Caspase-2 Can Trigger Cytochrome c Release and Apoptosis from the Nucleus*

The cysteine proteases specific for aspartic residues, known as caspases, are localized in different subcellular compartments and play specific roles during the regulative and the executive phase of the cell death process. Here we investigated the subcellular localization of caspase-2 in healthy cells and during the execution of the apoptotic program. We have found that caspase-2 is a nuclear resident protein and that its import into the nucleus is regulated by two different nuclear localization signals. We have shown that in an early phase of apoptosis caspase-2 can trigger mitochondrial dysfunction from the nucleus without relocalizing into the cytoplasm. Release of cytochrome c occurs in the absence of overt alteration of the nuclear pores and changes of the nuclear/cytoplasmic barrier. Addition of leptomycin B, an inhibitor of nuclear export, did not interfere with the ability of caspase-2 to trigger cytochromec release. Only during the late phase of the apoptotic process can caspase-2 relocalize in the cytoplasm, as consequence of an increase in the diffusion limits of the nuclear pores. Taken together these data indicate the existence of a nuclear/mitochondrial apoptotic pathway elicited by caspase-2.

The cysteine proteases specific for aspartic residues, known as caspases, are localized in different subcellular compartments and play specific roles during the regulative and the executive phase of the cell death process. Here we investigated the subcellular localization of caspase-2 in healthy cells and during the execution of the apoptotic program. We have found that caspase-2 is a nuclear resident protein and that its import into the nucleus is regulated by two different nuclear localization signals. We have shown that in an early phase of apoptosis caspase-2 can trigger mitochondrial dysfunction from the nucleus without relocalizing into the cytoplasm. Release of cytochrome c occurs in the absence of overt alteration of the nuclear pores and changes of the nuclear/cytoplasmic barrier. Addition of leptomycin B, an inhibitor of nuclear export, did not interfere with the ability of caspase-2 to trigger cytochrome c release. Only during the late phase of the apoptotic process can caspase-2 relocalize in the cytoplasm, as consequence of an increase in the diffusion limits of the nuclear pores. Taken together these data indicate the existence of a nuclear/mitochondrial apoptotic pathway elicited by caspase-2.
Members of the aspartate-specific cysteine family of proteases known as caspases play a critical role in apoptosis (1). Caspases can be divided into initiator caspases and effector caspases based on the presence of a large prodomain at their amino-terminal region (2). Initiator caspases generally act at the apex of a proteolytic cascade, whereas effector caspases act downstream and are involved in the cleavage of specific cellular proteins (death substrates) (3). Once processed, death substrates modulate the morphological changes characterizing the apoptotic process (4,5). The long prodomains of the initiator caspases trigger/facilitate the activation of the proenzymes through the interaction with specific adaptor molecules (6). Caspase-2, caspase-8, caspase-9, and caspase-10 are the long prodomain caspases involved in the apoptotic process.
A plethora of studies have demonstrated that the large prodomain caspases-8, -9, -10 play a fundamental role in trans-ducing specific apoptotic stimuli (7). Caspase-9 is involved in apoptotic pathways that relay on mitochondrial dysfunction (8), whereas caspase-8 and -10 are involved in the apoptotic pathways mediated by death receptors (9,10).
The role of caspase-2 in a specific apoptotic pathway is controversial. Caspase-2-deficient mice exhibit apoptotic deficit in the oocytes, following exposure to chemotherapeutic drugs, in B lymphoblasts following incubation with perforin and granzyme B, in neurons when apoptosis is induced by ␤-amyloid, and in macrophages during Salmonella-induced death (11)(12)(13). In addition activation of caspase-2 could be part of a compensatory pathway of caspase activation in the absence of the Apaf-1/ caspase-9-based apoptosome (14). More recently it has been reported (15) that compensatory pathways can be activated also in caspase-2 null neurons, thus raising a cautionary note regarding the interpretation of findings obtained with caspasenull animals.
A regulative function for caspase-2 is also suggested by the ability of caspase-2-truncated isoforms, generated via alternative splicing, to antagonize cell death induced by serum deprivation and DNA damage (16,17). Finally, a role for caspase-2 as regulative caspase is reinforced by its ability to induce mitochondrial dysfunction and subsequent activation of the apoptosome through direct or indirect Bid processing (18).
Regulated localization of proteins to specific subcellular compartments is a common event in multiple transduction pathways including apoptosis. Therefore, the definition of the subcellular localization of important regulators of the apoptotic process is an important step toward the understanding of the mechanisms regulating cell death. The demonstration of caspase-12 localization in the endoplasmic reticulum has supported its role in mediating apoptosis induced by signals perturbing endoplasmic reticulum structure/function (19).
Caspase-2 has been shown to be localized in various subcellular compartments including cytosol, mitochondria, Golgi, nucleus, and plasma membrane (20 -25). In addition, when overexpressed as a GFP 1 fusion protein, the catalytically inactive caspase-2 mutant or the prodomain can form higher order structures mostly in the nucleus (22,26), similar to the filamentous structures formed by death effector domains of Fasassociated death domain and caspase-8 or by other proteins (27)(28)(29)(30). In the case of caspase-2 these structures are due to CARD-mediated oligomerization of the protein (22). Overall these data, sometimes controversial, still do not clarify the subcellular localization of caspase-2 and the relative implication for its death activity.
The goals of this work were to clarify the subcellular localization of caspase-2 in healthy cells and to identify the regions responsible for its subcellular targeting. Furthermore, we have investigated whether during apoptosis caspase-2 could translocate in different subcellular compartments and the relationship between its subcellular localization and its ability to trigger mitochondrial dysfunction.
Transfection, Microinjection, and Time Lapse Analysis-Transfections were performed using the calcium phosphate precipitation method. Cells were seeded 24 h before transfection at 1.2 ϫ 10 4 cells/ml and analyzed 12 h after removal of the precipitates. For caspase-2 killer activity 2 g of each expression plasmid were co-transfected with 400 ng of pEGFPN1 (Invitrogen) to score the transfected cells in vivo.
Microinjection was performed using the Automated Injection System (Zeiss Germany) as described previously (32). Cells were injected with the reported amount of expression vector for 0.5 s, at a constant pressure of 50 hectoPascal. TRITC-dextran, 66 kDa (Sigma), was used at a concentration of 1 mg/ml. Time-lapse studies were performed using a laser scan microscopy Leica TCS NT in a 5% CO 2 atmosphere at 37°C.
