Importance of poly(ADP-ribose) polymerase and its cleavage in apoptosis. Lesson from an uncleavable mutant.

We have studied the apoptotic response of poly(ADP-ribose) polymerase (PARP)-/- cells to different inducers and the consequences of the expression of an uncleavable mutant of PARP on the apoptotic process. The absence of PARP drastically increases the sensitivity of primary bone marrow PARP-/- cells to apoptosis induced by an alkylating agent but not by a topoisomerase I inhibitor CPT-11 or by interleukin-3 removal. cDNA of wild type or of an uncleavable PARP mutant (D214A-PARP) has been introduced into PARP-/- fibroblasts, which were exposed to anti-CD95 or an alkylating agent to induce apoptosis. The expression of D214A-PARP results in a significant delay of cell death upon CD95 stimulation. Morphological analysis shows a retarded cell shrinkage and nuclear condensation. Upon treatment with an alkylating agent, expression of wild-type PARP cDNA into PARP-deficient mouse embryonic fibroblasts results in the restoration of the cell viability, and the D214A-PARP mutant had no further effect on cell recovery. In conclusion, PARP-/- cells are extremely sensitive to apoptosis induced by triggers (like alkylating agents), which activates the base excision repair pathway of DNA, and the cleavage of PARP during apoptosis facilitates cellular disassembly and ensures the completion and irreversibility of the process.

Apoptosis or programmed cell death is a fundamental biological process that plays an important role in early development, cell homeostasis, and in diseases such as neurodegenerative disorders and cancer (1)(2)(3). Programmed cell death can occur in response to many stimuli such as genotoxic insult when DNA repair is saturated, removal of growth factors, or activation of the CD95 antigen by CD95 ligand or anti-CD95 antibodies. Morphologically it is characterized by the appearance of membrane blebbing, cell shrinkage, chromatin condensation, and DNA cleavage, and finally the cell is fragmented into membrane-bound apoptotic bodies. At the biochemical level, there is increasing evidence for a central role of the family of cysteine proteases, the caspases, in the pathway that mediates the highly ordered process leading to cell death (4). Caspases have been identified as the enzymes responsible for the proteolysis of key proteins to be selectively cleaved at the onset of apoptosis. It appears that the role of these proteases in cell suicide is to disable critical homeostatic and repair enzymes as well as key structural components. A discrete but increasing number of specific proteins appears to be targeted for proteolytic cleavage during apoptosis, including poly(ADP-ribose) polymerase (PARP, 1 EC 2.4.2.30), which was first described in Ref. 5. In the last years, cleavage of PARP has been used extensively as a marker of apoptosis. However, the reason for the cell to inactivate this protein during the execution phase of apoptosis is not fully understood.
PARP is a nuclear zinc finger DNA-binding protein that detects and binds to DNA strand breaks. PARP has a modular organization comprising a NH 2 -terminal DNA-binding domain, a central regulatory domain, and a COOH-terminal catalytic domain. At a site of DNA breakage, PARP catalyzes the transfer of the ADP-ribose moiety from its substrate, NAD ϩ , to a limited number of protein acceptors (heteromodification) involved in chromatin architecture (histones H1 and H2B, lamin B) or in DNA metabolism (topoisomerases, DNA replication factors) including PARP itself (automodification) (6 -9). Mice lacking PARP have been generated by homologous recombination to assess the biological consequences of PARP deficiency (10 -12). We have reported that PARPϪ/Ϫ mice are hypersensitive to genotoxic agents, like ␥-rays and monofunctional alkylating agents, compared with their ϩ/ϩ litter mates. Mutant mice displayed genomic instability as shown by an increased rate of sister chromatid exchanges and an increased occurrence of chromosomal breaks in their bone marrow cells. Using PARPϪ/Ϫ mice-derived cells, we and others have established that apoptosis occurs in the absence of PARP (11)(12)(13). However, we have shown that PARPϪ/Ϫ splenocytes exposed to a monofunctional alkylating agent underwent much more rapid apoptosis than wild-type (wt) cells (11).
