Combinatorial Efficacy of Olaparib with Radiation and ATR Inhibitor Requires PARP1 Protein in Homologous Recombination–Proficient Pancreatic Cancer

This article is featured in Highlights of This Issue, p. 217

Pancreatic cancer is notoriously resistant to most cytotoxic chemotherapies, targeted therapies, and immunotherapy (1, 2). Recently, however, the POLO (Pancreas Cancer Olaparib Ongoing) trial demonstrated the activity of the PARP inhibitor olaparib in patients with germline BRCA1/2-mutant pancreatic cancer leading to its FDA approval (3). While the findings of this trial are a major advancement in therapy for these patients, BRCA1/2 mutations or homologous recombination (HR) repair deficiencies only occur in about 10% of all pancreatic cancers (4). Combination strategies, however, may extend the therapeutic indications of PARP inhibitors to patients with BRCA1/2 wild-type (wt), HR-proficient tumors.

The combinatorial efficacy of PARP inhibitors with radiation is well established in HR-proficient cancers (5–7). Early studies demonstrated that this synergy between radiation and PARP inhibition requires DNA replication (7). In addition to radiation, agents that target the DNA damage response (DDR), in particular, those which target replication fork progression, stability and/or restart, have been shown to overcome PARP inhibitor resistance and improve therapeutic efficacy (8–11). Given that both radiation and agents that target the DDR (such as CHK1 or WEE1 inhibitors) interact with olaparib by exacerbating replication stress and DNA damage, subsequent studies assessed olaparib-radiotherapy in combination with DDR inhibitors and found these therapies are highly efficacious in HR-proficient pancreatic cancer (12, 13). Furthermore, mechanistic studies confirmed the importance of replication stress for sensitization to combined olaparib-radiotherapy by DDR inhibition (e.g., WEE1; ref. 14). Therefore, we hypothesized that more direct targeting of DNA replication fork initiation and stability with an ataxia telangiectasia and Rad3-related (ATR) inhibitor would further improve the efficacy of olaparib-radiotherapy.

The primary therapeutic target of olaparib is PARP1, a DDR protein that promotes the repair of both DNA single-strand (SSB) and double-strand breaks (DSB). After binding to sites of DNA damage, PARP1 acts primarily through the poly(ADP-ribosyl)ation (PARylation) of several proteins, including PARP1 itself. This PARylation signals the subsequent recruitment of DDR proteins such as XRCC1 to the damage site and facilities release of the PARP1 protein from DNA (15). Although early preclinical work defined the role of PARP in base excision repair, more recent studies have focused on the ability of PARP1 to regulate both DNA replication–associated repair, including stabilization and restart of stalled DNA replication forks (16–18), as well as DNA replication fork speed (19, 20).

Although the replication-associated stress that results from PARP inhibition may determine the combinatorial efficacy of PARP inhibitors with a variety of DNA-damaging agents, the cytotoxic activity of PARP inhibitor monotherapy has been attributed to a distinct process termed PARP trapping (21). PARP trapping is hypothesized to result from both inhibition of PARP catalytic activity, and thus the autoPARylation required for dissociation of PARP from DNA, as well as drug-induced allosteric changes in the PARP1 protein (22–24). We and others have reported that persistent PARP1-DNA binding, or trapping, is a key mechanism of radiosensitization by PARP inhibitors (5, 14). Furthermore, it has been posited that trapped PARP1 interferes with DNA replication, causing fork stalling, which, if left unresolved because of PARP catalytic inhibition or impaired HR, ultimately leads to replication fork collapse and lethal DSBs (5, 14, 25). While more recent work has called this specific hypothesis into question (19), there is strong evidence that DNA replication is mechanistically important for olaparib-mediated radiosensitization (26, 27).

The importance of replication and replication stress to the efficacy of DDR inhibitors such as AZD1775 (adavosertib) in combination with PARP inhibitors and radiation (14, 28), suggests that ATR would also be a promising target for further improvement of therapeutic efficacy. ATR is a central mediator of the DNA replication stress response that is activated by replication protein A (RPA)-coated single-stranded DNA at stalled replication forks (29, 30). In cooperation with ATRIP and CHK1, ATR promotes the stabilization and restart of these stalled forks and pauses cell-cycle progression to allow for the completion of DNA repair. In addition, ATR regulates and maintains the timely firing of DNA replication origins (31, 32). The ability of ATR inhibitors to overcome either acquired (11) or innate (33) resistance to PARP inhibitors further supports the hypothesis that ATR inhibitors sensitize HR-proficient cancers to olaparib-radiation combination therapies (10, 11, 33).

