Introduction

Top Abstract Introduction Materials & Methods Results Discussion References

The process of apoptosis can be subdivided into three functionally distinct phases: (a) the initiation phase, during which a number of "private" signal transduction or damage pathways are activated in a stimulus-dependent fashion; (b) the common effector/decision phase, during which the cell "decides" to die; and (c) the degradation phase, beyond regulation, during which the cells acquire the morphological and biochemical hallmarks of apoptosis (1). Much of the data obtained by us (4) and our colleagues (11) are compatible with the notion that the disruption of mitochondrial membrane barrier function constitutes the decisive event of apoptosis (see also references cited in 18-20). This suggests a scenario in which the initiation phase is mainly premitochondrial, the effector/ decision phase is essentially mitochondrial, and the degradation phase is postmitochondrial.

A wealth of data suggests that the mitochondrion can function as an integrator of very different proapoptotic stimuli. Thus, unfavorable metabolic conditions facilitate opening of the mitochondrial permeability transition (PT)¹ pore: reactive oxygen species, depletion of glutathione, depletion of NAD(P)H₂, depletion of ADP/ATP, or supraphysiological levels of cytosolic Ca²⁺. Opening of the PT pore then causes a dissipation of the inner mitochondrial transmembrane potential (m), an increase in the matrix volume, and the mechanical disruption of the outer mitochondrial membrane (18). These mitochondrial changes also occur during apoptosis in an order that may depend on the cell type and on the death trigger (4). Multiple proapoptotic molecules have been shown to permeabilize the inner and/or outer membrane of purified mitochondria in vitro. This applies to apical caspases (10, 24), ceramide-induced cytosolic factors (24), amphipathic peptides (e.g., mastoparan, reaper, -amyloid [25-27]), and proapoptotic members of the Bcl-2 family (Bax, Bak, BH3 peptides [28- 31]). In contrast, Bcl-2 and Bcl-X_L stabilize the mitochondrial membrane barrier function (5, 10). One prominent target of several of these agents is the PT pore complex (PTPC), a multiprotein structure formed in the contact site between the inner and outer mitochondrial membranes. The purified PTPC permeabilizes membranes in response to Ca²⁺, oxidative stress, thiol cross-linking, and caspases (10). Recombinant Bcl-2 or Bcl-X_L has a direct inhibitory effect on PTPC (10). Thus, the mitochondrial PT pore (and perhaps additional mitochondrial structures) can integrate distinct apoptosis-inducing and apoptosis- inhibitory pathways.

Although there is no doubt that mitochondrial membrane function is perturbed during cell death, the mechanisms linking these perturbations to apoptosis are not completely elucidated. Cytochrome c, which is normally sequestered in the mitochondrial intermembrane space, has been shown to be released through the mitochondrial outer membrane early during apoptosis (13) and to interact with Apaf-1 and dATP (or ATP), as well as procaspase-9 in a reaction that culminates in the proteolytic activation of the caspase-9 zymogen (32, 33). Caspase-9 then can process and activate procaspase-3 (32), and caspase-3 can activate DNA fragmentation factor (DFF)/caspase-activated DNAse (CAD), a DNAse which would be responsible for DNA fragmentation (34, 35). In some experimental systems, cytochrome c (14 kD) appears to be the sole stable rate-limiting protein necessary for the activation of this or similar cascades leading to caspase-3 activation and DNA fragmentation in vitro (13, 15). However, at least one labile factor, apoptosis-inducing factor (AIF; ~50 kD), has been purified from the mitochondrial intermembrane space and has been defined by its capacity to induce nuclear apoptosis in cell-free systems in which no other mitochondrial or cytosolic proteins are present (7, 8, 24). Moreover, it must be stressed that changes in mitochondrial membrane permeability are lethal for the cell, even in conditions in which caspases are inhibited. In such conditions, PT pore opening causes a nonapoptotic pattern of cytolysis (9, 36), shedding doubts on the role of caspases as principal "executioners" of the death process (37, 38). Undoubtedly, however, caspases are important for the acquisition of apoptotic morphology (9, 36).