In the present report, we have studied the specificity of the response to different inducers of apoptosis of PARP null cells and the biological meaning of PARP cleavage during apoptosis. To that purpose, we generated a mutant of PARP in which the specific caspase-3 cleavage site 211 DEVD 214 has been mutated.
A fully active, nuclear, and uncleavable protein has been produced and when expressed in PARPϪ/Ϫ cells delays loss of cell viability and the morphological changes associated with apoptosis upon activation of the CD95 receptor.
Site-directed Mutagenesis-The BS(Ϫ) SHT construct bearing the full-length human PARP sequence (14) was used to generate the mutation D214A. The sequence of the oligonucleotide is 5Ј-GCCACT-TCATCCACGCCGGCCACCTCATCGC-3Ј. The new restriction site NaeI (underlined) generated allowed us to verify the presence of the mutation in further subcloning. Oligonucleotide-directed mutagenesis was performed essentially according to the Amersham kit. After mutagenesis, cDNA was sequenced by the dideoxynucleotide chain termination method.
The mutant sequence was recloned in the prokaryotic expression vector pTG 161 PARP, which expressed the wt PARP (15) by replacing the BssHII-NcoI wt fragment by the mutated one resulting in pTG 161 D214A-PARP. These vectors were used directly for transformation of the Escherichia coli TGE 900 strain (16). A fragment PstI-PstI containing the whole cDNA from the PARP and D214A-PARP was introduced into the XhoI site of the eucaryotic expression vector pECV (17) with PstI-XhoI adapters resulting either in pECV PARP or pECV D214A-PARP, respectively.
Indirect Immunofluorescence Staining-This was done essentially as described by Schreiber et al. (18). Briefly, cells grown on coverslips were washed three times with PBS and fixed with methanol/acetone (v/v) for 10 min at 4°C. After washing three times with PBS supplemented with Tween 0.1% (v/v), cells were incubated overnight at 4°C with polyclonal anti-PARP antibody (1:100 dilution). After washing, the coverslips were incubated with a secondary antibody (Texas Red-conjugated) for 4 h at room temperature. Immunofluorescence was evaluated using a Zeiss Axioplan microscope equipped with a C5985 chilled CCD camera (Hamamatsu).
Apoptotic Cell-free Extracts-These were prepared from PARPϪ/Ϫ splenocytes treated with 2 mM MNU during 6 h following previously reported procedures (19). The cell-free reaction comprised the in vitro translated 35 S-labeled PARP and 35 S-labeled D214A-PARP (3 l of 50 l final volume of the TNT reaction), 5 l of cytoplasmic extract (10 mg/ml protein), and 4 l of the following buffer: 10 mM HEPES, 50 mM NaCl, 2 mM MgCl 2 , 5 mM EGTA, 1 mM dithiothreitol, 1 mM ATP, 10 mM phosphocreatine, and 50 g/ml creatine kinase.
PARP Activity Assay-Assay of PARP activity was carried out as described previously (20). Samples corresponding to 200 ng of wild-type or mutant PARP proteins were incubated for 10 min at 25°C in assay buffer (100 l) consisting of 50 mM Tris-HCl, pH 8.0, 4 mM MgCl 2 , 0.2 mM dithiothreitol, 2 g/ml total histones, 2 g/ml DNase I-activated calf thymus DNA, 200 mM NAD ϩ , and 5 Ci/ml [␣-32 P]NAD ϩ . The reaction was stopped by addition of 25 ml of 10% trichloroacetic acid, 1% inorganic pyrophosphate, and the acid-insoluble radioactivity was determined in a scintillation counter (21).