In this study, we set out to extend the efficacy of PARP inhibitors to otherwise resistant HR-proficient pancreatic cancers through their combination with radiation and ATR inhibitor. We investigated the efficacy of olaparib combined with radiation and AZD6738 (ceralasertib; ref. 34) in both HR-proficient and HR-deficient pancreatic cancer cells as well as the role of PARP1-DNA binding as a mechanism of therapeutic activity. Given the importance of PARP and ATR in coordinating the cellular response to replication stress, we then used DNA fiber combing and a neutral comet assay to investigate the contributions of replication stress and DSBs to the efficacy of this combination therapy. Finally, we confirmed the tolerability and therapeutic efficacy of combined olaparib with radiation and AZD6738 in animal tumor models. Our results show that PARP1 protein and its association with DNA in response to trapping concentrations of olaparib are required for the therapeutic activity of olaparib combined with radiation and ATR inhibitor in HR-proficient pancreatic cancer and support the clinical development of this combination as a strategy to overcome the intrinsic resistance of HR-proficient pancreatic cancer to PARP inhibitor monotherapy

MiaPaCa2 and Panc1 cells were obtained from and authenticated by the ATCC. Capan1.NEO, a clonal cell line expressing the neomycin resistance gene, was a gift from S. Powell (Memorial Sloan Kettering Cancer Center, New York, NY; ref. 35). Cells were grown in either DMEM (MiaPaCa2 and Panc1; Invitrogen), or IMDM medium (Capan1.NEO; Invitrogen) supplemented with 10% FBS (Premium Select; Atlanta Biologicals). These cell lines have been previously characterized as HR proficient (MiaPaCa2, Panc1) and HR deficient (Capan1; refs. 35–37). To establish a conditional HR-deficient model system, MiaPaCa2 cells were stably transduced with the Tet-pLKO-puro vector containing short hairpin (shRNA)-targeting RAD51 (sense 5′-CCGGAAGCTGAAGCTATGTTCGCCACTCGAGTGGCGAACATAGCTTCAGCTTTTTTTG; antisense 5′-AATTCAAAAAAAGCTGAAGCTATGTTC GCCACTCGAGTGGCGAACATAGCTTCAGCTT). Tet-pLKO-puro was a gift from Dmitri Wiederschain (Addgene plasmid 21915; ref. 38). Stable clones were selected with puromycin (1 μg/mL; Sigma) and screened for RAD51 depletion following doxycycline treatment (Supplementary Fig. S1B). To generate stable PARP1-knockout clones (PARP1^KO), MiaPaCa2 cells were transfected with a PARP1 CRISPR/Cas9 KO Plasmid (sc-419018, Santa Cruz Biotechnology) using Lipofectamine 2000 transfection reagent (Thermo Fisher Scientific) per manufacturer's protocol. Stable clones were selected with puromycin and isolated by limited dilution in 96-well plates. Wild-type PARP1 complemented clones (PARP1^KO + wtPARP1) were generated by transfecting two different PARP1^KO clones with a pcDNA4/V5-HisA plasmid expressing full-length wtPARP1 (1–1014; ref. 39). Stable clones were selected with zeocin (100 μg/mL; Thermo Fisher Scientific) and screened by immunoblotting for expression of PARP1 protein and PAR. For in vitro experiments, AZD6738 (34) and olaparib (AstraZeneca) were each dissolved in dimethyl sulfoxide (Sigma) and stored in aliquots at −20°C. For in vivo studies, AZD6738 and olaparib were dissolved in DMSO (final concentration 10%) and then further diluted in 40% propylene glycol (Sigma) or 10% 2-hydroxypropyl-β-cyclodextrin (Cayman Chemical), respectively, and stored at room temperature for up to 3 days.

Cells treated with drugs and/or radiation were processed for clonogenic survival as described previously (12, 40). Unless otherwise indicated, AZD6738 and olaparib were given for 25 hours, beginning 1 hour prior to radiation. Radiation survival curves were normalized for drug toxicity and the radiation enhancement ratio (RER) was calculated as the ratio of the mean inactivation dose under control conditions divided by the mean inactivation dose after drug exposure (41). A value significantly greater than 1 indicates radiosensitization. Cytotoxicity in the absence of radiation treatment was calculated by normalizing the plating efficiencies of drug-treated cells to nondrug-treated cells.

