Ataxia-telangiectasia and Rad3-related and DNA-dependent protein kinase cooperate in G₂ checkpoint activation by the DNA strand-breaking nucleoside analogue 2′-C-cyano-2′-deoxy-1-β-d-arabino-pentofuranosylcytosine Free

2′-C-Cyano-2′-deoxy-1-β-d-arabino-pentofuranosylcytosine (CNDAC), the prodrug (sapacitabine) of which is in clinical trials, has the novel mechanism of action of causing single-strand breaks after incorporating into DNA. Cells respond to this unique lesion by activating the G₂ checkpoint, affected by the Chk1-Cdc25C-cyclin-dependent kinase 1/cyclin B pathway. This study aims at defining DNA damage checkpoint sensors that activate this response to CNDAC, particularly focusing on the major phosphatidylinositol 3-kinase–like protein kinase family proteins. First, fibroblasts, deficient in ataxia-telangiectasia mutated (ATM), transfected with empty vector or repleted with ATM, were arrested in G₂ by CNDAC to similar extents, suggesting ATM is not required to activate the G₂ checkpoint. Second, chromatin associations of RPA70 and RPA32, subunits of the ssDNA-binding protein, and the ataxia-telangiectasia and Rad3-related (ATR) substrate Rad17 and its phosphorylated form were increased on CNDAC exposure, suggesting activation of ATR kinase. The G₂ checkpoint was abrogated due to depletion of ATR by small interfering RNA, and impaired in ATR-Seckel cells, indicating participation of ATR in this G₂ checkpoint pathway. Third, the G₂ checkpoint was more stringent in glioma cells with wild-type DNA-dependent protein kinase catalytic subunit (DNA-PKcs) than those with mutant DNA-PKcs, as shown by mitotic index counting. CNDAC-induced G₂ arrest was abrogated by specific DNA-PKcs inhibitors or small interfering RNA knockdown in ML-1 and/or HeLa cells. Finally, two phosphatidylinositol 3-kinase–like protein kinase inhibitors, caffeine and wortmannin, abolished the CNDAC-induced G₂ checkpoint in a spectrum of cell lines. Together, our data showed that ATR and DNA-PK cooperate in CNDAC-induced activation of the G₂ checkpoint pathway. [Mol Cancer Ther 2008;7(1):133–42]

Because cells are threatened by endogenous or exogenous genotoxic stresses, they have evolved delicate protective machineries to maintain the genomic integrity. In response to environmental perturbations/toxicants such as ionizing radiation (IR), UV light, and DNA-damaging agents, cells delay cell cycle progression by initiating cell cycle checkpoints to allow time for repair of damaged DNA. According to current models, the DNA damage checkpoints conceptually have four functional components: sensors, mediators, transducers, and effectors (1). Sensors of DNA damage, typically ataxia-telangiectasia mutated (ATM), ataxia-telangiectasia and Rad3-related (ATR), and DNA-dependent protein kinase (DNA-PK), which belong to the phosphatidylinositol 3-kinase–like protein kinase (PIKK) family and share homology of the phosphatidylinositol 3-kinase domain harboring the catalytic site of the active protein kinases (2), detect various insults to the genome and initiate signal transduction cascades to affect the S-phase and G₂ checkpoint pathways (3).

Although some functional redundancy exists among the PIKK sensors, the detection of different types of DNA damage resulting from specific forms of genotoxic stress is generally addressed by unique sensors. For example, ATM is the major sensor responding to the double-strand breaks (DSB) caused by IR (2, 4). Mediated by the Mre11-Rad50-Nbs1 complex (5–7), ATM is rapidly activated through intermolecular autophosphorylation and subsequent dimer dissociation (8). The activated ATM phosphorylates the DNA damage checkpoint kinase Chk2, which transduces signals downstream. In contrast, ATR plays a more crucial role when cells are exposed to UV radiation, stalled replication forks, or chemicals that make bulky base adducts (1, 2). ssDNA generated by these events becomes coated with replication protein A (RPA), which attracts the ATR-interacting protein (ATRIP). ATR associates with this complex and initiates cell cycle checkpoints via phosphorylation of another DNA damage checkpoint kinase, Chk1. Recent studies from different laboratories provided evidence for ATM activity upstream of ATR recruitment to IR-damaged chromatin (9–11), revealing cross-talk between these two important sensors.

