Calcium-dependent Modulation of Poly(ADP-ribose) Polymerase-1 Alters Cellular Metabolism and DNA Repair*

After genotoxic stress poly(ADP-ribose) polymerase-1 (PARP-1) can be hyperactivated, causing (ADP-ribosyl)ation of nuclear proteins (including itself), resulting in NAD+ and ATP depletion and cell death. Mechanisms of PARP-1-mediated cell death and downstream proteolysis remain enigmatic. β-lapachone (β-lap) is the first chemotherapeutic agent to elicit a Ca2+-mediated cell death by PARP-1 hyperactivation at clinically relevent doses in cancer cells expressing elevated NAD(P)H:quinone oxidoreductase 1 (NQO1) levels. β-lap induces the generation of NQO1-dependent reactive oxygen species (ROS), DNA breaks, and triggers Ca2+-dependent γ-H2AX formation and PARP-1 hyperactivation. Subsequent NAD+ and ATP losses suppress DNA repair and cause cell death. Reduction of PARP-1 activity or Ca2+ chelation protects cells. Interestingly, Ca2+ chelation abrogates hydrogen peroxide (H2O2), but not N-Methyl-N′-nitro-N-nitrosoguanidine (MNNG)-induced PARP-1 hyperactivation and cell death. Thus, Ca2+ appears to be an important co-factor in PARP-1 hyperactivation after ROS-induced DNA damage, which alters cellular metabolism and DNA repair.

Alterations in the initiation and regulation of caspase-mediated apoptosis are associated with an array of pathological disease states, including chemotherapy resistance in cancer (1). Therefore, elucidating mechanisms that initiate non-caspasemediated cell death are crucial for the development and use of novel anticancer agents.
A growing number of chemotherapeutic approaches focus on targeting specific DNA repair enzymes. In particular, inhib-itors of poly(ADP-ribose) polymerase-1 (PARP-1) 2 that sensitize cells to DNA-damaging agents are under extensive investigation (2). PARP-1 functions as a DNA damage sensor that responds to both single-and/or double-strand DNA breaks (SSBs, DSBs), facilitating DNA repair and cell survival. After binding to DNA breaks, PARP-1 converts ␤-NAD ϩ (NAD ϩ ) into polymers of branched or linear poly(ADP-ribose) units (PAR) and attaches them to various nuclear acceptor proteins, including XRCC1, histones, and PARP-1 for its autoregulation (3). However, in response to extensive DNA damage, PARP-1 can be hyperactivated, eliciting rapid cellular NAD ϩ and ATP pool depletion. PARP-1-mediated NAD ϩ and ATP losses have affects on mitochondrial function by decreasing the levels of pyruvate and NADH. Loss of mitochondrial membrane potential (MMP) ensues, causing caspase-independent cell death by as yet unknown mechanisms (3). PARP-1 hyperactivation was documented in the cellular response to trauma, such as ischemia-reperfusion, myocardial infarction, and reactive oxygen species (ROS)-induced injury (3). In each case, inhibition of PARP-1 was necessary for the long-term survival of damaged cells (4).
Cell Culture-Human MCF-7 and MDA-MB-468-NQϩ breast cancer cells were maintained and used as described (5). Human MDA-MB-231 (231) breast cancer cells that contain a 609CϾT polymorphism in NQO1 (9) and are deficient in enzyme activity, were obtained from the American Type Culture Collection (Manassas, VA). Cells were stably transfected with a CMVdriven NQO1 cDNA or the pcDNA3 vector alone as described (5). All cells were grown in high glucose-containing RPMI 1640 tissue culture medium containing 5% fetal bovine serum, 2 mM L-glutamine, 100 units/ml penicillin, and 100 mg/ml streptomycin at 37°C in a 5% CO 2 , 95% air humidified atmosphere (6). 231-NQO1ϩ (NQϩ) and ϪNQO1 (NQϪ) cells were maintained in medium containing 400 g/ml geneticin (8), but all experiments were performed without selection. All tissue culture components were purchased from Invitrogen (Carlsbad, CA) unless otherwise stated. All cells were routinely tested and found free of mycoplasma contamination.
Relative Survival Assays-Relative survival assays were performed as previously described (5). MCF-7 cells were pretreated or not with BAPTA-AM (5 M, 30 min) then treated with or without 2-h pulses of ␤-lap at the doses indicated, in the presence or absence of 40 M dicoumarol. In some experiments, cells were exposed to 5 M ␤-lap followed by delayed (t ϭ 0 -2 h) addition of 5 M BAPTA-AM. After drug addition, media were removed and drug-free media added. Cells were then allowed to grow for an additional 6 days and relative survival, based on DNA content (Hoechst 33258 staining), was determined (5). Prior studies using ␤-lap showed that relative survival assays correlated directly with colony forming ability assays (5). Data were expressed as treated/control (T/C) from separate triplicate experiments (means, Ϯ S.E.), and comparisons analyzed using a two-tailed Student's t test for paired samples.