Whole-cell lysates were prepared in ice-cold RIPA buffer (1 mol/L NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, 50 mmol/L Tris, pH 7.4) supplemented with both PhosSTOP phosphatase inhibitor and complete protease inhibitor cocktails (Roche; ref. 42). To assess PARP1 bound to chromatin, cells were fractionated using a series of salt stringency buffers as described previously (43). Briefly, nuclei from approximately 3 × 10⁶ irradiated and drug-treated cells were isolated by gentle lysis in 100 μL ice-cold hypotonic buffer (50 mmol/L HEPES, pH 7.9, 10 mmol/L KCl, 1.5 mmol/L MgCl₂, 0.34 mol/L sucrose, 10% glycerol, 1 mmol/L DTT, 0.1% Triton X-100, protease inhibitors), followed by slow-speed centrifugation (1,300 × g at 4°C for 4 minutes). Washed nuclei were then lysed in ice-cold buffer B (50 mmol/L HEPES, pH 7.9, 100 mmol/L KCl, 2.5 mmol/L MgCl₂, 0.05% Triton X-100, protease inhibitors) and insoluble chromatin was collected by centrifugation (15,000 × g at 4°C for 10 minutes) and washed once in buffer C (50 mmol/L HEPES, pH 7.9, 250 mmol/L KCl, 2.5 mmol/L MgCl₂, 0.05% Triton X-100, protease inhibitors) and once in buffer D (50 mmol/L HEPES, pH 7.9, 500 mmol/L KCl, 2.5 mmol/L MgCl₂, 0.1% Triton X-100, protease inhibitors). The chromatin pellet was then resuspended with buffer B supplemented with 5 mmol/L CaCl₂ and incubated at 37°C for 10 minutes with 3 units of micrococcal nuclease, centrifuged, and processed for Western blot analysis (44).

The following antibodies were used: rabbit polyclonal anti-PARP1 (#9542), mouse monoclonal anti-Histone H3 (96C10; #3638), and rabbit monoclonal anti-GAPDH (14C10; #2118) from Cell Signaling Technology; mouse monoclonal anti-PAR (10H; ab14459) from Abcam. Densitometric analyses of immunoblots were carried out using Image J software (NIH, Bethesda, MD).

Cells were treated with AZD6738 and/or olaparib for 1 hour prior to and 24 hours post-8 Gy radiation. Neutral comet assay was performed according to the manufacturer's protocol (Trevigen). Briefly, cells were scraped, mixed 1:10 with 1% molten LMAgarose, pipetted onto CometSlides, and submerged in neutral lysis buffer overnight at 4 °C. Slides were rinsed three times with TBE (90 mmol/L Tris base, 90 mmol/L boric acid, 2 mmol/L EDTA, pH 8.0) and subjected to neutral electrophoresis for 35 minutes at 25 V. Slides were then incubated in 2.5 μg/mL propidium iodide for 20 minutes, rinsed in 70% ethanol, and allowed to dry overnight. Images were acquired using an Olympus ix73 inverted microscope with a 10× objective. The Olive tail moment (OTM) from at least 75 cells for each experimental condition was determined using Comet Assay IV software (Instem). Individual OTM measurements were first normalized to the mean OTM value from an internal control for each independent experiment (cells collected immediately following irradiation with 8 Gy on ice) and then to the mean normalized OTM for the radiation alone sample.

MiaPaCa2 cells were pulse-labeled with 25 μmol/L 5-iododeoxyuridine (IdU) for 30 minutes, followed by two gentle washes with prewarmed PBS and a second pulse with 250 μmol/L 5-chloro-2′-deoxyuridine (CldU) for 30 minutes. Labeled cells were collected and fibers spread on silane-coated slides (Lab Scientific 7801B) as described previously (45) and consistent with the methodology with modifications noted below. Two slides were spread and stained for each experimental condition. IdU was detected with a mouse anti-bromodeoxyuridine (BrdU) antibody (Becton Dickinson, 347580) and CldU with a rat anti-BrdU antibody (BIORAD, OBT0030G). Secondary antibodies were Alexa Fluor 594 anti-mouse (Invitrogen, A11062) used at a 1:1,000 dilution and Alexa Fluor 488 anti-rat (Invitrogen, A21470) used at a 1:200 dilution. Images of well-spread fibers were acquired using an Olympus ix73 inverted microscope with a 60× objective. For each experimental condition, double-labeled replication forks from 3 to 8 fields per slide were analyzed manually using ImageJ software (NIH, Bethesda, MD), consistent with previously described methodology (46). Fork measurements from two independent experiments were pooled and the total number of DNA fibers from both experiments is presented.

