Identification of ATR–Chk1 Pathway Inhibitors That Selectively Target p53-Deficient Cells without Directly Suppressing ATR Catalytic Activity

Genotoxic stresses, including ionizing radiation (IR), UV irradiation, and many chemotherapeutic agents, elicit DNA damage checkpoints that promote cell-cycle arrest and DNA repair (1). ATM (ataxia telangiectasia mutated) and ATR (ataxia telangiectasia and Rad3 related) are key protein kinases that sense DNA damage and phosphorylate their downstream targets to activate DNA damage checkpoints. IR and doxorubicin cause DNA double-strand breaks and primarily activate the ATM–Chk2–p53 pathway. UV and hydroxyurea cause stalled replication forks and primarily activate the ATR–Chk1 pathway. However, there is extensive crosstalk between these two pathways (2). The resultant signaling is affected by diverse factors, including types and doses of stress as well as timing. Most types of genotoxic stress activate both pathways to some degree.

Targeting DNA damage checkpoints could be beneficial to cancer treatment (3–5). In normal cells, DNA-damaging agents activate ATM and/or ATR pathways that mediate G₁–S, intra-S, and G₂–M checkpoints to arrest cell-cycle progression and allow extra time for DNA repair. Many cancers are deficient in the G₁–S checkpoint, including those with defective p53 (6, 7). This results in reliance on later checkpoints for DNA repair and cell survival, including the ATR pathway. If the ATR pathway is inhibited in G₁ checkpoint-deficient cells in the presence of DNA damage, cells will not elicit cell-cycle arrest for DNA repair, leading to cancer cell death (8). Thus, inhibiting the ATR pathway has been an appealing strategy to selectively sensitize p53-deficient cells to chemotherapeutic agents that damage DNA (3, 4, 9).

Because ATR inhibition can sensitize cancer cells to DNA-damaging agents (8, 10, 11), inhibitors that target the ATR–Chk1 pathway are of great interest for cancer therapy, and some Chk1 inhibitors have entered clinical trials (4, 5, 9). Caffeine is widely used as a nonselective ATM/ATR inhibitor (IC₅₀ of 200 μmol/L for ATM; IC₅₀ of 1,100 μmol/L for ATR; ref. 12), but more selective and potent ATR inhibitors have been long sought to better understand DNA damage response pathways and to use as chemosensitizers in combination with chemotherapeutic agents (13). Recently, schisandrin B, a dibenzocyclooctadiene derivative isolated from an herb, has been identified as a potent ATR-selective inhibitor (IC₅₀ of 7.25 μmol/L for ATR; ref. 14). In addition, aminopyrazine derivatives, compound 45 (IC₅₀ of 0.012 μmol/L for ATR; ref. 15) and VE-821 (IC₅₀ of 0.026 μmol/L for ATR; ref. 16), have been identified as highly potent ATR-selective inhibitors. These compounds sensitized cancer cells to cisplatin, while sparing normal cells (15, 16), further supporting the potential utility of ATR inhibition to sensitize cancer cells to death.

The mechanism of ATR activation is complex, requiring recruitment of more than 10 proteins at sites of DNA damage for optimal kinase activation (17). Briefly, a replication fork stalls at DNA lesions, while helicases continue to separate DNA strands, generating single-stranded DNA (ssDNA) coated with RPA (replication protein A; ref. 17). Multiple proteins, including the 9–1–1 complex, TopBP1, and Claspin, are then recruited to RPA-coated ssDNA. Thus, not only ATR kinase but also a handful of known proteins involved in ATR pathway activation could be targets for suppression of the ATR pathway. Most compounds that target this pathway are believed to act as ATP-competitive inhibitors of ATR kinase activity, such as caffeine, schisandrin B, or VE-821. Compounds that target other components of the ATR pathway are few in number and typically affect nonspecific processes such as protein chaperones and stability. For example, Hsp90 inhibitors (geldanamycin and 17-allylamino-17-demethoxygeldanamycin) deplete Chk1 and hence block ATR pathway function, as well as many other targets (18, 19).