Transfections and Morphological Evaluation of Apoptosis-Cells were transiently cotransfected by electroporation with 3 g of LacZexpressing pCHK reporter vector (22) and 15 g of wt PARP, D214A-PARP, or the parental vector pECV. They were induced to apoptosis by anti-CD95 (1 g/ml) or 0.5 mM N-methylmethanesulfonate (MMS) 36 h later. Cells were fixed with 2% paraformaldehyde in PBS for 10 min at 4°C and stained with 100 g/ml X-gal, 5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, and 1 mM MgCl 2 in PBS for 5-18 h. After X-gal staining, the number of viable (flat) and non-viable (round) blue cells was counted. Viable cells remaining after treatment were repre-sented as a percentage of untreated viable blue cells (23).
For evaluation of the nuclear morphology of apoptosis, MEFs were plated onto glass slides in 6-well plates and transfected by calcium phosphate the day after. For transfection, 1 g of a plasmid coding for the green fluorescence reporter protein (pE-GFP) and 10 g of pECV, pECV PARP, or D214A were used. Cells were treated with anti-CD95 36 h following transfection. Detection of apoptotic nuclei was performed according to Jacobson et al. (24). Briefly, the cells were fixed in 4% paraformaldehyde (in 0.1 M phosphate buffer, pH 7.4) for 10 min at room temperature, incubated for 10 min with 5% acetic acid, 95% ethanol at Ϫ20°C, and then stained with 0.05 g/ml propidium iodide in PBS containing 50 g/ml RNase A for 30 min at 37°C. The cells were mounted with Mowiol and examined with a Zeiss fluorescence microscope.
DNA Laddering during Apoptosis-Primary bone marrow cells were treated with 2 mM MNU for 30 min, washed, and incubated in fresh medium. Cells were harvested at different times and lysed with 20 mM Tris-HCl, pH 7.7, 2 mM EDTA, 0.4% Triton X-100 for 15 min at 4°C. Centrifugation at 13,000 ϫ g eliminated cell debris and high molecular weight DNA. The supernatant was treated with phenol/chloroform (v/ v), and the upper phase was precipitated twice with ethanol and analyzed on 1% agarose gel.

PARP-deficient Cells Are Highly Sensitive to Apoptosis
Induced by Alkylating Agents-IL-3-dependent bone marrow (BM) cells are a model system to study biochemical events leading to apoptosis driven by growth factor removal (25). BM cells were isolated from PARPϩ/ϩ and PARPϪ/Ϫ mice (11) and maintained in culture in the presence of IL-3. As inducers of apoptosis, the genotoxic agent MNU, which specifically activates the catalytic function of PARP, the topoisomerase I inhibitor irinotecan (CPT-11, a semisynthetic camptothecin derivative), and the removal of IL-3 have been used. DNA damage induced by MNU activates the base excision repair (BER) pathway whereas the genotoxic effect produced by CPT-11 does not involve this pathway. On the other hand, IL-3 removal-induced apoptosis is independent of DNA damage and involves inactivation of survival signals coming from the IL-3 receptor (26). We have then compared the responsiveness of BM cells of both genotypes to these inducers of apoptosis by detection of DNA fragmentation into oligonucleosomes. As shown in Fig. 1A, primary cultured bone marrow cells exposed to MNU undergo apoptosis much more rapidly than their counterparts. In parallel (Fig. 1B), PARPϪ/Ϫ BM cells showed an increased sensitivity to MNU as determined by a dose-dependent effect on cell viability. When we induced DNA damage by CPT-11, not involving base excision repair, we find a similar dose-dependent effect on DNA laddering, independent of the cell genotype (Fig.  1C). Finally, when the cells were deprived of IL-3, there were no significant differences in the kinetic appearance of DNA laddering (Fig. 1D). These cells were also identically sensitive to the cytotoxicity induced by activation of the CD95 receptor and tumor necrosis factor-␣ in the presence of cycloheximide (not shown). Similar results have been described using primary MEFs (12) or hepatocytes (13) from PARP null animals, implying that the presence of PARP is not an absolute requirement for completion of apoptosis. In contrast, a more recent study on the role of poly(ADP-ribosylation) during CD95-induced apoptosis, finds that the absence of PARP makes the cells resistant to cell death following CD95 treatment (27). This conflicting result could arise from the use of clones of immortalized/stably transfected PARPϪ/Ϫ MEFs in the above mentioned study, which might influence the responsiveness to CD95.
