Active caspase-8 translocates into the nucleus of apoptotic cells to inactivate poly(ADP-ribose) polymerase-2.

Caspase-8 is the prototypic initiator of the death domain receptor pathway of apoptosis. Here, we report that caspase-8 not only triggers and amplifies the apoptotic process at cytoplasmic sites but can also act as an executioner at nuclear levels. In a murine model of acute ischemia, caspase-8 is relocated into the nucleus of apoptotic neurons, where it cleaves PARP-2, a member of the poly(ADP-ribose) polymerase family involved in DNA repair. As indicated by site-directed mutagenesis, PARP-2 cleavage occurs preferentially at the LQMD sequence mapped between the DNA binding and the catalytic domains of the protein. This is close to the cleavage sequence found in Bid, the cytoplasmic target of caspase-8. Activity assays confirm that cleavage of PARP-2 results in inactivation of its poly(ADP-ribosylation) property, proportional to the efficiency of the cleavage. Our findings add to the complexity of proteolytic caspase networks by demonstrating that caspase-8 is in turn an initiator, amplifier, and effector caspase.

Caspases are aspartate-specific cysteine proteases that are instrumental in the apoptotic process (1)(2)(3). They exist in all nucleated cells as inactive zymogens and are activated by proteolytic cleavage. Caspases are categorized into initiators and effectors on the basis of their position in apoptotic proteolytic cascades (4,5). Initiator caspases trigger and amplify apoptosis by activating effector caspases and proapoptotic factors. Effector caspases act at cytoplasmic sites, and eventually at other organelles, including the nucleus, to execute the proteolytic program that finalizes cell destruction (6 -8).
Caspase-8 (Flice/MACH/mch 5) is the prototypical caspase of so-called death receptor pathways of apoptosis, activated by ligand binding to members of the tumor necrosis factor-␣ (TNF-␣) 1 receptor superfamily (9,10). The interaction of Fas/ Apo-1, TNF receptor, or TNF-related apoptosis-inducing ligand receptor with their ligands induces receptor trimerization and the association of the death domain of the receptor with cyto-plasmic adapter proteins (11). This results in the recruitment, cleavage, and activation of pro-caspase-8, -2, and -10 (12,13). Active caspase-8 propagates the apoptotic signal by activating downstream effector caspases, such as caspase-3 and -7 (14). Besides this direct initiator role, caspase-8 amplifies the proteolytic cascade by cleaving the cytoplasmic factor Bid (15), a proapoptotic member of the Bcl-2 family. Bid interacts with Bax to initiate the so-called "mitochondrial pathway" of apoptosis, mediated by activation of caspase-9 in the apoptosome (7). Caspase-8 therefore contributes to cell death both by direct activation of effector caspases and by the initiation of another major cascade of the apoptotic process.
Caspase-8 was traditionally considered a strictly cytoplasmic protein, activated in the cytoplasm and cleaving cytoplasmic substrates, but this simple schema has been challenged by two recent studies. First, measurable levels of pro-caspase-8 have been observed in the nucleus of cultured fibroblasts and striatal neurons (16). Second, active caspase-8 was shown to bind aggregates of truncated hungtintin in the nucleus of cultured neurons overexpressing the pathogen protein (17). Caspase-8 activation is involved in a number of neurodegenerative pathologies (17)(18)(19), including stroke (17,18), and in Fas-or TNF␣induced apoptosis in virtually all cell types (20). We therefore investigated the possibility of a nuclear role for active caspase-8 in focal cerebral ischemia, a paradigm known to trigger massive cytoplasmic activation of caspase-8 in apoptotic neurons (18,21). Our results reveal that nuclear caspase-8 targets and inactivates PARP-2, a recently identified member of the poly(ADP-ribosyl) polymerase (PARP) family, which plays an active role in base excision repair mechanisms (22). Thus, caspase-8 has distinct cytoplasmic and nuclear roles in the processing of ischemia-induced apoptotic cell death.

EXPERIMENTAL PROCEDURES
Middle Cerebral Artery Occlusion Model-Studies were performed on 8-week-old male C57Bl/6 mice (25-30 g, Janvier, Le genest-St-Isles, France). Permanent focal cerebral ischemia was performed by electrocoagulation and section of the left middle cerebral artery (MCAO) according to Ref. 23 under 4% chloral hydrate anesthesia. Sham-operated animals were subjected to the same procedure without MCA electrocoagulation and section. Animals were sacrificed at 1, 3, or 6 h post-MCAO.
