Mitochondrially Localized Active Caspase-9 and Caspase-3 Result Mostly from Translocation from the Cytosol and Partly from Caspase-mediated Activation in the Organelle LACK OF EVIDENCE FOR Apaf-1-MEDIATED PROCASPASE-9 ACTIVATION IN THE MITOCHONDRIA*

Active caspase-9 and caspase-3 have been observed in the mitochondria, but their origins are unclear. Theo-retically, procaspase-9 might be activated in the mitochondria in a cytochrome c /Apaf-1-dependent manner, or activated caspase-9 and -3 may translocate to the mitochondria, or the mitochondrially localized procaspases may be activated by the translocated active caspases. Here we present evidence that the mitochondrially localized active caspase-9 and -3 result mostly from translocation from the cytosol (into the intermembrane space) and partly from caspase-mediated activation in the organelle rather than from the Apaf-1-medi-ated activation. Apaf-1 localizes exclusively in the cytosol and, upon apoptotic stimulation, translocates to the perinuclear area but not to the mitochondria. In most cases, the mitochondrially localized procaspase-9 and -3 are released early during apoptosis and translocate to the cytosol and/or perinuclear area. Cytochrome c and the mitochondrial matrix protein Hsp60 are also rapidly released to the cytosol early during apoptosis. Both the early release of proteins like cytochrome c and Hsp60 from the mitochondria as well as the later translocation of the active caspase-9/-3 are partially inhibited h of protein using a standard curve. The results were generally presented as fold activation over the control (30). In some experiments, the DEVDase activity was continuously monitored over a period of 2 h.

Apoptosis plays an essential role in animal development and in maintaining the homeostasis of adult tissues (1). The family of caspases (cysteine aspartic acid-specific protease) is the key effector in the execution of apoptotic cell death (2). Caspases are synthesized as inactive proenzymes, which become proteolytically cleaved during apoptosis to generate active enzymes. The active caspases cleave cellular proteins such as poly(ADP-ribose) polymerase (PARP) 1 to dismantle the apoptotic cells (3). The exact mechanism(s) whereby various caspases become activated are still unclear. Two pathways leading to caspase activation are relatively better understood. In the first pathway, apoptotic stimuli cause the release of cytochrome c (cyt. c) from the intermembrane space (IMS) of the mitochondria to the cytosol. The released cyt. c binds to and activates the adaptor protein Apaf-1, which in turn activates the initiator procaspase-9 in the presence of ATP, leading to the formation of apoptosome and subsequent activation of "executioner" caspases such as caspase-3, -6, or -7 (4). In the second pathway, the FADD/TRADD adaptor proteins recruit the initiator procaspase-8 (or -10) to cell surface death receptors, leading to death receptor-induced signaling complex formation, caspase-8 activation, and, subsequently, activation of executioner caspases (5). Recent evidence also implicates caspase-2 activation upstream of the cyt. c/Apaf-1 apoptosome-initiated apoptosis (6 -10) or cyt. c/Apaf-1-independent apoptotic pathways (11)(12)(13).
Apart from their cytosolic residence, procaspases are also localized in other subcellular compartments. For example, procaspase-2 has been found in the Golgi apparatus and nucleus (14,15). Procaspase-12 is mainly expressed in the endoplasmic reticulum where it serves as a major sensor of local stress (16). Depending on cell types, procaspase-2 (17), procaspase-3 (18,19), procaspase-8 (20), and procaspase-9 (17,21) have been reported to be present in the IMS of the mitochondria. Active caspases have also been found in different subcellular compartments. For example, in response to tumor necrosis factor-␣, procaspase-1 translocates to the nucleus where it is proteolytically activated (22,23). Upon treatment with tunicamycin, some activated caspase-12 translocates to or around the nuclei of apoptotic cells (24). In some experimental systems, the mitochondrially localized procaspase-2 and -3, upon activation, translocate to the nucleus (14,15,17). Activated caspase-7 has been shown to be associated exclusively with the mitochondrial/microsomal fractions (25). Activated forms of caspase-2 and caspase-9 have been detected in the mitochondria (17). Acti-vated caspase-2 has also been detected in apoptotic nuclei (26). Upon induction of apoptosis, mitochondrial procaspase-9 translocates to the cytosol and nucleus in both cell culture and animal model systems (17,21). Similarly, activated caspase-3 has been detected in normal and apoptotic nuclei (27)(28)(29). In all these apoptotic systems, it is generally thought that procaspases "stored" in various subcellular compartments are released, upon apoptosis induction, and become activated in the cytosol or that the cytosolically activated caspases translocate to various organelles to participate in apoptosis. It is still unclear whether the procaspases localized in the organelles (e.g. mitochondria) can ever be activated in situ. A pertinent philosophical question is why these procaspases have to be "imported" into these organelles if they can only be activated in the cytosol?
Recently, we have found that apoptosis induced by many stimuli involves an early mitochondrial activation, which is characterized by up-regulation of the mitochondrial respiratory chain (MRC) proteins and many other mitochondrially localized non-MRC proteins (30). A cardinal feature of this mitochondrial activation-dependent apoptotic pathway (or MADAP) is the early up-regulation and enrichment of cyt. c in the mitochondria, which precede its release. Because procaspase-9 is present in the IMS of the mitochondria (14,17,21), we hypothesized that the increased mitochondrial cyt. c might lead to the activation of procaspase-9 and, subsequently, of procaspase-3 inside the organelle. In this study, we utilized our MADAP model to test this hypothesis. Our results show no evidence of cyt. c/Apaf-1-mediated procaspase-9 activation inside the mitochondria. Instead, the results suggest that the mitochondrial active caspase-9 and -3 result mostly from the translocation from the cytosol and partly from caspase-mediated activation in the mitochondria.

MATERIALS AND METHODS
Cells and Reagents-GM701.2-8C (GM701) cells were kindly provided by Dr. M. King (Thomas Jefferson University) and cultured in Dulbecco's minimum essential medium (Invitrogen) supplemented with 10% heat-inactivated fetal bovine serum and antibiotics. Human prostate cancer cells, PC3 and LNCaP, were purchased from ATCC (Manassas, VA) and cultured in RPMI 1640 supplemented with 5 and 10% fetal bovine serum, respectively.
The primary antibodies used are listed in Table I. All secondary antibodies, i.e. goat, donkey, or sheep anti-mouse or rabbit or goat IgG conjugated to horseradish peroxidase, fluorescein isothiocyanate, or rhodamine, together with enhanced chemiluminescence (ECL) reagents were acquired from Amersham Biosciences. Biotinylated goat anti-rat or anti-rabbit antibodies were obtained from Jackson ImmunoResearch (West Grove, PA). Fluorogenic caspase substrates DEVD-AFC and LEHD-AFC, Pan-caspase inhibitor Z-VAD-fmk, caspase-3/7 inhibitor Z-DEVD-fmk, and recombinant caspase-3 were bought from Biomol (Plymouth Meeting, PA). Streptavidin conjugated to AlexaFluor 594 or 488 and mitochondrial dyes were purchased from Molecular Probes (Eugene, OR). All other chemicals were purchased from Sigma unless specified otherwise.
Subcellular Fractionation and Western Blotting-Mitochondria were prepared as described previously (30 -32). Briefly, cells were treated with various chemicals, inhibitors, or vehicle (ethanol or Me 2 SO) control. In some experiments, cells were pretreated with cyclosporin A (CsA), Z-VAD-fmk, or Z-DEVD-fmk before apoptosis induction. At the end of the treatment, cells were harvested with trypsin, washed twice in ice-cold PBS, and resuspended in 600 l of homogenizing buffer (20 mM HEPES-KOH, pH 7.5, 10 mM KCl, 1.5 mM MgCl 2 , 1 mM sodium EDTA, 1 mM sodium EGTA, and 1 mM dithiothreitol) containing 250 mM sucrose and a mixture of protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 1% aprotinin, 1 mM leupeptin, 1 g/ml pepstatin A, and 1 g/ml chymostatin). After 30 min of incubation on ice, cells were homogenized in the same buffer using a glass Pyrex homogenizer (type A pestle, 140 strokes). Unbroken cells, large plasma membrane pieces, and nuclei were removed by centrifuging the homogenates at 500 ϫ g for 5 min at 4°C. The resulting supernatant was centrifuged at 10,000 ϫ g for 20 min to obtain mitochondria. The remaining supernatant was again subjected to centrifugation at 100,000 ϫ g for 1 h to obtain the cytosolic fraction (i.e. S100 fraction). Mitochondrial pellet was washed three times in homogenizing buffer and then solubilized in TNC buffer (10 mM Tris acetate, pH 8.0, 0.5% Nonidet P-40, 5 mM CaCl 2 ) containing protease inhibitors. Protein concentration was determined by Micro-BCA kit (Pierce).