Cells were trypsinized, washed with ice-cold PBS, and fixed at a concentration of 2 × 10⁶ cells/mL in ice-cold 70% ethanol. For γH2AX analysis, samples were incubated with a mouse anti-γH2AX–specific antibody (clone JBW301; Millipore) overnight at 4°C followed by incubation with a FITC-conjugated secondary antibody (Sigma) as described previously (47). Samples were stained with propidium iodide to measure total DNA content and analyzed on a FACScan flow cytometer (Becton Dickinson) with FlowJo software (Tree Star). γH2AX positivity was quantified by setting a gate on the control, untreated sample to define a region of positive staining for γH2AX of 5%–10%. This gate was then overlaid on the drug/radiation-treated samples.

Irradiations were performed using a Philips RT250 (Kimtron Medical, Woodbury, CT) at a dose rate of approximately 2 Gy/minute at the University of Michigan Comprehensive Rogel Cancer Center Experimental Irradiation Core. Dosimetry was performed using an ionization chamber connected to an electrometer system that is directly traceable to a National Institute of Standards and Technology calibration. For tumor irradiation, animals were anesthetized with isoflurane and positioned such that the apex of each flank tumor was at the center of a 2.4 cm aperture in the secondary collimator, with the rest of the mouse shielded from radiation.

Animals were handled according to a protocol approved by the University of Michigan Committee for Use and Care of animals. MiaPaCa2 cells (5 × 10⁶) were suspended in a 1:1 mixture of DMEM+10% FBS:Matrigel (BD Biosciences) and injected subcutaneously, bilaterally into the flanks of 3–5 weeks old, male, athymic nude mice (Envigo). Treatment was initiated when the average tumor volume reached 100 mm³ and consisted of AZD6738 (25 mg/kg) and/or olaparib (50 mg/kg), once daily, 1 hour preradiation (1.8 Gy/fraction) Monday–Friday for one cycle. AZD6738 and olaparib were administered via oral gavage. Tumor size was measured two times per week. Tumor volume (TV) was calculated according to the equation: TV = π/6 (ab²), where a and b are the longer and shorter dimensions of the tumor, respectively. Measurements were made until day 80 or until the tumor volume increased by approximately a factor of 4.

Statistically significant differences for the clonogenic survival and flow cytometry assays were determined by one-way ANOVA with the Tukey postcomparison test in GraphPad PRISM version 8.0 (GraphPad software). Comet assay and fiber combing data were also analyzed in PRISM with the Kruskal–Wallis one-way ANOVA. For tumor growth experiments, the time required for tumor volume tripling was determined for each xenograft by identifying the earliest day on which it was at least three times as large as on the first day of treatment. The Kaplan–Meier method was used to analyze the tripling times and the log-rank test was used to compare any two treatment groups in PRISM.

Before assessing the efficacy PARP inhibitors in combination with radiation and ATR inhibitor, we first determined the cytotoxicity of olaparib and AZD6738 in the absence of radiation. We hypothesized that this combination would be preferentially cytotoxic to HR-deficient cells due to the synthetic lethality between HR defects and PARP inhibition (48, 49). Because silencing of RAD51 leads to functional inhibition of HR, we constructed a doxycycline-inducible shRAD51-MiaPaCa2 cell line to test this hypothesis (Supplementary Fig. S1). As expected, RAD51-depleted, HR-deficient MiaPaCa2 cells (MiaPaCa2.HRD) were sensitive to olaparib-mediated loss of clonogenicity compared with HR-proficient MiaPaCa2 cells (MiaPaCa2.HRP) with IC₅₀ values of 13.2 ± 0.04 nmol/L and > 10 μmol/L, respectively (Fig. 1A and B). We found, however, that both HR-proficient and HR-deficient models were sensitized to olaparib by AZD6738, though much higher, albeit clinically relevant (50) concentrations of olaparib were required to achieve sensitization in HR-proficient cells. These findings extended to other HR-proficient (Panc1 cells) and HR-deficient (BRCA2 mutant; Capan1) pancreatic cancer cell lines (Fig. 1C and D) and support the hypothesis that ATR inhibition can overcome PARP inhibitor resistance, regardless of HR status (10).