Taken together, these results clearly indicate that PARPdeficient cells specifically have a decreased apoptotic threshold when the trigger is a DNA-damaging agent activating the BER pathway, such as the alkylating agent MNU.
D214A-PARP Mutant Is Nuclear, Fully Active, and Resists the Cleavage by Caspase-3 and by an Apoptotic Cell-free Ex-tract-During apoptosis PARP (116 kDa) is cleaved to fragments of relative molecular size 85 and 27 kDa. Amino-terminal sequencing of the purified 85-kDa fragment revealed that cleavage occurs between Asp-214 and Gly-215 (5) inside the bipartite nuclear localization signal of PARP (18), leading to a loss of function. Recently it has been confirmed that the tetrapeptide DEXD is the optimal recognition motif for caspase-2, -3, and -7 (4).
To further analyze the biological consequences of PARP cleavage on apoptosis, the aspartic acid at position 214 was changed to an alanine in the caspase-3 cleavage site, giving rise to the mutant D214A-PARP ( Fig. 2A). As the mutation takes place in the nuclear localization signal, the subcellular distribution of the mutated protein was assessed. For this purpose, MEF PARPϪ/Ϫ cells were transfected with D214A-PARP cDNA, and the expressed protein was immunostained using a polyclonal antibody raised against the whole protein. We found that D214A-PARP protein was completely nuclear, indicating that the Asp-214 residue is not involved in the nuclear targeting of PARP (Fig. 2B). Moreover, cells transiently transfected with D214A-PARP cDNA were positive for poly(ADP-ribose) production following treatment with 0.1 mM H 2 O 2 , as revealed by double immunofluorescence using an anti-polymer antibody (results not shown). The same specific activity was found for both wt PARP and D214A-PARP overproduced in bacteria (20) (Fig. 2C), indicating that the mutation located in the NH 2terminal domain of PARP does not affect the catalytic activity of the protein as some other mutations do (28).
That the D214A mutation gave rise to a non-cleavable protein was confirmed by an in vitro cleavage assay using purified human caspase-3 and the [ 35 S]methionine-labeled PARP prepared by in vitro transcription/translation. Incubation of 35 Slabeled PARP with purified caspase-3 resulted in the time-dependent cleavage of the 116-kDa PARP into the 85-kDa fragment, similar to that observed during apoptosis. On the contrary, mutant D214A-PARP remains uncleaved even after long term incubation with purified caspase-3 (Fig. 2D). Similarly, mutant D214A-PARP was resistant to the cleavage by a cytosolic extract obtained from PARPϪ/Ϫ splenocytes (Fig. 2E) induced to apoptosis by treatment with 2 mM MNU, which is a potent trigger of apoptosis in these cells (11), indicating that mutation D214A impaired the cleavage by caspase-3 and other caspases acting during apoptosis.
Cell Death Triggered by CD95 Activation Is Delayed in Cells Expressing an Uncleavable PARP Mutant-PARP has been shown to be an important substrate for caspase-3 (29), activation of which is essential for CD95-mediated apoptosis (30). CD95 (Apo1/Fas) is a cell death-promoting receptor that belongs to the tumor necrosis factor receptor family (31) (reviewed in Ref. 32). Triggering the CD95 molecule by either agonistic antibodies or by the natural CD95 ligand induces apoptosis. CD95 receptor molecule is directly linked to caspase-8 protease, which cleaves and activates caspase-3, a protease responsible for triggering the execution phase of apoptosis (4). On the other hand, upon activation of caspase-3 by proteolysis, PARP protein is also eliminated (33). PARP cleavage is, therefore, a hallmark of CD95-dependent programmed cell death.
