Targeting the ATR/CHK1 Axis with PARP Inhibition Results in Tumor Regression in BRCA-Mutant Ovarian Cancer Models

Ovarian cancer survival has improved minimally over the past decade (1) despite the unprecedented progress in understanding the genetics of ovarian cancer (2). There is a critical need to develop better therapeutic strategies that exploit the biology and genetics of high-grade serous ovarian cancer (HGSOC). Approximately 50% of HGSOCs have defects in genes involved in homologous recombination (HR) repair (2, 3). BRCA 1 and 2 (breast cancer susceptibility gene 1 and 2) mutant HGSOCs have a deficiency in the repair of double-strand DNA breaks (DSB) by HR (4). PARP inhibitors (PARPi) impair the repair of single-stranded DNA breaks (SSB), leading to DNA DSB, which cannot be repaired efficiently in BRCA1/2-mutant (BRCA^MUT) cancers capitalizing on synthetic lethality (5). PARPis, such as olaparib, have demonstrated a 31% overall response rate, leading to its FDA approval for recurrent germline BRCA^MUT HGSOCs (6). Rare complete responses (CR; 3%) are seen with PARPi monotherapy in the clinic (6–8). Our goal was to optimize PARPi therapy in BRCA^MUT HGSOC by evaluating scientifically rational combinations.

Another approach to modulate DNA repair activity and improve the therapeutic index of PARPi in HR-deficient HGSOCs is to interfere with cell-cycle checkpoint signaling. ATR (ataxia telengiectasia and Rad3-related) and its downstream kinase CHK1 (checkpoint kinase 1) are activated by DNA replication stress and DNA damage, thereby arresting cell-cycle progression allowing time for appropriate damage repair and completion of replication (9, 10). ATR/CHK1 blockade prevents DNA damage–induced cell-cycle arrest, resulting in inappropriate entry into mitosis, chromosome aberrations, unequal partitioning of the genome, and ultimately apoptosis (9). In addition, because the ATR–CHK1 pathway stabilizes replication forks and prevents collapse into DNA DSBs, inhibition of ATR/CHK1 is expected to increase reliance on HR to reform the replicatoin fork structure and complete replication.

Indeed, ATR inhibition is synthetic lethal with numerous cancer-associated changes, including oncogenic stress (oncogenic RAS mutations, MYC and CCNE1 overexpression), deficiencies in DNA repair (TP53, BRCA1/2, PALB2, and ATM loss), and other defects (9, 11–15). CHK1 inhibition, similarly, is synthetically lethal with TP53 or BRCA1/2 loss (16, 17). Almost all HGSOCs harbor a mutation in TP53 (2) and thus have lost G₁ checkpoint control, significantly increasing reliance on S and G₂ checkpoints for survival (11, 18). Targeting S and G₂ checkpoints by inactivation of the ATR/CHK1 pathway will inhibit the DNA damage–induced G₂ checkpoint arrest, leading to mitotic catastrophe and tumor cell death in contrast to normal cells, which maintain an intact G₁ phase checkpoint (19). A variety of metabolic perturbations in cancers cause a reliance on ATR/CHK1 to facilitate DNA synthesis and prevent the formation of DNA DSBs at replication forks (20). These breaks can increase to toxic levels in cancer cells when ATR or CHK1 is inhibited (9, 11–14, 20–22). Thus, ATR or its downstream effector, CHK1, is a reasonable target for treating HGSOCs, all of which have loss of functional TP53, and approximately 50% have defects in HR (2). Drugs targeting ATR (AZD6738, VX-970) and CHK1 (MK8776, SCH 900776, LY2606368, CCT245737) are in early phase I/II clinical trial development (ClinicalTrials.gov).

Although PARPi is active as monotherapy, it rarely leads to complete tumor responses (6–8, 23), emphasizing the need for alternative strategies capitalizing on synthetic lethality. We have developed a BRCA^MUT HGSOC orthotopic patient-derived xenograft (PDX) platform with more than 15 models that is molecularly annotated to strategize synthetic lethal approaches in BRCA^MUT HGSOC (24). Because we observed that PARPi caused ATR–CHK1 pathway activation, we reasoned that PARPi treatment alone may increase dependence on the ATR/CHK1 pathway for survival and that inhibition of ATR or CHK1 would increase DNA replication fork instability and promote cell death in BRCA^MUT HGSOC models. We show that PARPi treatment results in early activation of ATR/CHK1 and that combination PARPi with either CHK1i or ATRi is synergistic in suppressing BRCA^MUT HGSOC growth in culture and in the PDX model.

