Cleavage of Automodified Poly(ADP-ribose) Polymerase during Apoptosis

The abundant nuclear enzyme poly(ADP-ribose) polymerase (PARP) synthesizes poly(ADP-ribose) in response to DNA strand breaks. During almost all forms of apoptosis, PARP is cleaved by caspases, suggesting the crucial role of its inactivation. A few studies have also reported a stimulation of PARP during apoptosis. However, the role of PARP stimulation and cleavage during this cell death process remains poorly understood. Here, we measured the stimulation of endogenous poly(ADP-ribose) synthesis during VP-16-induced apoptosis in HL60 cells and found that PARP was cleaved by caspases at the time of its poly(ADP-ribosyl)ation. In vitro experiments showed that PARP cleavage by caspase-7, but not by caspase-3, was stimulated by its automodification by long and branched poly(ADP-ribose). Consistently, caspase-7 exhibited an affinity for poly(ADP-ribose), whereas caspase-3 did not. In addition, caspase-7 was activated and accumulated in the nucleus of HL60 cells in response to the VP-16 treatment. Furthermore, caspase-7 activation was concommitant with PARP cleavage in the caspase-3-deficient cell line MCF-7 in response to staurosporine treatment. These results strongly suggest that, in vivo, it is caspase-7 that is responsible for PARP cleavage and that poly(ADP-ribosyl)ation of PARP accelerates its proteolysis. Cleavage of the active form of caspase substrates could be a general feature of the apoptotic process, ensuring the rapid inactivation of stress signaling proteins.

Apoptosis is a conserved mechanism of cell death controlling the development and homeostasis of multicellular organisms. In the last few years, a family of cysteine proteases named caspases, highly related to the interleukin-1␤-converting enzyme and the proapoptotic CED-3 gene of Caenorhabditis elegans, have emerged as important mediators of the apoptotic process (1,2). All caspases exist in the cytoplasm in the form of inactive proenzymes that are processed to a large and a small subunit to form the active enzyme (1,2).
Three of these caspases have been implicated in the execution phase of apoptosis (1,2) and are shown to cleave specific substrates as the cell begins to present the characteristic morphological changes of apoptosis (nuclear condensation, cell blebbing, and formation of the apoptotic bodies). Caspase-3 was shown to cleave a wide range of cytoplasmic and nuclear proteins (1), which suggests an important role for this protease in apoptosis. Strikingly, caspase-3 knock-out mice, while showing major defects of apoptosis in the brain, seemed to have normal apoptotic responses otherwise (3). Caspase-7 is highly related to caspase-3 and shows the same synthetic substrate specificity in vitro (4). It is believed that caspase-7 cleaves caspase-3 substrates in caspase-3 knock-out mice. A recent caspase-3 knock-out report suggested that caspase-3 and -7 have distinct but possibly overlapping roles in apoptosis, because some caspase substrates are not cleaved in the knock-out cells, but the overall process is not altered (5). The third execution phase caspase, caspase-6, is responsible for the cleavage of the lamins (6).
Poly(ADP-ribose) polymerase (PARP) 1 synthesizes poly-(ADP-ribose) from NAD in response to DNA strand breaks and is involved in many genomic processes including DNA base excision repair (7), DNA replication (8) and transcription (9). PARP is thought, along with DNA protein kinase, ATM, and p53, to be part of the cascade signaling DNA damage in the cell (10). PARP was one of the first substrates that was shown to be cleaved by caspases (11,12). Although almost all caspases, including caspase-1, can cleave PARP in vitro (2,13), it is most likely that caspase-3 and -7 are responsible for the in vivo processing of PARP to its apoptotic 24-and 89-kDa fragments (1,2). The 89-kDa fragment, carrying the automodification domain and the catalytic domain of the enzyme, retains only a basal activity because it loses its capacity to bind to damaged DNA (11). The 24-kDa fragment, which contains the two zinc fingers responsible for the DNA binding of PARP, is very likely to act as a transdominant inhibitor of active PARP, similar to the inhibition observed with the complete 46-kDa DNA-binding domain (14). Indeed, Smulson et al. (15) have shown recently that the 24-kDa apoptotic fragment of PARP irreversibly binds to DNA breaks.