A chemical genetic screen is a powerful approach to discover small-molecule compounds that perturb cellular processes and elucidate biologic systems (20, 21). Small molecules that suppress the ATR–Chk1 pathway without inhibiting ATR could be identified through pathway-targeted screening. Such compounds could uncover previously unknown but druggable targets in signaling pathways or reveal known targets for which no small-molecule inhibitors have been identified. To discover novel DNA damage response inhibitors that target the ATR pathway, we performed cell-based, high-content screening of small-molecule compounds. Following biologic assays for validation, we found four novel small-molecule compounds that inhibit the ATR–Chk1 pathway. Importantly, these four compounds did not directly inhibit ATR kinase catalytic activity in vitro but had chemosensitization effects, especially in p53-deficient cells in vitro. One compound, dihydrocelastryl, was also tested in a xenograft model and showed selective inhibition of p53-deficient tumor growth in vivo. Identification of these mechanistically distinct inhibitors of DNA damage checkpoints allows their use as probes to further elucidate checkpoint function.

HeLa and 293T cells were obtained from the ATCC, authenticated through short tandem repeat profiling by the ATCC, and cultured in DMEM (#11995; Invitrogen) supplemented with 10% heat-inactivated FBS (Invitrogen) and 1% penicillin–streptomycin (PS; Invitrogen). HCT116 p53^+/+ cells and the isogenic p53^−/− cells were gifts from Dr. Bert Vogelstein (Johns Hopkins University, Baltimore, MD; ref. 22) and cultured in McCoy's 5A Medium (#16600; Invitrogen), supplemented with 10% FBS and 1% PS. p53 expression in HCT116 p53^+/+ and p53^−/− cells was validated by Western blot analyses. Upon receipt of all cell lines, frozen stocks were made at low passage and thawed for short-term culture (less than 3 months). All cells were maintained at 37°C in a humidified atmosphere of 5% CO₂.

Screening was performed at the Broad Institute of MIT and Harvard using its small-molecule libraries, comprising known bioactive and novel synthetic compounds. Primary screen data were deposited to ChemBank (http://chembank.broadinstitute.org) under the project name “DNADamageImagingScreen.” HeLa cells (4,000 cells/well) were plated onto 384-well flat clear bottom black polystyrene tissue culture treated microplates (#3712; Corning) and incubated at 37°C overnight in a humidified atmosphere of 5% CO₂. Compound mixture of hydroxyurea and a test compound for cell treatment was prepared in a separate 384-well plate: 100 mmol/L hydroxyurea was diluted with cell culture medium (DMEM supplemented with 10% FBS and 1% PS) to a final concentration of 3 mmol/L in 45 μL medium in each well, and a single small-molecule compound was added to each well by 100 nL pin-transfer (CyBio) of a compound stock solution in DMSO. Culture medium was aspirated, and 45 μL of compound mixture was added to each well that also contained 5 μL of residual culture medium (total 50 μL/well). Because the average compound concentration of stock solution was 10 mmol/L in DMSO, the final compound concentration for cell treatment was 20 μmol/L in test wells (final DMSO 0.2%). Cells were incubated with 3 mmol/L hydroxyurea and a single small-molecule compound in each well at 37°C for 2 hours in a humidified atmosphere of 5% CO₂. For immunofluorescence, cells were fixed by adding 15 μL per well of 16% paraformaldehyde (formaldehyde) aqueous solution (#15710; Electron Microscopy Sciences) to 50 μL medium in each well (final 4% formaldehyde) and incubating at 4°C overnight. Fixed cells were treated with 2N HCl (50 μL/well) for 10 minutes at room temperature. Cells were washed twice with 50 μL per well of TBS (50 mmol/L Tris, 150 mmol/L NaCl, pH 7.4) and treated with 50 μL per well of TBS supplemented with 0.5% Triton X-100 for 15 minutes at room temperature. Cells were washed with 50 μL per well of TBS and treated with 50 μL per well of TBS supplemented with 0.1% sodium borohydride (#452882; Sigma-Aldrich) for 5 minutes at room temperature. Cells were washed twice with 50 μL per well of TBS and treated with 50 μL per well of TBS supplemented with 3% BSA (A9647; Sigma-Aldrich; 3% BSA–TBS) for 10 minutes at room temperature. Cells were incubated with 20 μL per well of 1:150 dilution of anti–phospho-Chk1 (pChk1–Ser345) [133D3] rabbit mAb (#2348; Cell Signaling Technology) in 3% BSA–TBS at 4°C overnight. Cells were washed with 50 μL per well of 3% BSA–TBS and incubated with 20 μL per well of 1:150 dilution of Alexa Fluor 488 goat anti-rabbit IgG antibody (A-11034; Invitrogen) in 3% BSA–TBS for 1.5 hours at room temperature. Cells were washed with 50 μL per well of 3% BSA–TBS and incubated with 20 μL per well of 1:10 dilution of VECTASHIELD Mounting Medium with DAPI (1.5 μg/mL DAPI; H-1200; Vector Laboratories) in 3% BSA–TBS for 30 minutes at room temperature. Cells were washed twice with 50 μL per well of 3% BSA–TBS and twice with 50 μL per well of TBS. Immunofluorescent images of cells stained with anti-pChk1 antibody and DAPI in each well were acquired by automated fluorescence microscopy using DAPI and FITC filters (ImageXpress; Molecular Devices) and analyzed by MetaXpress (Molecular Devices). To detect pChk1 signals only from nuclei and eliminate background signals from cytoplasm and the surface of cell culture wells, pChk1 nuclear signal intensity in each well was quantitated within all nuclei of HeLa cells (determined by DAPI-positive areas). To normalize cell number differences among wells of screening plates, average pChk1 nuclear signal intensity for one test compound in each well was calculated by dividing pChk1 nuclear signal intensity from all nuclei in each well by the number of nuclei (DAPI-positive areas). For hit compound selection, the average of pChk1 intensities among all 384 wells of each plate (whole plate average intensity) was defined as 100% (because most compounds had a little or no effect on hydroxyurea-induced Chk1 phosphorylation), and compounds whose average pChk1 nuclear signal intensities were <85% were chosen for follow-up.