For Western blotting, various amounts of mitochondrial or cytosolic proteins were loaded in each lane of a 15% SDS-polyacrylamide gel. After gel electrophoresis and protein transfer, the membrane was probed or reprobed, after stripping, with various primary and corresponding secondary antibodies. Western blotting was performed using ECL as described previously (30).
Characterization of Mitochondrial Preparation by Transmission Electron Microscopy-The mitochondrial fractions prepared from PC3 cells as described above were fixed for 30 min at 4°C in 1% acrolein (v/v) in homogenizing buffer. At the end of fixation, the mitochondria were pelleted by centrifugation at 10,000 ϫ g and then resuspended in homogenizing buffer containing 0.5% dimethyl sulfoxide (v/v). Mitochondria were pelleted and resuspended in the same buffer four times for 15 min each, post-fixed for 30 min in 1% osmium tetroxide at room temperature, washed in homogenizing buffer, dehydrated in graded ethanol (20,40,60,80,90,95, and 100% 2 times), and then passed through propylene oxide, followed by infiltration and embedding in epoxy resin. Ultrathin (80 -100-nm) sections were cut with a Reichert Ultracut E ultramicrotome, picked up on 200 mesh copper grids, and stained with 2% (w/v) aqueous uranyl acetate followed by lead citrate. Grids were examined and photographed in a Zeiss 10 C transmission electron microscopy at an accelerating voltage of 80 kV.
Preparation of Percoll Gradient-purified Mitochondria-PC3 cells (12 ϫ 10 6 ) were treated with ethanol or BMD188 (40 M) for 1 h and then harvested for homogenization and differential centrifugation as described above. The resulting 10,000 ϫ g mitochondrial pellet was washed and then resuspended in an EDTA-free medium and layered on a Percoll gradient consisting of four layers of 10, 18, 30, and 70% (by volume) Percoll in 0.3 M mannitol and 5 mM MOPS, pH 7.2 (33). After centrifugation for 35 min at 13,500 ϫ g, the purified mitochondria were separated from non-mitochondrial membranes and non-functional or- ganelles and collected at the 30/70% interface and washed with 0.3 M mannitol, 5 mM MOPS, pH 7.2, containing 1 mg/ml bovine serum albumin to remove the Percoll. PARP Cleavage-PARP cleavage assays were performed as described previously (30 -32).
Quantification of Apoptotic Nuclei Using DAPI Staining-Cells were plated on glass coverslips (4 ϫ 10 4 cells/18-mm 2 coverslip) the day before treatment. The next day, cells were treated with vehicle control (i.e. ethanol or Me 2 SO) or various inducers. Thereafter, cells were incubated live with DAPI (0.5 g/ml) for 10 min at 37°C followed by washing. The percentage of cells exhibiting apoptotic nuclei, as judged by chromatin condensation or nuclear fragmentation, was assessed by fluorescence microscopy (34). An average of 600 -700 cells was counted for each condition.
DEVDase and LEHDase Activity Measurement-Cells were washed twice in PBS and the whole cell lysate was made in the lysis buffer (50 mM HEPES, 1% Triton X-100, 0.1% CHAPS, 1 mM dithiothreitol, and 0.1 mM EDTA). For activity measurement, 30 or 100 g whole cell lysate, mitochondria, or cytosol was added to a reaction mixture containing 10 or 30 M fluorogenic peptide substrates, Ac-DEVD-AFC or Ac-LEHD-AFC, 50 mM HEPES, pH 7.4, 10% glycerol, 0.1% CHAPS, 100 mM NaCl, 1 mM EDTA, and 10 mM dithiothreitol, in a total volume of 1 ml and incubated at 37°C for 1 h. Production of 7-amino-4-trifluoromethylcoumarin (AFC) was monitored in a spectrofluorimeter (Hitachi F-2000 fluorescence spectrophotometer) using excitation wavelength 400 nm and emission wavelength 505 nm. The fluorescent units were converted into nanomoles of AFC released per h per mg of protein using a standard curve. The results were generally presented as fold activation over the control (30). In some experiments, the DEVDase activity was continuously monitored over a period of 2 h.
Cell-free Caspase Activation in the Mitochondrial Fractions-Mitochondria and cytosol were prepared from treated or untreated PC3 cells as described above. Various amounts of freshly isolated, unlysed or lysed (using the TNC buffer) mitochondria were co-incubated with 100 g of cytosol freshly prepared from untreated PC3 cells for 1 h at 37°C. At the end, either 10 or 30 M DEVD-AFC substrate was added to the mixture, which was further incubated for 1 h at 37°C. Subsequently, the DEVDase activity in the co-incubates was measured.
Kinetic Study of Caspase Activation-Two sets of PC3 cells (10 7 each) were simultaneously treated with 40 M BMD188 for 0, 30, 45, 60, 90, and 120 min. At the end, one set of cells was harvested for the preparation of whole cell lysate and the other set harvested for subcellular fractionation, as described above. Subsequently, the three preparations, i.e. whole cell lysate, cytosol, and mitochondria, all derived from 10 7 cells, were used to measure the caspase activity by incubating with 10 M of the Ac-DEVD-AFC substrate, as detailed above.
Immunostaining of Caspases and Apaf-1-Cells grown on glass coverslips were treated for various time intervals. Fifteen min prior to the end of the treatment, cells were incubated live with DAPI to label apoptotic nuclei (30). Then cells were fixed in 4% paraformaldehyde for 10 min followed by permeabilization in 1% Triton X-100 for 10 min. After washing in PBS, cells were first blocked in 10% goat whole serum for 30 min at 37°C and then probed with Apaf-1, caspase-9, or caspase-3 (all at 1:500) for 1 h at 37°C. After washing in PBS, cells were incubated with biotinylated goat anti-rat or anti-rabbit IgG antibody (1:1000). Finally, cells were incubated with streptavidin-AlexaFluor 488 or 594 (1:500) for 1 h at 37°C. After thorough washing, coverslips were mounted on slides using Vectashield mounting medium (Vector Laboratories, Inc., Burlingame, CA) and observed under an Olympus BX40 epifluorescence microscope. Images were captured with Magna-Fire software and processed in Adobe Photoshop.
Proteinase K Digestion of Isolated Mitochondria-Freshly isolated mitochondria (100 g) were incubated in the homogenizing buffer (without protease inhibitors) alone or in the presence of proteinase-K (0.1 g/ml) only or proteinase K plus Triton X-100 (1% final concentration). After 10 min of incubation on ice, 2 l of phenylmethylsulfonyl fluoride (100 M) was added to terminate proteolysis followed by addition of 6ϫ SDS-loading buffer. Samples were then boiled for 5 min and analyzed immediately by Western blotting (20).
Alkali Extraction of the Mitochondria-The mitochondrial pellet (100 g) was suspended in 0.1 M Na 2 CO 3 , pH 12.0, and incubated on ice for 30 min. At the end of the incubation, the sample was centrifuged at 100,000 ϫ g for 1 h. The resulting pellet was then lysed in the TNC buffer. Both supernatant and pellet were then subjected to Western blotting (35).