PEO1 (BRCA2^MUT; c.C4965G) and PEO4 (BRCA2 reversion mutation) serous ovarian cancer cell lines were grown in RPMI media with 10% FBS and penicillin/streptomycin (generous gift from Dr. Andrew Godwin, University of Kansas, Kansas City, KS). JHOS4 (BRCA1; c.5278-1G>A) ovarian cancer cells were grown in DMEM/F12 media with 10% FBS and penicillin/streptomycin. The WO-20 primary ovarian culture was generated in our laboratory from a patient with HGSOC (UPCC 17909), and the cells were cultured in OCMI-E media (Live Tumor Culture Core at Sylvester Comprehensive Cancer Center, Miller School of Medicine, Miami, FL). Mutation profiles for all cell lines were evaluated using a targeted panel of genes by whole-exome sequencing (24). All cell lines were confirmed negative for mycoplasma. Authenticity was confirmed by short tandem repeats analysis by the Wistar Genomics Core.

Cells (5 × 10³) were seeded on 96-well plates and treated with the indicated doses of PARPi (AZD2281), CHK1i (MK8776), and ATRi (AZD6738) for 5 days. At the end of the treatment period, the relative cell viability was determined by an MTT colorimetric assay. Cells were incubated with 10 μL of MTT at 5 mg/mL (Sigma Aldrich, St. Louis, MO) for 2 hours at 37°C. DMSO was added and the absorbance was measured in a microplate reader at a wavelength of 570 nm. IC₅₀s were calculated using GraphPad Prism (GraphPad Software).

Cells (1–2 × 10⁴) were plated onto 12-well plates and incubated at 37°C. Cells were treated for 10 to 14 days. Media and drugs were refreshed every 3 to 4 days. Colonies were washed with PBS, fixed with 4% paraformaldehyde, and then stained with 0.2% crystal violet. Whole-well images were scanned and colony-forming area was quantitated using ImageJ (NIH, Bethesda, MD). For each sample, the results from three replicates were averaged (25).

NSG mice were purchased from The Jackson Laboratory (NOD/SCID IL2Rγ^−/−). All mice were housed according to the policies of the Institutional Animal Care and Use Committee of the Wistar Institute (Philadelphia, PA). Five- to 8-week-old female mice were used for tumor transplantation. PDXs are generated by sectioning of fresh tumor tissue and engrafting pieces (2 × 2 × 2 mm³) orthotopically to the mouse fallopian tube fimbria/ovary. Tumor was obtained from debulking surgeries conducted at the Hospital of University of Pennsylvania (Philadelphia, PA; IRB# 702679). Once the transplanted tissue reaches approximately 700 to 1,000 mm³, it is harvested, analyzed by genomic and proteomic studies, expanded, and banked for preclinical studies (24). For preclinical studies, cryopreserved tissue is thawed, washed with Hank's Balanced Salt Solution, and retransplanted to the fallopian tube fimbria/ovary for evaluation of in vivo drug response. Tumor length and width were measured by ultrasound (SonoSite Edge II Ultrasound System) and used to calculate tumor volume. Once tumor volume reached 70 to 100 mm³, animals (n = 80) were randomized to seven treatment groups: vehicle (10% 2-hydroxylpropyl-b-cyclodextrin), MK8776 (50 mg/kg i.p. every 3rd day; Selleckchem), AZD2281 (50 mg/kg/day × 6 days weekly by oral gavage; AstraZeneca), AZD2281 (100 mg/kg/day × 6 days weekly by oral gavage; AstraZeneca), AZD6738 (25 mg/kg/day × 6 days weekly by oral gavage, AstraZeneca), MK8776 + AZD2281 (MK8776 50 mg/kg i.p. every 3rd day; and AZD2281 50 mg/kg/day × 6 days weekly), and AZD6738 + AZD2281 (AZD6738 25 mg/kg/day day 1–3 weekly and AZD2281 50 mg/kg/day × 6 days weekly). Tumor volume and body weight were measured weekly. Animals were euthanized according to Institutional Animal Care and Use Committee guidelines. Tumors were collected and snap frozen for protein analysis and IHC.

