MNNG-induced Cell Death Is Controlled by Interactions between PARP-1, Poly(ADP-ribose) Glycohydrolase, and XRCC1*

PARP-1 (poly(ADP-ribose) polymerases) modifies proteins with poly(ADP-ribose), which is an important signal for genomic stability. ADP-ribose polymers also mediate cell death and are degraded by poly(ADP-ribose) glycohydrolase (PARG). Here we show that the catalytic domain of PARG interacts with the automodification domain of PARP-1. Furthermore, PARG can directly down-regulate PARP-1 activity. PARG also interacts with XRCC1, a DNA repair factor that is recruited by DNA damage-activated PARP-1. We investigated the role of XRCC1 in cell death after treatment with supralethal doses of the alkylating agent MNNG. Only in XRCC1-proficient cells MNNG induced a considerable accumulation of poly(ADP-ribose). Similarly, extracts of XRCC1-deficient cells produced large ADP-ribose polymers if supplemented with XRCC1. Consequently, MNNG triggered in XRCC1-proficient cells the translocation of the apoptosis inducing factor from mitochondria to the nucleus followed by caspase-independent cell death. In XRCC1-deficient cells, the same MNNG treatment caused non-apoptotic cell death without accumulation of poly(ADP-ribose). Thus, XRCC1 seems to be involved in regulating a poly(ADP-ribose)-mediated apoptotic cell death.

Poly(ADP-ribosylation) of proteins is involved in the regulation of basal cellular processes and seems to be crucial for genomic integrity and cell survival. Responsible for the synthesis of poly(ADP-ribose) (PAR) 2 are poly(ADP-ribose) polymerases (PARPs) (1). The most abundant and active PARP enzyme is PARP-1, a predominantly nuclear protein of 113 kDa. PARP-1 rapidly binds to DNA breaks, is thereby activated, and covalently automodifies itself under consumption of NAD ϩ . To a lesser extent, some other nuclear proteins are also modified with PAR polymers (2). The primary structure of PARP-1 is well conserved between species. The N-terminal DNA binding domain, which contains two zinc finger motifs, is linked to a nuclear localization signal, the main acceptor sites of automodification are located within the central domain, and the 55-kDa C-terminal domain of the enzyme contains the catalytic site (1). PARP-1 appears to physically interact with multiple proteins involved in DNA metabolism, such as histones, transcription factors, replication factors, and DNA repair enzymes (3). Among DNA repair proteins, PARP-1 interacts with x-ray repair cross-complementing protein 1 (XRCC1) (4). A mutant line of Chinese hamster ovary (CHO) cells, which displays hypersensitivity to a broad range of genotoxins was established and termed EM9 cells (5). It turned out that EM9 cells have a reduced ability to rapidly repair DNA single-strand breaks and are genetically unstable as a consequence of XRCC1 deficiency. Different analyses revealed that XRCC1 physically interacts with several DNA repair enzymes, thereby regulating their corresponding activities (6). XRCC1, a polypeptide of 70 kDa, contains two breast cancer C-terminal domains and a nuclear localization signal but is lacking any known enzymatic activity (6). Notably, recruitment of XRCC1 to single-strand breaks strictly depends on PARP-1 activity (7,8). PAR polymers are synthesized in response to DNA breaks, which can arise directly or indirectly, for example, after treatment with alkylating agents. Produced PAR modifications are rapidly degraded by poly(ADP-ribose) glycohydrolase (PARG), which cleaves the polymers with high specificity at the glycosidic bonds, generating free ADP-ribose. PARG is the physiological counterpart for all PARP enzymes, encoded by a unique gene (9). Human PARG, encoded by 18 exons is 110 kDa in size (10) and the catalytic domain resides in the C-terminal part of the enzyme (11,12). Interestingly, recently a 39-kDa protein termed ARH3 has been isolated (13), which possesses a glycohydrolase activity although is structurally unrelated to the conventional 110-kDa PARG. In addition, several PARG isoforms with different sizes, resulting either from alternative initiation events or from posttranslational proteolysis, have been described in mammalians (14). Human PARG contains several putative localization signals: nuclear localization signal, nuclear export signals, and a mitochondrial localization signal (15,16). Whereas PARG activity is detected predominantly in the cytoplasm, full-length PARG is localized to the nucleus (17,18). Knockout of the fulllength isoform of PARG in mice resulted in an increased sensitivity to genotoxic and endotoxic stress (19) and the loss of PARG activity in Drosophila melanogaster caused progressive neurodegeneration (20). Finally, it was demonstrated that after complete abrogation of PARG expression, murine embryonic cells were only viable in the presence of PARP inhibitors. After withdrawal of these inhibitors an accumulation of PAR polymers was observed and cells underwent apoptosis (21). Thus, the metabolism of PAR plays a fundamental role for the decision of the cell to survive or die (22).