Cell-free Caspase Activation and Translocation Experiments-All cell-free reactions were performed in homogenizing buffer in a total volume of 100 l. Freshly isolated cytosol and/or mitochondria were incubated at 37°C for 2 h with addition of bovine cyt. c (50 g/ml) (36,37). At the end, samples containing co-incubated mitochondria were centrifuged at 10,000 ϫ g for 20 min to obtain the mitochondrial pellet. The resulting pellet was washed twice and suspended in homogenizing buffer. For translocation studies, cyt. c-activated cytosol was further incubated with untreated mitochondria for 1 or 3 h at 37°C followed by centrifugation (10,000 ϫ g) for 20 min. Following two washes in homogenizing buffer, the mitochondrial pellet, together with the supernatant, was subjected to Western blotting. In separate sets of experiments, isolated cytosol or mitochondria from BMD188 (40 M) treated or untreated cells were co-incubated in various combinations for 1 or 2 h at 4, 25, and 37°C. After co-incubation, the pelleted mitochondria were washed twice with homogenizing buffer, and then equal amounts of proteins were separated followed by Western blotting. To determine whether the mitochondrially localized procaspases can ever be activated in the organelle, mitochondria were incubated in the homogenizing buffer for 2 h at 37°C by addition of recombinant caspase-3 (1 M) or caspase-9 (1 M). At the end of the incubation, the mitochondria were pelleted, washed twice in homogenizing buffer, and resuspended in 100 l of homogenizing buffer. The supernatant and pellet were then subjected to SDS-PAGE and Western blotting analysis.

Activated Mitochondria Possess Caspase Activity-Apoptosis
induced by multiple stimuli is preceded by a rapid up-regulation and accumulation of cyt. c in the mitochondria (30). Because many procaspases are localized in the mitochondria (see Introduction), we reason that some of these mitochondrially localized procaspases, in particular procaspase-9 and procaspase-3, might become activated in the organelle in a cyt. c-dependent manner, analogous to apoptosome-mediated activation of procaspase-9 and -3 in the cytosol. To test this possibility, we used BMD188, a cyclic hydroxamate that activates MADAP in multiple cell types (30 -32), to treat PC3 prostate epithelial cancer cells. Following treatment for various time intervals, whole cell lysate, and mitochondrial and cytosolic fractions were prepared from equal numbers of cells (i.e. 10 7 cells for each fraction), and the DEVDase activity, which measures caspase-3 and -7 activation, was determined. The quality of subcellular fractionation was monitored by Western blot analysis of known compartment-specific proteins: cytochrome c oxidase subunit II for mitochondria, lamin A/C for nuclei, and lactate dehydrogenase for cytosol. We did not detect any crosscontamination among different fractions (see Refs. 30 and 32, and data not shown). Electron microscopy analysis also revealed pure mitochondria in our preparations (not shown). As shown in Fig. 1A, 45 min post-BMD188 treatment of PC3 cells, about 4-and 2-fold DEVDase activity was observed in the whole cell lysate and mitochondria, respectively. Only marginal DEVDase activity was detected in the cytosol by this time (Fig. 1A). By 60 -120 min, increased DEVDase activity was observed in the cytosol, but the corresponding mitochondrial DEVDase activity was still higher (Fig. 1A). As expected, the DEVDase activity was always the highest in the whole cell lysate (Fig. 1A).
Next, we further examined the DEVDase activity in the mitochondrial fractions using a cell-free system. Mitochondria and cytosol were isolated from 1.5-h BMD188-treated or untreated PC3 cells and used in DEVDase activity measurements. Untreated mitochondria, either lysed (with Nonidet P-40) or unlysed, had negligible caspase activity (Fig. 1B, lanes  1 and 2). Untreated cytosol consistently showed slightly higher basal level DEVDase activity (Fig. 1B, lane 5), which did not change when co-incubated with untreated mitochondria (Fig.  1B, lanes 6 and 7). When the mitochondria from BMD188treated PC3 cells were co-incubated with untreated cytosol, significantly increased DEVDase activity was observed (Fig.  1B, lanes 8 and 9; note that higher DEVDase activities were observed in these experiments than in Fig. 1A because higher amounts (i.e. 100 g) of protein and substrate (30 M) were used). Our initial interpretation was that cyt. c released from the BMD188-activated mitochondria caused caspase activation in the cytosol. Surprisingly, however, when the activated mitochondria were directly used in DEVDase activity measure-ment, in the absence of untreated cytosol, essentially similar levels of increased DEVDase activity were observed ( As expected, both Z-DEVD-fmk, a caspase-3/7-specific inhibitor, and Z-VAD-fmk, a pan-caspase inhibitor, inhibited the mitochondrial and cytosolic DEVDase activity (Fig. 1C), just as both inhibited the BMD188-induced apoptosis (30,32). Collectively, the data in Fig. 1 suggest that the BMD188-activated mitochondria possess DEVDase activity, which most likely re-

FIG. 2. Caspase activation in vitro.
A, caspase activity expressed as the raw fluorescence unit as a result of Ac-DEVD-AFC hydrolysis. One hundred g of cytosol or mitochondria (mito) from untreated (control) or BMD188-treated (40 M; 90 min) PC3 cells was incubated with 30 M Ac-DEVD-AFC substrate at 37°C for 0 (i.e. immediate measurement following mixing), 30, 60, 90, and 120 min. The fluorescence measured at A 505 was directly plotted as a function of time. Triplicate samples were run for each time point, and the data represent mean Ϯ S.D. Note that the mitochondrial samples had lower background fluorescence levels (i.e. the values at time 0) than the cytosolic samples; the former was 0.4 -0.9 and the latter was 5-7.5. B, caspase activity expressed as fold activation. The data obtained in A were plotted as fold activation that was derived by dividing the mean fluorescence value at each time point by the mean fluorescence value at time 0. Thus, the fold activation at time 0 was considered as 1. C, caspase activity expressed as corrected fluorescence, which was derived by dividing the mean fluorescence value at each time point in BMD188-treated mitochondria or cytosol by the mean fluorescence value at corresponding time point in control mitochondria or cytosol, respectively. Therefore, DEVDase activities at all time points in the control mitochondria and cytosol were 1 (not plotted).
FIG. 1. Activated mitochondria possess caspase activity. A, mitochondria, cytosol, and whole cell lysate were prepared from equal numbers of PC3 cells (i.e. 10 7 cells for each fraction) treated with BMD188 (40 M) for the indicated time intervals and used to measure DEVDase activity using 50 g of protein and 10 M Ac-DEVD-AFC substrate. The DEVDase activity was expressed as fold activation over the control, i.e. time 0. The data represent mean Ϯ S.E. derived from three independent experiments. B, cell-free reconstitution experiment. One hundred g of mitochondria (mito) was prepared from either control or BMD188-treated (40 M; 90 min) PC3 cells. The mitochondria, either lysed (using Nonidet P-40) or unlysed, were either directly utilized in the DEVDase measurement (lanes 1-4) or first co-incubated with 100 g of cytosol prepared from untreated PC3 cells (lanes 6 -9) for 1 h at 37°C before incubation with 30 M DEVD-AFC substrate (1 h at 37°C). Lane 5 was the untreated cytosol alone. Lane 10 was caspase-3 activity measured with 100 g of whole cell lysate from the same batch of PC3 cells treated with BMD188. The data represent mean Ϯ S.E. To characterize further the DEVDase activity associated with the BMD188-activated mitochondria, we purified mitochondria using the Percoll gradient and the corresponding cytosol from control or 1.5-h BMD188-treated PC3 cells and incubated them with DEVD-AFC substrate in vitro at 37°C for a continuous period (i.e. from 0 to 120 min). The caspase activity was then plotted as either raw fluorescence, fold activation, or corrected fluorescence as a function of incubation time. As shown in Fig. 2, BMD188-activated mitochondria demonstrated a time-dependent increase in caspase activity. When expressed as raw fluorescence, the cytosol samples possessed higher fluorescence values than the corresponding mitochondrial samples at all time points ( Fig. 2A). However, cytosol samples also possessed higher background (i.e. time 0) fluorescence values (see Fig. 2 legend). Therefore, when plotted as fold activation (Fig. 2B) or corrected fluorescence (Fig. 2C), the mitochondrial samples demonstrated higher DEVDase activities than did the corresponding cytosolic samples.
Similar DEVDase activity was also detected in the mitochondrial fractions in several other MADAP models (30) including PC3 cells treated with butyrate, GM701 cells treated with BMD188, LNCaP cells subjected to serum starvation, and MDA-MB-231 cells treated with etoposide (not shown).