Cells and tissues were harvested and lysed in a Laemmli sample buffer (Bio-Rad Inc., Hercules, CA) containing a protease and phosphatase inhibitor cocktail (Calbiochem, San Diego, CA). Following protein concentration determination (Bio-Rad Inc., Hercules, CA), cell lysates were separated on reducing SDS-PAGE gels and immunoblotted with phospho-ATR (cat. # ABE462, EMD Millipore, Billerica, MA), total ATR (cat. # sc1887, Santa Cruz Biotechnology, Inc., Dallas, TX), phospho-CHK1 (Ser³⁴⁵), (cat. # 2348, Cell Signaling Technology, Inc., Danvers, MA), total CHK1 (cat. # sc8408, Santa Cruz Biotechnology, Inc., Dallas, TX), and γH2AX (cat. # 9718, Cell Signaling Technology, Inc., Danvers, MA). The species-appropriate horseradish peroxidase (HRP)-conjugated secondary antibody was used, followed by detection with chemiluminescent substrate (Thermo Fisher Scientific, Waltham, MA). Odyssey Quantitative Fluorescent Imaging System (LI-COR Biotechnology, Lincoln, NE) was used for image generation. Anti-β-actin (cat. # 3700, Cell Signaling Technology, Inc., Danvers, MA) was used as an internal control. Band intensity was quantitated using ImageJ (NIH, Bethesda, MD).

Cell cycle was analyzed using a FITC-BrdU Flow Kit (BD Biosciences, Franklin Lakes, NJ). Cells (5 × 10⁵) were plated in 10-cm dishes. At 48 hours after the initial seeding of the cells, the cells were incubated with drugs for an additional 48 hours. Bromodeoxyuridine (BrdUrd; 10 μmol/L) was added to culture medium and incubated for 2 hours before harvest. Cells were fixed and labeled with FITC-conjugated anti-BrdUrd and propidium iodide (PI) solution. Cell suspensions were incubated for 15 minutes at room temperature and immediately analyzed in a flow cytometer (BD FACSCalibur, BD Biosciences, Franklin Lakes, NJ). Data were analyzed by FlowJo (Tree Star, Inc., Ashland, OR).

Cells were harvested for chromosome preparations using colcemid (50 ng/mL for 90 minutes followed by an 18-minute incubation in 0.075 mol/L potassium chloride (KCl) at 37°C and dropwise addition of Carnoy fixative (3:1 methanol:acetic acid). Cells were incubated in fixative for one hour, pelleted at 1,000 g, and fixative was replenished. After cells were incubated at 4°C overnight, the fixative was again replenished. Fixed cells were dropped onto uncoated microscope slides and dried for at least 24 hours at room temperature. Dropped slides were stained in Giemsa staining solution (Sigma Aldrich, St. Louis, MO) for 4 minutes. Stained slides were analyzed for total gaps and breaks in a blinded fashion using a 100× objective and a Nikon Eclipse 80i microscope. Fifty metaphases were scored for each sample in two independent experiments for a total of 100 metaphases scored for every sample.

Cells (5 × 10⁵) were plated in 10-cm dishes. At 48 hours after the initial seeding, the cells were incubated with drugs for 48 hours. Apoptosis was detected by using an Annexin V Flow Kit (BD Biosciences, Franklin Lakes, NJ) according to the manufacturer's instructions. Annexin V–labeled cells were analyzed in a flow cytometer (FACSCalibur, BD Biosciences, Franklin Lakes, NJ). The data were analyzed by FlowJo (Tree Star, Inc., Ashland, OR).