Supplementary Materials and Methods include chemicals, siRNA transfection, treatment with DNA-damaging agents, Western blot analyses, flow-cytometry analyses, in vitro ATR kinase assay, cell viability assay, human tumor xenografts in athymic nude mice, and statistical analyses.

To discover DNA damage response inhibitors that suppress the ATR pathway and to gain insight into replication checkpoint signaling, we performed a cell-based, high-content screen of small-molecule libraries using an automated cell imaging platform. The specific small-molecule screen involved detection of hydroxyurea-induced phosphorylation of Chk1 at Ser345, a downstream target of ATR, as an indicator for ATR pathway activation (23, 24). Hydroxyurea inhibits DNA synthesis and stalls replication forks. This replication stress activates the ATR pathway, leading to phosphorylation of Chk1. In this screen, HeLa cells were plated onto 384-well plates and treated with hydroxyurea and a single small-molecule compound in each well (Fig. 1A). HeLa cells were selected for this cell imaging assay, because this cell type demonstrated a robust signaling response and remained adherent during washing steps required for immunofluorescence. Cells were stained with anti–phospho-Chk1 (pChk1–Ser345) antibody and DAPI for nuclear staining, and the immunofluorescent images of cells in each well were acquired by automated fluorescence microscopy. Subsequently, the acquired images were analyzed to quantitate pChk1 signal intensities within nuclei that were defined by DAPI-positive areas. Compounds that suppressed hydroxyurea-induced Chk1 phosphorylation were selected as putative ATR pathway inhibitors.

Using this approach, hydroxyurea-induced pChk1 was reliably detected as a significant increase in pChk1 nuclear signal intensity (middle bar in Fig. 1B) compared with untreated cells (left bar). Addition of 3 mmol/L caffeine (right bar) suppressed hydroxyurea-induced phosphorylation of Chk1 to 85% of pChk1 intensity of hydroxyurea-treated cells. This effect on pChk1 signal intensity in cells was consistent with Western blot results using the same pChk1 antibody (Fig. 1B), although untreated cells showed no pChk1 signal in a Western blot analysis. Despite relatively high nonspecific background immunofluorescence in the microscopy assay, cell imaging at the single-nucleus level was reliable in detecting phospho-Chk1 signals in a high-throughput manner.