Nevertheless, the induction of PARP-1 activity by irreparable amounts of DNA breaks can deplete the cell of NAD and ATP, finally leading to cell death (23). For example, PARP-1-dependent necrosis can be triggered by treatment with 1 mM H 2 O 2 (24). Yu et al. (25) described yet another cell death program depending on PARP-1 activity. In response to an exposure to 0.5 mM N-methyl-NЈnitro-N-nitrosoguanidine (MNNG), PAR polymers accumulate, instantly provoking the translocation of apoptosis inducing factor (AIF) from mitochondria to the nucleus. This death stimulus then induces nuclear shrinkage and finally caspase-independent cell death (25,22). How an accumulation of PAR is accomplished, whether by overactivation of PARP-1 or by repression of PARG activity, is not known. Therefore, the analysis of the relationship between PARP-1 and PARG is necessary to unravel the role of PAR metabolism in cell death.
Previously we reported that PARG interacts with human PARP-1 from HeLa cell extracts (26). In line with our findings, an affinity of PARG for PARP-1 was shown in a recent proteomic approach (27). Here we characterize the functional relationship between PARP-1 and PARG. Both enzymes interact directly and PARG has the ability to modulate PARP-1 activity. In addition, PARG also interacts with XRCC1. Above all, we provide evidence that the interplay between PARP-1, PARG, and XRCC1 regulates apoptotic cell death induced by supralethal MNNG doses.
PAR Metabolism-[␣-32 P]NAD ϩ was obtained from Amersham Biosciences and gallotannins were from Sigma. Poly-(ADP-ribosylation) activities were determined as described previously (26) with minor modifications. Purified recombinant proteins as specified in the legends to the figures or 20 g of nuclear proteins were incubated in 25 l of phosphate buffer (50 mM potassium phosphate, pH 7.2, 200 M EDTA, 10 mM ␤-mercaptoethanol, 100 g/ml bovine serum albumin) with 10 M [␣-32 P]NAD ϩ and 10 g/ml nicked DNA at 30°C for 20 min. All reactions were performed in the presence of 200 M EDTA and absence of Mg 2ϩ ions to reduce endogenous phosphodiesterase and/or ADP-ribose pyrophosphatase activities that convert PAR and ADPR to AMP (34). Reactions were stopped by trichloroacetic acid precipitation and incorporation of [ 32 P]PAR was determined. For determination of relative PARG activity [ 32 P]PAR was synthesized in vitro as described earlier (33) and incubated with purified proteins or nuclear extracts (20 g of proteins per 30 l reactions) in phosphate buffer for 30 min at 30°C. Reactions were stopped by precipitation with acetone. Precipitated nucleotides were dissolved in TE buffer. Samples containing equal amounts of radioactivity were applied to cellulose-coated plates. Thin layer chromatography was performed using the solvent system 0.3 M LiCl, 1 M acetic acid. After separation, dried cellulose plates were subjected to autoradiography and quantified using a phosphorimager or by Cerenkov counting of excised thin layer slices. Protein modifications with [␣-32 P]ADP-ribose were analyzed by PAGE and autoradiography. Alternatively, reactions were stopped by precipitation with acetone and nucleotides were separated by thin layer chromatography.
Antibodies and Immunostaining Analyses-Antibodies against PARG were raised in rabbits immunized with the purified Histagged catalytic domain of PARG (26). These PARG antibodies recognize the catalytic 65-kDa PARG fragment and to a lesser extent also the human full-length PARG. Anti-pentahistidine antibodies were obtained from Qiagen. Antibodies directed against XRCC1 (H-300, from rabbit), AIF (H-300, from rabbit), GST (Z-5, from rabbit), PARP-1 (A-20, from goat), and YY 1 (C-20, from rabbit) were from Santa Cruz. Monoclonal ␣-tubulin antibodies (DM1A, from mouse) were from Sigma (Taufkirchen, Germany), polyclonal anti-PAR antibodies were from Alexis (96-10Ϫ04 from rabbit). For immunofluorescence analyses, cells were fixed with 3% formaldehyde, 0.25% Triton X-100, and antibodies against PAR or AIF were used and visualized using Alexa Fluor-conjugated secondary antibodies (Invitrogen). Nuclei were counterstained with DAPI (3 M, Invitrogen). Necrotic cells were stained by incubating unfixed cells with 100 g/ml propidium iodide (Molecular Probes) in BBS (3.1 mM KCl, 134 mM NaCl, 1.2 mM CaCl 2 , 1.2 mM MgSO 4 , 0.25 mM KH 2 PO 4 , 15.7 mM NaHCO 3 , 2 mM glucose, pH 7.2) for 30 min at 37°C and analyzed immediately using a fluorescence microscope. Photomicrographs were obtained at room temperature with a microscope (Leica, Wetzlar, Germany) equipped with a digital camera. The relative amount of PAR accumulation was quantified using imaging software (Matrix Vision, Oppenweiler, Germany). Localization of AIF in cells and nuclear shrinkage of cells were estimated by visual inspection. At least 300 cells were counted for each sample and all experiments were repeated three times.