Here, we have studied the temporal association between PARP activation, its cleavage, and the appearance of the DNA ladder during the course of VP-16-induced apoptosis in HL60 cells. In these cells, we show that PARP is cleaved when it is poly(ADP-ribosyl)ated. In vitro analysis of the cleavage kinetics of PARP as a function of its state of automodification reveals that PARP automodification stimulates its cleavage by caspase-7 but not by caspase-3. We also show that caspase-7 is activated and accumulates in the nucleus in response to the VP-16 treatment in HL60 cells.
Cell Culture and Induction of Apoptosis-HL60 human leukemia cells and MCF-7 cells were maintained in RPMI medium supplemented with 10% heat-inactivated fetal bovine serum and 2 mM glutamine for MCF-7 cells. Apoptosis was induced with 68 M VP-16 and 1 M staurosporine for HL60 and MCF-7, respectively. For the Western blots, cells were centrifuged for 5 min at 830 ϫ g, washed with HEPES-saline buffer (140 mM NaCl, 7 mM KCl, 6 mM glucose, 10 mM HEPES, pH 7.4), and resuspended in reducing loading buffer. The cells were sonicated prior to loading on a 8% polyacrylamide gel, resolved, and transferred on a nitrocellulose membrane. PARP cleavage was detected using the monoclonal antibody C II 10, which recognizes full-length PARP and its 89-kDa apoptotic fragment, as described by Lazebnik et al. (17). The 116-and 89-kDa bands were quantified using a cooled CCD camera equipped with a Chemi Imager 4000, and the data were analyzed with the Digital Imaging Analysis Systems (Alpha Innotech Inc.). The apoptotic internucleosomal cleavage of the DNA was performed as described by McGahon et al. (18). For the separation of nucleus from the cytoplasm, cells were resuspended in 10 mM Tris, pH 7.4, 1 mM EDTA, 300 mM sucrose, 2 mM ␤-mercaptoethanol, 0.1% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride, and antiprotease mixture (Roche Molecular Biochemicals). Cells were then homogenized with a Dounce homogeneizer and centrifuged at 10,000 ϫ g for 3 min. Pellets containing the nucleus were redissolved in the same buffer. Cytoplasm and nucleus were then diluted in reducing loading buffer prior to loading on a 15% polyacrylamide gel and submitted to Western blotting with an anti-caspase-7 p19 subunit antibody.
Analysis of Cellular NAD and Poly(ADP-ribose) Levels-Purification of NAD and poly(ADP-ribose) from cells and analysis of the NAD levels were done essentially as described by Shah et al. (19), except that the formic acid precipitation step, including addition of bovine serum albumin and cold poly(ADP-ribose), was omitted. Poly(ADP-ribose) was measured, using a dot blot technique with the LP96 -10 antibody as described previously (20). To determine the sizes of the polymers in the apoptotic cells, 5 ϫ 10 7 HL60 cells were treated for 3 h with 68 M VP-16. After purification, the polymer was resolved on a 20% polyacrylamide gel, transferred on a Hybond N ϩ membrane, and revealed using the LP96 -10 antibody (21).
Enzyme Purification-PARP was purified to homogeneity from bovine thymus essentially as described by Zahradka and Ebisuzaki (22) and modified by Huletsky et al. (23). The specific activity of the purified enzyme was 1341 units/mg protein. Human recombinant caspase-3 and -7 were purified to homogeneity from overexpressing bacteria, as described previously (6). Poly(ADP-ribose) glycohydrolase was purified up to the polyethylene glycol precipitation step according to Thomassin et al. (24).