Using concentrations of olaparib chosen to inhibit PARP catalytic activity or induce PARP trapping (10 nmol/L or 3 μmol/L, respectively; ref. 25), we next assessed the combinatorial efficacy of olaparib with radiation and AZD6738 in HR-deficient and HR-proficient cells. Although 10 nmol/L olaparib in combination with AZD6738 was sufficient to radiosensitize HR-deficient cells, it had no effect on the radiosensitivity of the HR-proficient cell lines (Supplementary Fig. S2; Table 1A). In those cells, radiosensitization required higher concentrations of olaparib, either alone or in combination AZD6738 (Fig. 2; Table 1A). Furthermore, the combined efficacy of 3 μmol/L olaparib with AZD6738 in HR-proficient cells was significantly greater than the radiosensitizing activity of either drug alone. These data demonstrate the marked differences in the olaparib concentration required for radiosensitization in HR-proficient and HR-deficient cells, especially in the context of combined therapy with ATR inhibitor. Furthermore, these data suggested that while catalytic inhibition of PARP1 with low concentrations of olaparib (in combination with AZD6738) was sufficient to radiosensitize HR-deficient pancreatic cancer cells, radiosensitization in the HR-proficient cell lines required higher concentrations of inhibitor that would cause PARP trapping at sites of radiation-induced DNA damage. To further test the hypothesis that PARP trapping is required for the combinatorial efficacy of olaparib with radiation and AZD6738 in HR-proficient pancreatic cancer cells, we first compared the abilities of low (10 nmol/L) and high (3 μmol/L) concentrations of olaparib to inhibit PARP activity. In both Panc1 and MiaPaCa2 cells, both concentrations of olaparib, alone or in combination with AZD6738, effectively inhibited PAR formation, even in the presence of radiation-induced DNA damage (Fig. 3A and B). This result suggests comparable inhibition of PARP catalytic activity by either concentration of olaparib.

We next evaluated the effects of olaparib alone or in combination with AZD6738 on PARP1-DNA binding in response to radiation-induced DNA damage. Relative to untreated control cells, we found that radiation alone caused a modest increase in chromatin-associated PARP1, an effect which likely reflects the transient association of PARP1 with sites of DNA damage (Fig. 3C–E). Furthermore, 3 μmol/L olaparib, but not 10 nmol/L olaparib, either alone or in combination with AZD6738, increased radiation-induced PARP1-DNA binding with minimal change in the amount of total PARP1 in whole-cell lysates, a finding that is consistent with prior studies demonstrating a lack of PARP1 trapping with lower (<100 nmol/L) concentrations of olaparib (Fig. 3C–E; Supplementary Fig. S3; ref. 25). These data suggest that the therapeutic activity of olaparib in combination with radiation and/or AZD6738, is not only a function of PARP catalytic inhibition but is also associated with increased PARP1-DNA binding.

Much of the data in support of the PARP trapping model are derived from PARP1 knockout studies which have demonstrated not only that deletion of PARP1 protein does not replicate the cytotoxic effects of pharmacologic PARP inhibition, but additionally that PARP1 protein is required for the therapeutic efficacy of PARP inhibitors (25, 51). To test the requirement for PARP1 protein for the activity of olaparib in combination with radiation and AZD6738, we used CRISPR-Cas9 to establish MiaPaCa2 PARP1 null cells (PARP1^KO). As expected, PARP1^KO cells displayed reduced levels of PAR comparable with that observed following treatment of Cas9 control cells with olaparib (Supplementary Fig. S4A). Furthermore, PARP1^KO cells were resistant to the cytotoxic effects of the potent PARP trapping agent talazoparib (Supplementary Fig. S4B), and to the cytotoxicity of combined olaparib and AZD6738 (Supplementary Fig. S4C), sensitivities that were partially rescued by reintroduction of wtPARP1. To test the hypothesis that PARP trapping is required for the combinatorial efficacy of olaparib and radiation in HR-proficient cells, we assessed the radiosensitivity of two independent MiaPaCa2 PARP1^KO clones (#1 and #2) and found that PARP1 deletion did not replicate the sensitizing effects of 3 μmol/L olaparib (Fig. 4A). This result is consistent with the lack of radiosensitization by 10 nmol/L olaparib in parental MiaPaCa2 cells (Supplementary Fig. S2) and our prior study demonstrating minimal radiosensitization by PARP1 depletion with siRNA (14). Furthermore, while MiaPaCa2 Cas9 cells were sensitized to radiation by olaparib alone or in combination with AZD6738 (Fig. 4B; Table 1B), these effects were lost in PARP1^KO cells (Fig. 4C) and partially restored by complementation with wtPARP1 (Fig. 4D). Taken together, these data demonstrate that PARP1 protein is required for the activity of olaparib in combination with radiation and AZD6738, and support the hypothesis that the combinatorial efficacy of PARP and ATR inhibitors with radiation is mediated by PARP1-DNA binding rather than PARP catalytic inhibition.