Tissue samples were fixed in 10% formalin. Tissues were dehydrated in graded ethanol solutions, cleared in xylene, and embedded in paraffin. Paraffin blocks were cut into 4- to 6-μm sections and placed onto slides. After deparaffinization and rehydration, antigen retrieval was done via pressure cooker. Slides were pressure cooked in 1× target retrieval solution at 120°C at 18 to 20 psi. Endogenous hydrogen peroxidase activity was blocked with hydrogen peroxide for 10 minutes, followed by rinsing with wash buffer. Slides were incubated with pCHK1 (Cell Signaling Technology, cat# 2348) antibody at 1:1,000 titer for 40 minutes. Alternatively, slides were incubated with appropriate isotype controls and diluted similarly. Slides were washed and incubated with anti-rabbit HRP polymer for 30 minutes, followed by a further wash. Slides were developed using 3,3′-diaminobenzidine (DAB) + chromogen for 5 minutes and washed with water. After staining, slides were counterstained, dehydrated, and mounted with mounting reagent.

MTT, colony formation assays, FACS, and Western assays were done at least twice, and means ± SEM are displayed in bar graphs. One- or two-way ANOVA was conducted to assess differences among means. Following a significant ANOVA result (P ≤ 0.05) rejecting the null hypothesis that means are the same across the treatment groups, the Tukey honest significant difference test was used for all pairwise mean comparisons. This multiple comparison procedures ensure actual family-wise error rates no greater than prespecified 5%. Stata MP Version 14.0 (StataCorp) or GraphPad Prism version 5.00 for Windows (GraphPad Software, La Jolla, CA) was used for statistical analyses. Tumor growth data were analyzed using two-way ANOVA with Tukey posttest. All other data were analyzed using Student t test. To analyze the drug interaction between ATRi and CHK1i and PARPi combined with either agent, the coefficient of drug interaction (CDI) was calculated from an in vitro study (26). CDI is defined by the following formula; CDI = AB/(A × B). According to the absorbance of each group, AB is the ratio of the two-drug combination group to the control group, and A or B is the ratio of the single-drug group to the control group. CDI <1 indicates synergism, CDI <0.7 significant synergism, CDI = 1 additivity, and CD>1 antagonism. Analysis of potential synergy between drug A and drug B on tumor xenograft growth used the combination ratio (27). Fractional tumor volume (FTV) is defined as the ratio of mean final tumor volume in drug-treated animals divided by the mean final tumor volume in untreated controls. The combination ratio compared the FTV expected if there was no synergy with the observed FTV. The combination ratio was calculated as: (FTV of drug A × FTV of drug B)/observed FTV of combination. Observed and expected FTVs are described as follows: expected FTV = (mean FTV of drug A) × (mean FTV of drug B); observed FTV = final tumor volume combined therapy/final tumor volume control; combination ratio = expected FTV/observed FTV. A combination ratio greater than 1 indicates drug synergy, whereas a ratio less than 1 indicates a less than additive effect.

Increasing concentrations of PARPi, olaparib (AZD2281), were more cytotoxic in BRCA^MUT cells (PEO1, JHOS4) compared with HR-proficient cells (PEO4, BRCA2 reversion mutation; WO-20 primary tumor cultures, BRCA wild type). PARPi did not result in complete cell death even in the BRCA^MUT with 30% to 45% of cells still viable after 5 days (Fig. 1A). Colony-forming ability after treatment with increasing concentrations of PARPi similarly decreased more in the BRCA^MUT compared with the HR-proficient cells (Fig. 1B). BRCA2^MUT PDXs (BRCA2 8945delAA) were established, and when they reached 70 to 100 mm³, they were treated with prolonged PARPi (olaparib) at the maximum tolerated dose (MTD; 100 mg/kg/day). Tumor suppression but not regression was seen for 21 weeks (Fig. 1C), after which resistance emerged (not shown). Given the lack of complete cell killing and tumor regression in vivo, we investigated ways to improve the antitumor effects of PARPi.

In addition to their role of blocking repair of SSBs leading to DSBs (4), PARPi increased G₂ arrest (28). We thus sought to evaluate how PARPi affects the ATR/CHK1 cell-cycle checkpoint pathway. PARPi treatment at 1 μmol/L increased pATR, pCHK1, and γH2AX protein within 2 to 6 hours in both BRCA^MUT (PEO1, JHOS4; Supplementary Fig. S1) and HR-proficient cells (PEO4), but more so in the BRCA^MUT cells suggesting activation of ATR/CHK1 for survival (Fig. 1D). At higher concentrations of PARPi (5 μmol/L), pCHK1 increased within 2 hours of PARPi treatment in HR-proficient cells (Supplementary Fig. S4). DNA damage was increased with PARPi treatment in both the BRCA^MUT and HR-proficient lines but more so with the HR-deficient cells.