DNA-PK is another PIKK family member known to be a nuclear serine/threonine kinase and a molecular sensor in the DNA damage response. Composed of a catalytic subunit (DNA-PKcs) and two regulatory subunits (Ku70 and Ku80; ref. 12), DNA-PK is one of the core components of mammalian repair of DSBs through the nonhomologous end-joining pathway. It has been proposed to sense DNA damage and enhance signal transduction via phosphorylation of downstream targets, including H2AX and p53 (13). However, a role for DNA-PK in the activation of transducers of the DNA damage signaling pathway in cellular responses to genotoxic stress has not been established.

Designed on a basis of a DNA self-strand–breaking mechanism, 2′-C-cyano-2′-deoxy-1-β-d-arabino-pentofuranosylcytosine (CNDAC) is incorporated within the DNA of proliferating cells and induces a single-strand break (SSB) by a β-elimination process (14, 15). The DNA nick results from the rearrangement of the CNDAC molecule to form 2′-C-cyano-2′,3′-didehydro-2′,3′-dideoxycytidine (CNddC), which is a de facto DNA chain terminator (16). Unlike the nucleoside analogues 1-β-d-arabinofuranosylcytosine, gemcitabine, and fludarabine that cause S-phase arrest, cells respond to this damage by activating the G₂ checkpoint (17). Although CNDAC-induced cell cycle arrest is instituted through activation of the Chk1-Cdc25C-cyclin-dependent kinase 1 (Cdk1)/cyclin B checkpoint pathway, participants in the signaling events upstream of Chk1 kinase remain obscure.

The aim of this study was to elucidate the sensor proteins that detect CNDAC-induced DNA damage and activate the G₂ checkpoint but not to focus on cell survival mechanisms. Strategies taking advantage of genetically defined cell lines and inhibition of PIKK family proteins by small-molecule inhibitors or RNA interfering technology were used to determine the apical molecules in the cellular responses to DNA damage generated after CNDAC incorporation. Our results show that both ATR and DNA-PK participate in activation of the G₂ checkpoint but that ATM is not essential for this response.

Acute myelogenous leukemia cell line ML-1 (gifted by Dr. Michael B. Kastan, St. Jude Children's Research Hospital, Memphis, TN) was maintained in exponential growth phase in RPMI 1640 supplemented with 10% heat-inactivated fetal bovine serum at 37°C in a humidified atmosphere with 5% CO₂. AT22IJE-T (AT-C), a fibroblast cell line derived from an ataxia-telangiectasia patient, and lines stably transfected with either an episomal expression vector (AT22IJE-TpEBS7, AT-V) or full-length ATM cDNA (AT22IJE-TpEBS7-YZ5, AT-AT) were gifts from Dr. Yosef Shiloh (Tel Aviv University, Tel Aviv, Israel; ref. 18) and were cultured in DMEM with high glucose and 20% fetal bovine serum. Glioma-derived cell lines M059-K (wild-type) and M059-J (DNA-PKcs deficient) were obtained from Dr. M.J. Allalunis-Turner (Brookhaven National Laboratory, Upton, NY; refs. 19, 20) and grown in α-MEM supplemented with 20% fetal bovine serum. Colon carcinoma cell line HCT116 (gifted by Dr. Minoru Koi, National Institute of Environmental Health Sciences, Research Triangle Park, NC) was maintained in DMEM with high glucose and 10% fetal bovine serum. Cervical cancer cell line HeLa CCL2 was purchased from the American Type Culture Collection and cultured in MEM with nonessential amino acids, sodium pyruvate, and 10% fetal bovine serum. EBV-transformed lymphoblastoid cell lines, wild-type control line GM02188, and Seckel syndrome lines GM09703 and GM18367A were from Coriell Cell Repositories and were grown in RPMI 1640 supplemented with 20% heat-inactivated fetal bovine serum. All cells were free of Mycoplasma, as determined by an ELISA kit (Life Technologies MycoTest kit).

CNDAC was synthesized as described (21). A stock solution (15–25 mmol/L) was prepared in PBS (pH 6.5), sterilized by filtration, stored at −20°C, and diluted in sterile PBS just before use. Camptothecin and PIKK inhibitor wortmannin were obtained from Sigma-Aldrich, Inc., and DNA-PK inhibitors NU7026 and IC86621 were obtained from Calbiochem/EMD Biosciences and stored as 10 mmol/L in DMSO aliquots at −70°C. Caffeine was also from Sigma-Aldrich and prepared freshly to 100 mmol/L in warm water and filtered before use. All other chemicals were reagent grade.