Figure 1C shows a representative result of one 384-well plate from the primary screen. Average pChk1 intensity of 384 wells was set to 100% (whole-plate average shown as a solid horizontal line), presuming that the majority of compounds were not ATR pathway inhibitors, and thus the whole-plate average was virtually the same as the level of pChk1 induced by hydroxyurea alone. Compounds that decreased the nuclear pChk1 signal below 85% (dotted horizontal line) of the whole-plate average were more potent than 3 mmol/L caffeine, and thus were selected for follow-up. From the example plate shown in Fig. 1C, dihydrocelastryl was selected for follow-up whereas teniposide, thiram, and bleomycin showed high levels of pChk1 intensity (Fig. 1C). These latter three compounds are indeed known DNA-damaging agents that activate DNA damage response pathways, suggesting that this assay could appropriately detect both activating and inhibitory compounds.

In a secondary screen, we performed dose–response experiments in triplicate by the same automated cell imaging assay (Fig. 2A). Increasing doses of caffeine inhibited hydroxyurea-induced pChk1 as a positive control. The four novel compounds (anthothecol, dihydrocelastryl, erysolin, and MARPIN: “ATM and ATR pathway inhibitor”) also inhibited phosphorylation in a dose-dependent manner, all of which were significantly more potent than caffeine.

Because we performed a small-molecule screen detecting hydroxyurea-induced pChk1, we needed to exclude the possibility that these novel compounds could be merely modulators of hydroxyurea-induced signaling, not ATR pathway inhibitors. In a tertiary screen, we performed Western blot analyses to validate the inhibitory effects of novel compounds on pChk1 induced by two mechanistically distinct ATR pathway activators: hydroxyurea (acts by nucleotide depletion) and UV irradiation (acts by generating polymerase-blocking thymine dimers). Hydroxyurea-induced pChk1 was inhibited by these novel compounds in a dose-dependent manner (Fig. 2B) as was UV-induced pChk1 (Fig. 2C). These findings confirmed that these compounds are indeed ATR pathway inhibitors, not merely modulators of hydroxyurea-induced signaling.

In total, we screened 9,195 compounds from two distinct small-molecule libraries, comprising known bioactive and novel synthetic compounds (Fig. 2D). The average compound concentration in this cell-based screen was 20 μmol/L following pin transfer of the compounds into test wells. Test compounds that did not show dose-dependent inhibition of hydroxyurea-induced pChk1 in the secondary and tertiary screens were eliminated from further investigation. After these screens by cell imaging and Western blot analyses, we eventually identified four novel ATR pathway inhibitors from 9,195 compounds.

Chemical structures of these novel ATR pathway inhibitors and caffeine are shown in Fig. 2E. The four novel ATR pathway inhibitors were structurally different from each other and from caffeine. Anthothecol, dihydrocelastryl, and erysolin were derived from libraries of known bioactive compounds and are commercially available. The fourth inhibitor, identified from a diversity-oriented synthetic library, was named MARPIN (ATM and ATR pathway inhibitor) based on its ability to inhibit both the ATM and ATR pathways, as determined in subsequent experiments. MARPIN's chemical name is methyl (1R,6aR)-2-benzoyl-1-(4-fluorobenzyl)-6-methylene-5-oxo-1,2,3,5,6,6a-hexahydrocyclopenta[c]pyrrole-1-carboxylate [ChemBank ID 1965360, PubChem Substance ID 11253916, PubChem Compound ID 3247230]. To the best of our knowledge, none of these four compounds have been previously reported to be DNA damage response pathway inhibitors. The hit rate from the known bioactive compounds library (0.12%) was much higher than that of the synthetic library (0.015%). This difference in hit rates supports the idea that the known bioactive compounds are heavily biased toward biologically active compounds and that the cellular activities of these agents have not yet been exhaustively described.

All four novel ATR pathway inhibitors suppressed hydroxyurea-induced pChk1 at concentrations below 30 μmol/L in the immunofluorescence imaging assays, whereas caffeine required 3,000 μmol/L for the same degree of pChk1 inhibition (Fig. 2E). Thus, the novel compounds were at least 100 times more potent than caffeine for ATR pathway inhibition in cells.