Affinity Precipitation and Pulldown Assays-For GST pulldown experiments, GST or GST-PARG 65 (7.5 g) together with recombinant PARP-1 or XRCC1 constructs (20 g) were incubated with glutathione-Sepharose (50 l) in 0.5 ml of BP (10 mM Tris/HCl, 7 mM MgCl 2 , 150 mM NaCl, 50 M ZnCl 2 , 0.05% (v/v) Nonidet P-40, 1 mM dithiothreitol, pH 8.0) by head over head rotation at 4°C for 45 min. After 5 washing steps with BP, bound proteins were extracted with SDS gel loading buffer and subjected to Western blot analysis. Precipitation of XRCC1-bound proteins from EM9-XH cells was performed as described by Caldecott et al. (35). For immunoprecipitation, 40 l of nuclear extracts from HeLa S3 cells (75 g of protein) were diluted with 120 l of immunoprecipitation buffer (20 mM HEPES, 1 mM EGTA, 1 mM EDTA, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, pH 7.5) and pre-cleared by incubation with 50 l of protein A-Sepharose (Sigma), covalently coupled with control goat antibodies, at 4°C for 30 min. Thereafter, the unbound fraction was incubated with 10 g of goat anti-PARP-1 antibodies (PARP-1 (A-20), Santa Cruz) covalently coupled to 35 l of protein A-Sepharose at 4°C for 30 min. The Sepharose beads were washed three times with immunoprecipitation buffer including 100 mM NaCl, and bound proteins were extracted subsequently with SDS gel loading buffer and subjected to Western blot analysis.
Yeast Two-hybrid Analysis-Human full-length PARG was cloned into the two-hybrid vector pGADT7 (Clontech). pAS-XRCC1 (36) was kindly provided by K. W. Caldecott. Plasmids were transformed into the yeast strain PJ69-4A (37) and diploids were selected using synthetic medium lacking leucine and tryptophan. For validation of protein-protein interaction, colonies were transferred to histidine-or adenine-free medium.
Treatment of CHO Cells-Stock solutions of MNNG, 3-aminobenzamide (3-ABA), staurosporine, and Z-VAD(OMe)-fmk (Bachem, Weil am Rhein, Germany) were dissolved in Me 2 SO and serially diluted with BBS or growth medium immediately before use. Nonconfluent CHO cells were treated with 0.5 mM MNNG in BBS for 10 min. Then cells were washed with medium and incubated in fresh medium at 37°C for time periods as indicated in the figure legends. Perchloric acid extracts were prepared from CHO cells and the cellular ATP level was determined as described earlier (32) using the luciferase assay. The cellular NAD level was determined from perchloric acid extracts as described by Jacobson and Jacobson (38).

RESULTS
PARG Interacts with PARP-1-Previously, we demonstrated an interaction between full-length PARG immobilized to tannin-Sepharose and endogenous PARP-1 from HeLa cell extracts (26). We also detected interaction of PARP-1 with the immobilized 65-kDa C-terminal fragment of PARG (amino acids 378 -976), in the following termed PARG 65 , which contains the catalytic activity (data not shown). To further characterize the interaction between PARP-1 and PARG we performed GST pulldown experiments with purified recombinant constructs (Fig. 1A). We expressed catalytically active GST-PARG 65 in E. coli, purified it, and incubated it with His-tagged full-length PARP-1. Fractions from GST pull-down experiments were separated by SDS-PAGE and subjected to Western blot analysis using anti-polyhistidine and anti-GST antibodies, respectively (Fig. 1A). As evident, after GST pulldown fulllength PARP-1 was detected only in fractions eluted from GST-PARG 65 beads (Fig. 1A, uppermost panels, fourth lane). If Histagged constructs of isolated domains of PARP (33) were applied, only the automodification domain showed an affinity to GST-PARG 65 , whereas no interaction was observed with the other PARP domains (Fig. 1A). These experiments were performed in the absence of NAD ϩ or PAR, thus demonstrating a direct protein-protein interaction between catalytically active PARG 65 and the automodification domain of PARP-1. Next, purified full-length PARP-1 and GST-PARG 65 were incubated as described before and the elution fraction from the GST pulldown was analyzed for PARP activity by incubation with [␣-32 P]NAD and nicked DNA. Eluted PARP-1 was catalytically active (Fig. 1B, second lane), indicating that native PARP-1 bound to the GST-PARG 65 beads. Menard et al. (39) reported that PARG was no acceptor for PAR modifications in vitro, when PARP-1, PARG, histones, DNA, and NAD were incubated together at physiological ratios. Similarly, we observed that the predominant reaction in elution fractions from pulldown with GST-PARG 65 was automodification of PARP-1, whereas [ 32 P]ADP-ribosylation of PARG 65 was only detectable if GST-PARG 65 was added in excess (data not shown).