Sources of antibodies are as follows: rabbit antibodies against phosphorylated Ser³¹⁷ (polyclonal) and Ser³⁴⁵ (monoclonal) of Chk1 and lamin A/C and rabbit polyclonal antibodies against phosphorylated Thr⁶⁸ of Chk2 and phosphorylated Ser⁶⁴⁵ of Rad17 (Cell Signaling Technology); mouse monoclonal antibodies to Chk1 and RhoA and rabbit polyclonal antibodies to Chk1, H2A, and Rad17 (Santa Cruz Biotechnology); rabbit polyclonal antibody to ATM (Novus Biologicals); mouse anti-DNA-PKcs (Lab Vision Corp.); mouse monoclonal antibody to phosphorylated Ser¹³⁹ of H2AX (γ-H2AX) and rabbit polyclonal antibodies to ATRIP (Upstate Biotechnology); mouse anti-RPA70 (Oncogene Research Products/EMD Biosciences); rabbit anti-RPA32 and mouse anti-ATR (GeneTex, Inc.); mouse anti-Orc2 (BD PharMingen International); mouse monoclonal antibody to β-actin and phosphorylated Ser¹⁰ of histone H3 (Sigma-Aldrich); rabbit anti-XRCC4 (AbD Serotec); anti-mouse or anti-rabbit IgG horseradish peroxidase–conjugated antibody (Amersham Biosciences); and FITC-conjugated goat anti-mouse IgG (Jackson ImmunoResearch Laboratories, Inc.).

The CNDAC and camptothecin concentrations were chosen based on dose-finding studies for the maximal G₂ arrest after drug exposure for one normal population doubling time. Cells were washed with ice-cold PBS (pH 7.4) and fixed in 70% ethanol. Fixed cells were washed with PBS before incubation with 25 μg/mL of propidium iodide (Sigma-Aldrich) and 2.5 μg/mL of DNase-free RNase A (Roche). Fluorescence was measured on a Becton Dickinson FACSCalibur flow cytometer. At least 20,000 cells were measured for each sample.

Cells were immunostained for phosphorylated Ser¹⁰ of histone H3, nuclear stained with propidium iodide, and subjected to detection by flow cytometry, as described previously (17).

After centrifugation of cells to slides by Cytospin (Thermo Electron Corp.), cells were fixed with 4% paraformaldehyde in PBS (pH 7.4), stained with 4′,6-diamidino-2-phenylindole (2 μg/mL in PBS), and mounted in Vectashield mounting medium (Vector Laboratories, Inc.). Mitotic morphology was identified by the appearance of nuclear DNA condensation under an epifluorescence microscope (Nikon). At least 400 randomly selected cells on each slide were counted.

Cell lysates were subjected to isolation by SDS-PAGE and immunoblotting, as described previously (17). Subcellular fractionation was done according to the method of Zou et al. (22) and Wang et al. (23). Fractions (S1, cytoplasmic; S2, nuclear soluble; P2, chromatin bound) were loaded on 4% to 12% Bis-Tris gradient gels (Bio-Rad Laboratories) or 10% SDS-polyacrylamide gels. Immunoblots were quantitated by Li-Cor Odyssey software in indicated experiments.

Small interfering RNA (siRNA) transfections were done with Opti-MEM I (Invitrogen Corp.) containing Lipofectamine 2000 (Invitrogen) or DharmaFECT I (Dharmacon Research, Inc.) and 100 nmol/L final concentrations of oligos, following experimental optimization of the manufacturers' instructions. siRNA duplexes of 21-nucleotide RNAs with a 2-nucleotide 3′-overhang were purchased from Dharmacon Research. The siRNA target sequences of ATR (24) are 5′-AACCTCCGTGATGTTGCTTGA-3′ (ATR-2 duplex). Lamin A/C (25), whose target sequences are 5′-AACTGGACTTCCAGAAGAACA-3′, and siCONTROL RISC-free siRNA were used as positive and negative control siRNAs, respectively. The siRNAs directed against DNA-PKcs were a SmartPool of four duplexes (PRKDC). Transfection mixtures were removed after 6 h and cells were incubated with complete medium for additional 18 or 42 h before introduction of CNDAC. Cells were harvested 24 h after drug treatment and subjected to flow cytometric analysis.