The amplification products were inserted in pEGFPN-1 or pEGFPC3 (Invitrogen) as BglII fragments.
All constructs generated were sequenced to check for the respective introduced mutations, deletions, and the translating fidelity of the inserted PCR fragments.
Subcellular Fractionation-Subcellular fractionation was performed using detergent lysis or Dounce homogenization. For detergent lysis, cells were washed twice in PBS, scraped, washed, and resuspended in 3 volumes of ice-cold extraction buffer A (20 mM Hepes, pH 7.5, 1.5 mM MgCl 2 , 1 mM EDTA, 1 mM EGTA, 0.05% Nonidet P-40, 10 g/ml cytochalasin B, 1 mM DTT, 1 mM PMSF, and 10 g/ml each of chymostatin, leupeptin, antipain, and pepstatin). After 10 min of incubation on ice, lysates were centrifuged at 800 ϫ g for 10 min. The post-nuclear supernatant was then centrifuged at 8000 ϫ g for 5 min to obtain the crude cytosolic fraction. The nuclear pellet was resuspended in extraction buffer A to the same final volume as the cytosolic fraction.
For subcellular fractionation with the Dounce homogenizer, after washing, cells were resuspended in extraction buffer B (20 mM Hepes, pH 7.5, 1.5 mM MgCl 2 , 1 mM EDTA, 1 mM EGTA, 1 mM DTT, 250 mM saccharose, cytochalasin B 10 g/ml, 1 mM DTT, 1 mM PMSF, and 10 g/ml each of chymostatin, leupeptin, antipain, and pepstatin), incubated for 20 min on ice, and then subjected to 40 strokes in a glass Dounce homogenizer type B. The obtained cell lysate was centrifuged at 500 ϫ g for 10 min. The obtained pellet was considered as nuclear fraction. The supernatant was centrifuged at 6000 ϫ g for 20 min, and the resultant pellet was considered enriched in mitochondria (P2). Finally the supernatant was further centrifuged at 100,000 ϫ g ϫ 60 min. The resultant supernatant (S100) was considered as cytosol and the pellet as microsomal fraction (P100). Pellets of each fraction were resuspended in a volume of extraction buffer B equal to the S100 fraction. All steps were performed on ice or at 4°C.
After addition of the SDS sample buffer, equal amounts of each sample were loaded on a 15% SDS-PAGE and subjected to the immunoblotting analysis.
Immunofluorescence Microscopy-For indirect immunofluorescence microscopy, cells were fixed with 3% paraformaldehyde in PBS for 20 min at room temperature. Fixed cells were washed with PBS, 0.1 M glycine, pH 7.5, and then permeabilized with 0.1% Triton X-100 in PBS for 5 min. The coverslips were treated with the following antibodies: anti-caspase-2 (18), anti-HA (Sigma), anti-cytochrome c (Promega), anti-Ran (Transduction Laboratories), anti-p62 (Transduction Laboratories), and anti-␤-galactosidase (Promega) diluted in PBS for 45 min in a moist chamber at 37°C. They were then washed with PBS twice and incubated with the relative TRITC-conjugated secondary antibodies (Sigma), fluorescein isothiocyanate (Sigma), or Alexa Fluor 633 (Molecular Probes) for 30 min at 37°C. Cells were examined with a laser scan microscope (Leica TCS NT) equipped with a 488 -534 argon laser and a 633 He-Ne laser.
Expression and Purification of GST Fusion Protein-pGEX-prodomain caspase-2 was transformed into Escherichia coli strain BL21. GST and GST fusion proteins were prepared as described previously (33).
In Vitro Binding Assays-Radiolabeled RAIDD/CRADD, caspase-2, and its deleted derivatives were generated using TNT T7-coupled reticulocyte lysate system (Promega). GST prodomain caspase-2 fusion protein immobilized onto glutathione-Sepharose beads was incubated with 200 l of binding buffer (20 mM Hepes, pH 7.5, 150 mM NaCl, 0.05% Nonidet P-40, 10% glycerol, 1 mM PMSF) at 4°C for 3 h in the presence of the appropriate amounts of the 35 S-labeled proteins. The beads were separated by brief centrifugation and washed four times with washing buffer containing 20 mM Hepes, pH 7.5, 150 mM NaCl, 0.05% Nonidet P-40, and 10% glycerol. Beads were boiled in SDS sample buffer, and proteins were resolved on a 15% SDS-PAGE.

RESULTS
Subcellular Localization of Caspase-2-Caspase-2 localization was investigated by subcellular fractionation of IMR90-EIA cells. Nuclear and cytosolic extracts were prepared by detergent lysis and were used to visualize caspase-2 in the different fractions by Western analysis. As shown in Fig. 1a, caspase-2 was revealed both in the nuclear and in the cytosolic fractions. The quality of the fractionation was verified by probing the lysates for well defined nuclear and cytoplasmic markers as illustrated in Fig. 1a. To confirm the observed caspase-2 subcellular distribution, fractionation was performed by a dif-ferential centrifugation (see "Experimental Procedures"). As shown in Fig. 1b, caspase-2 was detected both in the nuclear (P1) and in the cytosolic fractions (S100). In addition, residual amounts of caspase-2 were also detected in P100 and P2 pellets that should represent microsomal and mitochondrial fractions, respectively (Fig. 1b). Here again fractionation quality was verified by controlling the distribution of specific subcellular markers. The subcellular localization of caspase-2 was next investigated by confocal microscopy using the affinity-purified anti-caspase-2 antibody (18). In IMR90-E1A endogenous caspase-2 was prevalently revealed as a nuclear staining, with the exclusion of the nucleoli (Fig. 1c). The nuclear staining was undetectable after pre-absorption of the antibody with recombinant caspase-2 (Fig. 1c, PA).
Confocal analysis of the overexpressed caspase-2 by using the specific antibody revealed the same diffuse nuclear staining as obtained for the endogenous caspase-2 (Fig. 1d). Nuclei were visualized by propidium iodide staining (Fig. 1d, PI). When caspase-2wt was co-expressed with its previous identified adaptor RAIDD/CRADD (21, 34) again, it was prevalently observed as a diffuse nuclear staining, whereas RAIDD/CRADD showed both a nuclear and a cytoplasmic distribution (data not shown).