To study the impact of the interaction between PARP-1 and PARG in vivo, we performed co-immunoprecipitation experiments. First, Western blot analyses showed that PAR modifications were virtually absent in extracts of untreated HeLa cells (Fig. 1C). PARG 65 was detected after co-imunoprecipitation of these extracts with anti-PARP-1 antibodies (Fig. 1D). Thus, a preformed complex of PARG and PARP-1 might exist in the nucleus of HeLa cells, independent of PARP-1 activation. To further study the interplay of these enzymes, we supplemented nuclear extracts of HeLa cells with increasing amounts of recombinant GST-PARG 65 and incubated the extracts with [␣-32 P]NAD and nicked DNA. After reaction for 20 min, modifications with [ 32 P]poly(ADP-ribose) were visualized in SDS-PAGE (Fig. 2B). Amounts of released [ 32 P]ADP-ribose were determined by evaluation of thin layer chromatograms ( Fig.  2A), and polymer chain lengths were analyzed by gel electrophoresis (Fig. 2C). As expected, the addition of increasing amounts of catalytically active GST-PARG 65 enhanced the levels of released [ 32 P]ADP-ribose ( Fig. 2A, bars 1-4), whereas the sizes of the produced [ 32 P]PAR polymers were decreased (Fig.  2C, lanes 1-4). Previously we presented evidence that tannins elevate the level of PAR in HeLa cell extracts by inhibition of PARG (26). In the presence of nuclear proteins 150 M tannins had no modulating influence on PARP-1 activity (26). If similar experiments as shown in Fig. 2, bars/lanes 1-4, were performed in the presence of 150 M tannins, the levels of [ 32 P]ADP-ribose were not affected by the addition of PARG 65 ( Fig. 2A, bars 5-8), but, notably, the sizes of the polymers were significantly reduced (Fig. 2, B and C, lanes 5-8). In conclusion, even inde-pendent of its hydrolyzing activity PARG 65 may affect the catalytic activity of PARP-1 by direct protein-protein interaction, resulting in the synthesis of only short polymers.
PARG Interacts with XRCC1-PARG 65 interacts with the automodification domain of PARP-1 (Fig. 1A), which contains a breast cancer C-terminal domain and is known to be the region important for interactions with other proteins (33) such as for instance XRCC1 (4). To analyze a potential direct connection between PARG and XRCC1, we investigated a direct interaction of these proteins using the following set of experiments. First of all, human full-length PARG was cloned into the yeast two-hybrid reporter plasmid pGADT7 and interaction with an appropriate construct containing XRCC1 (pAS-XRCC1; Ref. 36) was studied in yeast cells (Fig. 3A). Indeed, only in the presence of both PARG and XRCC1 was a supplementation of histidine biosynthesis observed, whereas controls using empty vectors were negative (Fig. 3A). The interaction between PARG 65 and XRCC1 was further verified by in vitro GST pulldown experiments. For that purpose, His-tagged XRCC1 was purified from E. coli and incubated with purified GST-PARG 65 . After GST pulldown, fractions were analyzed by Western blot as described above for PARP-1 (cf. Fig. 1A). Fig. 3B shows that His-tagged XRCC1 bound to GST-PARG 65 . The interaction between XRCC1 and GST-PARG 65 was stable toward salt concentrations of up to 150 mM NaCl (data not shown). To analyze the interaction between PARG and XRCC1 in vivo, we made use of EM9 cells, complemented with either empty vector (EM9-V) or a vector containing His-tagged, human wild-type XRCC1 (EM9-XH) (29). Whole cell extracts of exponentially growing CHO cells were incubated with Ni-NTA-Sepharose. After washing the Ni-NTA beads, elution fractions were subjected to Western blot analysis using specific antibodies against PARG 65 . As shown in Fig. 3C, PARG clearly bound to His-tagged XRCC1. In conclusion, PARG has the abil-  Controlled by PARP-1, PARG, and XRCC1 NOVEMBER 10, 2006 • VOLUME 281 • NUMBER 45 ity to interact with PARP-1 and also with XRCC1, even in the absence of its substrate PAR.