Drug-treated adherent cells were immunostained for γ-H2AX as described previously (17). γ-H2AX and nuclear staining were viewed with a FV500 laser scanning confocal microscope (Olympus) using a 63× objective. The projections of multiple slices were saved as TIF files.

General linear model was used to compare the mean difference of mitotic indices between or among three cell lines and different interventions. All statistical tests are two sided, and P < 0.05 was considered statistically significant. Statistical analysis was done using Statistical Analysis System Release 9.1 (SAS Institute). Comparison of mitotic indices of different treatments in individual cell lines was conducted by Student's t test.

The first approach to define the role of PIKKs in G₂ checkpoint activation by CNDAC was to test if cells that lack ATM function would arrest in G₂ in response to CNDAC, as do wild-type cells. Treatment of human fibroblast lines that were either lacking ATM function (AT-C), transfected with an empty vector (AT-V), or repleted for ATM (AT-AT) with 4 to 8 μmol/L CNDAC for 24 h increased the G₂ population equally in all lines (Fig. 1A). Second, the stringency of the G₂ checkpoint activated by CNDAC was compared between ATM-deficient and ATM-repleted cells by scoring cells that had escaped from G₂ and were trapped by nocodazole in M phase. Mitotic indices of both lines in response to a series of concentrations of CNDAC were comparable, suggesting similar stringency of the G₂ checkpoint (Fig. 1B). Third, the short-term effect of CNDAC on the G₂-M checkpoint was compared with that of IR. Consistent with previous findings (26), cells deficient in ATM function failed to arrest progression into mitosis within 1 to 4 h after irradiation to the same extent as wild-type cells (Fig. 1C). When AT-AT fibroblasts were exposed to 10 Gy γ-irradiation, the mitotic ratio (percentage of total cells in mitosis, determined by positive staining for histone H3 phosphorylation, relative to that of untreated cells) dropped from 100% to 20% at 1 h and decreased further to 6% at 2 h and maintained this low level by 4 h (Supplementary Fig. S1).³

In contrast, when incubated with CNDAC for up to 4 h, the mitotic ratios of both lines reduced at similar rates to ∼60% by 4 h (Supplementary Fig. S1);³ therefore, the relative mitotic ratio of AT-AT to AT-C remained constant (0.98–1.08; Fig. 1C). The difference in short-term response of cells to these two agents reflects their distinct mechanisms of action in that IR has an immediate effect on cells in all phases of the cell cycle, whereas incorporation of CNDAC triphosphate during S phase is required for subsequent effects of the drug. Finally, changes in biochemical markers of checkpoint activation and DNA damage were also compared among these isogenic paired lines. Chk1 kinase became phosphorylated on Ser³⁴⁵ in all lines but to a greater level in ATM-complemented cells (Fig. 1D). The CNDAC-induced increases in the ratios of phosphorylated Chk2/total Chk2 were similar in the three lines, suggesting that this response is independent of ATM. Likely, this response was secondary to CNDAC-induced DNA damage. Another ATM substrate, p53, was phosphorylated on Ser¹⁵ in response to CNDAC to similar extents in all three lines (Fig. 1D). Phosphorylation of p53 on Ser²⁰ and Ser³⁷ was also similarly positive in all lines (data not shown). In addition to immunoblotting, immunostaining revealed augmentation of H2AX phosphorylation on Ser¹³⁹ in all lines, as manifested by nuclear foci, which was more remarkable in ATM-repleted cells (Supplementary Fig. S2A).³ Together, the data suggest that ATM is not required to activate the G₂ checkpoint pathway in response to CNDAC.

To determine the role of ATR in activation of the CNDAC-induced G₂ checkpoint, factors that are required for ATR recruitment and activation were investigated. One such protein, RPA, a ubiquitous ssDNA-binding protein, consisting of three subunits of 70, 32, and 14 kDa, associates with DNA at sites of damage (27–29). ATRIP, the partner of ATR, is attracted to RPA-ssDNA complexes, thereby initiating ATR-dependent DNA damage sensing processes (30, 31). These factors were evaluated in HeLa cells, which were arrested in G₂ after a 24-h incubation with 4 μmol/L CNDAC. Although no changes in cellular RPA protein level were detectable in whole-cell lysates after CNDAC treatment, densitometric analysis indicated that chromatin-associated RPA70 and RPA32 increased about 2- and 5-fold, respectively, in response to CNDAC (Fig. 2A). In similar experiments with HCT116 cells, analysis of RPA70 and RPA32 during 3 days of exposure to 3 μmol/L CNDAC showed that, as with HeLa cells, both proteins increased their association with chromatin during checkpoint activation. Chromatin-bound Rad17, an ATR substrate, also accumulated during CNDAC treatment, as did its phosphorylated form (Fig. 2B), suggesting activation of ATR kinase in response to CNDAC.