To clarify the apparent discrepancy between the nuclearcytoplasmic localization of caspase-2 observed after biochemical fractionation and its exclusively nuclear localization detected by immunofluorescence analysis, we decided to analyze the subcellular localization of the overexpressed caspase-2 by subcellular fractionation. We used GFP-tagged caspase-2 to exclude epitope masking in the immunofluorescence assay. To reduce the killer activity of caspase-2, low amounts of the expression plasmid were transfected, and the analysis was performed soon after transfection (within 10 h). Under these conditions few cells entered apoptosis (data not shown), and caspase-2-GFP was mainly revealed as unprocessed proenzyme (see Fig. 1f).
Similarly to the untagged protein, caspase-2-GFP was mainly detected as a diffuse nuclear staining with the exclusion of the nucleoli by confocal microscopy (Fig. 1e). On the other hand ectopically expressed caspase-2-GFP was detected both in the nucleus and in the cytoplasm, by means of subcellular fractionation and Western blotting. Similarly endogenous caspase-2 was detected both in the nucleus and in the cytoplasm.
In conclusion these results suggest that the accumulation of caspase-2 in the cytosolic compartment observed after subcellular fractionation might be ascribed to a dissociation of caspase-2 from the nuclei occurring during the experimental procedure.
Dissection of the Caspase-2 Prodomain, Implications for Nuclear Localization and Killer Activity-In order to map the regions of caspase-2 required for its targeting to the nuclear compartment, a number of different deleted constructs were generated, as outlined in Fig. 2a. We created a caspase-2 lacking the prodomain-(⌬1-152), because it has been demonstrated that murine caspase-2 lacking this region is impaired in nuclear localization (22). We also created a caspase-2 lacking the first 27 aa (⌬1-27), because a nuclear localization sequence is present in this region, and it has been shown previously (22,26) that this region is required for the nuclear localization of the caspase-2 prodomain. Finally, a caspase-2 version lacking both the amino-terminal region and the CARD domain-(⌬1-104) and a caspase-2 consisting only of the prodomain-(⌬153-435) were also generated. All the deleted derivatives were cloned in pcDNA 3 expression vector containing an HA epitope tag at the amino terminus.
The subcellular localization of the overexpressed proteins was analyzed using both the anti-caspase-2 and the anti-HA antibodies with similar results. Representative cells are shown in Fig. 2b. Overexpressed caspase-2wt was detected as a dif-FIG. 1. Subcellular localization of caspase-2. a, distribution of caspase-2 among subcellular fractions. Crude nuclear (N) and cytosolic (C) fractions were obtained using detergent lysis as described under "Experimental Procedures." Protein samples were prepared for Western blotting, and membranes were probed with the indicated antibodies. b, distribution of caspase-2 among subcellular fractions. Fractions were obtained using Dounce homogenizer as described under "Experimental Procedures." P1 fraction was designed as enriched nuclear fraction; P2 was designed as enriched mitochondrial fraction; and P100 was designed as enriched microsomal fraction and S100 as cytosol. Protein samples were prepared for Western blotting, and membranes were probes with the indicated antibodies. c, immunolocalization of caspase-2. IMR90-E1A cells were fixed, permeabilized, and labeled with anti-caspase-2 antibody or labeled with anti-caspase-2 antibody preabsorbed on recombinant caspase-2 (PA). Bar, 15 m. d, IMR90-E1A cells were transfected with pGDSV7-caspase-2, fixed, permeabilized, and double-stained with anti-caspase-2 antibody and with propidium iodide (PI) to visualize nuclei. Bar, 5 m. e, IMR90-E1A cells were transfected with pEGFP-N1-caspase-2, fixed and analyzed by confocal microscopy. Bar, 15 m. f, distribution of ectopic expressed caspase-2-GFP among subcellular fractions. Crude nuclear (N) and cytosolic (C) fractions were obtained using detergent lysis, whereas P1 and S100 fraction were obtained by differential centrifugation as described under "Experimental Procedures." Protein samples were prepared for Western blotting, and the membrane was probed with anti-caspase-2 antibody. C2-GFP, caspase-2-GFP; C2, endogenous caspase-2.
FIG. 2. Deletion constructs of caspase-2. a, schematic representation of caspase-2 and deleted derivatives. Prodomain, large subunit (LS), and small subunit (SS). b, immunofluorescence analysis of IMR90-E1A cells overexpressing caspase-2 wt and its deleted derivatives. IMR90-E1A fused nuclear staining. Caspase-2-(⌬1-27), lacking the bipartite NLS, was mainly detected as a diffuse staining in the cytoplasm, but some staining was also observed in the nucleus. Surprisingly, a larger deletion of the amino terminus that removes the bipartite NLS and the CARD-(⌬1-104) showed a reduced cytoplasmic staining and a consistent nuclear accumulation. At last, deletion of the prodomain-(⌬1-152) resulted, prevalently, in a diffused cytoplasmic localization of the enzyme. A complex picture emerged when an isolated caspase-2 prodomain was overexpressed. Cells showed a diffuse staining both in the nucleus and in the cytoplasm or present higher order structures such as filaments and dots highly positive for caspase-2 (Fig. 2b, ⌬153-435), as reported previously (22,26).
By using the GST-caspase-2 prodomain, we developed an in vitro pull-down assay to evaluate the ability to heterodimerize the different deleted versions of caspase-2. GST-prodomain caspase-2 was incubated with different caspase-2 deleted forms obtained by in vitro translation. In these experiments RAIDD/ CRADD was used as a positive control. As represented in Fig.  2, caspase-2 and the prodomain-(⌬153-435) were able to interact with GST-prodomain caspase-2. On the contrary, neither one of the other deleted versions tested interacted with caspase-2 in this assay (Fig. 2c). Because the shorter aminoterminal deletion (⌬1-27) also alters the CARD by removing the first ␣-helix (35), it is possible that a functional CARD is required for caspase-2 dimerization.
We next evaluated the killer activity of these caspase-2 deletions that exhibit different subcellular localizations. Cell death activity was analyzed in IMR90-E1A cells by co-transfecting caspase-2 or its deleted versions together with GFP as a reporter, and the appearance of apoptotic cells in vivo was analyzed 24 h later.
As outlined in Fig. 2d, caspase-2 wt and the deleted version lacking the prodomain-(⌬1-152) efficiently induced cell death. All the other deleted versions of the prodomain only modestly induced cell death, just like the catalytically inactive caspase-2, thus suggesting that this minor apoptotic response might result from some secondary effects due to the overexpression of an altered caspase-2.