A pivotal function of DNA damage response kinases is to elicit cell-cycle checkpoint signaling and stop cell-cycle progression, allowing time for DNA repair (4). To determine whether novel inhibitors functionally suppress cell-cycle arrest, the effect of these compounds on the IR-induced G₂–M checkpoint was assessed by measuring the percentage of cells positive for phospho-Histone H3 (Ser10) with 4N DNA content as an indicator of mitosis. After 2 hours, 2 Gy of IR decreased the percentage of mitotic cells from 2.2% to 0.1% (Fig. 3A), indicating that IR activated the G₂–M checkpoint and inhibited mitotic entry. Known inhibitors of ATM and/or ATR abrogated the IR-induced G₂–M checkpoint, including caffeine (dual ATM/ATR inhibitor), KU-55933 (ATM-selective inhibitor), and VE-821 (ATR-selective inhibitor; Fig. 3). These data suggest that the IR-induced G₂–M checkpoint is mediated via both ATM and ATR, consistent with prior studies (25, 26). Furthermore, all four novel ATR pathway inhibitors we discovered abrogated IR-induced G₂–M checkpoint in a dose-dependent manner (Fig. 3), supporting that these inhibitors functionally block the induction of cell-cycle arrest, a pivotal role for DNA damage response pathways.

To investigate whether novel compounds directly inhibit the catalytic activity of ATR kinase, we performed in vitro ATR kinase assays using immunopurified ATR proteins (Fig. 4). Known nonspecific ATR inhibitors, caffeine (12) and wortmannin (27), suppressed ATR kinase activity, validating the ATR kinase assay. However, none of the four novel compounds suppressed ATR kinase activity in vitro even at far higher concentrations of these compounds than were required to inhibit hydroxyurea-induced pChk1 in cells. These results suggest that the novel inhibitors are mechanistically distinct from caffeine in their mode of suppressing ATR signaling in cells.

To examine whether these novel inhibitors suppress ATR- or ATM-dependent phosphorylation events or both, we carefully selected conditions under which signaling is predominantly dependent on ATR or ATM, as there is often crosstalk between these two pathways (2). To this end, we used siRNA against ATR or ATM to determine the extent to which a signaling event was dependent on either of these kinases. This survey examined approximately 700 phospho-specific events under various conditions or time points: Three different cell types (HeLa, 293T, and HCT116), five DNA-damaging agents (hydroxyurea, UV, IR, bleomycin, and doxorubicin), three doses of each agent, five time points, and six different phosphorylation sites on four DNA damage response proteins [Chk1 (Ser317 and Ser345), p53 (Ser15 and Ser20), Chk2 (Thr68), and ATM (Ser1981)] for their dependence on either of these kinases. We ultimately identified hydroxyurea-induced phospho-Chk1 (Ser345) and phospho-p53 (Ser15) at 2 hours in 293T cells as the most ATR-selective phosphorylation events (Fig. 5A). The most ATM-selective signaling event was phospho-Chk2 (Thr68) induced by low dose (2 Gy) of IR at an early time point (15 minutes) in HCT116 p53^+/+ cells (Fig. 5B). We then used these optimized, pathway-selective conditions to test the extent to which the novel inhibitors suppressed signaling via the ATR and ATM pathways. All four compounds inhibited ATR-selective phosphorylation events (hydroxyurea-induced phospho-Chk1 and phospho-p53) in a dose-dependent manner (Fig. 5C). Erysolin and MARPIN inhibited IR-induced phospho-Chk2 in a dose-dependent manner, but anthothecol or dihydrocelastryl did not inhibit this ATM-selective phosphorylation event (Fig. 5D). Taken together, all four compounds are ATR pathway inhibitors, two of which also have the ability to suppress the ATM pathway at an early time point.

G₁ checkpoint-deficient cells, particularly those deficient in p53 function, are resistant to DNA-damaging agents but can be sensitized by ATR (8) or Chk1 (28, 29) inhibition. Therefore, we tested whether these novel ATR pathway inhibitors would also sensitize cancer cells to DNA damage. To investigate whether these ATR pathway inhibitors can overcome the chemoresistance of p53-deficient cells, we examined cell viability after treating cells with combinations of DNA-damaging agents with novel ATR pathway inhibitors. We used HCT116 cells (human colorectal cancer cell lines), because the isogenic p53^−/− cells are available (22), and we can readily compare p53 wild-type (p53^+/+) and p53-deficient (p53^−/−) cells.