Increasing concentrations of CHK1 inhibitor (CHK1i; MK8776) were more cytotoxic in BRCA^MUT cells (PEO1 and JHOS4) compared with the HR-proficient cells (PE04; BRCA^REV). Notably, BRCA2^MUT cell line (PEO1) was more sensitive to CHK1i than BRCA1^MUT cell line (JHOS4). With increasing doses of ATR inhibitor (ATRi; AZD6738), there was a significant decrease in cell viability among both BRCA^MUT (PEO1, JHSO4) and HR-proficient cells (PEO4) beginning at 0.5 μmol/L after 5 days of treatment (Fig. 2A). Given PARPi increases ATR/CHK1 signaling and both cause replication fork collapse into DSBs using different mechanisms (20, 29), we hypothesized that the combination would be more effective in decreasing cell survival.

Combination therapy with PARPi and CHK1i was significantly more cytotoxic and decreased colony formation ability more than either drug alone in the BRCA2^MUT cells compared with wild type. Drug synergy was demonstrated by PARPi–CHK1i combination in BRCA^MUT cells but not in wild type (Fig. 2B; Supplementary Fig. S2). ATRi in combination with PARPi was significantly more cytotoxic than either drug alone in both BRCA2-deficient and wild-type cells. PARPi–ATRi combination demonstrated synergy (Fig. 2C; Supplementary Fig. S2). In BRCA^MUT cells, PARPi in combination with ATRi decreased the PARPi upregulation of pATR and pCHK1. pCHK1 increased with CHK1i treatment as expected, given the inhibition of CHK1 phosphatase site (30). In HR-proficient cells, addition of ATRi to PARPi decreased PARPi upregulation of pATR and pCHK1 (Fig. 2D). There was an increase in γH2AX in BRCA2^MUT cells relative to wild type, and in the BRCA2^MUT cells, there was an increase with combination therapy compared with monotherapy.

We reasoned that the function of ATR–CHK1 activation in PARP-inhibited cells may be to prevent cell-cycle progression in the context of PARPi-induced DNA damage. Thus, the effects of ATRi/CHK1i added to PARPi treatment on cell cycle were evaluated. In HR-deficient cells, PARPi treatment alone (1 μmol/L) increased G₂–M phase from 13% to 44%. ATRi and CHK1i each alone increased G₂–M but less so than PARPi from 13% untreated to 30% with ATRi and 19% with CHK1i, respectively. When PARPi-treated cells were exposed to ATRi or CHK1i, 13% and 22% of the G₂–M arrested population was released, respectively (Fig. 3; Supplementary Fig. S3). A defect in nucleotide incorporation, consistent with replication fork collapse, was observed in ATRi and PARPi-ATRi treatment groups (Fig. 3, upper panel, see y axis). In HR-proficient cells, a higher dose of PARPi (5 μmol/L) was used because 1 μmol/L had minimal effects alone or in combination on cell cycle (data not shown). PARPi (5 μmol/L) treatment did have an effect on HR-proficient cells, but it was different from what was observed in BRCA^MUT cells. PARPi at 5 μmol/L increased G₂–M from 11% to 17%. ATRi and CHK1i alone had a similar modest effect on G₂–M (ATRi, 11%–15%; CHK1i, 11%–16%). When PARPi-treated cells were exposed to CHK1i, G₂–M remained at 16%. However, with ATR exposure, a 24% increase in G₂–M was noted (Supplementary Fig. S4). Interestingly, the increase in G₂–M phase cells, as determined by DNA content, could represent either an increase in DNA damage with ATRi/CHK1i addition that activates alternative checkpoint proteins that recognize DSB (ATM), leading to G₂ arrest, or aberrant progression into the M-phase and stalling therein. Either mechanism provides insight into the mechanism of ATRi/CHKi synergy with PARPi.