Immunofluorescence and Confocal Microscopy-Immunofluorescence analysis was performed as described on perfusion-fixed brain sections (18). Primary antibodies were rabbit anti-pro-caspase-8 and anti-active caspase-8 (SK441 and SK440, respectively, GlaxoSmithKline, (21)), and rabbit anti-C-terminal-PARP-2 (YUC (22)). DNA was visualized by staining with Hoescht 33258 (Sigma) for light microscopy or with Yo-Pro-3 (Molecular Probes) for confocal microscopy. Images were recorded on a Zeiss LSM confocal microscope or an Axioplan II microscope (Zeiss) equipped with a CoolScan (Photometrics) video camera. * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Caspase Activity Assay and Cleavage Studies-Protein fractions were isolated from MCA-supplied regions dissected from the hemicortex of ischemized or sham animals. For nuclear protein extraction, tissues were treated as described in Ref. 24. Caspase-8 catalytic activity was measured with the synthetic substrate acetyl-Ile-Glu-Thr-Asp-7-amino-4-trifluoromethylcoumarin (ac-IETD-AFC, Biomol Research Laboratories) as described (18). For PARP cleavage studies, 25 ng of human PARP-1 or mouse PARP-2 were incubated with increasing quantities of purified recombinant human (rh) active caspase-8 (Upstate Biotechnology) in the caspase assay buffer (18) for 2 h at 37°C. Controls were performed with 1 unit of rh-caspase-9 (Biomol Research Laboratories). The entire assay sample was then resolved on an SDS-PAGE gel. For inhibition studies, 200 nM c-IETD-fluoro-methyl-ketone (ac-IETD-FMK) were preincubated with the active caspase-8 for 30 min at room temperature.
Glutathione S-Transferase (GST) Pull-down Assay and Immunoblot Analysis-To test PARP-2 cleavage in vivo, rh-caspase-8 (1 unit) or 100 g of nuclear proteins from cortical tissues were incubated with 5 l of GST⅐PARP-2 fusion proteins (22) attached to glutathione-agarose beads in caspase assay buffer for 2 h at 37°C. The beads were washed extensively and collected by centrifugation, and the supernatant was subjected to immunoblotting for GST (anti-GST, Sigma). For quantification of in vivo levels of active caspase-8 and PARP-2, membranes containing 25 g of proteins per sample were probed with primary antibodies (anti-caspase-8, Santa-Cruz Biotechnology; anti-PARP-2, YUC antibody). Antigens were revealed by enhanced chemiluminescence reaction (Amersham Biosciences, ECLϩ). Densitometric values expressed in arbitrary units were estimated by the ImageQuant software (Molecular Dynamics) and were normalized to the corresponding Coomassie Blue control densitometric values (also estimated by Image-Quant) to avoid variations due to inequal loading of the sample, before statistical comparison. Multiple film exposures were used to verify the linearity of the samples analyzed and avoid saturation of the film.
Microsequencing and Site-directed Mutagenesis-For microsequencing of the C-terminal fragment of PARP-2, an amplification scale of the cleavage reaction was resolved with SDS-PAGE and transferred onto a polyvinylidene difluoride membrane. The membrane was stained with Coomassie Blue, and the 37-kDa band was cut out and was subjected to microsequencing at the custom service of INSERM U184 (Strasbourg, France). For site-directed mutagenesis, mPARP-2 cDNA was subcloned between the SmaI and the NotI sites of the PBS-KS vector (Stratagene). Aspartic acid residues at positions 119 and 186 were individually mutated using the QuikChange site-directed mutagenesis kit (Stratagene) following the manufacturer's instructions. Mutated sites were verified by DNA sequencing. The corresponding proteins were produced using the TNT-coupled reticulocyte lysate system (Promega) with T3 as the polymerase. Mutated proteins were then subjected to the PARP-2 cleavage assay as described above.