Cell Death Is
XRCC1 Enhances the Synthesis of Poly(ADP-ribose) in MNNGtreated Cells-XRCC1 interacts with PARP-1 (4) and also with PARG (cf. Fig. 3). Thus, XRCC1 might have the capacity to regulate cellular PAR metabolism. Therefore, in the following experiments we addressed the impact of XRCC1 on PAR synthesis using EM9 cells. First, EM9-V and EM9-XH cells were treated with 0.5 mM MNNG, a supralethal dose to induce poly-(ADP-ribosylation). Interestingly, immunofluorescence analyses clearly showed that higher amounts of PAR polymers were produced in XRCC1-containing EM9-XH cells compared with XRCC1-deficient EM9-V cells (Fig. 4A). Evaluation of quantified PAR signals revealed that in EM9-XH cells, PAR accumulation was increased about 3-4-fold compared with the accumulation observed in EM9-V cells (Fig. 4, B and C). Western blot analyses of cell extracts confirmed this (Fig. 4D). Even 30 min after MNNG treatment, considerably higher amounts of PAR and larger PAR modifications were still detectable in EM9-XH cells, whereas only lower amounts and smaller sizes of PAR modifications remained in EM9-V cells (compare lanes 3 with 6 in Fig. 4D). Hence, after treatment with toxic MNNG doses XRCC1 influences PAR metabolism in living CHO cells.
To further elucidate the regulation of PAR metabolism by XRCC1, PARP and PARG activities were analyzed and compared in different CHO lines deficient or proficient in XRCC1. Western blot analyses revealed no significant differences in PARP-1, or PARG 110 /PARG 65 levels, whereas levels of DNA ligase III (Lig III) were considerably reduced in XRCC1-deficient cells (data not shown). Consequently, whereas Lig III is stabilized by XRCC1 (35), cellular XRCC1 does not appear to be essential for the stability of PARP-1 or PARG. Previously, Masson et al. (4) proposed that XRCC1 down-regulates PARP-1 activity. Accordingly, we observed an inhibition of poly(ADP-ribosylation) in a reconstituted system with recombinant proteins, when XRCC1 was incubated with PARP-1 in a molar ratio of 8:1 (Fig. 5A). Furthermore, we found that PARG activity was unaltered, even if an excess of XRCC1/Lig III was added to GST-PARG 65 (Fig. 5B). Analyzing relative PARP and PARG activities in nuclear extracts of different CHO lines deficient or proficient in XRCC1 resulted in comparable PARP and PARG activities in all cell lines (Fig. 5, C and D). Similarly, it has been reported that cellular NAD content and relative PARP activity appeared normal in EM9 cells (40). These seemingly contradictory observations are explainable because in living cells the level of XRCC1 is not higher than the level of PARP-1. A direct decreasing effect on PARP-1 activity was only observed if an excess of XRCC1 was added to isolated PARP-1 (Fig. 5A).
In response to treatment with genotoxic agents, XRCC1 is recruited to DNA single-strand breaks in a PARP-1-dependent fashion (7). In comparison to PARP-1, the cellular concentration of XRCC1 is lower (41). Therefore, recruitment might represent an increase of the local concentration of a limited factor at sites of recruitment. Because neither PARP-1 nor PARG activities are directly regulated by XRCC1 (Fig. 5, C and D), XRCC1 recruitment might be required for the observed PAR accumulation in MNNG-treated XRCC1-proficient cells (Fig.  4). Therefore, in the next experiment, we reconstituted XRCC1 recruitment in vitro by supplementation of XRCC1-deficient nuclear extracts with an excess of recombinant XRCC1. For that purpose, nuclear extracts from XRCC1-deficient EM9 cells were complemented with recombinant XRCC1 during monitoring of PARP activity for a period of 20 min using nicked DNA and [␣-32 P]NAD ϩ . [ 32 P]Poly(ADP-ribosylation) was visualized by SDS-PAGE (Fig. 6A) and polymer chain lengths were analyzed in parallel (Fig. 6B). Indeed, we observed that PAR syn- thesis was dramatically enhanced after supplementation with XRCC1 (Fig. 6). Both, the amounts and the sizes of formed polymers increased significantly after supplementing XRCC1 to EM9 extracts (Fig. 6, A and B, right lanes). Large and branched PAR polymers, which were unable to enter the gel, were only produced when XRCC1 was present (Fig. 6, A and B, top of right lanes). The fractions of [ 32 P]poly(ADP-ribosylated) proteins with large modifications were quantified as insoluble fractions after acetone precipitation (PAR* in Fig. 6C). Soluble [ 32 P]ADP-ribose metabolites PAR, ADP-ribose, AMP, and NAD ϩ were separated by thin layer chromatography and relative ratios of all [ 32 P]ADP-ribose metabolites were determined (Fig. 6C). The evaluation