We hypothesized that cells treated with the ATRi/CHK1i–PARPi combination sustain significant DNA damage, and some of these cells were permitted to progress through G₂ and M-phase due to ATR–CHK1 pathway inhibition. Indeed, the mechanism of synergy may at least partly involve progression into the M-phase with chromosome breaks. If so, then a synergistic increase in breaks and chromosome abnormalities in mitosis should be observed when ATRi/CHK1i treatments are added to PARPi treatment. Thus, we tested the effects of these drugs alone and combined on chromosomal breaks, gaps, and aberrations by metaphase chromosome spreads in BRCA^MUT cells (Fig. 4).

ATRi treatment alone significantly increased gaps and breaks (5/cell) relative to untreated BRCA^MUT cells (2/cell), consistent with prior reports of the effect of ATR suppression (15, 31, 32). PARPi or CHK1i alone had minimal effects at the doses tested. However, chromosomal aberrations, in which the DSBs have been incorrectly repaired, were increased with both ATRi–PARPi and CHK1i–PARPi combinations (Fig. 4B). Moreover, the combination PARPi–ATRi treatment caused 3 times more gaps and breaks than ATRi monotherapy (Fig. 4B), and such breaks appearing in mitosis are indicative of unrepaired DNA DSBs entering inappropriately into the M-phase. Therefore, particularly in the case of ATRi, PARPi treatment in combination with checkpoint abrogation increases the incidence of chromosome damage in metaphase, which causes cell lethality through abnormal partitioning of damaged and underreplicated DNA into daughter cells, a process known as mitotic catastrophe (33).

Given that combination therapy resulted in increased DNA damage compared with monotherapy in BRCA^MUT cells, the effects on apoptosis using Annexin V, PI, and cleaved caspase-3 were then evaluated. PARPi and CHK1i each alone increased early (Annexin V positive) and late apoptosis (PI positive) by approximately 2-fold in the BRCA^MUT cells without an additional increase when in combination (Fig. 5A and B). This combination did not induce apoptosis (by Annexin V or PI) in HR-proficient cells. PARPi and ATRi each alone increased early/late apoptosis approximately 2- and 4-fold from control, respectively in BRCA^MUT cells (Fig. 5A and C). Combination PARPi–ATRi treatment increased apoptosis 2-fold from PARPi alone in BRCA^MUT cells. When similar drug concentrations were tested in HR-proficient cells, apoptosis increased minimally with monotherapy but 2-fold with combination PARPi–ATRi (Fig. 5C). When higher concentrations of PARPi were tested (5 μmol/L), apoptosis increased 3-fold to 64% with the addition of ATRi to PARPi compared with PARPi alone (23%) and ATRi alone (19%; Supplementary Fig. S4) correlating with cell-cycle findings where G₂–M is increased, suggesting cells are unable to repair DNA. Caspase-3, a protein activated in the apoptotic cell both by extrinsic (death ligand) and intrinsic (mitochondrial) pathways, is another marker that was evaluated (34, 35). In BRCA^MUT and HR-proficient cells, treatment with CHK1i and ATRi increased cleaved caspase-3 compared with control (Fig. 5D). Combination treatments did not substantially increase this protein compared with monotherapy.

We next tested whether the synergistic increases in genomic instability and cell death resulting from ATRi/CHK1i combinations with PARPi would be reflected in increased therapeutic efficacy. To test this, we utilized the best known animal model of human ovarian cancer progression, genetics, and response to therapy: the orthotopic PDXs (24, 36, 37). Although some tumor growth suppression was observed with PARPi and CHKi as single agents, the addition of ATRi/CHK1i to PARPi in a BRCA2^MUT PDX model led to a statistically significant decrease in tumor volume relative to single-agent therapies (Fig. 6A). Notably, significant differences were observed in responsiveness to the PARPi–ATRi and PARPi–CHK1i combinations. Although PARPi–CHK1i combination indeed led to significantly increased tumor suppression over single-agent treatments, PARPi–ATRi led to a significant increase in the incidence of tumor regression. When looking at individual responses in each group using the Response Evaluation Criteria in Solid Tumors (RECIST) 1.1 score (38), 57% of mice had a Complete Remission (CR) in the PARPi and ATRi combination group compared with only 14% (1 mouse) in the PARPi and CHK1i combination group (Fig. 6C). There were no CRs using single-agent therapy (Fig. 6B). Toxicity was acceptable as mouse weights were comparable in the PARPi–ATRi and PARPi–CHK1i treatment arms to the control vehicle arm. Although not associated with obvious gastrointestinal symptoms, such as weight loss, abdominal distension, or death, we did observe increased bowel dilatation at necropsy for the PARPi–CHK1i group compared with the control arm. These findings indicate that checkpoint abrogation, particularly ATRi, synergizes with PARPi to promote tumor suppression and regression in BRCA1^MUT tumors in an orthotopic PDX model.