The comparable cell death activity of caspase-2 wt and caspase-2-(⌬1-152) was unexpected. As described above, caspase-2-(⌬1-152) was unable to interact with GST-caspase-2 prodomain and thus supposed to be unable to oligomerize. Therefore, in order to study if the apoptotic pathway triggered by caspase-2 was similarly activated by caspase-2-(⌬1-152), we investigated if caspase-2-(⌬1-152) was able to trigger cytochrome c release from mitochondria as recently demonstrated for caspase-2 wt (18). IMR90-E1A cells, containing a dominant negative form of caspase-9, were used (31) because they are resistant to caspase-2 overexpression. Cytochrome c translocation was evaluated by immunofluorescence analysis. As summarized in Fig. 2e, IMR90-E1A caspase-9 DN cells overexpressing caspase-2 wt or ⌬1-152 showed the same percentage of cytochrome c release in the cytoplasm. This result confirms that a caspase-2 lacking the prodomain induces cell death by acting upstream of cytochrome c release, similarly to caspase-2 wt.
The Prodomain of Caspase-2 Contains a Second Nuclear Localization Signal-Previous studies have suggested that caspase-2 can also be targeted to the nuclear compartment independently from the previous identified bipartite NLS (22). We have demonstrated that caspase-2-(⌬1-104), which lacks the bipartite NLS and the CARD, still showed a prevalent nuclear localization. This observation prompted us to investigate whether other potential NLS were present in the prodomain of caspase-2. Other types of NLS described include those found in the proto-oncogene c-myc (PAAKRVKLD) where proline and aspartic acid residues flank a central basic cluster (36). Sequence analysis revealed that this type of NLS is present in the prodomain of caspase-2 between aa 131 and 143 (see Fig. 3a). To investigate the role of this second NLS (NLS-II) in the nuclear localization of caspase-2, the two lysines at position 135 and 136 were replaced by two alanines (K135A/K136A).
Caspase-2 K135A/K136A showed a predominant diffuse cytosolic staining when overexpressed in IMR90-E1A cells, even though some nuclear staining, as above described for ⌬1-27, was still detectable (Fig. 3b). In contrast, caspase-2 wt was exclusively nuclear under the same experimental conditions. This residual nuclear localization could be both ascribed to a simple diffusion in the nuclear compartment of the overexpressed protein, which size is behind the exclusion limits of the nuclear pore, or to a residual nuclear import activity of the remnant NLS.
To determine whether the newly identified NLS-II could target a cytosolic protein to the nucleus, a peptide containing NLS-II was fused to the reporter protein ␤-galactosidase. The immunofluorescence analysis demonstrated that ␤-galactosidase containing the caspase-2 NLS-II was localized prevalently in the nucleus, and on the contrary ␤-galactosidase lacking the caspase-2 NLS-II was localized prevalently in the cytoplasm.
To characterize further the caspase-2 K135A/K136A mutant, we evaluated its ability to dimerize in an in vitro pull-down assay using GST-caspase-2 prodomain. Caspase-2 K135A/ K136A was able to interact with caspase-2 similarly to the wt form (Fig. 3c).
By having demonstrated that caspase-2 K135A/K136A was still able to dimerize with caspase-2 and that it was impaired in the nuclear localization, the next question was to study its killer activity.
A cell death assay was performed as described above, and the results obtained are summarized in Fig. 3d. Caspase-2 K135A/ K136A was able to induce cell death similarly to caspase-2 wt in IMR90-E1A cells, and moreover, caspase-2 wt and caspase-2 K135A/K136A showed similar capacity to induce cytoplasmic relocalization of cytochrome c in IMR90-E1A caspase-9 DN cells (Fig. 3e). cells after transfection with the relative constructs were fixed and processed for the immunofluorescence analysis to visualize caspase-2. Bar, 7 m. c, in vitro caspase-2 binding. A GST-caspase-2-prodomain (GST-caspase-2) and GST as control, immobilized on glutathione-Sepharose beads, were incubated with the in vitro translated products of the indicated constructs. After washing, proteins bound to the beads were evaluated by SDS-PAGE. d, IMR90-E1A cells were co-transfected with the indicated constructs and pEGFPN1 as a reporter. The appearance of the apoptotic cells was scored after 20 h from transfection. Cells showing a collapsed morphology and presenting extensive membrane blebbing were scored as apoptotic. Data represent arithmetic means Ϯ S.D. of four independent experiments. e, IMR90-E1A caspase-9 DN cells were co-transfected with the indicated constructs and pEGFPN1 as a reporter. After 20 h immunofluorescence assay was performed using anti-cytochrome c antibody. Cells co-expressing GFP and caspase-2 or caspase-2-(⌬1-152) were scored for cytochrome c release from mitochondria. Data represent arithmetic means Ϯ S.D. of three independent experiments.

Subcellular Localization of Caspase-2 Fused to GFP in
Vivo-The demonstration that the caspase-2 mutants (K135A/ K136A and ⌬1-152), which showed a predominant cytoplasmic localization, were equally able to induce cell death and cyto-chrome c release raises the question of a possible cytoplasmic function of caspase-2 during apoptosis. To address this point, we decided to study if ectopically expressed caspase-2 can relocalize from the nucleus into the cytoplasm during cell death. FIG. 3. Caspase-2 contains a second nuclear localization sequence (NLS-II). a, schematic representation of caspase-2 where the putative nuclear import sequences (NLSs) are underlined. Prodomain, large subunit (LS), small subunit (SS). b, immunofluorescence analysis of IMR90-E1A cells overexpressing caspase-2 wt and its deleted or point-mutated derivatives. IMR90-E1A cells after transfection with the relative constructs were fixed and processed for the immunofluorescence analysis to visualize caspase-2. Bar, 7 m. c, immunofluorescence analysis of IMR90-E1A cells overexpressing ␤-galactosidase or ␤-galactosidase containing caspase-2 NLS-II. IMR90-E1A cells after transfection with the relative constructs were fixed and processed for the immunofluorescence analysis to visualize ␤-galactosidase. Bar, 7 m. d, in vitro caspase-2 binding. A GST-caspase-2-prodomain (GST-caspase-2) and GST as control, immobilized on glutathione-Sepharose beads, were incubated with the in vitro translated products of the indicated constructs. After washing, proteins bound to the beads were evaluated by SDS-PAGE. e, IMR90-E1A cells were co-transfected with the indicated constructs and pEGFPN1 as a reporter. The appearance of the apoptotic cells was scored after 20 h from transfection. Cells showing a collapsed morphology and presenting extensive membrane blebbing were scored as apoptotic. Data represent arithmetic means Ϯ S.D. of four independent experiments. f, IMR90-E1A caspase-9 DN cells were co-transfected with the indicated constructs and pEGFPN1 as a reporter. After 20 h immunofluorescence assay was performed using anti-cytochrome c antibody. Cells co-expressing GFP and caspase-2 or caspase-2 K135A/K136A were scored for cytochrome c release from mitochondria. Data represent arithmetic means Ϯ S.D. of three independent experiments. Because this relocalization cannot be unambiguously analyzed by biochemical fractionation, we decided to monitor the subcellular localization of active caspase-2 during apoptosis by confocal microscopy. We used a fusion of caspase-2 with GFP at the carboxyl terminus (caspase-2-GFP) to visualize the proenzyme and, after its processing, the active tetramer and a fusion of caspase-2 with GFP at the amino terminus (GFP-caspase-2) to visualize the proenzyme and, after its processing, the prodomain.