Poly(ADP-ribosyl)ation and PARP Cleavage Assay-Bovine PARP (150 ng/reaction) was poly(ADP-ribosyl)ated in a modified caspase-3 assay buffer (50 mM HEPES, pH 7.4, 100 mM NaCl, 10% sucrose, and 10 mM dithiothreitol) (12,25) in the absence or the presence of 6 g/ml activated calf thymus DNA and 200 M of NAD for 15 min at 25°C. The automodification reaction was stopped by the addition of 100 M DHQ (26). Upon addition of 0.1% CHAPS and the indicated amount of caspase-3 or -7, the substrate was digested for specific times at 37°C. The caspase-3-like protease inhibitor DEVD-CHO was then added to a final concentration of 100 nM to stop the digestion, followed by the addition of 0.25 units of glycohydrolase for 30 min at 25°C to remove the polymer. The reaction was stopped by adding an equal volume of reducing loading buffer and heating at 65°C for 15 min. The apoptotic fragment was separated from the full-length PARP and visualized as described above. To produce polymers of different lengths, PARP was automodified for 1 min in the presence of activated DNA and the indicated concentrations of [ 32 P]NAD. Poly(ADP-ribose) lengths were then determined as described (25). One unit of caspase activity was defined as the amount of caspase necessary to produce one pmol of AFC/min from the substrate DEVD-AFC, in the same reaction conditions that for PARP.

Noncovalent Interactions between Caspases and Poly(ADP-ribose): Polymer Blot Assay-
The experiments were carried out essentially as described by Althaus et al. (27). Caspases and total histones were resolved on a 15% acrylamide gel and transferred on a nitrocellulose membrane. The membrane was washed three times with TBS-T (10 mM Tris, pH 7.4, 150 mM NaCl, and 0.05% Tween 20), blocked with 3% bovine serum albumin in TBS-T, and washed again three times with TBS-T prior to incubation with the indicated quantities of [ 32 P]poly-(ADP-ribose) for 1 h. After washing the membranes with TBS-T, the radioactive polymer still bound to the membrane was analyzed by autoradiography. The same experiments were also carried out with radiolabeled genomic DNA (Sigma). The DNA was radiolabeled by [ 32 P]dCTP using T4 DNA polymerase, as described by Legault et al. (28).

PARP Cleavage by Caspases at the Time of Its Activation-
During apoptosis, a drop in cellular NAD was observed in some cellular systems. This NAD depletion was partially inhibited by the PARP inhibitor 3-aminobenzamide, thus implicating PARP activation in apoptosis (11,29,30). However, enzymes other than PARP have recently been shown to decrease NAD levels following DNA damage (31). Furthermore, studies on polymer levels during apoptosis have relied on permeabilized cells, which do not allow an accurate measurement of endogenous poly(ADP-ribose) levels. These techniques have thus failed to evaluate the poly(ADP-ribose) metabolism in intact cells. To determine the extent of PARP activation during apoptosis and the relation to its cleavage, HL60 cells were treated with the topoisomerase II inhibitor, VP-16. PARP cleavage occurred 3 h after treatment, concomitantly with a transient polymer synthesis (Fig. 1). The 89-kDa apoptotic fragment of PARP was poly(ADP-ribosyl)ated as recognized by the antipolymer antibody, whereas full-length PARP was not (Fig. 1B). This result suggests that automodified PARP is a natural substrate for caspases. The band at 66 kDa is likely to be the bovine serum albumin from the culture medium, which is recognized by the antibody.
An immunological method recently developed in our laboratory (20) was then used to directly measure the endogenous polymer levels. This technique relies on affinity chromatography purification of the polymers followed by their specific immunodetection. PARP cleavage was measured in parallel using Western blotting with C II 10 antibody. The time course experi- ment presented in Fig. 2A shows the polymer peak at 3 h after the addition of VP-16, a time at which most PARP cleavage also occurred. This result further supports the notion that PARP is cleaved once it is activated. NAD was also measured during the course of the treatment to determine the relationship between polymer synthesis and NAD levels. A cellular drop in NAD occurred in apoptotic cells at the time at which the polymer peak was detected (Fig. 2B). However, NAD levels continued to drop after the polymer peak, even though more than 50% of PARP was cleaved. A concentration of 1 mM of the PARP inhibitor DHQ, known to completely inhibit PARP in vivo (26), only partially prevented the loss of NAD (Fig. 2B). These results indicate that the drop in NAD does not result only from PARP activation. During this time, 100% of the cells were still viable as determined by the exclusion of trypan blue dye.