Consistent with our observations (Figs. 1–5) and prior reports of the stimulatory effect of CHK1i on CHK1 phosphorylation (10, 30), PDX tumors, evaluated after 1 week of treatment, exhibited an increase in p-CHK1 in mice treated with PARPi and CHK1i as single agents. Furthermore, PARPi–ATRi decreased pCHK1 compared with PARPi monotherapy (Fig. 6D; Supplementary Fig. S5). Thus, the drugs recapitulated our cell culture observations and expected effects, indicating that they maintained access to orthotopic tumors in vivo.

Capitalizing on synthetic lethality, PARPis have proven their clinical potential in treating cancer, both in BRCA-mutated and wild types. Olaparib, currently the only FDA-approved PARPi, results in a 40% and 30% response rate for recurrent BRCA^MUT and wild-type HGSOC, respectively, after first-line carboplatin–taxane standard of care (7, 39). Unfortunately, such responses are short lived, most lasting only 5 to 7 months with CRs occurring rarely (2%; refs. 7, 39). Using our orthotopic mouse model, we have demonstrated that PARPi treatment alone can suppress tumor growth at the MTD in a BRCA^MUT PDX model, but, similar to the clinical setting, it does not completely eliminate tumor burden despite prolonged treatment (Fig. 1C). Thus, strategies to optimize PARPi therapies for ovarian cancer are needed. The purpose of our study was to increase the efficacy of PARP inhibition by targeting critical cell-cycle checkpoints that are relied upon for cell survival during PARPi treatment.

Herein, we demonstrate that PARPi treatment increases reliance on the ATR–CHK1 pathway for genome stabilization and survival of BRCA^MUT cells. Indeed, combination of PARPi with ATRi or CHK1i treatment synergistically decreases cell viability and colony formation of BRCA^MUT cells, and to a less extent, HR-proficient cells. We propose that the synergistic effect of CHK1i or ATRi when combined with PARPi results both from (i) an increase in replication fork collapse by loss of two independent fork-stabilizing mechanisms that are controlled by CHK1/ATR and PARPi; and (ii) loss of the G₂–M phase checkpoint, which permits cells with this high level of DSBs to enter mitosis prematurely. The inability to appropriately partition broken chromatid fragments symmetrically dramatically increases cell death, a process known as mitotic catastrophe (Supplementary Fig. S6). Apoptosis either in G₂ or after mitotic catastrophe can be activated by a variety of DSB-sensing mechanisms, including those regulated by ATM.

However, key differences were observed between the PARPi–ATRi and PARPi–CHK1i combinations on genome stability and survival of BRCA^MUT tumor cells and PDX tumors. Although the PARPi–CHK1i combination was well tolerated in PDX mice and resulted in tumor suppression in BRCA2^MUT orthotopic transplant model (Fig. 6), this combination did not lead to tumor regression. In contrast, the PARPi–ATRi combination resulted in tumor regression and eradication of BRCA^MUT ovarian cancer PDX tumors. The dosing regimen studied, continuous PARPi with day 1 to 3 ATRi, was well tolerated in vivo, as evidenced by weight stability over the treatment course. In contrast, apoptosis was significantly increased with the PARPi-ATRi combination compared with monotherapy in both BRCA^MUT and HR-proficient cell models (Fig. 5; Supplementary Fig. S4).