Detection of Proteolytically Activated Caspase-9 and Caspase-3 in the Mitochondria during Apoptosis-In principle, the DEVDase activity in the activated mitochondria may result from either procaspases becoming activated in the mitochondria or activated caspases translocating to the organelle, or both. We explored these possibilities by focusing on the activation of procaspase-9 and -3, two pivotal caspases in the apoptosome pathway initiated by cyt. c and Apaf-1.
In PC3 cells, significant amounts of procaspase-9 and procaspase-3 were localized in the mitochondria (Fig. 3A). Procaspase-9 (ϳ46 kDa) is activated mostly through auto- . Thirty (cytosolic) or 60 (mitochondrial) g of proteins was used in Western blotting for caspase-9, caspase-3, Apaf-1, and actin, as indicated. PARP cleavage and apoptosis were determined as described under "Materials and Methods." The asterisks on the left of A and B are novel or nonspecific bands detected by the antibodies against caspase-9 or caspase-3. All data are representative of at least three independent experiments. activation via Apaf-1-mediated oligomerization, resulting in proteolytic cleavage at Asp 315 to generate a 35-kDa fragment (p35) containing the caspase recruitment domain and the large subunit (38) and to expose the ATPF motif in the N terminus of the linker region to interact with XIAP and Smac/Diablo (39,40). Partially activated caspase-9 activates caspase-3, which in turn can cleave procaspase 9 at Asp 330 to generate a 37-kDa fragment (p37) (containing the caspase recruitment domain, large subunit, and the linker region) and ϳ10-kDa small subunit (38 -41). In our experiments, we utilized an antibody that recognizes both the proform and the p37/p35 bands (Table I).
As shown in Fig. 3A, 45 min post-BMD188 treatment, the mitochondrial procaspase-9 protein level was significantly reduced without the appearance of the p37/p35 bands, suggesting that, by this time point, the majority of the mitochondrial procaspase-9 was probably released from rather than activated in the organelle. Indeed, at 45 min, increased procaspase-9 was observed in both cytosol (Fig. 3A) and perinuclear area (see below). Also, the LEHDase assays did not detect caspase-9 activity (not shown).
In the cytosol, caspase-9 cleavage was observed at 45 min (Fig. 3A). By 1 h, the maximum level of caspase-9 cleavage products was observed in the cytosol (Fig. 3A). In the meantime, the p37/p35 bands began to appear in the mitochondria (Fig. 3A), suggesting that, in BMD188-treated PC3 cells, procaspase-9 became activated first in the cytosol and then the activated caspase-9 translocated to the mitochondria. In support, between 1 and 4 h, decreased amounts of activated caspase-9 were observed in the cytosol (Fig. 3A). At the same time, increasing levels of the p37/p35 products in the mitochondria were observed, whereas the mitochondrial procaspase-9 bands remained fairly constant (Fig. 3A). Interestingly, a prominent ϳ36-kDa protein band was detected by the anticaspase-9 antibody only in the mitochondria, which began to decrease at 1 h and disappeared by 4 h after BMD188 treatment (Fig. 3A). Whether this band represents a caspase-9 isoform (42,43) or just a nonspecific protein remains to be determined.
Procaspase-3 (ϳ32 kDa) is generally first cleaved by caspase-8, caspase-9, caspase-10, or granzyme-B at Asp 175 to generate the p12 small subunit and the ϳp24 large subunit that still contains the pro-domain. Then the p24-p12 complex is further cleaved at Asp 9 , possibly through its auto-catalytic activity, to generate p20, which could be further cleaved at Asp 28 to produce the p17 fragment (41,44,45). Frequently, a p31 fragment can also be detected, resulting from the procaspase-3 removed of the first 9 amino acids in the pro-domain. Various experiments have demonstrated that the p20/p17/p12 bands represent catalytically active caspase-3 (44,45). We utilized an antibody that recognizes both procaspase-3 and most cleavage products (Table I). In BMD188-treated PC3 cells, similar to caspase-9, the mitochondrial procaspase-3 was significantly reduced at 45 min (Fig. 3A). Different from caspase-9 by 45 min, the p20 and the p17 caspase-3 bands were detected simultaneously in the cytosol and mitochondria (Fig. 3A). By 1 h when significant amounts of caspase-3 activation occurred in both cytosol and mitochondria ( Fig. 1 and Fig. 3A), PARP cleavage was observed, and 55% of the cells became apoptotic (Fig. 3A). Caspase-3 cleavage overall paralleled the DEVDase activity (Fig. 1). Of interest, the anti-caspase-3 antibody also detected an unknown ϳ19-kDa band in untreated cytosol and mitochondria, which decreased as apoptosis proceeded (Fig.  3A).
To determine whether appearance of the activated caspases in the mitochondrial fraction is cell type-specific, we treated GM701 fibroblasts with BMD188. As in BMD188-treated PC3 cells, the mitochondrial procaspase-9 and, in particular, procaspase-3 were decreased by 30 min (Fig. 3B). Whereas the maximum decrease of procaspase-9 was observed by 30 min, the procaspase-3 level continued to decrease until the end of the treatment, i.e. 4 h (Fig. 3B). Activated forms of caspase-9 and caspase-3 were simultaneously detected at 2 h post-BMD188 treatment when PARP was cleaved and 56% cells were apoptotic (Fig. 3B). At 2 h, the p37/p35 caspase-9 fragments were detected in the mitochondria without an obvious decrease in the intensity of the mitochondrial procaspase-9 band (Fig. 3B), suggesting that, as in the BMD188-treated PC3 cells, the mitochondrial activated caspase-9 resulted, most likely, from translocation from the cytosol. In support, 4 h post-treatment, most procaspase-9 became cleaved in the cytosol, but the p37/35 bands actually decreased compared with those at 2 h (Fig. 3B). In the meantime, the p37/p35 bands in the mitochondria significantly increased without a corresponding decrease of the proform (Fig. 3B), strongly suggesting that the activated caspase-9 in the cytosol had translocated to the mitochondria. Again, as observed in PC3 cells, caspase-3 activation was observed simultaneously in the cytosol and mitochondria (Fig. 3B). Although the mitochondrial procaspase-3 decreased at 30 min and further decreased at 1 h, no cleavage products were observed at these time points (Fig. 3B). The anti-caspase-9 and anti-caspase-3 antibodies again detected the ϳ36and ϳ19-kDa bands, respectively, more obviously in the mitochondrial fractions (Fig. 3B). Accompanying the cleavages of procaspase-9 and procaspase-3, increased LEHDase (ϳ2-fold increase) and DEVDase (up to 8-fold increase) activities were detected in both cytosolic and mitochondrial fractions (not shown).
In both PC3 epithelial cells and GM701 fibroblasts treated with BMD188, we observed an early decrease of procaspase-9 and/or -3 in the mitochondria without concomitant cleavage (Fig. 3, A and B), suggesting that the mitochondrial procaspases may have been released from the organelle into the cytosol or other subcellular compartments, as observed in other apoptosis models (17-19, 21, 22). We carried out an immunofluorescence experiment in GM701 (Fig. 4) and PC3 (not shown) cells to explore this possibility. Procaspase-9 and -3 were homogeneously distributed in GM701 cells (Fig. 4, A and E; note that the mitochondrial localization of these procaspases was not very obvious due to their overwhelming expression in the cytosol). By 30 min (not shown) or 1 h (Fig. 4, B and F) post-BMD188 treatment, there was a prominent increase of both caspase-9 and caspase-3 staining in the nuclear area in most of the cells. Since by 30 min to 1 h there was no caspase-9 or caspase-3 activation in BMD188-treated GM701 cells revealed either by Western blotting (Fig. 3B) or by activity assays (Ref. 30; data not shown), the increased staining in the nuclear area therefore most likely resulted from the procaspase, rather than the active caspase, translocation from the mitochondria. Another piece of supporting evidence for this conclusion was that all cells that showed perinuclear staining of caspases were healthy and alive (Fig. 4, A-H), and only ϳ10% of the cells were apoptotic at this time frame (Fig. 3B) (30), thus suggesting that the caspases had not been activated. The increased perinuclear staining of procaspases-9 or -3 generally was slightly larger than the DAPI staining (e.g. compare Fig. 4, B and F versus D  and H), suggesting that the translocated procaspases were perhaps clustered outside the organelle.