Stimulated by these premises, we decided to characterize the mitochondrial factors released after opening of the PT pore. Here, we show that at least four potentially apoptogenic proteins are released from mitochondria after opening of the PT pore in a Bcl-2-regulated fashion. In addition to cytochrome c and AIF, mitochondrial supernatants contain a DNAse activity and protease cleaving the caspase substrate Z-VAD.afc. These proteins are different from each other, based on their chromatographic separation and on their response to inhibitors. Most of the Z-VAD.afc- cleaving activity is due to the presence of caspase-2 and -9, which are present in mitochondria of different organs and which redistribute to the cytosol during apoptosis induction. These findings suggest the existence of several independent pathways linking opening of the PT pore to the commencement of apoptotic degradation.

	Materials and Methods

Top Abstract Introduction Materials & Methods Results Discussion References

Animals and Cell Lines.

Induction of PT.

Immunodepletion and Immunoblots.

Quantitation of AIF Activity, Z-VAD.afc Cleavage, and DNAse Assay.

Submitochondrial Fractionation, and Immunoelectron Microscopy.

Purification of Biological Activities Contained in the Mitochondrial Intermembrane Space.

Microinjection of Cells.

Immunofluorescence Analysis.

	Results

Top Abstract Introduction Materials & Methods Results Discussion References

Addition of Atr, a ligand of the adenine nucleotide translocator, causes isolated liver mitochondria to open the PT pore and physically disrupts the outer but not the inner mitochondrial membrane (6, 23). The supernatant of Atr-treated mitochondria thus contains water-soluble intermembrane proteins, including cytochrome c (23, 45; Fig. 1, D and H). In addition, it contains three different biological activities potentially relevant for the apoptotic degradation phase: (a) an AIF activity which causes isolated nuclei to undergo chromatin condensation and DNA loss (8, 24); (b) an activity which cleaves several caspase substrates, the optimal substrate being Z-VAD.afc (Z-VAD.afc  Z-DEVD.afc  Z-YVAD.afc > Z-VDVAD.afc >> Z-VEID.afc) (Fig. 1 B); and (c) a DNAse capable of digesting purified plasmid DNA (Fig. 1 C). Of note, it appears that the as yet uncharacterized mitochondrial DNAse capable of digesting purified DNA differs from previously described mitochondrial DNAse, such as endonuclease G, which acts in a sequence-specific fashion (46). None of these biological activities is affected by immunodepletion of cytochrome c (Fig. 1, A-D). These data indicate that the presence of cytochrome c is not rate-limiting for the activation of these molecules. Cytochrome c and the three biological activities (AIF, Z-VA-Dase, and DNAse) are released by Atr (Fig. 1, E-H) as well as by other PT pore opening agents such as Ca²⁺ or the reactive oxygen species donor tert-butylhydroperoxide (not shown). Cytochrome c, AIF, Z-VADase, and DNAse have been detected in the supernatant of Atr-treated mitochondria from different sources, including liver (Fig. 1, A-K), T cells (Fig. 1, E-H), heart, kidney, and brain (not shown). When the effect of Atr on the PT pore is blocked by BA or CsA, no such release occurs (Fig. 1, E-H). Similarly, transfection-enforced overexpression of Bcl-2 can prevent the release of all three biological activities (Fig. 1, E-H). Thus, the release of these factors occurs in a coordinate, Bcl-2- regulated fashion.

View larger version (80K): [in this window] [in a new window]   Fig. 1.   Three biological activities not requiring cytochrome c are coordinately released into the supernatant of mitochondria. (A) AIF activity not requiring cytochrome c. Supernatants of Atr-treated liver mitochondria were immunodepleted of cytochrome c or sham-immunodepleted using an anti-IL-2 antibody and tested for their capacity to induce hypoploidy in isolated HeLa nuclei in a cell-free system of apoptosis. (B) A Z-VAD.afc- cleaving activity not relying on the presence of cytochrome c. The same supernatants as in A were incubated with the fluorogenic caspase substrate Z-VAD. afc, and afc-dependent fluorescence was measured. (C) A DNAse activity not requiring cytochrome c. The mitochondrial supernatant (60 µg/ml) was incubated with purified supercoiled plasmid DNA (pUC), followed by horizontal agarose gel electrophoresis and detection of DNA with ethidium bromide. (D) Control of cytochrome c depletion by immunoblot. Supernatants of control mitochondria (Co.) or Atr-treated mitochondria (lanes 1-6) were left untreated (lane 1) or subjected to immunodepletion with antibodies specific for IL-2 (lane 2) or cytochrome c (lane 3). In addition, the immunocomplexes immobilized by beads were analyzed for the presence of cytochrome c (lane 4, anti-IL-2; lane 5, anti-cytochrome c; lane 6, no antibody). (E-H) Mitochondria purified from liver or T cell hybridoma cells expressing Bcl-2 or a Neo control vector were treated with the indicated combination of Atr, BA, and/or CsA and tested for AIF activity (E), Z-VAD.afc-cleaving activity (F), DNAse activity (G), or cytochrome c (H). (I-K) Inhibitory effect of Z-VAD.fmk. The supernatant of Atr-treated hepatocyte mitochondria was treated with Z-VAD.fmk (50 µM, 15 min) and tested for AIF activity (I) and for Z-VAD.afc-cleaving activity at two different protein concentrations (J) and for DNAse activity (K) at 40 µg protein/ml. In addition, the effect of ATA (5 mM) on the DNAse activity was assessed (K).