The underlying causes of this clinically relevant difference may be best surmised from the distinct signaling roles of these kinases and their effects on genome stabilization when combined with PARPi. The ATR kinase lies upstream of CHK1 and phosphorylates numerous factors that may help preserve replication fork stability and control cell-cycle progression. The direct substrates of ATR include RPA, CLSPN, MCM2, p53, and many other factors that play roles in replication fork progression, DNA repair, and the cell cycle (19, 20). Thus, ATR may be able to stabilize replication forks independent of CHK1 (40) and permit cell survival when CHK1 is inhibited (41). In addition, ATR can suppress origin firing and the intra-S checkpoint independent of CHK1 (42, 43). Consistent with these interpretations, ATRi in combination with PARPi caused a substantial increase in chromatid breaks in the M-phase, a phenotype that represents unrepaired DSBs being permitted to enter the M-phase inappropriately (Fig. 4). In contrast, the appearance of chromosome aberrations in either PARPi–ATRi or PARPi–CHK1i implies inappropriate repair of DSBs before entry into mitosis, and such capping of DSB ends would be expected to suppress alternative DSB-stimulated checkpoint pathways. Therefore, the more substantial effects of the PARPi–ATRi combination on tumor progression likely result from the combination of increased replication fork collapse and abrogation of the G₂–M phase pathways, as described in more detail in the following paragraph. Additional research is required to further dissect the effect of PARPi–ATRi on genome stability and cancer cell survival, which may also depend on the genetics of the tumor.

Although differences in the efficacy of PARPi–ATRi and PARPi–CHK1i were observed, each of these combinations demonstrated significantly improved treatment efficacy over the application of any single agent. The mechanism behind this improvement is likely rooted in the distinct functions of PARP and ATR–CHK1 in preserving genome integrity. PARP helps ligate SSBs, which occur spontaneously at 20,000 to 50,000 sites per genome per day (5, 29). When left unrepaired because of PARPi treatment, these SSBs are converted into DSBs during DNA replication (5, 29). In contrast, ATR prevents DSB formation by making the replication fork less vulnerable to endonuclease attack (20, 42, 44). The additive, or possibly synergistic, effects of inhibiting these distinct pathways are further exacerbated by the suppression of G₂–M phase cell-cycle control by ATR–CHK1 pathway inhibition, leading to mitotic catastrophe (Figs. 4 and 5). Therefore, inhibition of ATR–CHK1 and PARP together increases DSB generation from fork collapse, which results either in elevated apoptosis in S–G₂ phase from other DSB-sensing mechanisms or mitotic catastrophe through cell-cycle checkpoint abrogation through ATR–CHK1 suppression. These mechanisms help explain the effects of PARPi–CHK1i and PARPi–ATRi combinations on tumor suppression, and in the case of PARPi-ATRi, tumor regression.

In summary, we have shown that PARPi increases reliance on ATR/CHK1 for genome stability and that the combination of PARPi with ATRi leads to complete ovarian tumor regression in an HR-deficient PDX model. Such responsiveness is not achievable with the maximum dose of PARPi alone, which is in accord with response rates to PARPi single-agent therapy in the clinic. Our goal is to convert the partial tumor responses typically seen with PARPi monotherapy into durable complete regressions using the combination of PARPi plus ATRi. AZD6738, a selective and bioavailable ATRi, is being investigated in early-phase clinical trials as monotherapy or in combination with chemotherapy or radiotherapy (ClinicalTrials.gov). Preliminary studies investigating AZD6738 as a monotherapy in the clinic show it is tolerable and demonstrates antitumor efficacy (45). PARPi in combination with ATRi will be evaluated in ovarian cancer patients in the near future.

No potential conflicts of interest were disclosed.

Conception and design: H. Kim, E. George, R.L. Ragland, R. Zhang, M. Morgan, M. Herlyn, E.J. Brown, F. Simpkins

Development of methodology: H. Kim, E. George, C. Krepler, E.J. Brown, F. Simpkins

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Kim, E. George, R.L. Ragland, E.J. Brown, F. Simpkins

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Kim, E. George, R.L. Ragland, S. Rafail, C. Krepler, E.J. Brown, F. Simpkins

Writing, review, and/or revision of the manuscript: H. Kim, E. George, R.L. Ragland, S. Rafail, R. Zhang, M. Morgan, M. Herlyn, E.J. Brown, F. Simpkins

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Rafail, F. Simpkins

Study supervision: C. Krepler, E.J. Brown, F. Simpkins

This work was supported from K08-CA151892-04, 1R01CA189743, Basser Team Science, and the Department of Defense OC150336 grants.

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.