of the overall [ 32 P]ADP-ribose metabolism revealed that in the presence of XRCC1 a significant and reproducible accumulation of large [ 32 P]PAR modifications (PAR*) was achieved (Fig. 6C). Moreover, the levels of [ 32 P]ADPribose decreased when XRCC1 was added (Fig. 6C). Thus, the observed enhanced formation of [ 32 P]PAR polymers was partly caused by a suppression of PARG activity. Neverthe- less, XRCC1 stimulated PARP activity because the consumption of [ 32 P]NAD ϩ was significantly increased in reactions supplemented with XRCC1 (Fig. 6C). Similarly, supplementation of nuclear extracts from HeLa cells with recombinant XRCC1 resulted in increased poly(ADP-ribosylation) (data not shown). A recruitment of or supplementation with XRCC1 appears to be necessary for efficient PAR accumulation. One possible explanation for this is that XRCC1 displaces PARG from binding PAR or from associations with other proteins. However, because PARG and PARP-1 interact (cf. Fig. 1) and PARG thereby down-regulates PAR syntheses (cf. Fig. 2), it is conceivable that the observed accumulation of PAR polymers is a result of alternating interactions between PARG, PARP-1, and XRCC1.
XRCC1-induced Accumulation of PAR Leads to Apoptotic Cell Death-It is known that an accumulation of PAR in response to high doses of MNNG triggers several further apoptotic events, including translocation of AIF from mitochondria to the nucleus and nuclear shrinkage within a few hours (25). Because we observed increased PAR accumulation in EM9-XH cells compared with EM9-V cells after treatment with supralethal MNNG doses (cf. Fig. 4), in the next experiments we monitored translocation of AIF and nuclear shrinkage in CHO cells, 3 or 6 h after MNNG treatment (Fig. 7). The cellular content of AIF was comparable in EM9-V and EM9-XH cells (Fig.  7C). Translocation of AIF after MNNG treatment could be detected in only a few EM9-V cells and in most EM9-XH cells, but not in the presence of the PARP inhibitor 3-ABA (Fig. 7, A  and B). Because XRCC1 increases the amount of PAR formation (Figs. 4 and 6), it appears that XRCC1 might as well regu-late PARP-1-dependent translocation of AIF from mitochondria to the nucleus. Notably, 6 h after MNNG treatment, shrunken nuclei were detected in more than 80% of EM9-XH cells but only in 10% of EM9-V cells (Fig. 7D). When the XRCC1 proficient parental cell line AA8 was treated with MNNG, similar apoptotic features as obtained with the EM9-XH cells were observed (data not shown). The caspase inhibitor Z-VA-D(OMe)-fmk (100 M) failed to block MNNG-induced nuclear shrinkage in EM9-XH cells (Fig. 7D). In a control experiment,  . Supplementation of EM9 nuclear extracts with recombinant XRCC1 increases the PAR level. Nuclear extracts (15 g of protein/reaction) from EM9 cells were supplemented with 1 g of bovine serum albumin (BSA) or recombinant XRCC1 as indicated. 10 M [ 32 P]NAD and 10 g/ml nicked DNA were added and reactions were stopped after 20 min. A, proteins were subjected to SDS-PAGE and autoradiography. Relative molecular weights of marker proteins and the positions of PARP-1 and XRCC1 are indicated. B, reactions were precipitated with trichloroacetic acid, polymers were detached and subjected to polymer chain length analysis using an 8% polyacrylamide gel. The autoradiogram of the gel is shown and the numbers of ADP-ribose units are indicated. C, reactions were precipitated with acetone. Amounts of insoluble precipitates, which contained proteins modified with large PAR polymers, were termed PAR* fractions and quantified. Soluble 32 P metabolites were separated by thin layer chromatography and signals of [ 32 P]PAR, -ADP-ribose, -AMP, and -NAD were quantified. The evaluation of relative ratios of all 32 P metabolites is shown. In the control reaction (diagram bottom right) no proteins were present. the impact of XRCC1 on induction of apoptosis was analyzed, using staurosporine, another trigger of cell death (Fig. 8). Staurosporine has two effects on nuclear structure, either causing caspase-independent partial nuclear condensation (stage I) or caspase-dependent advanced nuclear condensation and fragmentation (stage II) (42). We observed that AIF translocation and nuclear condensation induced by staurosporine were comparable in EM9-V and EM9-XH cells (Fig. 8). Furthermore, Z-VAD(OMe)-fmk did not prevent AIF translocation but effectively blocked the formation of shrunken nuclei (stage II) in both cell lines (Fig. 8). Thus, the impact of XRCC1 on apoptotic events appears to be restricted to MNNG-induced caspase-independent cell death.