Confocal microscopy was performed as previously described (7). Cells were fixed in methanol/acetone (1:1) and incubated with ␣-PAR (10H; Alexis, San Diego, CA) or ␣-␥-H2AX (Trevigen, Gaithersburg, MD) for 2 h at room temperature. Nuclei were visualized by Hoechst 33258 staining at a 1:3000 dilution. Confocal images were collected at 488 nm excitation from a krypton/argon laser using a Zeiss LSM 510 confocal microscope (Thornwood, NY). Images were representative of experiments performed at least four times. The number of PARpositive cells and ␥-H2AX foci/cell were quantified by counting 60 or more cells from four independent experiments (means Ϯ S.E.).
Formation of ROS was monitored by the conversion of nonfluorescent 6-carboxy-2Ј7Ј-dichlorodihydrofluorescin diacetate, di(acetoxymethyl ester) to fluorescent 6-carboxy-2Ј,7Јdichlorofluorescein diacetate di(acetoxymethyl ester) (DCF) as previously described (11,12). Briefly, MCF-7 cells were seeded at 2-3 ϫ 10 5 cells in 35-mm glass bottom Petri dishes (MatTek Corp., Ashland, MA) and allowed to attach overnight. Cells were loaded with 5 M DCF in media for 30 min at 37°C. After loading, cells were washed twice with phosphate-buffered saline, and incubated for an additional 20 min at 37°C to allow for dye de-esterification. Confocal images of DCF fluorescence were collected using 488-nm excitation from an argon/krypton laser, 560-nm dichroic mirror, and a 500 -550 nm band pass filter. Three basal images were collected before drug addition Single Cell Gel Electrophoresis (Alkaline Comet) Assays-DNA damage was assessed after different drug treatments by evaluating DNA "comet" tail area and migration distance (13). MCF-7 cells were pretreated with BAPTA-AM (5 M, 30 min) or Me 2 SO (1:1000 dilution), and then exposed to H 2 O 2 (500 M, 1 h), ␤-lap (4 M, various times), or vehicle alone, and harvested at various times. Cell suspensions (3 ϫ 10 5 /ml cold PBS) were mixed with 1% low melting temperature agarose (1:10 (v/v)) at 37°C and immediately transferred onto a CometSlide TM (Trevigen). After solidifying (30 min at 4°C), slides were submerged in prechilled lysis buffer (2.5 M NaCl, 100 mM EDTA pH 10, 10 mM Tris Base, 1% sodium lauryl sarcosinate, and 1% Triton X-100) at 4°C for 45 min, incubated in alkaline unwinding solution (300 mM NaOH, and 1 mM EDTA) for 45 min at room temperature and washed twice (5 min) in neutral 1ϫ TBE (89.2 mM Tris Base, 89 mM boric acid, and 2.5 mM EDTA disodium salt). Damaged and undamaged nuclear DNA was then separated by electrophoresis in 1ϫ TBE for 10 min at 1 V/cm, fixed in 70% ethanol, and stained using SYBR-green (Trevigen). Comets were visualized using an Olympus fluorescence microscope (Melville, NY), and images captured using a digital camera. Images were analyzed using ImageJ software (14,15) and comet tail length was calculated as the distance between the end of nuclei heads and the end of each tail. Tail moments were defined as the product of the %DNA in each tail, and the distance between the mean of the head and tail distributions in Equation 1, where TA is the tail area, TAI is the tail area intensity, HA is the head area, and HAI is the head area intensity. Importantly, tail moment and tail area calculations yielded similar experimental results. Each datum point represented the average of 100 cells Ϯ S.E., and data are representative of experiments performed three times.
Determination of NAD ϩ and ATP Levels-Intracellular NAD ϩ levels were measured as described (16) with modification. Briefly, cells were seeded at 1 ϫ 10 6 and allowed to attach overnight. Cells were pretreated for 2 h with 3-AB (25 mM), DPQ (20 M), BAPTA-AM (5 M), or Me 2 SO and then exposed to ␤-lap (2-20 M) for the indicated times. Cell extracts were prepared in 0.5 M perchloric acid, neutralized (1 M KOH, 0.33 M KH 2 PO 4 /K 2 HPO 4 (pH 7.5)), and centrifuged to remove KClO 4 precipitates. Supernatants or NAD ϩ standards were incubated 4:1 (v/v) for 20 min at 37°C with NAD ϩ reaction mixture as described (17). Measurements from extracts were taken at an absorbance of 570 nm and intracellular NAD ϩ levels were normalized to 1 ϫ 10 6 cells. Data were expressed as %NAD ϩ Ϯ S.E., for T/C samples from nine individual experiments.
ATP levels were analyzed using a luciferase-based bioluminescence assay as described (18). Data were graphed as means Ϯ S.E. of experiments performed at least three times. Results were compared using the two-tailed Student's t test for paired samples.
NQO1 Enzyme Activity Assays-NQO1 enzymatic assays were performed as described (19) using cytochrome c (practical grade, Sigma) in Tris-HCl buffer (50 mM, pH 7.5). NADH (200 M) was the immediate electron donor, and menadione (10 M) the electron acceptor. Changes in absorbance were monitored using a Beckman DU 640 spectrophotometer (Beckman Coulter, Fullerton, CA). Dicoumarol (10 M) inhibitable NQO1 levels were calculated as nmol of cytochrome c reduced/ min/g protein based on initial rate of change in absorbance at 550 nm and an extinction coefficient for cytochrome c of 21.1 nmol/liter/cm (20). Results were expressed as means Ϯ S.E. of three or more separate experiments.