To determine whether inducers other than BMD188 also can induce association of activated caspase-9 and -3 with the mitochondria, we subjected LNCaP cells to serum starvation, a MADAP model involving an early mitochondrial activation (30). As shown in Fig. 3C, activated caspase-9 and -3 were detected in the cytosol as well as in the mitochondria on day 6, when PARP was cleaved and 55% of the cells were apoptotic. In LNCaP cells there was only low levels of procaspase-9 in the mitochondria (Fig. 3C). When cytosolic procaspase-9 was cleaved on day 6, the p35 band was preferentially observed in the mitochondria, whereas the p37 band was most prominent in the cytosol (Fig. 3C). On day 8 when most cytosolic procaspase-9 was processed, essentially no p37/p35 bands were detected in the cytosol, whereas increased levels of both bands were seen in the mitochondria (Fig. 3C). These results once again suggest that the mitochondrial active caspase-9 most likely results from the translocation of the proteolytically activated caspase-9 in the cytosol. There was a slight decrease in the mitochondrial procaspase-9 in LNCaP cells starved for 8 days (Fig. 3C), raising the possibility that during late apoptosis the mitochondrial procaspase-9 might be activated in situ. Different from procaspase-9, LNCaP cells expressed abundant procaspase-3 in the mitochondria, which showed a time-dependent decrease, similar to cytosolic procaspase-3 (Fig. 3C).
By day 6 when PARP was cleaved, decreased procaspase-3 and increased active p17 fragments were observed in both cytosol and mitochondria (Fig. 3C). By day 8 when most PARP was degraded, a further decrease of procaspase-3 and increase in the p20/p17 fragments were observed in both compartments. However, the increase of the p17 active caspase-3 in the mitochondria was much more prominent than that in the cytosol (Fig. 3C).
Similarly, in MDA-MB-231 breast cancer cells treated with etoposide, yet another MADAP model (30), we also observed activated caspase-9 and caspase-3 in the mitochondria (not shown).
To determine whether only MADAP inducers could cause association of active caspases with the mitochondria, we treated GM701 cells with staurosporine, which causes cell death independently of MRC functions and without mitochondrial activation (30,32). As shown in Fig. 3D, although there was no early exodus of mitochondrially localized procaspase-9 or -3, activated forms of both caspases were detected in the mitochondria, simultaneously with their activation in the cytosol, PARP cleavage, and increased apoptosis.
The conclusions from the above experiments are as follows. 1) The presence of the activated caspase-9 and -3 seems to be a general phenomenon during apoptosis. 2) The mitochondrial caspase-9 activation generally appears either after the cytosolic caspase-9 activation (Fig. 3A) or concomitant with the cytosolic caspase-9 activation. In the latter case, the first appearance of the activated caspase-9 in the mitochondria is rarely accompanied by a decrease in the corresponding proform, whereas the first appearance of the activated caspase-9 in the cytosol is always accompanied by a decrease in procaspase-9 (Fig. 3, B-D). Late in apoptosis, however, the increased active caspase-9 levels in the mitochondria may (Fig. 3, C and D, the last lanes) or may not (Fig. 3, A and B, the last lanes) be accompanied by a decrease in the mitochondrial procaspase-9. These results suggest that the mitochondrial active caspase-9 may initially come from translocation from the cytosol but, late during apoptosis, the increased activated caspase-9 in the mitochondria may result from translocation and/or caspase-mediated activation in the organelle. 3) Activated caspase-3 is always detected simultaneously in the mitochondria and cytosol concomitant with the caspase-9 activation in the cytosol (Fig. 3,  A-D), raising a similar possibility that the activated caspase-3 in the mitochondria may result from translocation and/or caspase-mediated activation in the organelle.
Apaf-1 Localizes Exclusively in the Cytosol and Translocates to the Perinuclear Area during Apoptosis-The second point mentioned above suggests that the initiator caspase, caspase-9, may not be activated in the organelle in a cyt. c/Apaf-1-dependent manner. Theoretically, if procaspase-9 were activated inside the mitochondria in an apoptosome-dependent manner, Apaf-1 should be expressed in the mitochondria. Western blot analysis indicated that, in all cells studied, Apaf-1 was expressed exclusively in the cytosol (Fig. 3 and data not shown), consistent with an earlier report (46). Upon apoptosis induction, Apaf-1 showed a rapid and time-dependent decrease (Fig.  3, A-D). Immunofluorescent labeling revealed that Apaf-1, much like the mitochondrial procaspase-9 and -3 (Fig. 4, A-H), translocated to the perinuclear area in GM701 (Fig. 4, I-L) or PC3 (not shown) cells treated with BMD188 or in serumstarved LNCaP cells (not shown). In multiple Western and immunolabeling experiments, we failed to observe translocation of Apaf-1 to the mitochondria upon apoptosis induction ( Fig. 3 and 4; data not shown). These observations suggest that procaspase-9 is unlikely to be activated in the mitochondria in a cyt. c/Apaf-1-dependent manner. Active Caspases Are Located in the IMS of the Mitochondria-Next, we utilized the mitochondria from the BMD188-treated (3 h) or untreated GM701 cells to determine where the active caspases were located in the mitochondria, i.e. on the outer mitochondrial membrane (OMM) or in the IMS. When control mitochondria were treated with proteinase K alone, the procaspase-9 and procaspase-3 bands were detected (Fig. 5A, lanes  1 and 2). When the mitochondria were treated with proteinase K in the presence of Triton X-100 (to solubilize the organelle), procaspase-9 was completely degraded, whereas procaspase-3 was degraded into two specific fragments (Fig. 5A, lane 3). These results were consistent with these procaspases being located in the IMS (17-19, 21, 22). Likewise, when the BMD188-activated mitochondria were treated with proteinase K alone, the active caspase-9 (p37/p35) or caspase-3 (p17) bands were not significantly affected (Fig. 5A, lane 5). When these mitochondria were treated with proteinase K together with Triton X-100, both the pro-and active forms were degraded (Fig. 5A, lane 6; note that the bands marked by the dotted circles were a proteinase K degradation product rather than p17). These results suggest that the active caspase-9 and -3 were also present in the IMS. As an experimental control, voltage-dependent anion channel or VDAC, a protein mostly localized in the OMM with a small part of the molecule protruding to the cytosol (47,48), was not digested by proteinase K treatment alone (Fig. 5A, lane 2) but mostly degraded after solubilization with Triton X-100 (Fig. 5A, lanes 3 and 6). Note that proteinase K treatment alone appeared to have partially digested both caspase-9 and caspase-3 as well as VDAC in the BMD188-treated mitochondria (Fig. 5A; compare lanes 5 versus   lanes 4), which was consistent with the observations that the 3-h BMD188-activated mitochondria had significant mitochondrial membrane permeabilization or MMP (30) thus allowing partial access of proteinase K to the IMS.
In another set of experiments, we utilized a different strategy to confirm the IMS localization of the activated caspase-9 and caspase-3 in the mitochondria. Specifically, we adopted an alkali extraction protocol to strip the proteins localized either peripheral or integral to the OMM, depending on the harshness of the extraction conditions. As shown in Fig. 5B, under our experimental conditions, VDAC, an OMM protein, was extracted from the mitochondria but Hsp60, a mitochondrial matrix protein, was not. Under the same conditions, both pro-and the activated forms of caspase-9 and caspase-3 remained in the mitochondrial pellet (Fig. 5B), thus confirming that the activated caspases were localized in the IMS. Note that in both experiments BMD188 treatment resulted in an up-regulation of VDAC (Fig. 5A, lane 4 versus lane 1; Fig. 5B, lane 3 versus lane 1), consistent with our previous observations that MADAP inducers such as BMD188 systematically up-regulate MRC proteins as well as mitochondrially localized non-MRC proteins (30).