To further evaluate the role of ATR in G₂ checkpoint activation by CNDAC, the ATR protein level in HeLa cells was knocked down to <30% of control levels by siRNA at 72 h after transfection. Phosphorylation of Chk1 kinase on Ser³⁴⁵ and Cdk1 kinase on Tyr¹⁵ was markedly enhanced on CNDAC treatment (Fig. 2C), as indicated by increases in the phosphorylated protein levels normalized to total proteins. This is consistent with their functional roles in signaling CNDAC-induced G₂ checkpoint activation (17). This response was diminished for both proteins following siRNA knockdown of ATR. Flow cytometric analysis of portions of these same cultures showed a reduction of the CNDAC-induced G₂ arrest in siRNA-treated cells.

Finally, G₂ population accumulations in response to 20 to 60 μmol/L CNDAC treatment were observed in wild-type human lymphoblasts and two lines from Seckel syndrome patients with varying expression of ATR (Supplementary Fig. S3).³ Comparison of the stringency in the G₂ checkpoint showed that ∼17% of the control B lymphocytes derived from a healthy donor (GM02188) were in mitosis when incubated with colcemid; this percentage dropped to 2% or 1% when cells were preincubated with 40 or 60 μmol/L of CNDAC (11% or 6%, respectively, relative to colcemid alone; P < 0.01; Fig. 2D). Similarly, when the Seckel cell line GM09703 (32), which retains ∼50% of ATR protein level compared with GM02188, was treated with 40 or 60 μmol/L of CNDAC followed by colcemid, cells exhibited mitotic scorings of 9% or 14%, respectively, relative to that of colcemid alone (P < 0.01). In contrast, another cell line, GM018367, which is derived from a Seckel syndrome patient with a splicing mutation (2101A→G) in the ATR gene (33) and has little residual expression of ATR, presented a mitotic index of 26% in response to colcemid alone. There were 9% or 11% cells in mitosis on combination treatment with CNDAC and colcemid (35% or 41%, respectively, relative to colcemid alone; P < 0.01; Fig. 2D), indicating the G₂ checkpoint of Seckel cells with extremely low levels of ATR is impaired compared with ATR intact cells. Comparison of differences of mitotic indices of colcemid alone and CNDAC treatment between lines indicated that the wild-type and most severely affected Seckel cell line (GM18367A) were highly significant (P < 0.0001) in their ability to arrest in G₂, whereas the less severely affected Seckel line (GM09703) was marginally different (P = 0.042) from the wild-type. Thus, optimal activation and maintenance of the G₂ checkpoint by CNDAC is compromised in the absence of ATR.

To define the role of DNA-PK in activating the G₂ checkpoint by CNDAC, we used human glioma cell lines proficient (M059-K) and deficient (M059-J) in DNA-PKcs. Both M059-K and M059-J cells arrested in G₂ on CNDAC treatment (data not shown). Twenty-four–hour incubation of 16 μmol/L CNDAC, which caused maximal G₂ arrest, was associated with Chk1 phosphorylation on Ser³¹⁷ in both lines (Fig. 3A), as was H2AX phosphorylation. Each of these activities was relatively stronger in the wild-type cells. Confocal microscopy revealed more intense nuclear foci of γ-H2AX on CNDAC treatment in wild-type M059-K cells (Supplementary Fig. S2B).³ Significantly more M059-J cells progressed through G₂ following CNDAC treatment and were trapped in mitosis by nocodazole compared with the wild-type (Fig. 3B). M059-K cells incubated with CNDAC had a mitotic index of 4.2%, which was only one fifth of that in cells treated with nocodazole alone (21%), whereas the mitotic index of DNA-PKcs–deficient M059-J cells (16.1%) was about half of that in cells treated with nocodazole alone (30.3%). M059-K cells pretreated with IC86621, a specific inhibitor of DNA-PKcs, before CNDAC presented mitotic indexes of 7.5% (P < 0.05) and 8.4% (P < 0.01) at 20 and 40 μmol/L, respectively, significantly greater than that of cells not exposed to the inhibitor (4.2%). These results suggest that inhibition of DNA-PKcs could mimic the response of the DNA-PK mutant to CNDAC. Thus, the G₂ checkpoint activated by CNDAC becomes less stringent when DNA-PK is not functional. Subsequently, IC86621 and a second inhibitor of DNA-PKcs, NU7026, were used to further characterize the functional role of DNA-PK in the CNDAC-induced G₂ checkpoint. Although these compounds alone had little effect on cell cycle distribution in ML-1 and HeLa cells, they efficiently prevented CNDAC-induced G₂ arrest (Fig. 3C).