Poly(ADP-ribosylation) is an energy-consuming process. Consequently, cellular levels of ATP and NAD were considerably reduced within 60 min after MNNG treatment in all tested cell lines (Fig. 9, A and B). 13 h after MNNG treatment, EM9-V cells were permeable for staining with propidium iodide, whereas most of EM9-XH cells had already died (Fig. 9D). Finally, 24 h after MNNG treatment nearly all EM9-V cells had died, whereas about 20% of EM9-XH survived (Fig. 9C). Thus, in EM9-V cells cell death was presumably caused by energy depletion.

DISCUSSION
The investigation presented here suggests a new level of coordination of MNNG-induced cell death regulated by interactions between PARP-1, PARG, and XRCC1. Several conclusions can be drawn from our study. First of all, PARG interacts with PARP-1 and XRCC1. Second, PARG can regulate PARP-1 activity. Third, XRCC1 regulates PAR-mediated apoptotic cell death induced by supralethal doses of MNNG.
The creation of different PARG knockout models revealed the importance of PAR polymers for cellular survival. For example, genetic PARG inactivation in D. melanogaster resulted in a severe PAR accumulation in neuronal cells and lethality at the larval stage (20) and mouse cells lacking PARG showed an accumulation of PAR leading to cell death by apoptosis (21). Thus, PAR can induce apoptosis and PAR signaling appears to play an important role in embryonic development. In a recent proteomic approach, localization of a fraction of PARG to messenger ribonucleoparticles and an interaction of PARG with Fragile-X-related protein was discovered (27). Given the fact that PARG activity is involved in many different cellular events these observed interactions could be important also for the regulation of the activities of PARP enzymes other than PARP-1. All current models regarding the functions of poly-(ADP-ribosylation) are based on the interplay between PARP-1 and PARG, but a direct interaction between both enzymes has not been described before. In the study presented here we show for the first time that PARP-1 interacts with PARG, even independently of the substrates NAD or PAR. In addition, we characterize XRCC1 as another interaction partner of PARG. Remarkably, phosphorylation of XRCC1 by protein kinase CK2 influences its interaction with the DNA repair protein polynucleotide kinase, as recently shown (30). Furthermore, in a large scale characterization of nuclear phosphoproteins from HeLa cells, phosphorylated PARG was detected (43). Thus, it is tempting to speculate that PARG activity or its interaction with other proteins might be regulated by phosphorylation. Nevertheless, functional analyses of how PARG activity is modulated in living cells are lacking yet.
The biological significance of PARP-1 activity has been the subject of numerous publications but it is poorly understood how this catalytic activity is modulated (1-3). As it is known, PARP-1 activity is induced by DNA breaks, introduced either directly or indirectly, and depends on the availability of NAD. Among the proteins interacting with PARP-1, XRCC1 and DNA-dependent protein kinase have been reported to negatively regulate PARP-1 activity (4,44). On the other hand, we characterized transcription factors that have the ability to directly stimulate PARP-1 activity (45)(46)(47). Furthermore, an allosteric activation of PARP-1 automodification by Mg 2ϩ , Ca 2ϩ , histones H1 and H2, and polyamines has been demonstrated (48). Here, with PARG we present another PARP-1interacting protein that may function as a negative regulator. In line with these findings, Cortes et al. (19) reported that when the full-length isoform of PARG was knocked-out in mice, residual PARG 60 severely reduced the automodification activity  of PARP-1 in vivo. Thus, it is feasible that PARG, when interacting with PARP-1, inhibits the automodification reaction and thereby guarantees that large PAR modifications are not generated. This inhibition of PARP-1 activity might be achieved directly or indirectly by displacement of PARP-1 activators as histones or transcription factors. Thus, PARG appears to have a dual function in PAR metabolism, degradation of PAR and down-regulation of PAR synthesis. In contrast to PARG, PARP-1 is highly abundant in the nucleus (14). In situations of low levels of DNA damage, PAR synthesis and degradation are balanced and no accumulation of PAR occurs. It has therefore been assumed that the high specific activity of PARG compensates for its low cellular concentration (39). Based on our findings it can be speculated that at low levels of DNA damage, only a minor fraction of nuclear PARP-1 is catalytically activated and at the same time down-regulated by PARG. Hence, the mechanism of regulation proposed here ensures that in the absence of high amounts of DNA lesions, PAR polymers do not accumulate. Accordingly, it has been reported that in response to moderate doses of alkylating agents, such as 15 M MNNG, the production of PAR is limited, whereas higher MNNG doses trigger the formation of large PAR polymers in cells peaking within 10 -15 min (49).