To begin to understand the mechanisms of, and the contribution of PARP1 protein to, the effects of olaparib in combination with radiation, we first assessed γH2AX, a marker of DSBs. Given that ATR is a direct kinase for H2AX (52) and that AZD6738 attenuates radiation-induced γH2AX signaling in parental MiaPaCa2 cells (Supplementary Fig. S5A), AZD6738 was not included in these experiments. Despite the lack of PARP catalytic activity in PARP1^KO cells, PARP1 deletion had little effect on γH2AX induction or resolution following radiation (Fig. 4E), a result consistent with the inability of 10 nmol/L olaparib to delay resolution of radiation-induced γH2AX signaling in parental MiaPaCa2 or Panc1 cells (Supplementary Fig. S5A and S5B). Furthermore, while 3 μmol/L olaparib significantly increased the levels of γH2AX following radiation in Cas9 control cells, it did not affect radiation-induced γH2AX signaling in PARP1^KO cells (Fig. 4E). These data suggest that the loss of PARP catalytic activity itself neither inhibits the repair of radiation-induced DSBs nor promotes the formation of additional DSBs as cells reenter S-phase and initiate DNA replication. Furthermore, these data support the hypothesis that the physical interaction of PARP1 protein with damaged DNA contributes to the olaparib-mediated increase in γH2AX signaling.

To more directly assess the effects of combined ATR and PARP inhibition on radiation-induced DNA damage, we next used a neutral comet assay to measure DSBs 24 hours postradiation. In the absence of radiation, treatment with trapping concentrations of olaparib, alone or in combination with AZD6738, induced a small but significant level of DSBs relative to control (Supplementary Fig. S5C and S5D). Consistent with the γH2AX signaling patterns described above, 10 nmol/L olaparib also had no effect on the resolution of radiation-induced DSBs while 3 μmol/L olaparib significantly increased the extent of DSBs 24 hours postradiation (Fig. 5A–C; Supplementary Fig. S5E). Furthermore, while AZD6738 in combination with 10 nmol/L olaparib did not significantly affect repair of radiation-induced DNA damage, AZD6738 in combination with 3 μmol/L olaparib caused a significant increase in DSBs. This damage may reflect persistent, unrepaired DSBs or possibly de novo DSBs generated when the DNA replication machinery encounters unrepaired, previously stalled forks (53). Taken together, these data demonstrate that the elevated DSBs mechanistically associated with therapeutic efficacy require concentrations of olaparib that cause persistent PARP1-DNA binding and are potentiated by an ATR inhibition.

On the basis of our previous study which identified replication stress as a key factor in olaparib- and AZD1775-mediated radiosensitization (14), and the well-established role of ATR in regulating origin firing and replication fork stability (31, 54), we hypothesized that persistent PARP1-DNA binding creates an obstacle for active DNA replication forks that is further exacerbated by inhibition of ATR, ultimately resulting in fork stalling/collapse and lethal DSBs (25, 33). To test this hypothesis, we used DNA fiber combing to directly visualize and measure DNA replication fork progression in cells treated with olaparib, radiation, and AZD6738. In contrast to our initial hypothesis, combination treatment did not result in DNA replication fork stalling or collapse, which would be reflected in the accumulation of highly asymmetric replication forks (ref. 55; Supplementary Fig. S6A and S6B). Instead, the higher concentrations of olaparib associated with PARP1-DNA binding significantly increased the rate of replication fork progression (Fig. 5E and F; Supplementary Fig. S6C). Furthermore, we found that canonical PARP1 trapping by treatment with temozolomide and talazoparib also failed to slow DNA replication rates (Supplementary Fig. S7). This olaparib-mediated increase in replication rate was associated with activation of an ATR-mediated replication stress response marked by increased ATR (T1989) and CHK1 (S345) phosphorylation (Supplementary Fig. S8) and γH2AX (Supplementary Fig. S5A and S5B). In contrast, ATR inhibition slowed DNA replication fork speed, a finding consistent with the role of ATR in suppressing aberrant origin firing to preserve replication machinery and maintain normal DNA replication rates (32, 54, 56). Furthermore, ATR inhibition negated the accelerating effects of olaparib on DNA replication fork rates, perhaps due to the effects of AZD6738 on origin firing limiting the capacity of the cell to accelerate replication fork progression (31). These data suggest the DNA damage response to olaparib-mediated replication stress is ATR dependent, and, furthermore, that inhibition of ATR exacerbates this replication stress.