PARP Activity Assay-PARP-2 activity assay was performed according to Ref. 22. Five hundred ng of native or caspase-8-cleaved mPARP-2 were incubated for 10 min at room temperature in 100 l of assay buffer consisting of 50 mM Tris-HCl, pH 8.0, 4 mM MgCl 2 , 10 mM dithiothreitol, 1 g of calf thymus DNA pretreated with DNase I, 200 M AND, and 400 M [␣-32 P]NAD. The reaction was stopped by addition of 5% trichloroacetic acid containing 1% inorganic phosphate. The samples were filtered, the acid-insoluble radioactive residue was washed three times in the same solution and once in 95% ethanol, and radioactivity was measured with a Wallac 1409 liquid scintillation counter. The efficiency of the cleavage was assessed by quantitative immunoblotting.

RESULTS
Active Caspase-8 Is Translocated to the Nucleus during Apoptotic Cell Death-We first examined the presence of caspase-8 activity in the nucleus of apoptotic neuronal cells after MCAO by using immunofluorescence and confocal analysis performed with antibodies specific for either the pro-caspase-8 or the active protease (21). Active caspase-8 was observed at 3 h after ischemia in the nucleus of cortical neurons with condensed chromatin, localized in the ischemic core (Fig. 1a). Procaspase-8 was expressed in both the core and penumbra and localized exclusively in the cytoplasm, as confirmed by confocal microscopy (Fig. 1b).
The proteolytic activity of nuclear protein extracts from brain tissues was then measured at various time intervals following MCAO. Cleavage of IETD-AFC was significantly in-creased at 3 and 6 h post-MCAO (Fig. 1c). This indicates the presence of a caspase-8-like proteolytic activity in the nuclear fraction of apoptotic neurons. To confirm the participation of caspase-8 in this cleavage, nuclear levels of processed (p18) caspase-8 were estimated by immunoblotting (Fig. 1d). Densitometric quantification of the band (arbitrary units) revealed a 2-fold increase of p18 levels between sham (375 Ϯ 2.91, n ϭ 4 animals) and 3 h (699 Ϯ 62, n ϭ 4 animals) or 6 h (693 Ϯ 106, n ϭ 4 animals) post-MCAO. Levels were significantly increased only at 3 h post-MCAO (p Ͻ 0.05, one-factor analysis of variance with Student's t test). The high level of processed caspase-8 protein levels in the sham specimen may be explained by the fact that the sham surgery (anesthesia and surgery without occlusion of the artery) increases the activity of a number of caspases, including caspase-8, when compared with naive animals (18). Active caspase-8 levels increased with a time course parallel to the one determined by activity assays. The large discrepancy between the active caspase-8 content and the protease activity on IETD-AFC is in accordance with the catalytic properties of the caspase proteases on their substrates (25). Taken together, these studies provide evidence that caspase-8 activity is induced in the nucleus of ischemic neurons.
PARP-2 Is a Potential Target for Nuclear Caspase-8 -We then proceeded to find potential targets of nuclear caspase-8 by exploring members of the PARP family, which display crucial functions in apoptosis-induced DNA damage (26). PARP-1 is a nuclear substrate of caspases-3 and -7 (27,28), and cleavage of PARP-1 by caspase-3 is critical to ischemic cell death (29,30). Among PARP family members, PARP-2 participates along with PARP-1 in the DNA repair process by poly(ADP-ribosylation) of a number of substrates (31). To test the hypothesis that the PARPs are substrates of nuclear caspase-8, we first confirmed that PARP-2 was present in the nucleus of ischemic neurons (Fig. 2a) as described for other cell types (22). Confocal analysis of cells double-labeled with anti-PARP-2 and anti-caspase-8 antibodies showed that nuclear active caspase-8 was co-localized with PARP-2 in neurons at 3 h post-MCAO (Fig. 2b).
To investigate whether PARP-2 cleavage occurs during ischemia, we examined the expression pattern of the protein at 1, 3, and 6 h following arterial occlusion. Analysis of PARP-2 levels by immunoblotting was performed on nuclear protein extracts using the YUC antibody raised against the catalytic domain of the protein (22). Levels of a C-terminal fragment with an apparent molecular mass of 37 kDa increased as a function of time post-MCAO (Fig. 2c, d). This increase kinetically coincided with the activation of caspase-8 in the nucleus, suggesting a relationship between the two events.
mPARP-2 Is a Substrate of Caspase-8 -To determine whether the PARPs are substrates of caspase-8, we tested the ability of active rh-caspase-8 to cleave recombinant hPARP-1 and mPARP-2 in vitro. Both PARPs were cleaved in vitro. However, cleavage of PARP-1 by caspase-8 generated a fragment of 48 kDa (Fig. 3a), which differs from the 89-kDa fragment reported in vivo (32,33). In contrast, mPARP-2 was cleaved by rh-caspase-8 to generate a C-terminal fragment with the same electrophoretic mobility as the fragment produced in vivo (Fig. 3b). Since the recombinant protease was produced in Escherichia coli, potential contamination by endogenous bacterial proteases was excluded by performing the same assay in the presence of the caspase-8 inhibitor c-IETD-FMK (Fig. 3b). Inhibition of caspase-8 totally abolished the fragment. No cleavage of PARP-2 was observed with 1 unit of caspase-9 (Fig. 3b).