Because PARP is activated in response to DNA strand breaks, we expected the internucleosomal degradation of the DNA in the apoptotic cells to be responsible for the activation of PARP. Thus, the extent of DNA degradation in HL60 cells treated with VP-16 was measured. As shown in Fig. 2C, the DNA ladder appeared at the same time point as the poly(ADPribose) peak, strongly suggesting that PARP activation was caused by these strand breaks.
Poly(ADP-ribosyl)ation of PARP Stimulates Its Cleavage by Caspase-7, but Not by Caspase-3-The results obtained from the apoptotic cells suggested that PARP cleavage occurs in vivo when the enzyme is poly(ADP-ribosyl)ated in response to DNA strand breaks. To define more precisely the effect of PARP automodification on caspase activity, we investigated the in vitro kinetics of automodified PARP cleavage by purified caspase-3 and -7. PARP was automodified in the presence of NAD and activated DNA and then subjected to cleavage assay by caspase-3 and -7. Because highly modified PARP cannot be resolved by gel electrophoresis (23), poly(ADP-ribose) glycohydrolase, the enzyme responsible for poly(ADP-ribose) catabolism, was added to the cleaved poly(ADP-ribosyl)ated PARP to remove the polymers that impair its mobility on gel. Under these conditions, almost all of the polymers could be removed (data not shown), ensuring a precise determination of the PARP cleavage kinetics. As control, PARP was digested under the same conditions except that the NAD and DNA were omitted, because the binding of PARP on DNA delays its cleavage (25). Because the specific activity may vary between enzyme preparations, the same enzymatic activity was used for each caspase. The results shown in Fig. 3 indicates that caspase-7 cleaves PARP with a greater efficiency than caspase-3. Furthermore, they demonstrate that PARP automodification greatly stimulates its proteolytic cleavage by caspase-7 but not by caspase-3. Using higher amounts of caspase-3, PARP was cleaved efficiently independently of its state of automodification (data not shown). Thus, PARP cleavage by caspase-7 but not by caspase-3 is specifically increased by the poly(ADPribosyl)ated substrate. The effect of free poly(ADP-ribose) on caspase activity was also measured, using the synthetic substrate DEVD-pNA to ensure that the polymers would not interact with the caspase substrate. The presence of 10 M free poly(ADP-ribose) had no effect either on caspase-3 or on caspase-7 activity (data not shown). Thus, the stimulation of PARP cleavage by caspase-7 is not mediated by a direct effect of the free polymer on caspases. Other experiments showed similar caspase activity in the absence or in the presence of NAD, activated DNA and DHQ (data not shown). None of these compounds had an effect on caspase activity at the concentrations found in the PARP automodification reaction mixture.

FIG. 2. Poly(ADP-ribose) metabolism in apoptotic HL60 cells.
Cells were treated with 68 M VP-16 for the indicated times. A, poly-(ADP-ribose) synthesis and PARP cleavage. Poly(ADP-ribose) was purified by affinity chromatography and measured using an antibody specific to the polymer (LP96 -10) (q). PARP cleavage was measured in parallel using C II  To determine the length of poly(ADP-ribose) that stimulates caspase-7, PARP was automodified in presence of DNA and various concentrations of NAD, after which the automodified substrate was digested with caspase-7 for 20 min. Automodification with polymers up to 20 residues had only minimal effect on caspase-7 activity (Fig. 4, A and B). The presence of DNA, which has been shown to delay PARP cleavage by caspase-3 (25), could explain the absence of cleavage at concentrations of NAD below 0.1 M. However, PARP automodification with the long and branched polymers produced at 200 M NAD resulted in a strong stimulation of caspase-7 (Fig. 4, A-C). The length of the polymers was then measured in the apoptotic cells using a new technique consisting in the separation of the polymers on a polyacrylamide gel, followed by its transfer on a positively charged membrane and detection using an anti-poly(ADP-ribose) antibody (LP-96 -10). As shown in Fig. 4D, long and branched polymers are produced in these cells during apoptosis. These results indicate that principally long and branched polymers, which are produced in vivo, stimulate caspase-7mediated cleavage of PARP.