To test whether the presence of an uncleavable PARP mutant is functionally relevant during the cell's commitment to apoptosis, we have analyzed two different pathways to induce apoptosis in MEFs: treatment of the cells with either a CD95 receptor agonist, independent of DNA damage, or with the alkylating agent MMS, where lesions to DNA are repaired by BER and stimulate PARP activity.
PARPϪ/Ϫ MEFs were cotransfected with the ␤-galactosidase reporter plasmid together with the vectors overproducing the wt PARP or the D214A-PARP mutant or the empty vector. Apoptosis was quantified in cells expressing ␤-galactosidase either following an anti-CD95 treatment or exposure to an alkylating agent. Fig. 3A shows that 24 h following anti-CD95 treatment almost all cells were dead. Expression of wt PARP did not lead to protection against cell death. However, cells expressing the uncleavable PARP mutant showed a significant delay of about 10 h in the loss of cell viability, suggesting that PARP cleavage is necessary for the cell to achieve the apoptotic program in time. To confirm that loss of cell viability was because of activation of apoptosis, a specific inhibitor of caspase-3, benzyloxycarbonyl-DEVD-fluoromethylketone (34), was used. Preincubation of cells with this inhibitor at 200 M completely prevented anti-CD95-induced cell death (Fig. 3B) as has been previously shown (30). In contrast, PARP null cells are more sensitive to apoptosis induced by a monofunctional alkylating agent, and the overproduction of wt PARP restored the viability of PARPϪ/Ϫ cells after MMS treatment, thus demonstrating that the absence of PARP is directly responsible for the sensitization of cells to DNA alkylation. Cells transfected with the D214A-PARP mutant are also protected against the cytotoxic effects of MMS and restore the sensitivity of PARPϪ/Ϫ cells to approximately the same extent as wt PARP (Fig. 3C). Therefore, in this situation, the absence of PARP cleavage seems not to influence the recovery of cell viability.
Cytoplasmic Blebbings and Nuclear Disassembly during Apoptosis Are Delayed by the Presence of an Uncleavable PARP-To study at which step the non-cleavage of PARP interferes with the apoptotic process, we have analyzed the morphology of apoptotic PARPϪ/Ϫ MEF cells during anti-CD95 treatment. For this purpose, MEFs were cotransfected with vectors expressing the wt or the uncleavable PARP and a reporter plasmid expressing the green fluorescence protein (GFP), as described under "Materials and Methods." We established three different classes of cells expressing GFP, according to the morphology of the cytosol and nucleus: (i) intact cells (Fig. 4A, upper panels); (ii) cells with cytosolic blebbings but intact nucleus (Fig. 4A, middle panels); and (iii) cells with a fragmented/condensed nucleus (Fig. 4A, lower panels). All the cells in the third category had already lost the membrane integrity according to morphological criteria. In MEFs expressing wt PARP, 43% were intact cells, and almost all apoptotic cells had already lost the nuclear integrity (Fig. 4B). In clear contrast, cells expressing the D214A-PARP mutant were much more resistant to apoptosis; 78% were intact cells, and the fraction of apoptotic cells with nuclear disruption (4B) was low (1.6%) compared with their normal counterparts. In conclusion, the non-cleavage of PARP not only slows down the nuclear events of apoptosis but also interferes with the initial cytoplasmic events leading to cell membrane blebbing, as reflected by the elevated number of intact cells at this time. DISCUSSION Studies from different laboratories have addressed the role of PARP in the apoptotic response. We have previously shown that splenocytes isolated from PARPϪ/Ϫ mice undergo programmed cell death much more rapidly than their counterparts when treated with an alkylating agent (11). These results, together with those presented in Fig. 1, suggest that PARPϪ/Ϫ cells are exquisitely sensitive to agents whose DNA lesions are repaired by base excision but are equally sensitive to inducers of apoptosis that do not activate this pathway (12,13). To explain this specific sensitivity, we have to take into account that PARP participates in the BER pathway in association with XRCC1, DNA polymerase ␤, and ligase III (35)(36)(37) and in the organization of