Association of Active Caspases with the Mitochondria Is Inhibited by CsA-In the following experiments, we addressed how the activated caspase-9 and -3 might end up in the IMS of the mitochondria. First, we utilized the BMD188-treated PC3 cells to assess the relationship between mitochondrial protein release, caspase activation, and caspase association with the mitochondria. Western blotting using a monoclonal anti-cyt. c antibody that specifically recognizes holo-cyt. c (30) revealed significant release of holo-cyt. c from the mitochondria as early as 5 min post-treatment (Fig. 6). By 15 min cyt. c release reached the peak level (Fig. 6). By 2 h, cyt. c accumulation in the cytosol decreased (Fig. 6) probably due to extensive cell death by this time (Fig. 3A) leading to cyt. c leak into the culture medium (30,49). Note that this antibody somehow did not react well with the holo-cyt. c in the mitochondrial fraction in certain cell types ( Fig. 6 and data not shown). Surprisingly, the 60-kDa mitochondrial matrix protein Hsp60 was also rapidly released; the release was observed as early as 5 min and peaked at ϳ30 min (Fig. 6). The release of both cyt. c and Hsp60 occurred before the activated caspase-9 and -3 appeared in the mitochondria at ϳ1 h (Fig. 6). The release of both cyt. c and Hsp60 was inhibited (i.e. delayed) by CsA, an inhibitor of MMP (48), but not by Z-VAD, a general caspase inhibitor (Fig. 6). Interestingly, CsA also appeared to decrease the levels of active caspase-9 and caspase-3 specifically in the mitochondria, especially at 1 h post-treatment (Fig. 6). By contrast, Z-VAD inhibited activation of caspase-9 and, more prominently, of caspase-3, leading to significantly reduced levels in both cytosol and mitochondria (Fig. 6). Together, these results suggest that 1) perturbation of the mitochondrial integrity, as evidenced by cyt. c and Hsp60 release, occurs prior to caspase activation and their association with the mitochondria; 2) a CsA-inhibitable mechanism(s) is involved in the initial mitochondrial protein release; and 3) a similar CsA-inhibitable mechanism(s) might also be involved in the later association of the activated caspases in the mitochondria.
Mitochondrially Localized Active Caspases Result Mostly from Translocation and Partly from in Situ Activation-Next, we performed several sets of cell-free experiments to further address how the active caspases in the mitochondria may have originated. In the first set (Fig. 7, A and B, lanes 1-4), mitochondria or cytosol prepared from untreated GM701 cells were activated with cyt. c. As shown in Fig. 7A (lane 4), cyt. c caused complete proteolysis of procasapse-9 in the cytosol generating  1 and 2) and BMD188-treated (lanes 3 and 4) GM701 cells were subjected to alkali extraction as described under "Materials and Methods." At the end of treatment, the supernatant (S) and pellet (P) were collected by centrifugation and used in Western blotting for caspase-9 and -3, Hsp60, and VDAC. All data are representative of 2-3 independent experiments. prominent p37/p35 bands. Similarly, stimulation of the cytosol with cyt. c activated caspase-3 to generate the p24/p20/p17 bands (Fig. 7B, lane 4). In contrast to cytosolic procaspases, neither the mitochondrial procaspase-9 nor the mitochondrial procaspase-3 was proteolytically activated by cyt. c (Fig. 7, A  and B, lanes 2), confirming that there is no Apaf-1 in the mitochondria.
In the second set of cell-free experiments (Fig. 7, A and B,  lanes 5-8), we first combined the cytosol and the mitochondria (i.e. Cyto ϩ Mito) and then activated the mixture with cyt. c. At the end, the mitochondrial pellets were separated from the cytosol, and caspase activation was assessed by Western blotting. As shown in Fig. 7, A and B (lanes 5 and 6), simple co-incubation of the cytosol with the mitochondria did not result in procaspase-9 or procaspase-3 activation. However, stimulation with cyt. c resulted in complete procaspase-9 activation (Fig. 7A, lane 7), just as stimulating cytosol alone with cyt. c. Significantly, nearly equal amounts of the cleaved p37/p35 bands were now detected in the mitochondrial pellet (Fig. 7A,  lane 8), and both the cytosolic and the mitochondrial p37/p35 bands were roughly half those when cytosol alone was activated with cytochrome c (Fig. 7A, compare lanes 7 and 8 versus  lane 4). These results suggest that ϳ50% of the cyt. c-activated caspase-9 in the cytosol translocated to the mitochondria. Likewise, cyt. c alone activated procaspase-3 in the cytosol and the mitochondria to generate the p24 and p20 bands and a small amount of p17 (Fig. 7B, lane 7). Surprisingly, only the active p20 fragment translocated to the mitochondria (Fig. 7B,  lane 8).
Next, we ran a time course experiment in which we first made cyt. c-activated cytosol and then co-incubated it with the untreated mitochondria for 1 or 3 h. As shown in Fig. 7C (lanes  1-3), no caspase-9 or caspase-3 activation was observed in unactivated cytosol or unactivated cytosol ϩ mitochondria mix. However, the cytosolic procaspase-9 was again completely activated by cyt. c (Fig. 7C, lane 4). Co-incubation of the cyt. c-activated cytosol with the untreated mitochondria for 1 h led ϳ50% of the p37/p35 bands to translocate to the mitochondria (Fig. 7C, lanes 4 -6). Co-incubation for 3 h rendered all of the p37 band and most of the p35 band to translocate to the mitochondria (Fig. 7C, lanes 7 and 8 versus lanes 5 and 6). In the meantime, there was a decrease in the mitochondrial procaspase-9 band (Fig. 7C, compare lane 8 versus lane 6), suggesting that procaspase-9 may also have been activated in the mitochondria. Similarly, co-incubation of the cyt. c-activated FIG. 6. CsA inhibits both initial mitochondrial protein release and later caspase association with the mitochondria. PC3 cells were pretreated with vehicle (control), CsA (10 M), or Z-VAD-fmk (50 M) for 1 h followed by BMD188 (40 M) treatment in the presence of these inhibitors for the times indicated. At the end of the treatment, cells were harvested for subcellular fractionation. Twenty five g of cytosolic or mitochondrial proteins were used in Western blotting of holo-cyt. c, Hsp60, caspase-9, or caspase-3. The asterisks on the left indicate unknown caspase-9 or -3 bands. Data are representative of 2-3 independent experiments.
FIG. 7. Cell-free activation and translocation of caspase-9 and caspase-3. Mitochondria and cytosol were prepared from untreated GM701 cells. Caspase activation and translocation to mitochondria were assessed by Western blotting. A and B, cell-free activation of caspase-9 (A) or caspase-3 (B) initiated by cyt. c. Freshly prepared mitochondria (25 g) or cytosol (40 g) was individually incubated with none or bovine heart cyt. c at 37°C for 2 h (lanes 1-4). In a parallel experiment, cytosol (Cyto) and mitochondria (Mito) were co-incubated with none or cyt. c (lanes 5-8). At the end of the incubation, mitochondria were pelleted, and cytosolic and mitochondrial samples were separated and probed for caspase-9 (A) or caspase-3 (B) (lanes 5-8). C and D, time course of caspase activation and translocation. Cyt. c-activated (lanes 4 -8) (2 h at 37°C) cytosol (40 g) was co-incubated with fresh mitochondria (40 g) at 37°C for 1 or 3 h. After incubation, mitochondrial pellet (P) and supernatant (S) were subjected to Western blotting for caspase-9 (C) or caspase-3 (D). As controls, unactivated cytosol (lane 1), and the supernatant and the pellet obtained from co-incubated unactivated cytosol and unactivated mitochondria (lanes 2 and 3, respectively), were also run on the gel. Data are representative of at least 3 independent experiments. cytosol with the mitochondria for 1 h resulted in specific translocation of the p20 active caspase-3 band to the mitochondria with a concomitant decrease in the cytosolic p20 (Fig. 7D, lanes  4 -6). There was also a decrease in the mitochondrial procaspase-3 (Fig. 7D, compare lane 6 versus lane 3). Co-incubation for 3 h resulted in the accumulation of more p20 band as well as the p17 and p24 bands in the mitochondria (Fig. 7C,  lanes 7 and 8 versus lanes 5 and 6), without a further decrease in the mitochondrial procaspase-3 (Fig. 7D, lane 8 versus  lane 6).