Finally, DNA-PKcs was knocked down by siRNA in HeLa cells to a significantly lower level (10–20% of control cells) by 72 h (Fig. 3D). Cells treated with control siRNA exhibited increased phosphorylation on Chk1 kinase on Ser³⁴⁵ and Cdk1 kinase on Tyr¹⁵, reflecting the G₂ checkpoint response to CNDAC (17). In contrast, this G₂ checkpoint response was markedly reduced in DNA-PKcs–depleted cells. These molecular changes were associated with the loss of ability of DNA-PKcs knockdown cells to arrest in response to CNDAC treatment. Taken together, these results support the conclusion that DNA-PK is required for effective G₂ checkpoint activation by CNDAC.

Cellular responses to two PIKK inhibitors, caffeine and wortmannin, were used to investigate the actions of these kinases on initiation of the G₂ checkpoint by CNDAC (Fig. 4A). Camptothecin, which is a topoisomerase I poison and has known responses to these PIKK inhibitors (34), was incorporated in the experimental design as a positive control (Fig. 4B). Three tumor cell lines of different origins (ML-1, HCT116, and HeLa) and an ATM-deficient fibroblast line (AT-C) were evaluated. Caffeine alone had little effect on cell cycle distribution in the above four cell lines but abolished CNDAC-induced G₂ arrest in these lines. Relative to its activity against DNA-PKcs, caffeine is a 50- and 10-fold more potent inhibitor of ATM and ATR, respectively (35). Considering that ATM is likely not involved, these results support a role for ATR in G₂ checkpoint activation. Wortmannin (10 μmol/L) completely abolished CNDAC-induced G₂ arrest in ML-1 and ATM-mutant cells. At this relatively low concentration, DNA-PK and ATM are inhibited, but not ATR (36). As ATM protein was not detectable by immunoblot in ML-1 cells, even under the condition of IR (data not shown), these results indicate that DNA-PK is essential for activation of the G₂ checkpoint by CNDAC in these cells. CNDAC-induced G₂ arrest was reduced in HCT116 and HeLa cells by 10 μmol/L wortmannin from 54% and 42% to 33% and 27%, respectively, and further abolished to a level less than untreated cells (17% and 16%, respectively) by 100 μmol/L wortmannin, indicating participation of ATR in the G₂ checkpoint pathway in these lines. Overall, caffeine and wortmannin inhibited camptothecin-induced G₂ arrest in very similar patterns as for CNDAC.

Our results support a model in which ATR and DNA-PK likely cooperate to activate the G₂ checkpoint in response to CNDAC (Fig. 5). Incorporation of CNDAC into DNA leads through β-elimination to formation of CNddC, a de facto DNA chain terminator. As preliminary studies suggest that CNddC is a poor substrate for excision (37), it seems unlikely that the 3′-strand terminated by CNddC is a substrate for exonucleolytic processing. However, the resulting single-strand break or nick could be processed by nuclease(s) to forms recessed on the 5′-terminus and expanded to form a gap. Subsequently, the exposed ssDNA is coated by RPA proteins (Fig. 2A). This RPA-ssDNA complex recruits ATRIP and its binding protein, ATR, to the site of DNA damage. Rad17, which is independently recruited onto the damaged chromatin (Fig. 2B), is then phosphorylated by ATR, which may promote the activation of ATR (31). Recruitment of ATR onto damaged chromatin facilitates its phosphorylation of other substrates, such as Chk1. Activated Chk1 then phosphorylates Cdc25C, which inhibits its phosphatase activity (17). As a consequence, the equilibrium between the unphosphorylated and phosphorylated forms of Cdk1 is shifted to the inactive phosphorylated form and cell cycle progression is arrested in G₂ (Fig. 3D). However, approaches to test the central role of ATR in such a response pathway, including siRNA knockdown of ATR (Fig. 2C) and ATR-deficient cells (Fig. 2D), failed to completely abolish the G₂ checkpoint response to CNDAC, suggesting the participation of other sensor molecules. As cells deficient in ATM had a normal checkpoint activation to CNDAC (Fig. 1), we examined the possibility that DNA-PK may participate in this response.