We next determined whether caspase-8 was an effector of PARP-2 cleavage in vivo. A cell-free system based on nuclear protein extracts was combined with a GST pull-down assay. N-terminal GST-tagged mPARP-2 was incubated with nuclear extracts of sham or 3-h post-MCAO cortical tissue, with or without a caspase-8 inhibitor. Rh-caspase-8 generated a 52-kDa fragment that corresponds to the N-terminal GST-tagged cleavage product of PARP-2 (Fig. 3b). A similar fragment was generated by nuclear extracts from apoptotic tissue. Inhibition of caspase-8 by c-IETD-FMK totally abolished this process. These results indicate that PARP-2 is a target of nuclear caspase-8 in neurons undergoing apoptosis.
Caspase-8 Cleaves PARP-2 at a Specific, Bid-like, Site-The production of a single fragment and the high efficiency of cleavage suggest that PARP-2 contains a consensus sequence for recognition and cleavage by caspase-8. The mouse PARP-2 sequence (GenBank TM accession number AJ007780) displays two potential cleavage sites, 116 Leu-Glu-Asp-Asp 119 (LEDD) and 183 Leu-Gln-Met-Asp 186 (LQMD), which reside between the DNA binding domain and the catalytic site of the protein (Fig.  4a). These sites are conserved in human PARP-2 (GenBank TM accession number AJ236912). Both match the preferred cleavage sites of caspase-8, as determined by combinatorial peptide library screening (34) and by structural analysis (35). In addition, the LQMD site resembles the one cleaved by caspase-8 in the proapoptotic protein Bid (15). Cleavage at these sites would release C-terminal fragments with a theoretical molecular mass of 47 and 37 kDa, respectively (Fig. 4c), both of which are close to the apparent molecular mass of the fragment obtained in the in vitro assay. Microsequencing of the released C-terminal fragment obtained in the in vitro assay indicated that cleavage of PARP-2 by caspase-8 occurs at Asp 186 . This was confirmed by sitedirected mutagenesis. The two aspartic acid residues (Asp 119 and Asp 186 ) were mutated individually to glutamic acid in order to keep the negative charge. The two proteins generated, namely D119E and D186E, were subjected to cleavage by rhcaspase-8. Both mutations were effective in decreasing PARP-2 cleavage by caspase-8, but in correlation with the results of microsequencing, the second mutation had a more dramatic effect (Fig. 4b). This result points to the LQMD site as the preferential caspase-8 cleavage site of PARP-2.
Caspase-8 Cleavage Inactivates PARP-2-Cleavage of PARP-1 by caspase-3 results in a loss of its ability to be activated by DNA breaks and to synthesize ADP-ribose polymers, thus impairing DNA repair mechanisms (27,32). Similarly, the poly(ADP-ribosylation) activity of PARP-2 is dependent on the presence of DNA strand breaks (22), and impaired PARP-2 function results in a defective DNA repair process (31). To confirm that PARP-2 cleavage by caspase-8 leads to its inactivation, we evaluated the catalytic efficiency of cleaved PARP-2. The efficiency of cleavage of PARP-2 is directly correlated with the inhibition of its capacity to synthesize polymers (Table I).
These results show that caspase-mediated cleavage results in the loss of the main biological function of PARP-2.

DISCUSSION
The findings reported here indicate the existence of a nuclear pathway in caspase-8-mediated apoptotic neuronal death. Considered an initiator caspase until now, caspase-8 displays the anatomical and functional characteristics of an effector caspase. During ischemia, caspase-8 translocates into the nu-cleus of apoptotic neurons and targets at least one nuclear protein of the PARP family involved in cell maintenance and homeostasis. This evidence for a novel role for caspase-8 has a direct consequence on the way caspases should be regarded in the complex cascade of events that control apoptosis (Fig. 5).