chromatin architecture through the posttranslational modification of histones (38,39). The reason these cells lacking PARP are much more fragile and undergo programmed cell death faster than parental cells could be explained by the accumulation of unrepaired DNA damage (because of the absence of PARP), which makes the cell unable to cycle and engage the apoptotic pathway to avoid transmission of damaged DNA to a new generation of cells. This view is supported by the fact that PARPϪ/Ϫ cells have prolonged kinetics of DNA rejoining as measured by the comet assay (40) and an increased accumulation of p53 in PARP null splenocytes (11) upon treatment with an alkylating agent. A nearly universal biochemical hallmark of apoptosis is the proteolytic cleavage and inactivation of PARP in the execution phase of cell death; moreover, in all instances where apoptosis is inhibited by whichever mechanism (Bcl-2 overexpression, use of caspase-specific inhibitors, etc.) PARP cleavage is also inhibited. An obvious question to ask is why the cell needs to get rid of this protein to bring about apoptosis. Specific degradation of chromatin is a key component of the apoptotic process. DNA fragmentation during apoptosis is produced by numerous single-stranded nicks in the linker regions of chromatin (41), and PARP interacts preferentially with single-stranded DNA breaks (42). Through this property, PARP probably facilitates the recruitment and accessibility of the BER proteins to repair damaged DNA. Thus, inactivation of PARP by apoptotic caspases might serve to a double finality: first, disable key components of the genomic surveillance (together with DNA-PK (33)) to avoid unnecessary DNA repair during chromatin degradation and second, facilitate the accessibility of endonucleases to chromatin and nuclear disintegration. In support of this, it has been shown that inactivation of PARP by antisense mRNA delivery to the cells results in an increased accessibility of micrococcal nuclease to chromatin (43). The preferential localization of PARP in the vicinity of the nuclear envelope (36) also suggests that its cleavage during apoptosis participates in nuclear disassembly and facilitates downstream events, which are temporarily retarded upon expression of the uncleavable PARP mutant. In addition, the uncleavable PARP mutant allows for the persistence of PARP activity, which might positively influence cell viability upon anti-CD95 stimulation.
The morphological consequences of expressing uncleavable substrates of caspases on apoptosis have been so far examined for lamin (44) and p21-activated kinase (45). Mutant laminexpressing cells showed no signs of chromatin condensation or nuclear shrinkage and blocking p21-activated kinase cleavage/ activation during CD95-induced apoptosis inhibited membrane cell blebbing, whereas the nuclear modifications were unaffected; prevention of PARP cleavage has consequences at the cytosolic and nuclear level. We have observed in vivo both a delay in cell collapse/cytoplasmic blebbing formation and in chromatin condensation in cells expressing D214A-PARP following anti-CD95 treatment. These effects on cell morphology are strikingly similar to those observed following inhibition of caspase proteases with specific peptides. These inhibitors block only part of the classical apoptotic program such as the external display of phosphatidylserine and nuclear disintegration but not of cell blebbing (46). Obviously, preventing PARP cleavage is not sufficient to inhibit apoptosis as subsequent events take over and bring about cell death. Therefore, inhibiting PARP cleavage prevents loss of cell viability until perhaps proteolytic attack of other substrates promotes cell degradation leading to cell death. This attenuation of apoptosis upon expressing uncleavable PARP implicates PARP as a key substrate targeted for degradation relatively early in the apoptotic process.
In summary, the results presented in this study suggest that PARP is not just a "death substrate" but plays a positive role allowing the cell not to undergo apoptosis when the cell is able to repair the damaged DNA. Its cleavage might be a sign that the cell is unable to cope with a saturating DNA injury, arising from either apoptosis-derived chromatin fragmentation or an externally infringed genotoxic insult.