The supernatant of Atr-treated mitochondria was diluted in 50 mM Hepes-KOH at pH 6.75 and applied to a cation exchange (MiniS) column (Fig. 2). In these conditions, part of the AIF activity is retained in the column and can be eluted with 25 mM NaCl (Fig. 3, A and B). This latter preparation can be subjected to differential acetone precipitation, applied to a phenyl superose column, and eluted as ~50-kD AIF activity by reducing the salt concentration on a linear gradient, at a concentration of 400 to 200 mM (NH₄)2SO₄ (8; Fig. 2). Cytochrome c, the Z-VADase, and the DNAse were found in the flow-through of the MiniS column in the same conditions in which part of the AIF activity was retained (Fig. 3, C and D). The flow-through containing these activities can be applied to a cation exchange (MiniQ) column (Fig. 2, and Fig. 4 A), yielding a complete separation of residual AIF activity (Fig. 4 B), cytochrome c (not shown), the Z-VADase (Fig. 4 C), and the DNAse (Fig. 4 D), which elute at different levels of ionic strength (70, 200, 280, and 540 mM NaCl, respectively). These data indicate the presence of three biochemically distinct activities in the mitochondrial supernatant. Again, these activities are separate from cytochrome c.

View larger version (25K): [in this window] [in a new window]   Fig. 2.   Overview of the FPLC separation procedure. Mitochondrial intermembrane proteins were sequentially applied to MiniS cation exchange, MiniQ column anion exchange, or phenyl superose hydrophobicity columns as detailed in Materials and Methods and in Results. This methodology allows for the separation of all biological activities analyzed in this paper.

View larger version (43K): [in this window] [in a new window]   Fig. 3.   Selective retention of AIF activity on a MiniS cation exchange column. (A) Intermembrane proteins soluble in 50 mM Hepes-KOH (pH 6.75) were applied to a MiniS FPLC column and eluted by 25 mM NaCl. Fractions (40 µl) were recovered for further biological testing as in the legend to Fig. 1. Data are shown for two different experiments, one in which intermembrane proteins were obtained after addition of Atr (5 mM, 30 min; thin line) and one in which intermembrane proteins were obtained after preincubation of mitochondria with 100 µM Z-VAD.fmk for 15 min (thick line). (B) Profile of AIF activity obtained in the presence or absence of 100 µM Z-VAD.fmk. Z-VAD.fmk was either added to the mitochondria before treatment with Atr () or it was added to the supernatant (Sn) of mitochondria after treatment with Atr (). (C) Z-VAD.afc-cleaving activity measured in the presence or absence of Z-VAD.fmk added to the supernatant. (D) DNAse activity. Results are representative of ten independent experiments.

View larger version (69K): [in this window] [in a new window]   Fig. 4.   Biochemical and functional separation of biological activities. (A) The flow-through of the MiniS column shown in Fig. 3 (fractions 2-4) was recovered in 50 mM Tris-HCl (pH 8.5), applied to a MiniQ anion exchange column, and eluted by increasing the concentration of NaCl in fractions of 150 µl. These fractions were subjected to further evaluation as in the legend to Fig. 1, namely quantitation of AIF activity (B), Z-VAD.afc-cleaving activity (C), and DNAse activity (D). The numbers in B-D refer to the fractions eluted from the MiniQ column in A. Alternatively, supernatants of Atr-treated mitochondria were tested for AIF activity (E), Z-VAD.afc- cleaving activity (F), or DNAse activity (G) after addition of 1 mM ZnCl2 or 5 mM EGTA. Results are representative of five independent determinations.