Flow Cytometry and Apoptotic Measurements-Flow cytometric analyses of TUNEL-positive cells were performed as described using APO-DIRECT TM (BD Pharmingen) (5). Samples were analyzed in an EPICS Elite ESP flow cytometer using an air-cooled argon laser at 488 nm, 15 milliwatt (Beckman Coulter Electronics, Miami, FL) and Elite acquisition software. Experiments were performed a minimum of five times, and data expressed as means Ϯ S.E. Statistical analyses were performed using a two-tailed Student's t test for paired samples.
Glutathione Measurements-Disulfide glutathione and total glutathione (GSH ϩ GSSG) levels were determined using a spectrophotometric recycling assay (21,22). Following indicated treatments, pellets were thawed and whole cell homogenates prepared as described (21,22). Data were expressed as the %GSSG normalized to protein content, as measured using the method of Lowry et al. (23). Shown are means Ϯ S.E. for experiments performed at least three times.

Time and Ca 2ϩ Dependence of ␤-lap-induced Cell Death-
To elucidate the signaling events required for ␤-lap-induced cell death, log-phase human MCF-7 breast cancer cells, with high levels of endogenous NQO1 activity, were tested for their sensitivities to various concentrations of ␤-lap at various times. The purpose of these experiments was to determine the minimal time of ␤-lap exposure required to kill the entire cell population. Cells exposed to doses of Յ3 M ␤-lap required Ն4 h to elicit cell death, whereas 2 h exposures of ␤-lap at Ն4 M killed all MCF-7 cells (Fig. 1A).
Prior data from our laboratory demonstrated that Ca 2ϩ was released within 2-5 min from endoplasmic reticulum (ER) Ca 2ϩ stores after ␤-lap treatment (24), suggesting that this was a required initiating factor in ␤-lap-induced cell death (24). To test this, MCF-7 cells were pretreated with the intracellular Ca 2ϩ chelator BAPTA-AM (5 M, 30 min), then exposed to 4 M ␤-lap for various times (Fig. 1B). Under these conditions, we previously demonstrated that BAPTA-AM pretreatment was sufficient to block the rise in cytosolic Ca 2ϩ caused by ␤-lap treatment (24). BAPTA-AM abrogated ␤-lap-induced cytotoxicity and nuclear condensation ( Fig. 1B and supplemental Fig.  S1A). To determine the kinetics of Ca 2ϩ -dependent cell death, MCF-7 cells were treated with 5 M ␤-lap, and 5 M BAPTA-AM was added at various times thereafter, up to 2 h. A time-dependent decrease in survival was observed with delayed addition of BAPTA-AM (Fig. 1C), indicating that Ca 2ϩ release was a necessary event, occurring before cells were committed to death. Abrogation of ␤-lap cytotoxicity by BAPTA-AM was equivalent to that noted with NQO1 inhibition by dicoumarol (Fig. 1B). Addition of BAPTA-AM also prevented ␤-lap-induced, atypical proteolysis (e.g. ϳ60 kDa PARP-1 and p53 cleavage fragments), in a manner as effective as dicoumarol (supplemental Fig. S1B, lanes 3 and 4). Interestingly, Z-VADfmk (50 M), a pan-caspase inhibitor, did not block atypical PARP-1 and p53 proteolysis in ␤-lap-treated MCF-7 cells (lane 7). As expected, Z-VAD-fmk inhibited STS (1 M)-induced caspase-mediated proteolysis (25). However, BAPTA-AM did not affect STS-induced apoptotic proteolysis (supplemental Fig. S1C). These data, in conjunction with our prior data showing ␤-lap-induced ER Ca 2ϩ release (24), support a role for Ca 2ϩ in the initiation of cell death induced by this drug.
PARP-1 Hyperactivation after ␤-lap Treatment Is NQO1-dependent and BAPTA-AM-sensitive-Because ␤-lap-induced cell death was accompanied by a Ն80% loss of ATP within 1 h (24), we suspected PARP-1 hyperactivation played a role in the mode of action for this drug. To investigate this, we generated 231 NQO1-proficient (231-NQϩ) cells that are sensitive to ␤-lap (LD 50 : ϳ1.5 M), after a 2-h pulse, and compared these cells to vector alone, 231 NQO1-deficient (231-NQ-) cells, that are resistant to the drug (LD 50 : 17.5 M). Only ␤-lap-treated, 231-NQϩ cells exhibited an increase in PAR-modified proteins, mostly PARP-1, consistent with the role of PARP-1 as the predominant poly(ADP-ribosyl)ated species. This response peaked ϳ30 min during ␤-lap exposure ( Fig. 2A). In contrast, treatment of 231-NQ-cells with equal or significantly higher doses of ␤-lap did not induce PAR accumulation (data not shown). In contrast, both 231-NQϩ and 231-NQ-cells hyperactivated PARP-1 in response to H 2 O 2 . The NQO1-dependence of PARP-1 hyperactivation after ␤-lap exposure was confirmed in a number of other cell lines (e.g. breast, prostate, and lung cancers) that have elevated NQO1 activity demonstrating that the responses to ␤-lap were not cell type-specific. In all cases, dicoumarol suppressed ␤-lap-induced PAR formation (supplemental Fig. S2, A-C).