Although the nature of CNDAC-induced DNA damage has not been fully disclosed, it is clear that single-strand breaks induced by CNDAC can be transformed into DSBs (17). DNA ends, either DSBs or ssDNAs processed from nicks, are able to activate DNA-PK through Ku70 and Ku80 (38–41). Although the roles of these DNA-binding subunits remain to be investigated, it is expected that the sensing capabilities of the complex would be dependent on their function. Cells that lack DNA-PKcs had a less stringent G₂ arrest after CNDAC treatment than did wild-type (Fig. 3B). Similar results were obtained when DNA-PKcs was depleted with siRNA procedures (Fig. 3D) or after pharmacologic inhibition of the kinase (Figs. 3C and 4). The phosphorylation of Chk1 in response to CNDAC was reduced in cells deficient in DNA-PKcs (Fig. 3A) and after knockdown of the protein with siRNA (Fig. 3D). The phosphorylated form of Cdk1 was greatly diminished in DNA-PKcs–depleted cells, further suggesting the participation of DNA-PK in the Chk1-Cdk1–mediated checkpoint response to CNDAC. Although DNA-PK has a well-studied role in the error-prone DSB repair pathway nonhomologous end-joining, which is not cell cycle specific (42, 43), there are few reports of DNA-PK functioning in checkpoint activation (42).

Thus, both ATR and DNA-PK are clearly involved with activation of the G₂ checkpoint in response to CNDAC. Prior studies have shown that each is activated by different forms of damaged DNA. Therefore, the codependency for full checkpoint activation may reflect processing of the original nick to more complex forms. Nick modification may not include double-strand DNA break formation, as ATM does not seem to have a role in the G₂ checkpoint response. Alternatively, if DSBs do arise from CNddC-induced aberrant processing, DNA-PK and perhaps ATR may have a compensatory action in cells that lack ATM. DNA-PKcs possibly participates in activating Chk1, although phosphorylation of Chk1 by DNA-PKcs is not well documented. Our data presented above showed a substantial role of DNA-PKcs in this checkpoint activation. In addition, DNA-PK could contribute to CNDAC-induced G₂ arrest through p53, which has been reported to coimmunoprecipitate with and be phosphorylated by DNA-PKcs (44), and help to stabilize or maintain G₂ arrest after DNA damage (45, 46).

Currently, ATM and ATR are known to play distinct and cooperative roles in DNA damage responses (4). Although they share substrates, they show selectivity on substrates in response to different genotoxic stresses. Recent publications from different laboratories indicate that, during S and G₂ phases, ATR activation in response to DSBs is downstream of ATM (9–11, 47). Our data excluded ATM, but not ATR, from activating the G₂ checkpoint by CNDAC. It is likely that ATM is bypassed in the early events of damage sensing because the initial DNA damage caused by CNDAC is single-strand breaks. The exonucleases or endonucleases that process CNDAC-induced single-strand breaks or nicks and expand them into gaps recognized by RPA are unknown. Mre11 (48) and nucleases involved in excision repair pathways could be candidates for this action.

The current studies have provided more insights into the upstream signaling cascade of G₂ checkpoint activation by CNDAC. Our investigation has shown that ATR and DNA-PK are responsible for sensing DNA damage induced by CNDAC and subsequent activation of the G₂ checkpoint, whereas ATM is dispensable in this DNA damage response. Roles of these sensor proteins in cell survival after CNDAC treatment, which is also important and distinct from the goals of this article, are under investigation and will be addressed separately. As an orally bioavailable prodrug of CNDAC, sapacitabine is currently in clinical trials, these sensors and components of the pathways they regulate may be considered as potential targets for combination strategies. Inhibitors of both DNA-PK (49) and Chk1 (50) are currently undergoing clinical evaluation.

We thank Dr. Xin Wang (Department of Experimental Radiation Oncology), Drs. Walter Hittelman and Tao Lu (Department of Experimental Therapeutics), and Dr. Xiao Zhou (Department of Cytokine and Supportive Oncology, M. D. Anderson Cancer Center) for their expert technical assistance and valuable discussion.