The first result of this study is the demonstration of caspase-8 activity in the nucleus of ischemic cells. The comparison of these data with previous descriptions of caspase-8 activity in the same model (18) shows that caspase-8 is activated sequentially in the cytoplasmic and nuclear compartments. During focal ischemia, the cytoplasmic activation of caspase-8 is biphasic with two peaks separated by a period of lower activity at 3 h post-MCAO (18), which corresponds to the period of maximal activity of caspase-8 in the nucleus, suggesting that cytoplasmic and nuclear activations complement each other in the process of cell death.
Our second set of data assigned a crucial role for caspase-8 in the nuclear events of apoptosis. The growing PARP family is currently composed of five members (26,36). Among the PARPs, PARP-1 and PARP-2 respond to DNA strand breaks by poly(ADP-ribosylation) of nuclear proteins (30,34). They therefore play a critical role in the maintenance of genomic integrity by modulating DNA repair and cell survival programs (22). Cleavage of PARP-1 by caspase-3 is well documented in several models involving apoptotic cell death (37). In our study, caspase-8 cleavage of PARP-1 in vitro generated a fragment with an electrophoretic mobility very different from that observed in vivo in the same model (30,38) and in other models of neurodegenerative diseases (39 -42). This cleavage may have no physiological relevance, although the possibility of its occurrence in the case of caspase-3 deficiency cannot be ruled out. In contrast, fragments obtained after the cleavage of  PARP-2 by caspase-8 in vitro and in vivo are similar, indicating that this cleavage occurs under pathological conditions. Together, these results point to PARP-2 as the preferred substrate of caspase-8, although caspase-3 also cleaves PARP-1 and PARP-2. 2 The complementary, and eventually synergistic, actions of caspases on two members of the PARP family may act as a fail-safe system for the execution of the apoptotic program. As described for PARP-1 (43), cleavage of PARP-2 between the DNA binding domain and the catalytic site of the protein suppresses its ADP-ribose polymerization properties. Cleavage of PARP-2 by caspase-8 therefore amplifies DNA damage by adding to the DNA proteolysis resulting from PARP-1 inactivation by caspase-3 and -7. Inactivation of the PARPs also prevents depletion of ATP, which is needed to restore the PARP substrate nicotinamide adenosine diphosphate (44). The inactivation of PARP proteins therefore promotes the apoptotic process both by conserving essential ATP and by inhibiting poly(ADP-ribosylation) that would otherwise delay chromatin degradation (45).
Classical descriptions of the intracellular events that lead to cell apoptosis have differentiated two groups of caspases, referred to as initiator and effector caspases, on the basis of a set of functional criteria (46,47). In particular, initiator and effector caspases differ in the size, and consequently the function, of their prodomain (7), in their ability to translocate from the cytosol to other compartments, and in their sets of substrates (8,20,48). The demonstration of a dual role for caspase-8 as presented here concurs with recent data concerning other caspases, which suggests that established unidirectional models of caspase cascades in apoptotic pathways have to be modified. In hierarchical cascades of proteolytical activations (see Ref. 4 for a review), one type of apoptotic stimulus (e.g. death domain receptor activation, mitochondrial impairment, or reticular stress) leads to the activation of a unique cytoplasmic initiator caspase, which propagates the apoptotic signal by processing several effector or downstream caspases. However, new data have demonstrated that downstream caspases are able, in turn, to process the zymogens of initiator caspases, leading to their activation and thus creating a positive feedback loop that contributes to the amplification of the signal (49,50). In addition, caspases-2 and -9, originally considered as initiators, may exhibit effector-like characteristics (50,51). The result of this cross-talk is the parallel involvement of different caspase pathways, independent of the triggering stimulus. In agreement with these studies, although caspase-8 activation is classically described as linked to the activation of receptors of the TNF receptor family, some data suggest that it can be activated in a receptor-independent manner (52) and that caspase-8 activation may be secondary to the activation of other caspases (18,49,53). Our results add to this complexity, showing that caspase-8 simultaneously participates in multiple intracellular pathways with distinct goals.
In conclusion, the current models of apoptotic signaling pathways that involve parallel cascades of activation of specific caspases do not correctly reflect the data available on these proteases. Although the design of new signaling models will require much additional data, it is likely that they will take the form of interactive and complementary loops rather than cascades.