We then examined a possible connection between the involvement of Ca 2ϩ in lethality and PARP-1 hyperactivation. MCF-7 cells were pretreated with 5 M BAPTA-AM, then exposed to 5 M ␤-lap for the indicated times (Fig. 2B). The kinetics of PAR accumulation were faster in MCF-7 cells than in 231-NQϩ cells, (10 min versus 20 min Figs. 2, B and A, FIGURE 1. ␤-lap-induced cell death is time-and Ca 2؉ -dependent. A-C, cell death was examined using relative survival assays in NQO1ϩ MCF-7 cells. A, cells were exposed to various doses of ␤-lap for different lengths of time to determine the minimal exposure time required for cell death. After drug exposure, media were removed and drug-free media added. Cells were then allowed to grow for an additional 6 days and relative survival, based on DNA content was determined by Hoechst 33258 staining as described under "Experimental Procedures." Student's t test for paired samples, experimental group compared with 3 M ␤-lap (*, p Յ 0.001; **, p Յ 0.005). Prior studies using ␤-lap showed that relative survival assays directly correlated with colony forming ability.  Relative PAR levels were calculated by densitometric analyses by NIH ImageJ using PARP loading controls wherein controls were set to 1.0. B, immunoblots of PAR formation and steady state ␣-tubulin expression in extracts from MCF-7 cells treated with ␤-lap or H 2 O 2 (2 mM, control for PARP-1 hyperactivation). Other cells were pretreated with BAPTA-AM with or without ␤-lap (5 M). Samples were harvested at the indicated times. Relative PAR levels were calculated by densitometric analyses by NIH ImageJ using ␣-tubulin loading controls wherein controls were set to 1.0. C, assessment of PARP-1 hyperactivation, measured by PAR formation, in MCF-7 cells treated with ␤-lap alone or in cells pretreated with BAPTA-AM for 30 min prior to ␤-lap exposure. PAR formation was visualized using confocal microscopy. D, quantified percentages of PAR-positive cells from confocal microscopy analyses of at least 60 cells from four independent experiments (means Ϯ S.E.). respectively), consistent with higher NQO1 levels in MCF-7 cells. PAR levels decreased after 90 min, corresponding to auto-(ADP-ribosyl)ation of PARP-1, and efficient PAR hydrolysis by poly(ADP-ribose) glycohydrolase (PARG) (26). Interestingly, BAPTA-AM pretreatment abrogated PARP-1 hyperactivation induced by ␤-lap (Fig. 2B) as confirmed by confocal microscopy (Fig. 2C). Robust and extensive poly(ADP-ribosyl)ation occurred within 30 min (87% Ϯ 17 PAR-positive cells) of drug exposure and dissipated between 60 -90 min (Fig. 2, C and D).
However, in the presence of BAPTA-AM PAR accumulation in ␤-lap-treated MCF-7 cells was prevented (Fig. 2, C and D). To determine the global nature of this response, other cancer cell lines such as NQO1ϩ PC-3 human prostate cancer cells were examined and similar responses noted (supplemental Fig. S2D). Collectively, these data suggested a role for Ca 2ϩ in the modulation of PARP-1 hyperactivation after ␤-lap exposure.
␤-lap-induced PARP-1 Hyperactivation Alters Cellular Energy Dynamics Causing Cell Death-PARP-1 hyperactivation can elicit depletion of cellular NAD ϩ levels and cause cell death in situations of extreme DNA damage or ischemiareperfusion (27,28). Treatment of MCF-7 cells with doses of ␤-lap Ն5 M resulted in decreases (Ͼ80%) in NAD ϩ and ATP levels 1 h during treatment (Fig. 3, A-C). To determine if NAD ϩ and ATP losses were attributable to PARP-1 hyperactivation, MCF-7 cells were pretreated for 2 h with PARP inhibitors (i.e. 3-AB or DPQ), prior to 5 M ␤-lap exposure. Similar to pretreatment with BAPTA-AM, NAD ϩ and ATP losses in ␤-lap-treated MCF-7 cells were abrogated by 3-AB or DPQ (Fig. 3, B and C). 3-AB was more effective at preventing nucleotide loss than DPQ presumably because it has two distinct modes of PARP-1 inhibition, preventing NAD ϩ binding to the catalytic site and competing with the PARP-1 DNA binding domain (29), whereas DPQ is a competitive inhibitor of NAD ϩ (30). Similar effects of DPQ on NAD ϩ and ATP losses after DNA damage have been reported (31). Neither 3-AB nor DPQ (used at Ͼ2-fold higher doses than in the above experiments) altered NQO1 activity in enzyme assays in vitro. Finally, 3-AB did not affect ␤-lap-induced ROS formation (data not shown).