XRCC1, for which no enzymatic activity has been described, plays a central role in DNA singlestrand break repair (6). Here we suggest an unexpected role for XRCC1, influencing apoptotic cell death in response to treatment with supralethal doses of MNNG. XRCC1 binds preferentially to automodified PARP-1, most likely via a putative PAR recognition and binding motif within its internal breast cancer C-terminal I motif (4,50). Furthermore, it is established that XRCC1 is immediately recruited to DNA lesions by automodified PARP-1 (7,8). At physiological ratios a direct modulation of PARP-1 or PARG activity by XRCC1 was not detected (Fig. 5,  C and D). Notably, the effect of XRCC1 on PAR accumulation was only observed in response to treatment of cells with toxic doses of MNNG (Fig. 4). Because XRCC1 preferentially interacts with automodified PARP-1 (4, 51, 52) it may be suspected that PARG could be displaced from its interaction with the PARP-1 automodification domain.
Presumably, the repression of PARP-1 activity by PARG is abrogated by XRCC1, resulting in increased PAR synthesis. Based on the experiments presented here we propose the following hypothesis ( Fig. 10): PARP-1 interacts with PARG, even in the absence of significant automodification. In response to treatment with supralethal MNNG doses massive DNA lesions are introduced, PAR synthesis and degradation proceed, and automodified PARP-1 instantly recruits XRCC1. It can be speculated that if amounts of DNA lesions are beyond the cellular repair capacity, PARP-1 still remains catalytically active, whereas PARG might be displaced by XRCC1. As a result, the number and length of PAR polymers would increase further. Accumulated PAR polymers then trigger the translocation of AIF from mitochondria and nuclear shrinkage, ultimately leading to caspase-independent apoptotic cell death (25). In XRCC1-deficient cells, the combined activities of PARP-1 and PARG cause NAD and ATP depletion resulting in non-apoptotic cell death (Fig. 9). In addition, it has been reported that after methyl methanesulfonate treatment NAD depletion was augmented in EM9-V cells compared with EM9-XH cells (53). Cell killing effects of alkylating agents depend on the doses of agents and the growth state of CHO cells (54). Here we demonstrate that in response to treatment with a cytotoxic dose of MNNG, rapid accumulation of PAR and induction of apoptotic cell death occurred only in XRCC1 proficient cells. However, treatment with H 2 O 2 introduces oxidative DNA damage including direct single-strand breaks, so that in difference to treatment with MNNG, substan-tial amounts of PAR are still detectable 30 min after the treatment (55). Equal amounts of PAR foci triggered by 20 mM H 2 O 2 were detected in EM9-V and EM9-XH cells (55). Consequently, these PAR foci seem to arise independent of the presence of XRCC1. Presumably, H 2 O 2 -induced PAR foci persist because of the limiting amount of PARG molecules compared with the amount of damage-activated PARP-1 molecules. In line with this, we observed nuclear condensation also of most of EM9-V cells in response to a treatment with 1 mM H 2 O 2 (data not shown). Thus, the impact of XRCC1 on PAR accumulation and cell death depends on the type and amount of DNA lesions introduced and obviously seems to be specific for treatment with toxic MNNG doses.
When a cell is damaged, the cell has two options: repair or die. If the damages are too extensive the cell must also decide which cell death pathway to follow. Several ATP-dependent steps are required for apoptotic signal transduction and an excessive NAD/ATP depletion below 50% is believed to induce cell death by necrosis (56). Apparently, PARP-1 plays a dual role in triggering cell death. On the one hand PARP-1 consumes NAD, thus PARP-1 activation may cause energy deprivation. Accordingly, PARP-1 overactivation leads to necrotic cell death (24). Moreover, PARP-1-mediated necrotic cell death after treatment with 0.5 mM MNNG was detected in mice fibroblasts by HMGB1 exclusion (57), whereas AIF localization was not monitored in that study. On the other hand, accumulating PAR polymers appear to be an effective apoptotic stimulus signaling AIF release (25,22). Additionally, studies with cortical neurons revealed that an accumulation of PAR and not excessive NAD consumption was responsible for initiation of apoptosis (22,58).
Efficient progression of both DNA repair and apoptosis are essential for genome integrity. Poly(ADP-ribosylation) plays relevant roles in DNA damage sensing/repair and apoptosis. From our study we suggest that XRCC1 might be another important determinant regulating both processes.