The ability of caspase-2-GFP and GFP-caspase-2 to undergo autocatalytic activation was investigated by transient transfection in 293 cells. Active caspase-2 is a tetramer consisting of two large 18-kDa subunits and two small subunits of 14 or 12 kDa (Fig. 2a) (1, 37). We used an antibody specific for the small subunit of caspase-2 in immunoblot analysis to evaluate caspase-2 processing (Fig. 4a). Caspase-2-GFP and GFPcaspase-2 both were detected as processed forms. Moreover, in the case of caspase-2-GFP the molecular size of the small subunits was increased as a consequence of the GFP. This indicates that the GFP, when fused to the carboxyl terminus of caspase-2, assembles into the active tetramer.
We next performed a time-lapse analysis to evaluate the ability of GFP-caspase-2 and caspase-2-GFP to induce apoptosis. The time-lapse analysis was performed in IMR90-E1A cells 30 -60 min after microinjection, and frames were collected every 2 min for 12 h. Selected frames of a representative experiment, at the indicated times, are shown in Fig. 4, b and c.
Caspase-2-GFP was detected as a diffuse nuclear staining with nucleolar exclusion (Fig. 4b), similar to the untagged overexpressed caspase-2 as described above for fixed cells. Within 4 h after microinjection almost 90% of the overexpressing cells had died by apoptosis as judged by nuclear fragmen-tation and membrane blebbing (data not shown).
A different scenario was observed when GFP-caspase-2 was overexpressed. Initially, GFP caspase-2 was detected as a dif- fuse nuclear staining, but at later time points (see Fig. 4c, 1.51) small dots were observed in the nucleus containing GFPcaspase-2, which increased in number and size over time (Fig.   4c, 2.41). These dots resemble previously described (22) nuclear structures enriched in caspase-2 prodomain observed in cultured cells after overexpression of catalytic inactive caspase-2 FIG. 6. Effect of caspase-2 on nuclear pores and Ran gradient. a, IMR90-E1A cells were injected with pcDNA 3 HA-caspase-2 (50 ng/l) and pEGFP-N1-Bid (5 ng/l). After 3 h cells were fixed and processed for immunofluorescence to visualize Bid-GFP and the nuclear pore protein p62. Nuclei were stained with propidium iodide (PI). Images were obtained using a Leica TCS confocal microscopy and are displayed in pseudocolors. Bar, 7 m. b, IMR90-E1A cells were injected with pcDNA 3 HA-caspase-2 (50 ng/l) and pEGFP-N1-Bid (5 ng/l). After 3 h cells were fixed and processed for immunofluorescence to visualize Bid-GFP and the GTPase Ran. Nuclei were stained with propidium iodide (PI). Images were obtained using a Leica TCS confocal microscopy and are displayed in pseudocolors. Bar, 7 m. c, IMR90-E1A caspase-9 DN cells were injected with pcDNA 3 HA-caspase-2 (50 ng/l) and pEGFP-N1-Bid (5 ng/l). After 3 h cells were fixed and processed for immunofluorescence to visualize Bid-GFP and the GTPase Ran. Images were obtained using a Leica TCS confocal microscopy and are displayed in pseudocolors. Bar, 7 m. d, quantitative analysis of Ran distribution on IMR90-E1A caspase-9 DN cells overexpressing caspase-2 and Bid-GFP. 200 cells were analyzed, and the subcellular localization of Ran was scored in relation to Bid-GFP distribution. or of the prodomain alone in fusion with GFP. We have also noticed that these dots partially co-localize with the promyelocytic leukemia oncogenic domain (26) 2 and that their appearance anticipates cell death. The time dependence of apoptosis induced by caspase-2 in relation to the appearance of the nuclear dots containing caspase-2 is summarized in Fig. 4d. About 98% of the cells die by apoptosis 100 min after the appearance of caspase-2 in the nuclear dots. The occurrence of cell death shortly after the appearance of caspase-2 in dot-like structures could explain the failure to clearly observe these structures in transient transfection experiment with untagged caspase-2. Because localization of caspase-2-GFP in the nuclear dots was less evident, it is possible that such dots are enriched in a processed form of caspase-2 consisting of the prodomain but lacking the small subunit.
In summary, these experiments indicate that the fusion to GFP did not affect the subcellular localization of caspase-2, its catalytic activity, and its ability to induce cell death.
Subcellular Localization of Caspase-2 during Different Phases of the Apoptotic Response-At least two different phases can be identified during apoptosis induced by caspase-2. An initial apoptotic phase during which cytochrome c is released from the mitochondria, but no significant alterations of the nuclear morphology can be observed, and a late phase when drastic alterations of the nuclear morphology are evident. Therefore, in order to evaluate caspase-2 localization during early and late apoptosis, we performed a triple immunofluorescence analysis in IMR90-E1A cells to visualize caspase-2-GFP, cytochrome c, and nuclei. More than 300 cells were analyzed in triplicate experiments, and the results obtained are exemplified in Fig. 5. As expected in non-apoptotic cells both caspase-2-GFP and GFP-caspase-2 were revealed as a diffused nuclear staining (Fig. 5a). Again in cells at an early apoptotic stage, as judged by cytochrome c release, both caspase-2-GFP and GFPcaspase-2 were localized in the nucleus (Fig. 5b). However, whereas caspase-2-GFP was prevalently detected as a diffused nuclear staining, GFP-caspase-2 was observed prevalently in nuclear dots, thus suggesting a different subnuclear localization of the active enzyme with respect to the prodomain.