Caspase-7, but Not Caspase-3, Has Affinity for Poly(ADPribose)-The polymer blot assay described by Althaus et al. (27) was used to determine caspase-3 and -7 affinities for poly(ADPribose). Using 0.1 M of free [ 32 P]poly(ADP-ribose), we detected binding on the p19 subunit of caspase-7 as well as on total histones (Fig. 5A). A weak signal was observed from the p10 subunit of caspase-7. No binding was observed for caspase-3. The same experiment was repeated with 1 M of [ 32 P]poly-(ADP-ribose). At this concentration, the two subunits of caspase-7 appeared to bind the polymer almost equally, whereas in the case of caspase-3, only p17 showed a weak binding to the polymer (Fig. 5B). The binding of [ 32 P]poly(ADPribose) on native caspase-7 and its weak binding on native caspase-3 were also detected by dot blot (data not shown). Experiments using radiolabeled genomic DNA instead of poly-(ADP-ribose) were also performed to determine whether the binding was specific. We were able to detect the binding of DNA to histones but not to caspase-7 (Fig. 5C), indicating that the binding is not due to nonspecific ionic interactions.
Evidence for in Vivo Cleavage of PARP by Caspase-7-Because our results strongly suggest that caspase-7 is responsible for the cleavage of activated PARP in vivo, we verified the activation of caspase-7 in HL60 cells in response to VP-16 treatment. As shown in Fig. 6A, caspase-7 was effectively processed to its active form in the apoptotic cells as seen by the apparition of its p19 subunit. This activation occurred 3 h after the treatment (data not shown), the time at which PARP was cleaved, suggesting that caspase-7 could effectively cleave PARP in vivo. Furthermore, there was active caspase-7 located in the nucleus, where it could cleave and inactivate poly(ADPribosyl)ated PARP (Fig. 6A). Caspase-7 was also activated at the time of PARP cleavage in MCF-7 cells treated with 1 M staurosporine (Fig. 6, B and C). Because MCF-7 cells lack caspase-3 (32), these results further support that caspase-7 cleaves PARP in vivo.

DISCUSSION
During apoptosis, NAD depletion with partial restoration in the presence of PARP inhibitors has been reported (11,29,30). PARP activation has also been shown in the early (29,33,34) as well as in later stages of apoptosis (29,35,36). However, poly(ADP-ribose) levels resulting from the activation of PARP were measured using NAD incorporation in permeabilized cells. Major draw backs of this technique are the inability to measure polymer levels precisely and the presence of artificially high amounts of polymer, even in control cells (35). Thus, in this study, endogenous poly(ADP-ribose) was measured directly in intact HL60 cells to follow the activation of PARP during the course of VP-16-induced apoptosis. Using an immunological method for the quantification of polymer, we found that the activation of PARP coincided with the appearance of the DNA ladder. Furthermore, this endogenous polymer peak was sharp when compared with the one found using permeabilized cells (35,36), suggesting that PARP was activated only at the time of the appearance of the DNA ladder and not before. This suggests an activation of PARP by DNA strand breaks generated during internucleosomal DNA degradation, as proposed by Nosseri et al. (29). Activation of PARP in response to apoptosis inducers like anti-Fas antibody or campthothecin has also been shown by Western blot using an antibody against poly(ADP-ribose) (34).
During the course of VP-16-induced apoptosis of HL60 cells, PARP activation was low compared with the amount of DNA damage: 50 M of the alkylating agent 1-methyl-3-nitro-1-ni- trosoguanidine produces 144 pmol of poly(ADP-ribose)/10 8 cells (unpublished results), whereas we measured 18 pmol/10 8 apoptotic cells. Two mechanisms could account for this low activation. First, PARP is cleaved and inactivated by caspases at the time of its activation ( Figs. 1 and 2A). Second, the 24-kDa fragment of PARP generated by the caspases could act as a transdominant inhibitor of full-length PARP. This apoptotic fragment retains its ability to bind to degraded or fragmented DNA by the virtue of its zinc fingers. It cannot, however, be dislodged from the DNA breaks (15). A similar mechanism has been shown for the full-length 46-kDa DNA-binding domain (14). Our results also show that neither PARP activation or the residual activity of the 89-kDa apoptotic fragment are responsible for the apoptotic NAD depletion, because NAD levels dropped only by 20% at the time of PARP activation and that the PARP inhibitor DHQ did not prevent the NAD depletion occurring after 6 h of treatment. However, these results do not exclude the possibility that the absence of PARP cleavage could result in its overactivation and massive NAD depletion that could cause necrotic cell death (37).