First, we investigated whether novel ATR pathway inhibitors sensitize cancer cells to cisplatin, a DNA crosslinker. Cisplatin was selected because other studies have shown that it synergizes effectively with ATR inhibition in killing cancer cells (11, 30). In p53 wild-type cells, cisplatin alone (10 μmol/L) decreased cell viability (to 75%), whereas p53-deficient cells were resistant to cisplatin (92% cell viability; Fig. 6A). Cisplatin plus anthothecol (1 μmol/L) showed a marked synergistic decrease in cell viability, especially in p53-deficient cells (15% cell viability).

Because ATR inhibition potentiates the effects of various classes of chemotherapeutic drugs (11, 31), we also examined whether these novel ATR pathway inhibitors sensitize cancer cells to camptothecin, a Topoisomerase I poison that results in S phase slowing. Camptothecin alone (120 nmol/L) decreased cell viability, especially in p53 wild-type cells (53% cell viability; Fig. 6B), but p53-deficient cells were relatively resistant to camptothecin (73% cell viability). Although 1 μmol/L dihydrocelastryl had no effect in either cell type if given alone, it fully reversed the camptothecin-resistant phenotype of p53-deficient cells (73% cell viability with camptothecin alone vs. 21% when dihydrocelastryl was added).

To precisely examine whether novel ATR pathway inhibitors show synergistic effects on decreasing cell viability in combination with DNA-damaging agents, we extensively tested various concentrations of compounds (Supplementary Fig. S1). These cell viability data were used to calculate synergistic effects on decreasing cell viability (see example in Fig. 6C). All novel ATR pathway inhibitors and caffeine showed significant synergistic lethality with cisplatin (Fig. 6D) and camptothecin (Fig. 6E), especially in p53-deficient cells. In contrast, an ATM-selective inhibitor KU-55933 (32, 33) showed a modest synergistic effect on cisplatin treatment, implying that ATR inhibition plays a more important role in sensitizing cancer cells to cisplatin. These results demonstrate that chemoresistant p53-deficient cells can be sensitized by these novel ATR pathway inhibitors. As single agents, these compounds showed modest cytotoxicity in HCT116 cancer cells at concentrations that caused synergistic sensitization effects in HCT116 cells (Supplementary Fig. S1) but no cytotoxicity in normal cells (primary human keratinocytes at 24 hours).

The doses of novel inhibitors required for effective synergistic lethality at 48 hours (1–8 μmol/L; Fig. 6) were approximately 8- to 60-fold lower than doses required for inhibition of hydroxyurea-induced pChk1 at 2 hours (15–60 μmol/L; Fig. 5C). To verify that such very low doses of novel inhibitors that were used for synergistic lethality experiments can suppress cisplatin-induced pChk1, we used flow cytometry (a highly sensitive assay) to detect pChk1 status. Flow cytometry–based measurement of pChk1 was validated by detecting the inhibitory effect of VE-821 (established ATR-selective inhibitor) on hydroxyurea-induced pChk1 (Supplementary Fig. S2A and S2B). Cisplatin induced phosphorylation of Chk1 at Ser345 predominantly in S phase, although pChk1 induction by 10 μmol/L cisplatin was subtle, compared with pChk1 induced by 3 mmol/L hydroxyurea (Supplementary Fig. S2). At the very low doses that showed synergistic lethality, all four novel ATR pathway inhibitors weakly but reproducibly suppressed cisplatin-induced pChk1 (Supplementary Fig. S2D). This suggests that low levels of ATR–Chk1 pathway inhibition are sufficient to sensitize cells to DNA-damaging agents after 48 hours, exhibiting effective synergistic lethality.