Even in IMR90-E1A caspase-9 DN cells, release of cytochrome c following caspase-2 activation occurs in the absence of relocalization of the enzyme in the cytoplasm (data not shown).
The Subcellular Localization of Ran but Not Nuclear Pores Were Altered by Caspase-2-The above-described studies suggest that caspase-2 can trigger mitochondrial dysfunction from the nucleus and that, only during the final events of the apoptotic process, active caspase-2 can translocate, at some extent, into the cytoplasm. How can caspase-2 trigger cytochrome c release from the nucleus?
In eukaryotic cells macromolecular traffic between the nucleus and the cytoplasm can occur by simple diffusion, for particles of less than 50/60 kDa, or mediated by soluble transport receptors that shuttle through the nuclear pore complex (NPC) for larger molecules (38). Cargo molecules are recog-2 G. Paroni, manuscript in preparation. FIG. 7. Caspase-2 can induce cytochrome c release without altering the nuclear/cytoplasmic barrier. a, IMR90-E1A caspase-9 DN cells were injected with pcDNA 3 HA-caspase-2 (50 ng/l), pEGFP-N1-Bid (5 ng/l), and 66-kDa dextran (1 mg/ml). After 3 h cells were fixed and processed for immunofluorescence to visualize Bid-GFP and dextran. Images were obtained using a Leica TCS confocal microscopy. Bar, 1 m. b, IMR90-E1A caspase-9 DN cells were injected with pEGFP-N1-caspase-2 (20 ng/l) and 66-kDa dextran (1 mg/ml). After 3 h cells were fixed and processed for immunofluorescence to visualize caspase-2-GFP, dextran and cytochrome c. Images were obtained using a Leica TCS confocal microscopy. Bar, 1 m. nized via import or export targeting signals by the transport receptors, which are able to associate with components of the NPC (39). The small GTPase Ran is highly enriched in the nucleus in its GTP-bound form, and it regulates nuclear-cytoplasmic transport (40). Ran-GTP binds to importins, inducing the release of imported cargo, on the other side exportins interact with their substrates only in the nucleus in the presence of Ran-GTP, and they release the substrates in the cytoplasm after GTP hydrolysis (41).
The translocation of the apoptotic signal triggered by caspase-2 from the nucleus to the mitochondria could occur (i) by simple diffusion, (ii) by regulated nuclear export, and (iii) by increased diffusion (i.e. by increasing the diffusion limits of the nuclear pores).
Interestingly in apoptotic cells the diffusion limit of the nuclear pores is increased, thus allowing the redistribution throughout the cell of soluble proteins normally restricted to the nucleus or to the cytoplasm (42). In addition in apoptotic cells Ran subcellular localization is altered, thus suggesting a dysfunction of nuclear transport (42,43) Therefore, in order to investigate how nuclear localized caspase-2 could trigger cytochrome c release, we first evaluated if caspase-2 could regulate Ran localization and nuclear pores status. We used Bid translocation to the mitochondria as pcDNA3-caspase-2 and pGFPN1Bid were co-expressed by nuclear microinjection in IMR90-EIA. Cells were fixed 3 h later and processed for immunofluorescence to visualize Bid-GFP, nuclei, and p62, a component of the plug of the nuclear pores and a marker for nuclear pore integrity (42) (Fig. 6a) or Bid-GFP, nuclei, and Ran (Fig. 6b).
The antibody against p62 similarly detected the nuclear rim in cells presenting Bid-GFP as a diffuse staining or translocated to mitochondria, thus excluding overt alterations of the nuclear pores. As expected a large part of Ran was concentrated in the nucleus in cells presenting a diffuse localization of Bid-GFP. On the contrary Ran was distributed throughout the cell and therefore released from the nucleus in cells presenting Bid translocated to the mitochondria (Fig. 6b).
Caspase-2 can induce Bid translocation and cytochrome c release independently from caspase-9; therefore the relocalization of Ran was assayed in IMR90-E1A caspase-9 DN. As shown in Fig. 6c and summarized in Fig. 6d, caspase-2 can induce redistribution of Ran from the nucleus to the cytoplasm independently from caspase-9, and this event correlates with the translocation of Bid to the mitochondria.
Nuclear Caspase-2 Can Modulate Mitochondrial Integrity without Altering the Nuclear/Cytoplasmic Barrier-The abovedescribed studies do not exclude a possible effect of caspase-2 on the diffusion limits of the nuclear pore that could not be evidenced by p62 staining. To test this hypothesis we investigated the ability of caspase-2 to alter the diffusion limits of the nuclear pores. Fluorescent 66-kDa dextran, which does not diffuse passively through the nuclear pores, can be used as marker for nuclear breakdown in living cells (44). Here again, because caspase-9 is described to directly or indirectly disrupt the nuclear-cytoplasmic barrier, we decided to investigate if caspase-2 can specifically alter the permeability of the nuclear barrier in cells lacking caspase-9 activity. The nuclei of IMR90-E1A caspase-9 DN cells were injected with pcDNA3-caspase-2, dextran-TRITC 66-kDa, and pGFP-Bid. 4 h later cells were fixed to visualize Bid-GFP and dextran. As expected, in cells showing Bid-GFP in its unprocessed form (diffuse nuclear and cytosolic distribution), the 66-kDa dextran showed a nuclear localization (Fig. 7a). The same nuclear localization of the 66-kDa dextran was preserved in cells showing Bid-GFP translocated to the mitochondria. Similar results were obtained when caspase-2-GFP was co-expressed with dextran-TRITC 66-kDa, and caspase-2 activation was evaluated by cytochrome c distribution (Fig. 7b). Hence nuclear caspase-2 can trigger mitochondrial dysfunction without altering the nuclear/cytoplasmic barrier.
Caspase-2 Can Induce Cytochrome c Release and Apoptosis in the Presence of Leptomycin B-Nuclear pore complex protein CRM1 is able to recognize specific nuclear export sequences and mediate nuclear export from the nucleus (45). CRM1 is the target of the cytotoxic LMB. By direct binding to CRM1, LMB disrupts nuclear export sequence-dependent nuclear export. To understand if active nuclear export mediated by CRM1 could be implicated in triggering cytochrome c release once caspase-2 is activated in the nucleus, we investigated the effect of LMB on cytochrome c release in cells overexpressing caspase-2. IMR90-E1A caspase-9 DN cells were pretreated with LMB (5 ng/ml) 2 h before microinjection with plasmid encoding GFPcaspase-2. After further incubation for 3 h in the presence of LMB cells were fixed and processed for immunofluorescence. As shown in Fig. 8a, release of cytochrome c after activation of caspase-2 was independent from LMB treatment. Fig. 8b represents the quantitative analysis. Under the same experimen-tal conditions LMB was able to inhibit the nuclear export of IB-␣ (Fig. 8c) (45).