To confirm that the energetic consequences of PARP-1 hyperactivation (e.g. NAD ϩ and ATP losses) were necessary and sufficient for ␤-lap-induced cell death, the effects of 3-AB or DPQ, on apoptosis, was measured by TUNEL assay. Pretreatment with either inhibitor resulted in a reduction in apoptosis (2% and 15% total apoptosis, respectively for 3-AB and DPQ) compared with 55% in ␤-lap-treated cells (Fig.   FIGURE 3. PARP-1-dependent NAD ؉ and ATP pool depletion leads to cell death after ␤-lap exposure in MCF-7 cells. A and B, NAD ϩ pool depletion occurs immediately after ␤-lap treatment in MCF-7 cells. A, MCF-7 cells were exposed to varying concentrations of ␤-lap for 1 h, or B, 5 M ␤-lap alone with or without pre-and co-treatment of PARP inhibitors, (20 M DPQ or 25 mM 3-AB) or 5 M BAPTA-AM for various times and harvested for NAD ϩ content. Student's t test for paired samples, comparing experimental groups containing ␤-lap ϩ 3-AB or DPQ versus ␤-lap alone are indicated (*, p Յ 0.001; **, p Յ 0.05, respectively). C, intracellular ATP levels were monitored using a luciferase-based bioluminescence assay in MCF-7 cells treated with ␤-lap, or in cells pre-and co-treated with 20 M DPQ, or 25 mM 3-AB 2 h prior to ␤-lap addition. Differences were compared using two-tailed Student's t test. Groups having *, p Յ 0.05 values compared with ␤-lap alone are indicated. D, apoptotic DNA fragmentation was assessed using TUNEL assays in ␤-lap-exposed, log-phase MCF-7 cells with or without pre-and co-treatment with DPQ or 3-AB. NOVEMBER 3, 2006 • VOLUME 281 • NUMBER 44 JOURNAL OF BIOLOGICAL CHEMISTRY 33689 3D). Thus, PARP-1 inhibition by 3-AB or DPQ spared ␤-lapinduced apoptosis in NQO1ϩ MCF-7 cells, consistent with the effects of these inhibitors to prevent NAD ϩ and ATP losses. Cumulatively, these data strongly suggest that Ca 2ϩdependent PARP-1 hyperactivation caused NAD ϩ and ATP loss in NQO1ϩ human cancer cell lines after ␤-lap treatment.
Ca 2ϩ Chelation Modulates DNA Repair after ␤-lap Treatment-We postulated that metabolism of ␤-lap by NQO1 would generate superoxide, peroxide, and other ROS (36). We directly monitored intracellular ROS formation, using the conversion of non-fluorescent 6-carboxy-2Ј,7Ј-dichlorodihydrofluorescin to fluorescent 6-carboxy-2Ј,7Ј-dichlorodihydrofluorescein (DCF). Indeed, exposure of MCF-7 cells with 5 or 8 M ␤-lap treatment, caused an increase in fluorescence within 5 min compared with Me 2 SO-treated cells (Fig. 6A, left  panel). Region-of-interest analyses showed an ϳ2000-fold increase in fluorescence with ␤-lap alone over control cells, which could be abrogated by inhibiting NQO1 activity with dicoumarol (Fig. 6A, right panel). Because BAPTA-AM has moderate affinity for divalent cations other than Ca 2ϩ , we explored the possibility that BAPTA-AM may protect cells from DNA damage and subsequent cell death by interfering with Fenton chemistry. Cells pretreated with 5 M BAPTA-AM and then exposed to 5 or 8 M ␤-lap exhibited no significant difference in the rate or extent of ROS formation compared with ␤-lap alone-treated cells (Fig. 6A). These results were confirmed by examining the oxidative state of MDA-MB-468-NQϩ cells after treatment with 4 M ␤-lap in the presence or absence of 5 M BAPTA-AM. ␤-lap treatment caused an ϳ65% rise in disulfide glutathione (GSSG) levels, that persisted during drug exposure (Fig. 6B). Addition of BAPTA-AM did not alter the kinetics or levels of GSSG formation during ␤-lap exposure (Fig. 6B). These data suggest that the protective effects of BAPTA-AM on ␤-lap-treated NQO1ϩ cells were not caused by interference with ␤-lap-induced ROS formation. Similar results were found in 231-NQϩ cells (data not shown).
To assess the effects of Ca 2ϩ on DNA damage and repair, ␤-laptreated MCF-7 cells were analyzed by alkaline comet assays to monitor total DNA strand breaks with or without BAPTA-AM addition. ␤-lap-treated cells exhibited significant DNA strand breakage by 10 min, resembling the positive control (H 2 O 2 ), and after 30 min, ␤-lap-induced DNA damage exceeded those levels ( Fig. 6C and supplemental Fig. S3). Cells pretreated with BAPTA-AM exhibited less DNA damage compared with ␤-lap alone, and their repair of DNA damage correlated well with their ability (or lack thereof) to survive (Figs. 6C and 1B).
We then examined the kinetics of repair in MCF-7 cells following a 2-h ␤-lap exposure with or without BAPTA-AM pretreatment. After ␤-lap exposure, DNA damage persisted and gradually increased over time (Fig. 6C), indicative of inhibition of DNA repair and consistent with the drop in NAD ϩ and ATP levels (Fig. 3, B and C). Although cells treated with ␤-lap and BAPTA-AM at 2 h exhibited equivalent damage to 10 min of ␤-lap exposure alone (4.6 Ϯ 0.2 versus 4.7 Ϯ 0.4, p Ͼ 0.5 comet microns, respectively), BAPTA-AM pretreated cells were protected from PARP-1 hyperactivation (Fig. 2B), as well as decrements in NAD ϩ levels (Fig. 3B). BAPTA-AM-pretreated cells showed a time-dependent recovery from DNA damage (Fig. 6C  and supplemental Fig. S3). In contrast, ␤-lap-exposed cells showed extensive DNA damage with no signs of DNA repair. Collectively, these data suggest that NQO1-mediated metabolism of ␤-lap leads to the generation of ROS and subsequent DNA damage that hyperactivates PARP-1.