The kinetics of PARP stimulation reported here indicate that it occurs at the same time as that of its cleavage. This is supported by the fact that the 89-kDa apoptotic fragment of PARP, which retains only a basal activity (11), is automodified to a larger extent than the uncleaved full-length PARP (Fig.  1B). These results differ from those of Simbulan-Rosenthal et al. (34) who found that caspases were activated after the polymer peak. They used an in vitro translated PARP with cytosolic extracts, a system in which the kinetics could differ from ours, because we measured the cleavage of endogenous PARP. However, because the apoptotic DNase (caspase-activated DNase) is activated by caspases (38), it is likely that caspase activation occurs before PARP stimulation.
Contrary to proteolysis of PARP by papain, which is inhibited by its automodification, 2 its cleavage by caspase-7 is stimulated. We found that long and branched polymers attached to PARP are capable of stimulating its cleavage by caspase-7, possibly by increasing the affinity of the latter for PARP. In addition, we have demonstrated directly the difference in the specificity of the two caspases toward PARP, in vitro. We found that caspase-7 was more effective in cleaving PARP than caspase-3 (Fig. 3), despite a similar K m for the synthetic substrate DEVD-pNA (11 M and 12 M for caspase-3 and -7, respectively) (4). Because the K cat values of caspase-3 and -7 for DEVD-pNA are 10 s Ϫ1 and 1.3 s Ϫ1 , respectively, 3 and because PARP is cleaved faster by caspase-7 than by caspase-3, our results suggest that caspase-7 has a much higher affinity for PARP than caspase-3, although one study on caspase specificity in vitro, done using the 46-kDa DBD that lacks the automodification sites, showed a preferential PARP cleavage by caspase-3 (13). However, using different peptide inhibitors of caspases, Takahashi et al. (39) observed that caspase-7 had the highest specificity for the PARP cleavage site (GDEVD2GIDEV). They also showed that a caspase-3 specific inhibitor peptide, DMQD-CHO, prevented the cleavage of some of the caspase-3 substrates but not of PARP (40). In addition, PARP cleavage was observed in caspase-3 knock-out mice (3,5). PARP was also cleaved in the caspase-3 defective cell line MCF-7, cleavage that paralleled caspase-7 activation (Fig. 6, B  and C). These results all lend support to the notion that caspase-7 is the most efficient caspase for PARP cleavage. They also indicate that the tetrapeptide cleavage site is not the only important factor for efficient cleavage of caspase substrates. Furthermore, the poly(ADP-ribosyl)ation of PARP appears to target caspase-7 to the activated PARP to ensure its rapid inactivation.
To maintain cellular integrity and homeostasis, a set of proteins act as stress sensors to promote a cellular response to situations such as DNA damage or heat shock. Because apoptosis results in the complete dismantling of the cell, such stress signaling proteins may be stimulated and interfere with the apoptotic process. However, this could be prevented by the rapid inactivation of these proteins by caspases. Because DNA is rapidly degraded in apoptotic cells, inhibition of DNA repair is primordial in this view. Because PARP can act as a DNA break sensor to signal DNA damage, the rapid cleavage of the activated PARP molecules would ensure the interruption of this signal. Interestingly, similar reasoning could apply to another DNA break sensing molecule, DNA protein kinase. It has been reported that the active DNA protein kinase holoenzyme, bound to DNA breaks, would be the physiological target of caspase-3 (41). Such a preferential cleavage would permit a much more rapid execution phase of apoptosis, because only the active form of these stress sensing proteins would need to be inactivated before dismantling of the cell, although all the molecules would eventually be cleaved independently on their state of activation. The failure to cleave those apoptotic substrates would not prevent the death of the cell but could retard it sufficiently to render it dangerous for the organism. Such an inhibition of apoptotic morphology has recently been shown for PARP (42). were resolved on a 15% polyacrylamide gel and submitted to Western blotting using a polyclonal antibody against caspase-7 p19. PARP cleavage was also visualized in the MCF-7 cells using C II 10 antibody (C).