Disruption of the Nuclear/Cytoplasmic Barrier during the Late Phase of the Apoptotic Process Could Permit Caspase-2
Translocation into the Cytoplasm-To clarify the mechanism by which active caspase-2 can exit from the nucleus during the late apoptotic phase (see Fig. 5c), we analyzed the diffusion limits of the nuclear/cytoplasmic barrier in vivo throughout the apoptotic process triggered by caspase-2.
pGFPN1-caspase-2 and 66-kDa TRITC dextran were microinjected in the nucleus of IMR90-EIA cells, and soon afterward the cells were subjected to a time-lapse analysis. Frames were collected every 2 min during an 8-h period. Selected frames of a representative experiment, at the indicated times, are shown in Fig. 9a.
Caspase-2-GFP and dextran were well confined in the nucleus even though alterations of the nuclear morphology, indicative of the apoptotic process, were evident (starting from 2.30 h after microinjection). Localization of the 66-kDa dextran into the cytoplasm was first detected after 2.36 h and was more evident 2.42 h after microinjection. During the same time points caspase-2-GFP was more confined in the altered nucleus relative to the 66-kDa dextran. This observation might suggest that a fraction of caspase-2-GFP was not soluble at the indicated time points.
The time-lapse studies indicate that apoptosis triggered by caspase-2 in IMR90-E1A could induce alterations of the nuclear permeability, possibly as a consequence of caspase-9 activation (42). We next wanted to study if caspase-2 could alter nuclear permeability also independently from caspase-9. To this end pGFPN1-caspase-2 and 66-kDa TRITC dextran were microinjected in the nucleus of IMR90-EIA caspase-9 DN, and soon afterward cells were subjected to the time-lapse analysis. Frames were collected every 2 min during an 8-h period. Selected frames of a representative experiment at the indicated times are shown in Fig. 9b.
Caspase-2 and the 66-kDa dextran were well confined in the nucleus up to 4.00 h after microinjection. Some dextran staining can be observed in the cytoplasm 4.50 h after microinjection. This becomes more evident at later time points and parallels a decreased signal in the nucleoplasm (see Fig. 9b, 5.20 h after microinjection). Here again, although some caspase-2-GFP can relocalize into the cytoplasm, it is more confined in the nucleus with respect to the 66-kDa dextran during the same time course. Again, this observation might suggest that a fraction of nuclear caspase-2-GFP was not soluble at the indicated time points. It is important to note that the described time schedule can differ among cells. These differences probably reflect a differential responsiveness of IMR90-E1A cells to caspase-2 overexpression (data not shown).
To establish clearly the timing of caspase-2 activation in relation to the increased permeability of the nuclear barrier, we used Bid-GFP as a reporter for caspase-2 activation. pcDNA3-caspase-2, pGFPN1Bid, and dextran-TRITC were coexpressed by nuclear microinjection in IMR90-EIA caspase-9 DN cells, and soon after microinjection cells were subjected to time-lapse analysis. Frames were collected every 2 min during the period of 8 h. Selected frames of a representative experiment, at the indicated times, are shown in Fig. 9c.
We observed translocation of Bid-GFP to the mitochondria as early as 4 h after microinjection (see Fig. 9c, arrowhead 4. 01-4.15). During the same time course the first evidence of an effect of caspase-2 on the nuclear diffusion limits can detected 2 h later (see Fig. 9c, arrowhead 5.47 and 5.51) with the appearance of the dextran staining in the cytoplasm. Taken together these results demonstrate that the alteration of the nuclear/cytoplasmic barrier is a late event triggered by caspase-2.
Here again it is important to note that the schedule of Bid-GFP translocation and of dextran release from the nucleus can differ among cells (data not shown). DISCUSSION The proteolytic machinery controlling apoptosis includes numerous caspases that are localized in different subcellular compartments and play specific roles during the regulative and the executive phases of the cell death process (3)(4)(5).
The subcellular localization of caspase-2 is still debated. By means of immunofluorescence analysis endogenous caspase-2 was observed in the cytosol, in the nucleus, and in the Golgi (20,25), whereas ectopically expressed caspase-2 accumulated mainly in the nucleus (22,26). A different study, using subcellular fractionations, has indicated a prevalent cytosolic localization of endogenous caspase-2 (24). Our immunofluorescence studies support a nuclear localization for both endogenous and ectopically expressed caspase-2. In contrast, upon subcellular fractionation a large amount of caspase-2 was detected in the cytoplasm. We analyzed the subcellular localization of caspase-2-GFP both by fractionation and by confocal microscopy and present data strongly suggesting that the cytosolic localization of caspase-2, observed after fractionation, might be ascribed to a dissociation of caspase-2 from the nuclei during the experimental procedure. However, we cannot exclude that the nuclear import of caspase-2 could be regulated thus making possible its accumulation in the cytoplasm in certain cell types (20). Further studies are necessary to unambiguously answer this point.
The bipartite NLS present within the amino-terminal 44amino acid residues is necessary for the nuclear localization of caspase-2 (22,26). In the current study we have confirmed that this sequence is a critical NLS of caspase-2, and in addition we have identified a second, previously unrecognized NLS within the prodomain of caspase-2. The existence of a second NLS was suggested by Colussi et al. (22) to explain the failure of the amino-terminal 44-amino acid sequence to drive the nuclear localization of a reporter protein. This second NLS (NLS-II) is placed between aa 131 and 143 (PLYKKLRLSTD) and resembles the NLS of the proto-oncogene c-myc, where proline and aspartic acid residues define a central basic cluster (36). Our present study shows that NLS-II can drive the nuclear import of a cytoplasmic protein. The two separate NLSs present in the prodomain of caspase-2 could functionally collaborate with each other, and thus provide the cells with greater versatility in regulating caspase-2 nuclear import.