The notion that the three biological activities detected in the mitochondrial intermembrane space are distinct is confirmed by their disparate inhibitory profiles (Fig. 1, I-K, and Fig. 4, E-G). AIF is not inhibited by EGTA or by Zn²⁺. In contrast, EGTA or Zn²⁺ inhibit the DNAse (Fig. 4, E and G). Neither AIF nor DNAse are inhibited by ATA (Fig. 1 K), indicating that they are distinct from DFF/ CAD, which is inhibited by ATA (35, 47). Different caspase inhibitors (Z-VAD.fmk, Z-YVAD.fmk, and Z-DEVD.fmk, ID₅₀ < 10 µM) and thiol-reactive agents (iodoacetamide, ID₅₀ ~ 250 µM), as well as Zn²⁺ (ID₅₀ ~ 10 µM) (which inhibits caspases [48]), can completely inhibit the Z-VADase activity (Fig. 1 J, and Fig. 4 F). These agents also tend to reduce the AIF activity, though with a higher ID₅₀, when added to the mitochondrial supernatant after induction of PT pore opening (Fig. 1 I). Moreover, the inhibitory effect of Z-VAD.fmk can be overcome by increasing the dose of the supernatant (Fig. 1 I). The AIF activity that is retained in the MiniS column (but not that contained in the flow-through) is completely resistant to inhibition by Z-VAD.fmk (Fig. 4 B). However, when added before Atr, Z-VAD.fmk becomes more efficient in inhibiting AIF than when added after PT pore opening (Fig. 3 B). Preincubation of mitochondria with Z-VAD.fmk, before Atr is added, provokes the disappearance of the MiniS column-retained protein peak corresponding to AIF (Fig. 3 A). This suggests that a Z-VAD.fmk-inhibitable enzyme is required for the release and/or maturation of AIF.

In conclusion, the pharmacological data confirm that at least three distinct activities are present in the supernatant of mitochondria: (a) a caspase-like activity cleaving Z-VAD. afc, (b) AIF, and (c) a DNAse.

The Z-VAD.afc-cleaving activity is fully inhibited by biotinylated VAD.fmk (ID₅₀ < 1 µM), which covalently reacts with the large subunit of caspases (49, 50). Therefore, we incubated the supernatant of Atr-treated mitochondria or the partially purified Z-VAD.afc-cleaving activity with biotinylated Z-VAD, and purified Z-VAD.biotin-binding proteins by avidin affinity chromatography (Fig. 2, and Fig. 5 A). Consistently, two of the major Z-VAD.biotin-binding proteins contained in these preparations have ~33 and 18 kD (Fig. 5 A). Other Z-VAD. biotin-binding proteins found in the supernatant of Atr-treated mitochondria (lane 2 of Fig. 5 A) are lost during the purification process of the Z-VAD.afc-cleaving activity (lane 7) or after affinity purification of Z-VAD.biotin- binding proteins (lane 9), which might suggest that they reflect nonspecific or low-affinity binding (Fig. 5 A). Two closely related caspases, caspase-2 and -9, possess the largest potentially Z-VAD.biotin-reactive subunit among all known caspases, with ~33 kD (51, 52). Accordingly, the Z-VAD.biotin-binding activity found in liver mitochondria reacted with antibodies specific for the NH₂-terminal subunit of caspase-2 (which recognizes the 48-kD procaspase and the 33-kD intermediate form built up by the NH₂-terminal prodomain and the 18-kD large subunit of the mature caspase-2) (Fig. 5 B). In addition, we found that crude supernatants of Atr-treated mitochondria, purified Z-VADase, and the Z-VAD.biotin-binding activity reacted with an antiserum specific for the large 18-kD subunit of caspase-9 (which recognize the 48-kD procaspase, the 33-kD intermediate, and the 18-kD large subunit of the mature caspase-9) (Fig. 5 C). Crude mitochondrial supernatants contain the caspase-2 and -9 zymogens (48 kD) as well as their proteolytic activation products (Fig. 5, B and C, lane 2), whereas the purified Z-VADase activity (lanes 7 and 9 in Fig. 5, A-C) only contains the intermediate and mature forms of these caspases. When mitochondria were treated with Z-VAD.fmk before induction of PT (lane 3 in Fig. 5, A-C), only the procaspases were detected in the mitochondrial supernatant. Submitochondrial fractionation experiments indicate that the pro-forms of caspase-2 and -9 are present in the mitochondrial intermembrane space (Fig. 6, A and B, lane 4) but absent from the matrix or from purified inner and outer mitochondrial membranes. Immunoelectron microscopy confirmed that caspase-9 is located in the mitochondrial periphery rather than in the outer mitochondrial membrane (Fig. 6 C). Again, caspase-2 and -9 are found as zymogens (~48 kD) in the intermembrane space when submitochondrial fractionation is performed in the presence of Z-VAD.fmk (Fig. 6, A and B, lane 8). In contrast, they become proteolytically activated when intermembrane proteins are purified in the absence of Z-VAD.fmk (Fig. 6, A and B, lane 4), suggesting that the intermembrane space contains all factors required for caspase-2/9 (auto-)activation. The caspase-2- and caspase-9-specific antibodies (which do not cross-react; Fig. 5, F and G) both reduced the Z-VAD.afc-cleaving activity contained in the supernatant of Atr-treated liver mitochondria (Fig. 7 B). When combined, they reduce this activity to background levels, indicating that both caspase-2 and -9 account for the Z-VAD.afc-cleaving activity (Fig. 7 B). In contrast, no Z-VAD.biotin-binding protein or caspase-2/9 immunoreactivity was found in purified AIF (Fig. 5, A-C, lane 6) or semipurified DNAse preparations (not shown). Caspase-2 and -9 were detected in the supernatant of Atr-treated mitochondria from different organs, including kidney, brain, spleen, and heart (Fig. 5, D and E).

