tumor cells is replication dependent and mediated primarily through conversion of unrepaired SSB to DSB during DNA replication (11). Here, we show that sensitization to ionizing radiation and the alkylating agent MMS is enhanced in DSB repair–defective cells. Both findings predict that the sensitizing effects of PARP inhibitors will be more pronounced in tumors than in normal tissues. Furthermore, we show that the mechanisms responsible for sensitization differ between Ligase IV^−/− cells and Artemis^−/− or ATM^−/− cells and provide evidence for distinct mechanisms by which PARP inhibition increases unrepaired DSB after induction of single stranded DNA damage.

In this study, we examined the effect of PARP inhibition on MMS sensitivity and observed a markedly greater effect in Artemis^−/− or ATM^−/− cells than in DSB repair–proficient cells. The radiosensitizing effects of PARP inhibition were also enhanced in Artemis- and ATM-deficient cells than WT, but the differential effect was less marked, probably because ionizing radiation induces a spectrum of damage that includes SSB and DSB. Radiosensitization was replication dependent in WT, Artemis^−/−, or ATM^−/− cells. These observations are consistent with a model in which the primary effect of PARP inhibition is to abrogate SSB repair, leading to the replication-dependent generation of cytotoxic DSB, repair of which is inhibited in the absence of Artemis or ATM.

ATM might promote repair of replication-dependent DSB by a variety of mechanisms. These include initiation of cell cycle checkpoints, activation of homologous recombination repair of replication coupled DSB (29), and promotion by phosphorylation of the end-processing activity of Artemis (33). Likewise, Artemis has established roles in end-processing of complex DSB and resolving hairpin structures during variable (diversity) joining recombination (36). Because PARP inhibition causes persistent binding of PARP to damaged DNA (37), it is also possible that the nuclease activity of Artemis is required to remove the PARP-DNA complex and allow repair. Our data indicate that such a requirement may be particularly marked in the context of DNA replication.

It has been proposed that an alternative end-joining pathway functions in the absence of classic NHEJ (20). To consolidate the concept that this pathway is compromised by PARP inhibition, we analyzed induction and repair of γ-H2AX foci after low doses of radiation. To eliminate the effects of homologous recombination in S and G₂ phase, foci were enumerated in G₁ cells only. Consistent with previous findings, PARP inhibition blocked DSB repair in Ku80^−/− and Ligase IV^−/− cells. These data support the previously proposed model whereby Ku and PARP compete for binding at DSB ends (20) and provide an explanation for our observation that radiosensitization of Ligase IV–deficient cells by PARP inhibition is more pronounced than in WT cells and is independent of DNA replication.

In this study, we also show for the first time that PARP inhibition has no direct impact on DSB repair in Artemis- and ATM-defective G₁ cells, following exposure to ionizing radiation. This is consistent with our observation that sensitization to DNA damaging agents by PARP inhibition is replication dependent in Artemis^−/− and ATM^−/− cells and shows that the activity of the alternative end-joining pathway is not promoted by the absence of Artemis or ATM. It also supports the idea that Ku binding activity at DSB is unaffected by the absence of these repair factors. Of note, these results do not necessitate that PARP plays a functional role in alternative end joining. The data show only that inhibition of PARP can block the activity of this process.

PARP inhibition caused gross sensitization of Ligase IV^−/− cells to MMS. This was unexpected because MMS is not thought to induce significant numbers of DSB directly. To investigate whether MMS treatment generates DSB in the absence of DNA replication and the role of PARP inhibition in this process, γ-H2AX foci were enumerated in G₁ cells. Ligase IV–deficient cells exhibited significant accumulation of DSB (18 per G₁ nucleus), following exposure to 1 mmol/L MMS in the presence of a PARP inhibitor. This indicates that PARP inhibition promotes replication independent generation of DSB from MMS-induced lesions and that these DSB require core NHEJ for repair. Such DSB might arise from overlapping SSB or from interactions between transcription and obstructed SSB. Low numbers of γ-H2AX foci (7 per G₁ nucleus) were detected in WT cells exposed to MMS in combination with the PARP inhibitor, again indicating that MMS is capable of inducing replication-independent DSB when base excision repair is impaired. In NHEJ proficient cells, however, these DSB were rapidly repaired.

A previous study failed to detect induction of DSB following exposure to MMS (38). This discrepancy may be explained by the increased sensitivity of the γ-H2AX foci assay compared with pulsed-field gel electrophoresis, which has a detection limit of ∼300 DSB (equivalent to at least 10 Gy in G₁ cells). The γ-H2AX foci assay is highly sensitive but must be used with caution because of the ability of DNA lesions other than DSB to stimulate phosphorylation of H2AX. Its validity in these experiments is supported by the fact that induction of foci by MMS was entirely abolished by inhibition of ATM and DNA-PK because phosphorylation of H2AX by these two proteins occurs only when they are activated by the presence of DNA DSB (39). In addition, resolution of the γ-H2AX foci induced by MMS was Ligase IV dependent. In support of our observations, Woodhouse and colleagues (40) have shown that hydrogen peroxide treatment generates DSB in PARP-1–deficient but not control cells. Our study extends this finding by showing that SSB are converted into DSB in the absence of DNA replication and that the resulting DSB require NHEJ for repair. Artemis- and BRCA2-deficient cells exhibited normal repair of DSB under these conditions. Hence, in Ligase IV–deficient cells, inhibition of PARP acts through two distinct replication-independent mechanisms to increase the cytotoxic effects of SSB-inducing agents such as MMS and ionizing radiation.

The capacity of PARP inhibitors to increase tumor radiosensitivity has been shown in a number of in vitro and in vivo models (41, 42), and the low toxicity of these compounds predicts a beneficial clinical response. However, the major impediment to clinical use of radiosensitizing agents is parallel sensitization of normal tissues (43). In this study, we have shown that the sensitizing effects of PARP inhibitors are manifested only in replicating cells or in nonreplicating cells that are deficient in core NHEJ proteins. Sensitization to ionizing radiation, particularly to the alkylating agent MMS, was enhanced in cells deficient in the "noncore" NHEJ-repair proteins ATM and Artemis. Tumors exhibiting defective NHEJ repair may therefore be particularly sensitive to PARP inhibitors in combination with radiation or alkylating chemotherapy agents. In general, tumor cells are characterized by higher rates of replication than normal tissues and are much more likely to exhibit defective DNA repair (5, 44). Furthermore, specific defects in NHEJ have been documented in high-grade malignancies such as carcinoma of the bladder (45) and glioblastoma multiforme (46), and ATM deficiency is a feature of mantle cell lymphoma (47). Thus, it is likely that the radiosensitizing effects of PARP inhibitors such as olaparib will be more marked in tumor cells than in adjacent normal tissues. The resulting therapeutic benefit may be particularly apparent in the case of brain tumors, where the critical normal tissue, the brain, is composed predominantly of nonreplicating cells with intact DNA repair pathways (48).

No potential conflicts of interest were disclosed.

We thank Graeme Smith and Niall Martin at KuDOS Pharmaceuticals (AstraZeneca) for the helpful discussion and for providing olaparib and KU-55933.

Grant Support: Medical Research Council Clinician Scientist Fellowship (G108/589; A.J. Chalmers and D.A. Löser), Medical Research Council Programme Grant (G0500897; P.A. Jeggo and A. Shibata), and JSPS Research Fellowship for Young Scientists (A. Shibata).

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.