H 2 O 2 Causes Ca 2ϩ -dependent PARP-1 Hyperactivation-To examine the universality of Ca 2ϩ -modulated PARP-1 function in response to other DNA damaging agents, we examined responses to H 2 O 2 or MNNG. Unlike ␤-lap, H 2 O 2 treatment caused PARP-1 hyperactivation in both 231-NQϩ and 231-NQϪ cells (Fig. 7A). However, expression of NQO1 required higher doses of H 2 O 2 to cause PAR formation in 231 cells (Fig.  7A). These data suggest that NQO1 has a broader antioxidant role by protecting against ROS-induced damage as previously proposed (37)(38)(39). Consistent with ␤-lap, however, was the abrogation of H 2 O 2 -induced PAR formation by BAPTA-AM in 231 cells independent of NQO1 activity (Fig. 7B).
H 2 O 2 treatment also caused a dose-dependent increase in apoptosis in both 231-NQϪ and 231-NQϩ cells that was blocked by BAPTA-AM (Fig. 7C). However, 231-NQϩ cells were much less sensitive to H 2 O 2 than 231-NQ-cells. ATP loss was seen within minutes of H 2 O 2 exposure in 231-NQϪ, but not in NQO1-positive 231-NQϩ cells (supplemental Fig. S4). In addition, PARP-1 hyperactivation and cell death in response to equivalent doses of H 2 O 2 in 231-NQ-cells was much more robust than in 231-NQϩ cells (Fig. 7, B and C). Interestingly, as noted with ␤-lap exposure, treatment of MCF-7 cells with Ն200 M H 2 O 2 for 2 h resulted in formation of a 60-kDa PARP-1 and 40-kDa p53 fragments. This atypical proteolysis was effectively inhibited by BAPTA-AM pretreatment (supplemental Fig. S5).
Finally, BAPTA-AM had no effect on PARP-1 hyperactivity or cytotoxicity caused by treatment with the monofunctional DNA-alkylating agent, MNNG (Fig. 7E). Because MNNG does not cause Ca 2ϩ release like ␤-lap or H 2 O 2 , these data suggest that Ca 2ϩ modulation of PARP-1 hyperactivation is unique to ROS-producing agents.

DISCUSSION
The regulatory mechanisms controlling PARP-1 function to either promote cell survival or cell death in response to DNA damage remain enigmatic. PARP-1 facilitates DNA repair and cell survival in response to a variety of DNA-damaging agents. However, it also mediates programmed necrosis (17), as well as caspase-inde- FIGURE 5. ␤-lap-induced ␥-H2AX foci formation is abrogated by BAPTA-AM pretreatment. A, immunoblot analyses of ␥-H2AX, total H2AX, and ␣-tubulin protein levels in whole cell extracts from MCF-7 cells exposed to ␤-lap for various times, or IR (5 Gy) harvested after 15 min. B, visualization of ␥-H2AX foci in MCF-7 cells at various times after treatment with 5 M ␤-lap or 15 min post-IR (5 Gy) by confocal microscopy. C, BAPTA-AM pretreatment followed by ␤-lap exposure abrogates ␥-H2AX foci formation in MCF-7 cells as visualized by confocal microscopy. The number of ␥-H2AX foci per cell was determined from at least 60 cells for each treatment group from four independent confocal experiments (means Ϯ S.E.). D, immunoblot of ␥-H2AX, total H2AX, and ␣-tubulin protein levels in whole cell extracts from MCF-7 cells exposed to ␤-lap for various times with or without BAPTA-AM (5 M) pretreatment, or IR (5 Gy) harvested after 15 min. pendent apoptotic cell death following severe levels of DNA damage (40). The downstream pathways essential for the execution of cell death in response to PARP-1-mediated metabolic alterations remain poorly understood.
In elucidating the cell death pathway after exposure to ␤-lap, we uncovered a novel mechanism of PARP-1-mediated cell death. Our data suggest that this mechanism occurred selectively in response to ROS-generating agents. We demonstrated, for the first time, that Ca 2ϩmediated PARP-1 hyperactivation commits cells to death as a consequence of metabolic starvation without the involvement of caspases.
PARP-1 hyperactivation in response to ␤-lap treatment was not cell type-specific and has been observed in all cells that express elevated NQO1 levels ( Fig. 2A, and  supplemental Fig. S2, A-C). As a result, cells exposed to ␤-lap exhibited depletion of NAD ϩ and ATP, occurring 30 -60 min during drug exposure. NAD ϩ and ATP losses were, in part, PARP-1-mediated since PARP inhibitors (e.g. 3-AB and DPQ) partially abrogated nucleotide loss (Fig. 3, B and C). Chemical inhibition of PARP-1, or PARP-1 protein knock-down, not only prevented NAD ϩ and ATP losses, but also abrogated ␤-lap-induced apoptosis (Figs. 3D and 4, D-F). These data established PARP-1-mediated NAD ϩ and ATP losses as crucial upstream events in ␤-lap-mediated cell death.