Mitochondrial Procaspases Can Be Activated in the Organelle-The above cell-free reconstitution experiments have established that the mitochondrial active caspase-9 and -3 result mostly from translocation. Some evidence suggests that the mitochondrial procaspases may also become activated by translocated active caspases (see above). Therefore, we performed several additional cell-free experiments to address further this possibility. First, we performed a cell-free recombination experiment in which the freshly isolated mitochondria were co-incubated with either untreated or BMD188-treated GM701 cell cytosol and then separated the mitochondrial pellet from the cytosol. The rationale is that the BMD188-activated caspases in the cytosol may directly activate the mitochondrially localized procaspases, or translocate to the mitochondria, or both. As shown in Fig. 8A, the BMD188-treated cytosol had active caspase-9 (p37/p35) and caspase-3 (p20/p17) (lane 5), whereas the control cytosol had neither (lane 3). Correspondingly, the control cytosol had more procaspases than the BMD188-activated cytosol (Fig. 8A, compare lane 3 versus lane   5). After co-incubation for 1 h, nearly identical amounts of both procaspase-9 or -3 and their activated forms were recovered from the supernatant (Fig. 8A, compare lane 6 versus 5). Nevertheless, activated forms of both caspase-9 and caspase-3 were identified in the mitochondrial pellets concomitant with a decrease in the corresponding procaspases in the mitochondria (Fig. 8A, compare lane 2 versus lane 1). When the co-incubation was extended to 2 h, a further decrease in the mitochondrial procaspases was observed (not shown). These results suggest that, in this cell-free system, small amounts of the activated caspases in the cytosol may have permeabilized the OMM (48,49), gained access to and subsequently activated the mitochondrially localized procaspase-9 and -3 in the organelle.
To explore further this possibility, we directly incubated the intact mitochondria with recombinant active caspase-3 to see whether the mitochondrially localized procaspase-9 or -3 can ever be activated in the organelle. As shown in Fig. 8B, unactivated cytosol (lane 1) and mitochondria (lane 2) showed only procaspase-9 and procaspase-3 bands. The recombinant caspase-3 showed p20, p17, and p12 bands (Fig. 8B, lane 3). Co-incubation of the cytosol with active caspase-3 generated, as expected, the p37 active caspase-9, resulting from procaspase 9 cleavage at Asp 330 by caspase-3 (40). Similarly, co-incubation of the mitochondria with active caspase-3 also generated the p37 active caspase-9 concomitant with a decrease of procaspase-9, both of which were detected exclusively in the mitochondrial pellet (Fig. 8B, compare lane 6 with lane 2). These results suggest that exogenously added active caspase-3 entered the mitochondria and activated the procaspase-9 in the organelle. Surprisingly, co-incubation of the cytosol with active caspase-3 did not lead to obvious caspase-3 activation in the cytosol (Fig.  8B, compare lane 4 with lane 3). When the mitochondria were incubated with the active caspase-3, most p12 was recovered from the mitochondrial pellet (Fig. 8B, lane 6 versus 3), and little p17 was recovered from the supernatant (Fig. 8B, lane 5 versus lane 3), whereas slightly more p20 was recovered from the supernatant (Fig. 8B, lane 5 versus lane 3), suggesting that most of the exogenously supplied p17/p12 (but not p20) caspase-3 translocated to the mitochondria. However, there was a prominent increase in the level of p17 in the mitochondrial pellet concomitant with decreased procaspase-3 as well as p20 (Fig. 8B, lane 6 versus lanes 2 and 3). Since the p17 fragment comes mainly from caspase-3-mediated autocatalysis (44,45), these results, together, suggest the following: 1) the p17/12 caspase-3 enters the mitochondria to activate the procaspase-9 and -3 in the organelle, and 2) the activated mitochondrial caspase-9 (i.e. the p37) may further activate the mitochondrial procaspase-3, which autocleaves more procaspase-3 to generate more p17. DISCUSSION The current study focuses on the mitochondria-associated active caspase-9 and caspase-3. The multifaceted results allow us to present a model (Fig. 9) to explain their location, origin, relationship with the cytosolic counterparts, and potential biological functions. In response to apoptotic stimulation, several mitochondrial proteins such as the IMS protein cyt. c and the matrix protein Hsp60 are released, involving CsA-sensitive MMP. In certain apoptotic systems, the mitochondrial procaspases are similarly released into the cytosol, some of which translocate to the perinuclear area. In a similar fashion, Apaf-1, which is located exclusively in the cytosol, also translocates to the perinuclear region (Fig. 9). The released cyt. c triggers apoptosome formation in the cytosol leading to caspase-9 and, subsequently, caspase-3 activation. The activated caspase-9 and -3 rapidly translocate back to the mitochondria where they may further activate residual or other FIG. 8. Activation of the mitochondrial procaspases in the organelle. A, cell-free recombination experiments to show that the active caspases in the BMD188-actived cytosols may activate the mitochondrially localized procaspases. Cytosol (cyto) and mitochondria (mito) were isolated from control or BMD188-treated GM701 cells. Control or treated cytosol (30 g) was then co-incubated with control mitochondria (40 g) for 1 h at 37°C. At the end of incubation, cytosol (S) and mitochondria (P) were separated by centrifugation followed by Western blotting for caspase-9, and -3. The asterisk indicates the unknown p19 band. B, recombinant active caspase-3 can activate the mitochondrially localized procaspase-9 and procaspase-3. Cytosol (40 g; lane 1) and mitochondria (30 g; lane 2), freshly prepared from untreated GM701 cells, were incubated with recombinant active caspase-3 (lanes 4 -6, respectively). S and P refer to the supernatant and mitochondrial pellet, respectively, at the end of the incubation. As a control, an aliquot (5 ng) of the recombinant caspase-3 was also run on the same gel (lane 3). Data are representative of 3 independent experiments. resident procaspases in the organelle, degrade mitochondrial proteins, disintegrate mitochondria, and help drive the apoptotic process to completion.
Active Caspases in the Mitochondria: Unlikely Involvement of Apaf-1 in the Activation of Procaspase-9 in the Organelle-A multitude of studies have demonstrated the presence of proand activated caspases in the mitochondria (14, 17-19, 21, 27, 28, 50), although this phenomenon has been challenged recently (51). The reason for this discrepancy remains unclear, but the negative results may likely result from cell type-specific variations, different antibodies used, insufficient protein loading, and utilization of digitonin-permeabilized mitochondria (51). Our well controlled experiments using mitochondria prepared by either Percoll gradient purification (Fig. 2) or differential centrifugation (the rest) demonstrate the presence of procaspases in unstimulated mitochondria and, more importantly, of activated caspases in apoptosis-stimulated mitochondria. Like their procaspase counterparts, the activated caspase-9 and -3 are localized in the IMS of the mitochondria (Fig. 5).
Our recent observation (30) that a wide diversity of apoptotic inducers cause an early up-regulation and accumulation of cyt. c in the mitochondria raises the following intriguing question. Could this increased mitochondrial cyt. c lead to procaspase-9 activation right in the organelle analogous to apoptosomemediated procaspase-9 activation in the cytosol? Several pieces of evidence make our initial hypothesis unlikely. First, in all our apoptotic models the initial appearance of the activated caspase-9 in the mitochondria occurs either following or concomitant with the caspase-9 activation in the cytosol without proteolytic degradation of the mitochondrial procaspase-9 (Fig.  3). These observations suggest that caspase-9 activation is initiated in the cytosol. Second, Apaf-1, the key adaptor molecule in apoptosome-mediated caspase-9 activation, is exclu-sively expressed in the cytosol, consistent with the observations of other investigators (46), and is never found in the mitochondria. Third, even after apoptotic stimulation, no association of Apaf-1 with the mitochondrial fractions is observed (Fig. 3,  A-D). Finally, cell-free reconstitution experiments demonstrate that cyt. c does not directly activate the mitochondrial procaspase-9 although it completely activates the cytosolic procaspase-9 (Fig. 7A), thus confirming the lack of Apaf-1 in the mitochondria. These observations make it clear that procaspase-9 cannot be activated inside the mitochondria in an Apaf-1/apoptosome-dependent manner.