Because novel ATR pathway inhibitors indeed sensitized cancer cells to DNA damage in vitro, we investigated the in vivo antitumor activity of a novel ATR pathway inhibitor and its synergistic effect with cisplatin in a xenograft model. We selected dihydrocelastryl for testing its in vivo efficacy, because it was commercially available and preliminary in vivo data suggested efficacy at a concentration at which the compound could be dissolved. We used a subcutaneous tumor xenograft mouse model implanted with the human colon carcinoma cell line (HCT116), which we had also used for cell viability studies. This HCT116 xenograft model has been widely used to study antitumor activity of small-molecule compounds (34). We used both HCT116 p53-intact (p53^+/+) cells and the isogenic p53-null (p53^−/−) cells to elucidate p53 dependency on eliminating cancer cells. Because high doses (4 or 8 mg/kg/d) of cisplatin in our injection schedule resulted in severe body weight loss, we used 2 mg/kg/d of cisplatin that did not cause loss of body weight. Dihydrocelastryl at 0.1 mg/kg/d also did not cause loss of body weight. Cisplatin alone inhibited growth of p53^+/+ tumors (52% tumor volume compared with vehicle treatment), and the addition of dihydrocelastryl had no effect (Fig. 7A). Importantly, p53^−/− tumors were relatively resistant to cisplatin (85% tumor volume compared with vehicle treatment), and dihydrocelastryl sensitized p53^−/− tumors to cisplatin, suppressing tumor volume to 60% (Fig. 7B). Thus, dihydrocelastryl demonstrated synergistic suppression of chemoresistant p53^−/− tumor growth in vivo.

Because of potential chemotherapeutic (sensitization of chemoresistant cells to death; refs. 3, 4) and chemopreventive (skin cancer prevention; ref. 35) implications, ATR pathway inhibitors have long been sought. Caffeine has been widely used as an ATM/ATR inhibitor, but caffeine requires millimolar levels to inhibit ATR and has several other targets (12, 36). Previous attempts to discover novel DNA damage response inhibitors have focused on finding better ATP-competitive inhibitors of kinases in this pathway. In the present study, we used a phenotype-based screening approach to identify ATR pathway inhibitors, potentially mechanistically distinct from typical ATP-competitive kinase inhibitors. Because the ATR activation mechanism is complex and requires multiple proteins to be recruited to damaged DNA, any of these components of the ATR pathway could potentially be a target of the identified small-molecule inhibitors. Specifically, unlike traditional biochemical screening, this phenotype-based approach allowed us to discover novel ATR pathway inhibitors that are bioactive and do not act as direct ATR kinase catalytic inhibitors.

Mutations of p53 are the most common genetic abnormality in human cancers. The prevalence of p53 dysfunction is approximately 50% in major human cancers, including lung, colon, and skin cancers (6). p53-deficient cells are resistant to DNA-damaging agents (37), and there is, thus, a need to sensitize p53-deficient cells for effective cancer therapy. Because ATR inhibition can sensitize G₁ checkpoint-deficient cells to DNA-damaging agents (8), we investigated the chemosensitization abilities of novel ATR pathway inhibitors in the present study. The ATR pathway inhibitors demonstrated synergistic effects on cancer cell killing in p53-deficient cells. Despite the fact that these compounds did not directly inhibit ATR kinase catalytic activity, the compounds blocked ATR pathway function and promoted sensitization of drug-resistant p53-deficient cells. Previous studies demonstrated that loss of ATM sensitizes p53-deficient cells to camptothecin, but not cisplatin (38), whereas ATR inhibition can sensitize cancer cells to both camptothecin (31) and cisplatin (11). In particular, p53-deficient cells are preferentially sensitized to cisplatin by ATR inhibition (39). The inhibitors we describe here sensitized p53-deficient cells to both cisplatin and camptothecin. This wide range of chemosensitization effects of the ATR pathway inhibitors is potentially advantageous to cancer therapeutic applications.