PARP-1-mediated alterations in cellular metabolism caused by ␤-lap reported in NQO1-expressing cells in this study explain many of its purported effects in vitro and in vivo. These include, but are not limited to: (i) inhibition of NFB activation via inhibition of IKK-␣ (41), (ii) lack of caspase activation (6) and p53 stabilization (8), and (iii) inhibition of Topoisomerase (Topo) I and Topo II-␤ (42). Furthermore, ␤-lap can initiate cell death independently from Bax and/or Bak activation as changes in mitochondrial outer membrane permeabilization (MOMP) can occur via PARP-1mediated NAD ϩ loss. 3 Thus, the results reported here appear to explain all prior phenomena reported in cells exposed to ␤-lap.  A unique feature of PARP-1-mediated cell death stimulated by ␤-lap was that administration of BAPTA-AM abrogated PARP-1 hyperactivation (Fig. 2B, supplemental Fig. S2D), nucleotide pool loss (Fig. 3B), atypical proteolyses (assessed by p53 and 60 kDa PARP-1 cleavage) (supplemental Fig. S1B), and cell death (Fig. 1, B and C). When BAPTA-AM was added Ͼ20 min after ␤-lap treatment, cells were not protected from cell death (Fig. 1C), suggesting that events occurring within the first 20 min of drug exposure committed cells to death. BAPTA (free acid form) did not alter NQO1 activity in vitro (24). This appears to be confirmed by the inability of BAPTA-AM to prevent ROS generation, which arises from NQO1-mediated metabolism of ␤-lap. Instead, our data suggest that the ability of BAPTA-AM to prevent ␤-lap-induced lethality in NQO1ϩ cancer cells was caused by the specific prevention of PARP-1 hyperactivation. The observed differences in the amount of DNA damage and ␥-H2AX foci formation between ␤-lap alone and that of ␤-lap co-administered with BAPTA-AM suggest that preventing PARP-1 hyperactivation and subsequent changes in cellular metabolism can allow for cell recovery, noted by more rapid and extensive DNA damage repair (Figs. 5C and 6C). Recent data suggest that protein phosphatase 2A (PP2A) dephosphorylates ␥-H2AX and is required for DSB repair (43). It is possible that Ca 2ϩ chelation not only prevents PARP-1 hyperactivation, but also augments ␥-H2AX dephosphorylation through PP2A activity. ROS-induced ER Ca 2ϩ release may poison PP2A. However, we favor the theory that NQO1-mediated ␤-lap-induced ER Ca 2ϩ release has its predominant affect on PARP-1 hyperactivation, thereby inhibiting DNA repair and cell recovery.
The mechanism of cell death induced by ␤-lap could be recapitulated by treatment with high doses of ROS-generating agents, such as H 2 O 2 (Fig. 7, A-D). Notable similarities included: H 2 O 2 -mediated PARP-1 hyperactivation, Ca 2ϩdependent proteolytic cleavage of PARP-1 and p53, and apoptotic DNA fragmentation. Furthermore, H 2 O 2 -induced lethality was abrogated by BAPTA-AM (Fig. 7, A-D and supplemental Fig.  S5). Although ␤-lap and H 2 O 2 initiate a similar downstream death pathway, the compounds differed in their lethality in cells with respect to NQO1 expression. ␤-lap lethality was enhanced in cells that express NQO1, whereas H 2 O 2 caused greater cytotoxicity in NQO1-deficient cells (Fig.  7C). We noted striking similarities between ␤-lapor H 2 O 2induced cell death and the caspase-independent cell death induced by ischemia-reperfusion. ROS produced during ischemia-reperfusion induces DNA strand breaks beyond a normal threshold that lead to PARP-1 hyperactivation, metabolic catastrophe, and an increase in intracellular Ca 2ϩ levels leading to -calpain activation (44). These data suggest that programmed PARP-1-mediated cell death is a global response to these types of cellular insults. TUNEL assays were performed in H 2 O 2 -exposed, log-phase 231-NQ-and 231-NQϩ cells, with or without pretreatment with 5 M BAPTA-AM. D, MNNG-induced PARP-1 hyperactivation is not blocked by Ca 2ϩ chelation. Immunoblots of PAR and ␣-tubulin protein levels from whole cell extracts of MCF-7 cells treated with MNNG or in cells pretreated for 30 min with BAPTA-AM and then exposed to MNNG are shown. Cells were treated for the indicated times and immediately harvested.