On the basis of the combined efficacy of PARP inhibitors with radiation and ATR inhibitor in vitro, we next tested the therapeutic efficacy and tolerability of combined olaparib, radiation, and AZD6738 in animal tumor models. While radiation alone or in combination with olaparib or AZD6738 had a modest effect on tumor growth, the triple combination of olaparib, radiation, and AZD6738 had a profound antitumor effect (Fig. 6A). Tumors treated with this combination displayed a significantly increased time to tumor volume tripling relative to all other groups including radiation combined with either single agent (Fig. 6B). Noteworthy, 28% of the tumors treated with the combination of olaparib, radiation, and AZD6738 had complete responses that lasted for the duration of the study (80 days). This treatment regimen was well tolerated (Fig. 6C). In summary, these data validate the further development of PARP inhibitors in combination with radiation and ATR inhibitor as a therapeutic strategy for overcoming the intrinsic resistance of HR-proficient pancreatic cancers to PARP inhibitor monotherapy.

In this study, we show that the intrinsic resistance of HR-proficient pancreatic cancer cells to PARP inhibitor monotherapy can be overcome by combination with radiation and ATR inhibitor. Using both pharmacologic and genetic models, we further demonstrate that, while catalytic inhibition of PARP with low concentrations of olaparib is sufficient to radiosensitize HR-deficient models, radiosensitization in HR-proficient models requires both PARP1 protein and concentrations of olaparib that stabilize PARP1–DNA complexes. We also show that high concentrations of olaparib induce replication stress in the form of accelerated replication fork progression, and, in combination with radiation and ATR inhibitor, significantly increase DSBs in a PARP1 protein-dependent manner. While the effects of olaparib on replication fork progression are antagonized by AZD6738, olaparib-mediated DNA damage is amplified by ATR inhibition and correlates with radiosensitization. Finally, we show that combined treatment with olaparib, radiation, and AZD6738 has a profound antitumor effect in vivo with minimal toxicity.

The combinatorial efficacy of radiation with a variety of different PARP inhibitors (talazoparib, olaparib, veliparib) is well established in HR-proficient cancers where PARP inhibitor monotherapy has proven ineffective (57). In addition, both intrinsic (33) and acquired PARP inhibitor resistance (11) may be overcome by targeting the ATR-mediated DDR. The therapeutic potential of PARP and ATR inhibitors with radiation is further supported by a study from Carruthers and colleagues (28) which demonstrated that combined PARP and ATR inhibitors overcome the amplified DDR and intrinsic radioresistance of glioblastoma stem cells. While this study highlighted the potential of combined PARP and ATR inhibitors to enhance radiation-induced DNA damage, the contributions of PARP trapping and its consequences on DNA replication and radiosensitization were not investigated.

Although the precise biochemical mechanism of PARP trapping is incompletely understood, it is hypothesized to result from both inhibition of PARP catalytic activity, and thus autoPARylation, as well as drug-induced allosteric changes in the PARP1 protein (22–24). Our data, however, demonstrate an increase in chromatin-associated PARP1 in the absence of olaparib, postradiation (Fig. 3; Supplementary Fig. S7). Given that radiation alone does not inhibit PARP1 auto-PARylation (Fig. 3), this specific interaction does not reflect PARP1 trapping. With this conclusion in mind, we cannot definitively state whether or not the chromatin-associated PARP1 in the presence of radiation plus olaparib is trapped PARP, only that there is increased PARP1 associated with chromatin under these conditions. It is possible that this interaction reflects enhanced chromatin-associated PARP1 and not the bona fide PARP1 trapping that has been described in response to methyl methanesulfonate or temozolomide-induced DNA lesions (25, 43). Regardless, our studies suggest that the maximal therapeutic activity of olaparib in combination with radiation and ATR inhibitor requires the formation of these PARP1–DNA complexes.