The prodomain of caspase-2 also contains the CARD, which is composed by six antiparallel amphipathic ␣-helices, and mediates homophilic interactions, thus triggering proenzyme activation (35,46). Caspase-2, containing deletions or alterations of the CARD, was impaired in inducing cell death. Surprisingly, a caspase-2 without the prodomain-(⌬1-152) was able to induce cell death and cytochrome c release similarly to the wt form, whereas a shorter deletion of the prodomain-(⌬1-104) was impaired in inducing cell death. Previous studies (47) have shown that murine caspase-2-(⌬1-138), which corresponds to human caspase-2-(⌬1-121), was also unable to induce cell death. Therefore, it is possible that the prodomain contains an inhibitory domain, which should be located between aa 104 and 151 likely between aa 121 and 151. An inhibitory effect of the prodomain was previously suggested by the observation that active recombinant caspase-2 could be obtained by expressing in E. coli a fragment lacking the prodomain-(⌬1-152). Expression of full-length caspase-2 did not show proteolytic activity even though it was processed into FIG. 9. An increase of the diffusion limits of the nuclear pore is a late event during the apoptotic response elicited by caspase-2. a, time-lapse images of a representative IMR90-E1A cell double-stained for caspase-2-GFP and dextran. Frames at selected times after microinjection (as indicated) of a representative cell injected with pEGFP-N1-caspase-2 (20 ng/l) and 66-kDa dextran (1 mg/ml) are shown. Bar, 2 m. b, time-lapse images of a representative IMR90-E1A caspase-9 DN cell double-stained for caspase-2-GFP and dextran. Frames at selected times after microinjection (as indicated) of a representative cell injected with pEGFP-N1-caspase-2 (20 ng/l) and 66-kDa dextran (1 mg/ml) are shown. Bar, 2 m. c, time-lapse images of a representative IMR90-E1A caspase-9 DN cell double-stained for Bid-GFP and dextran. Frames at selected times after microinjection (as indicated) of a representative cell injected with pcDNA 3 HA-caspase-2 (50 ng/l), pEGFP-N1-Bid (5 ng/l), and 66-kDa dextran (1 mg/ml) are shown. Bar, 2 m. a prodomain large subunit and a small subunit (37). It is interesting to note that an autoinhibitory domain has been recently discovered in procaspase-3. Caspase-3 is under strict regulatory self-control by a "safety catch" Asp-Asp-Asp tripeptide, contained within the proenzyme itself (48).
The identity of apoptotic signals specifically activating caspase-2 is unknown. However, upon ectopic expression, one can artificially activate caspase-2, possibly as consequence of the forced oligomerization. This phenomenon has allowed us to study the apoptotic signals originated by caspase-2 in the nucleus.
We have observed that activated caspase-2 can trigger the release of cytochrome c from the nucleus, and only at a later phase can the active enzyme partially translocate into the cytoplasm.
Caspase-2 triggered the release of cytochrome c in the cytoplasm when (i) active caspase-2 was still confined in the nucleus, (ii) no changes of the diffusion limits of the nuclear pores were evident, (iii) Ran accumulated in the cytoplasm, and (iv) the CRM1-dependent nuclear export was inhibited by LMB treatment.
These observations can be explained with two hypothetical mechanisms as follows: a nuclear factor, regulated by caspase-2, may exit the nucleus by a CRM1-independent pathway, or alternatively, the nuclear factor may translocate to the cytoplasm by simple diffusion.
In a simple model system the nuclear factor should be cleaved directly or indirectly by caspase-2, and once cleaved it should be able to exit the nucleus and to propagate the signal to the mitochondria. Unfortunately, the number of known death substrates, cleaved by caspase-2, is limited. Besides caspase-2 autocatalysis, it has been reported that caspase-2 can cleave Golgin-160 and Bid (18,25,49). Golgin-160 is a peripheral Golgi membrane-associated protein, and therefore its involvement in transferring the apoptotic signal from the nucleus to the cytoplasm is unlikely. Bid has been localized in the cytosol by subcellular fractionation (50). Nevertheless, when overexpressed as GFP fusion or epitope-tagged (18,49,51), it can be localized throughout the cell including the nuclear compartment where it probably accumulates by simple diffusion. Interestingly the time-lapse analysis, reported in this work, has demonstrated that, after caspase-2 activation, nuclear localized Bid-GFP can efficiently relocalize to mitochondria (Fig. 9c). If endogenous Bid can localize, at some extent, in the nuclear compartment it might represent a caspase-2 substrate, which bears the requested characteristics for transducing a caspase-2 apoptotic signal to the mitochondria. The presence of Bid both in the nucleus and in the cytoplasm might also explain the ability of the cytoplasmic localized mutants of caspase-2 (K135A/K136A and ⌬1-152) to efficiently induce cell death and cytochrome c release. However, it is clear that it will be necessary to characterize new caspase-2 substrates to answer this question.
During the late apoptotic phase, coupled to nuclear fragmentation, caspase-2 can relocalize at some extent in the cytoplasm. When present in the cytoplasm active caspase-2 could cleave new death substrates such as Golgin-160, thus contributing to the dismantling of the apoptotic cell.
The nuclear pore complex spans both membranes of the nuclear envelope and allows the receptor-mediated transport of macromolecules and the passive exchange of ions, metabolites, and intermediate sized macromolecules (52). The NPC harbors an ϳ10-nm diameter diffusion channel that admit polypeptides of around 50-kDa (38,53). Processed caspase-2 is supposed to be a tetramer of about 60 -64 kDa in its uncomplexed form (37,47). This implies that processed caspase-2, even though not complexed to other proteins, cannot exit the nucleus by diffusion.
Caspase-9 directly or indirectly is responsible for increasing nuclear permeability during apoptosis, which is due to alterations of nuclear pores (42). Caspases can cleave some nuclear pore elements, thus suggesting a possible mechanism for the increase of the nuclear diffusion limit (42,43,54), but alternative mechanisms could also be evoked (55). We found that caspase-2 can increase the nuclear diffusion limits. This effect was shown to be independent from caspase-9, indicating that caspase-2 can activate alternative pathways to increase nuclear permeability.
One possibility is that caspase-2 might cleave directly some of the more than 50 proteins of the vertebrate nuclear pore complex (38,56). However the observed delay, ϳ2.00 h, between caspase-2 activation, as indicated by Bid-GFP processing, and the diffusion of dextran out of the nucleus, argues against a direct processing of the nuclear pore components by caspase-2. Hence we suggest an indirect effect of caspase-2 on the nuclear-cytoplasmic barrier.
In conclusion our data clarify the subcellular localization of caspase-2 and demonstrate the existence of two NLSs regulating its nuclear import. Active caspase-2 can mediate cytochrome c release from the nucleus thus indicating the existence of a nuclear/mitochondrial apoptotic pathway.