PARP-1 hyperactivation was also observed following high doses of MNNG, however, this response was not affected by BAPTA-AM (Fig. 7D). These data highlight two separate PARP-1 regulatory mechanisms. First, ROS-induced, PARP-1mediated cell death appears to require Ca 2ϩ as a cofactor, whereas alkylation-mediated PARP-1-induced cell death does not. We propose that Ca 2ϩ release following ROS-induced stress directly influences PARP-1 and PARG function. Both Mg 2ϩ and Ca 2ϩ exert significant (Ն3-fold increases) allosteric activation of PARP-1 auto(ADP-ribosyl)ation in vitro that is inhibited by EDTA addition (45). We, therefore, speculate that Ca 2ϩ chelation modulates PAR synthesis by dampening PARP-1 auto(poly-ADP)-ribosylation. Furthermore, since increases in [Ca 2ϩ ] can inhibit PARG function by up to 50% in vitro, maintenance of homeostatic Ca 2ϩ levels after drug treatment would, thereby, restore the normal turnover of PAR by PARG, lifting PARP-1 self-inhibition (46). Our data are consistent with the hypothesis that both PAR synthesis and degradation can be modulated by BAPTA-AM to spare the cell from metabolic catastrophe via Ca 2ϩ -mediated NAD ϩ and ATP losses (Figs. 2B, and 3, B and C) (47). The remaining PARP-1 activity would be necessary for DNA break repair, ultimately providing a survival advantage to damaged cells (Figs. 1B, 5C, 6C, and 7C). We are currently exploring the mechanism by which Ca 2ϩ modulates PARP-1 hyperactivation and subsequent DNA repair after H 2 O 2 or ␤-lap treatments versus MNNG.
There appears to be some disagreement as to the role of Ca 2ϩ in PARP-1-dependent cell death. Ca 2ϩ can hyperactivate PARP-1 in the absence of DNA breaks (48). In neuronal cells, glutamate caused Ca 2ϩ -mediated ROS production through mitochondrial dysfunction, leading to DNA damage, PARP-1 hyperactivation, and cell death. Furthermore, Ca 2ϩ chelators, such as BAPTA-AM, EGTA-AM, and Quin-2-AM, protected against other oxidative stress-induced apoptotic and necrotic cell death mechanisms (49,50). In these studies, Ca 2ϩ chelation did not directly inhibit PARP-1 activity, but rather prevented DNA damage by inhibiting ROS. Contrary to these observations, in our system BAPTA-AM did not alter the direct production of ROS or oxidative stress in H 2 O 2 -or ␤-lap-exposed cells (Fig. 6A). Therefore, there does not appear to be an interference with transition metal-mediated oxidant production by BAPTA-AM (e.g. Fenton reaction) as previously suggested after H 2 O 2 treatment (51,52). In fact, our data demonstrate that ␤-lap caused equivalent ROS production in both ␤-lap alone and ␤-lap ϩ BAPTA-AM treated cells. In contrast to ␤-lap alone-treated cells, cells pretreated with BAPTA-AM did not exhibit notable PARP-1 hyperactivation, associated NAD ϩ and ATP losses, and showed a decrease in DNA damage over time (supplemental Table S1). While these data are suggestive of ongoing DNA repair in the presence of Ca 2ϩ chelators, we cannot discount that these observations could also be the result of a decrease in the initial amount of DNA damage created in NQO1ϩ cells in response to ␤-lap. Initial DNA damage and active DNA repair would be indistinguishable in these experiments. Thus, although we believe it is unlikely, the Ca 2ϩ -dependence of PARP-1 hyperactivation could be an indirect consequence of a BAPTA-AM-mediated decrease (i.e. protection) in the initial amount of DNA lesions created in response to ␤-lap. Future studies will address this issue by utilizing DNA repair-compromised NQO1ϩ cells.
Collectively, our data suggest that PARP-1 is necessary for the initiation of cell death caused by ␤-lap. To date, however, the endonuclease responsible for the execution of cell death in response to ␤-lap treatment remains unknown. We therefore propose, that PARP-1-mediated NAD ϩ and ATP losses, in addition to PARG-liberated ADP-ribose, causes an influx of Ca 2ϩ from extracellular and intracellular sources. Impairment of ATP-dependent membrane/organelle transporters (e.g. plasma membrane Ca 2ϩ ATPases (PMCA) and sarcoplasmic/ endoplasmic reticulum Ca 2ϩ ATPases (SERCA)) by ATP loss, and activation of plasma membrane cation channels (e.g. transient receptor potential-melastatin-like (TRMP)) by ADP-ribose, leads to high intracellular Ca 2ϩ levels (53) sufficient to activate the Ca 2ϩ -dependent protease -calpain and commit the cell to death. Previous studies from our laboratory have demonstrated that ␤-lap causes the downstream activation of -calpain resulting in its translocation to the nucleus concomitant with nuclear proteolytic cleavage of p53 and PARP-1 (7). Studies from our laboratory indicate that ␤-lap treatment causes apoptosis-inducing factor (AIF) translocation from the mitochondria to the nucleus, leading to nuclear condensation following -calpain activation. 4 To date, the mechanism responsible for PARP-1-mediated AIF release remains unclear. We speculate that AIF release under conditions of DNA damage may be mediated through a concerted effort of both -calpain and PARP-1. Disruption of the mitochondrial membrane potential through PARP-1-dependent NAD ϩ and ATP losses, in conjunction with -calpain-mediated cleavage of Bid or of AIF itself, may mediate its release from the mitochondria (54,55).
In conclusion, our studies offer new insights into the signal transduction pathways necessary for PARP-1-mediated cell death, providing a connection between PARP-1 hyperactivation and cell death via fluctuations in Ca 2ϩ homeostasis. Knowledge of this pathway may be used to understand, and effectively treat, a large number of human pathologies (e.g. ischemia-reperfusion during heart attacks and stroke, and diabetes), as well as to enhance current cancer chemotherapeutic agents through modulation of PARP-1 hyperactivation.