Early studies on PARP inhibitor–radiation combination schedules identified DNA replication as a key factor in determining overall cytotoxicity (26). This finding suggested that the DSBs associated with therapeutic efficacy result from replication-associated events rather than directly from radiation-induced DNA damage. Furthermore, the kinetics of this model are consistent with our finding that radiation-induced PARP1–DNA complexes are a relatively late (24 hours postradiation) event compared with the rapid recruitment of PARP to SSBs caused directly by radiation (58). Taken together, these observations support the hypothesis that replication is a key component of olaparib-mediated radiosensitization, suggesting a more nuanced view of how PARP is recruited to and retained on chromatin in response to ionizing radiation that may have clinical ramifications for the scheduling of these combination therapies.

In contrast to the posited inhibitory effect of PARP1-DNA binding on ongoing replication forks (5, 14, 25), we found that DNA replication rates accelerate in cells treated with concentrations of olaparib that cause PARP1-DNA binding, even in the presence of radiation-induced DNA damage. This finding is consistent with a previously published study on the replication stress caused by olaparib monotherapy (19) and supports their finding that PARP1–DNA complexes do not impede DNA replication. Of note, however, is that this replication stress is unlikely to result from p53-dependent p21 processes as MiaPaCa2 cells lack wt-p53 function. We propose that, combined with radiation, these competing stresses on replication (olaparib-mediated fork acceleration and ATR inhibitor–mediated suppression of replication fork progression), ultimately lead to unrepaired DSBs and reduced survival, as well as profound inhibition of tumor growth.

At the time of writing, eight active and recruiting trials are investigating the combination of olaparib and AZD6738, including one in pancreatic cancer. Many of these trials are founded on work by Kim and colleagues (9) which demonstrated the preferential activity of combined ATR and PARP inhibition in BRCA1/2-mutant, HR-defective ovarian cancer. This strong emphasis on HR status is reflected in the patient stratification (NCT03330847) and inclusion criteria (NCT02576444) for subsequent trials investigating the combination of PARP and ATR inhibitors. The remarkable response we have seen in HR-proficient pancreatic tumors treated with olaparib in combination with radiation and ATR inhibitor however, suggests this combination strategy may be extended to resistant HR-proficient cancers. In addition, given that PARP and ATR inhibitors as well as radiation influence tumor immunity, their combination has the potential to enhance antitumor immune responses (59–61). We anticipate that the mechanisms highlighted in this study will help guide the choice of agents as well as their dose and schedule for future clinical trials combining PARP inhibitors with radiation and ATR inhibition.

L.A. Parsels reports grants from the National Institutes of Health and other from AstraZeneca during the conduct of the study. D.R. Wahl reports grants and personal fees from Agios Inc. and grants from Innocrin Inc. outside the submitted work. T.S. Lawrence reports personal fees from AstraZeneca outside the submitted work. M.A. Morgan reports grants, personal fees, and financial support from AstraZeneca during the conduct of the study. No disclosures were reported by the other authors.

L.A. Parsels: Conceptualization, data curation, formal analysis, investigation, writing–original draft, writing–review and editing. C.G. Engelke: Data curation, formal analysis, investigation, writing–original draft. J. Parsels: Data curation, formal analysis, investigation. S.A. Flanagan: Data curation, formal analysis, investigation. Q. Zhang: Investigation, methodology. D. Tanska: Investigation. D.R. Wahl: Conceptualization, writing–review and editing. C.E. Canman: Methodology. T.S. Lawrence: Conceptualization, writing–review and editing. M.A. Morgan: Conceptualization, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.

We would like to acknowledge AstraZeneca, AZD6738 and olaparib teams, for their support and thank Alan Lau and Mark O'Connor for their helpful comments and review of this article.

This work was funded by NIH grants R01CA163895 (to M.A. Morgan), R01CA240515 (to M.A. Morgan), P50CA130810 (to T.S. Lawrence), U01CA216449 (to T.S. Lawrence), R50CA251960 (to L.A. Parsels), Cancer Center Support Grant P30CA46592, and a research grant from AstraZeneca (to M.A. Morgan).

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.