Interestingly, in all the apoptotic models studied here, cytosolic Apaf-1 rapidly translocates to the perinuclear area, similar to recent findings (52) in different mammalian cells. Remarkably, the Apaf-1 homologue in Caenorhabditis elegans, CED-4, also translocates to the nuclear membrane during apoptosis (53). What could be the functions of this translocation? One clue is that, in addition to Apaf-1, cyt. c (52), procaspase-9 and -3 (this study), and many other caspases (22-24, 26, 28, 29) also become associated with the nucleus during apoptosis. It is possible that these molecules move together to assemble functional apoptosomes to induce caspase activation in or around the nucleus in order to disassemble the nuclear cytoplasmic barrier (54) and catalyze the systematic degradation of the nucleus.
Active Caspases in the Mitochondria Are Derived Mostly from the Translocation-If the Apaf-1-containing apoptosome is not involved, then how is the mitochondrial active caspase-9 brought about? Although Apaf-1-independent mechanisms recently emerged (11)(12)(13) could be involved, the most plausible explanation is that the active caspase-9 in the mitochondria comes from translocation from the cytosol. Supporting evidence comes from PC3 or GM701 cells treated with BMD188 in which increasing levels of the activated caspase-9 are seen in the mitochondria in the absence of corresponding decreases in the mitochondrial procaspase-9 levels (Fig. 3, A and B). Instead, the cytosolic procaspase-9 continuously decreases without a corresponding increase in the active caspase-9 (Fig. 3, A and B). More importantly, in PC3 cells the cytosolic caspase-9 activation occurs prior to the mitochondrial caspase-9 activation (Fig.  3A). Cell-free reconstitution experiments also provide clear-cut evidence that the cytosolically activated caspase-9 can translocate to the mitochondria. Translocation seems to be a temperature-and energy-dependent process; translocation is decreased at room temperature and stopped at 4°C. 2 The mitochondria-associated active caspase-3 generally appears at the same time when the cytosolic caspase-3 becomes activated (Fig. 3, A-D), but at least in some cases, an early translocation of the activated cytosolic caspase-3 to the mitochondria can be suggested. For example, in BMD188-treated PC3 cells, the mitochondrial caspase-3 activation is observed at 45 min post-treatment (Figs. 1, 2, and 3A). However, the active mitochondrial caspase-9 is not observed by this time (Fig. 3A), suggesting that the active caspase-3 in the mitochondria probably comes from translocation from the cytosol. In addition, increasing amounts of both active caspase-3 (i.e. the p20 fragment) as well as the procaspase-3 are observed in the mitochondria (Fig. 3A). Finally, cell-free reconstitution experiments show that the cytosolically activated caspase-3 can indeed translocate to the mitochondria (Fig. 7).
How may activated caspases translocate to the IMS of the mitochondria? It is still unclear how various procaspases, which generally lack an identifiable mitochondria-targeting sequence (51), get into the IMS. Evidence presented here suggests that a CsA-inhibitable mechanism(s) seems to be responsible, at least partially, for both the early release of the mitochondrial proteins such as cyt. c and Hsp60 as well as for the later translocation of the activated caspases back to the mitochondria. CsA, which has been shown to inhibit apoptosis in multiple model systems, is generally thought to block apoptosis by inhibiting MMP (48,49). CsA has been shown recently (55) to also inhibit the remodeling of the mitochondrial cristae and consequently inhibit cyt. c release. How mitochondrial proteins are released from the mitochondria is still controversial, but it seems to involve different types of channels consisting of proapoptotic Bcl-2 proteins such as Bax and Bak and some resident mitochondrial proteins such as VDAC (reviewed in Refs. 48 and 49). Consistent with our earlier data (30), CsA inhibits the early release of both cyt. c and Hsp60 (Fig. 6), thus suggesting the potential involvement of MMP in this process. Interestingly, CsA also appears to inhibit the later translocation of active caspases as it specifically reduces their levels in the mitochondria (Fig. 6). These results suggest that translocation of the activated caspases back to the mitochondria might use a CsA-sensitive mechanism(s) similar to those utilized for the initial mitochondrial protein release. It is unclear how the activated cytosolic caspase-9, most of which is thought to remain bound to the apoptosome (38 -40), gets to the mitochondria. The supramolecular openings formed by Bid, Bax, and the mitochondrial membrane lipids can release molecules of 2,000 kDa (56), which should be sufficient to "entertain" the 700-kDa to 1.4-MDa apoptosome. However, Apaf-1 is never detected in the mitochondria. It is possible that some activated caspase-9, especially those activated by caspase-3, may not be apoptosome-bound and thus be free to translocate to the mitochondria.
Recent evidence (6 -10) suggests that, in some apoptotic systems, caspase-2 activation precedes or may even be required for the compromise of the mitochondrial integrity such as cyt. c release. Our data suggest that, in our systems, caspase activation occurs after the disruption of the mitochondrial integrity and function (30) (Fig. 6 of this study). More importantly, Z-VAD does not affect the initial release of the mitochondrial proteins (Fig. 6). Since Z-VAD inhibits caspase activation, as expected, but the activated caspase-9 and -3 are still observed in the mitochondria, the reduced levels of the mitochondriaassociated active caspases (Fig. 6) mostly likely are caused by the overall inhibition of caspase activation rather than inhibition of translocation.
Active Caspases in the Mitochondria May Also Derive Partly from the Caspase-mediated Activation in the Organelle-Conceptually, translocated active caspases may subsequently activate the mitochondrial procaspases in the organelle. Several pieces of evidence support this possibility. First, late during apoptosis most mitochondrial procaspase-9 (Fig. 3D) and procaspase-3 (Fig. 3, B-D) becomes completely proteolyzed. Second, cell-free reconstitution experiments suggest that the mitochondrial procaspase-9 and -3 may become proteolytically activated in the organelle (Fig. 7, C and D). Third, cell-free recombination experiments also suggest direct procaspase activation in the mitochondria (Fig. 8A). Finally, both mitochondrial procaspase-9 and -3 can be activated by recombinant active caspase-3 (Fig. 8B). Various cell-free experiments using intact mitochondria (Figs. 7 and 8) suggest that a small amount of activated caspases may permeabilize the OMM (47,48), enter the IMS, and activate the mitochondrially localized procaspases. Under continuous apoptotic stimulation presumably the MMP persists, thus allowing the access of the cytosolically activated caspases to the IMS without the obvious need for caspase-mediated permeabilization, which explains why CsA but not Z-VAD inhibits caspase translocation to the mitochondria (Fig. 6).
The Potential Biological Functions of the Mitochondrially Localized Active Caspases-Multiple pieces of evidence suggest that the active caspases in the mitochondria may play a role in apoptosis. First, significant amounts of activated caspase-9 and -3 are accumulated in the mitochondria, especially late during apoptosis (Fig. 3, A-D). In support, significant caspase activities are detected in the mitochondria (Figs. 1 and 2; and data not shown). Second, in the reconstitution experiments, activated caspases preferentially translocate to the mitochondria (Fig. 7). Third, the preferential association of the active caspase-3 with the mitochondria seems to be essential for the complete activation of caspase-9. 2 Finally, the cytosolic procaspase-3 does not seem to be efficiently activated by exogenous active caspase-3 in the absence of the mitochondria (Fig.  8B).
Presumably, the activated caspase-9 and -3 in the mitochondria can play the following functions. First, they may further activate residual or other resident procaspases in the organelle (Fig. 9). Thus the expression of the active caspases in the mitochondria establishes a positive feedback amplification mechanism. Second, they may degrade some mitochondrial proteins. It is interesting to note that compared with the known caspase substrates in the cytosol and other organelles, few distinct mitochondrial substrates have been identified. Recently, caspase-9 is shown to be able to degrade certain cellular components other than caspases (57). Therefore, it is possible that active caspase-3 as well as caspase-9 in the mitochondria may specifically degrade some mitochondrial substrates such as the MRC protein complexes. Third, the mitochondrial active caspase-9 and -3 may further disintegrate mitochondrial integrity and function such as facilitating MMP, cyt. c release, and generation of reactive oxygen species (58). Together these functions help drive the apoptotic process to completion.