Certain barriers often block translation of a given small-molecule compound to the in vivo setting even if the compound has desired in vitro effects. These barriers include solubility of the compound, in vivo stability, drug delivery to tumors, and toxicity (40). To investigate the in vivo characteristics of a novel ATR pathway inhibitor, we tested one compound, dihydrocelastryl, using a xenograft model. Dihydrocelastryl was selected for xenograft experiments, because it showed strong chemosensitization for p53^−/− cells in the cell viability study and in an in vivo pilot study. Dihydrocelastryl overcame these barriers and synergistically sensitized p53-deficient tumors in combination with cisplatin in vivo without severe body weight loss or any other overt health concerns. This synergistic effect on antitumor efficacy in vivo is meaningful for reducing the dose of cisplatin to minimize undesired toxic effects of chemotherapeutic agents, because higher doses of cisplatin (4 or 8 mg/kg/d) were toxic to mice in this experimental setting, inducing cachexia. Thus, ATR pathway inhibitors have the potential to improve the efficacy of existing chemotherapeutic agents that cause DNA damage.

Three of the novel ATR pathway inhibitors that we discovered were from a library of known bioactive compounds. Anthothecol is a limonoid of Khaya anthotheca (the mahogany family Meliaceae) and has antimalarial activity (41). Dihydrocelastryl is structurally similar to celastrol, a triterpene extracted from the root bark of the Chinese medicine “Thunder of God Vine.” Celastrol is a potent proteasome inhibitor and suppresses human prostate cancer growth in nude mice (42). Erysolin is structurally similar to sulforaphane, an isothiocyanate that has antimicrobial activity and shows cancer-preventive effects (43). Sulforaphane is abundant in broccoli sprouts. Sulforaphane and erysolin induce detoxification enzymes (44). The induction of detoxification enzymes may be a significant component of the anticarcinogenic action of broccoli. Importantly, on the basis of PubMed/literature searches, we were unable to find prior studies, which suggest that any of these compounds may act as inhibitors of DNA damage checkpoint function.

Because these novel compounds inhibited the ATR pathway without directly affecting ATR kinase catalytic activity, it is plausible that the compounds may inhibit a mediator protein involved in this signaling pathway. Previous studies showed that ATM and ATR functions could be suppressed by downregulating a protein required for their signaling such as BRCA1 (45), protein phosphatase 5 (46, 47), or Timeless (48, 49). Thus, it is possible that the ATR pathway inhibitors target a nonkinase component of the ATM/ATR pathways to suppress signal transduction. Further investigation to elucidate direct molecular targets of these novel ATR pathway inhibitors is ongoing, with a particular focus on target identification for MARPIN using structure–activity relationship studies and MARPIN-immobilized beads to pull down relevant binding proteins (50).

In conclusion, a cell-based phenotypic screen yielded novel ATR–Chk1 pathway inhibitors that are mechanistically distinct from typical kinase inhibitors and effectively sensitize p53-deficient cells not only in vitro but also in an in vivo xenograft model. This phenotype-based approach has the advantage that it can identify bioactive small-molecule compounds that could target any aspect of the ATR pathway, rather than focusing only on ATR kinase catalytic activity. Importantly, these compounds that are not ATR catalytic inhibitors will be useful as probes to better understand DNA damage response pathways. Detailed characterization of the mechanisms of action will further elucidate ATR pathway function in response to various types of DNA damage, and may identify novel druggable proteins for cancer prevention and therapy.

No potential conflicts of interest were disclosed.

Conception and design: M. Kawasumi, J.E. Bradner, P. Nghiem

Development of methodology: M. Kawasumi, J.E. Bradner, P. Nghiem

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Kawasumi, J.E. Bradner, N. Tolliday, R. Thibodeau, H. Sloan, K.M. Brummond, P. Nghiem

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Kawasumi, J.E. Bradner, N. Tolliday, P. Nghiem

Writing, review, and/or revision of the manuscript: M. Kawasumi, J.E. Bradner, N. Tolliday, P. Nghiem

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Kawasumi, P. Nghiem

Study supervision: M. Kawasumi, P. Nghiem

The authors thank Dr. Bert Vogelstein (Johns Hopkins University) for the gift of HCT116 p53^+/+ and the isogenic p53^−/− cells. Allan Conney, PhD, passed away in 2013. We are deeply grateful for the inspirational role that Dr. Conney played in this study.

This work was supported by the NIH (R01-AR049832; P. Nghiem), a Dermatology Foundation Career Development Award (M. Kawasumi), and the Michael Piepkorn Endowment for Dermatology Research at University of Washington.

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.