View larger version (49K): [in this window] [in a new window]   View larger version (44K): [in this window] [in a new window]   Fig. 5.   Identification of the Z-VAD- binding activity as caspase-2 and -9. (A) Proteins reacting with biotinylated VAD.fmk in the intermembrane space of liver mitochondria. The supernatant of control mitochondria (lane 1) or of Atr-treated mitochondria (lanes 2 and 3), the flow-through of the MiniS column (see Fig. 2 and Fig. 3 A; lanes 4 and 5), purified AIF (lane 6), or the Z-VAD.afc-cleaving activity eluting at 280 mM from the MiniQ column (see Fig. 2 and Fig. 4 A; lanes 7 and 8) were allowed to react with biotinylated VAD.fmk, either without pretreatment (lanes 1, 2, 4, 6, and 7) or after preincubation with Z-VAD.fmk (lanes 3, 5, and 8). Note that Z-VAD.fmk has been added to mitochondria before Atr (line 3). In addition, the proteins reacting with biotinylated VAD.fmk retained on an avidin column were purified (lane 9). These proteins, which contained approximately similar levels of Z-VAD.afc-cleaving activity (10 U) or ~100 ng purified protein (lane 6) were separated by SDS-PAGE, blotted onto nitrocellulose, and subjected to the detection of biotinylated VAD.fmk using an avidin-based detection system. (B and C) The same blot as in A was subjected to immunodetection with antibodies specific for caspase-2 (B) or -9 (C). (D and E) Mitochondria from different organs were purified and cultured for 30 min in the presence or absence of 5 mM Atr, followed by immunoblot detection of caspase-2 (D) or -9 (E). Results are representative of two to four independent experiments. (F and G) Specificity control of caspase-2- and caspase-9-specific antisera. Recombinant caspase-2 (lane 1), -3 (lane 2), or -9 (lane 3) was immunoblotted (100 ng/lane), followed by immunodetection with the caspase-2 (F) or caspase-9 (G)-specific antibody. Similarly, caspase-2- and caspase-9-specific antibodies fail to recognize caspase-6 and -7 (not shown).

View larger version (26K): [in this window] [in a new window]   View larger version (109K): [in this window] [in a new window]   Fig. 6.   Submitochondrial localization of caspase-2 and -9. Purified mitochondria were either lysed by osmotic shock (total preparation, lane 1) or were subjected to fractionation into matrix (lanes 2 and 6), inner membrane (lanes 3 and 7), intermembrane (lanes 4 and 8), or outer membrane (lanes 5 and 9) proteins, in the absence (lanes 2-5) or presence (lanes 6-9) of 100 µM Z-VAD.fmk throughout each single step of the fractionation procedure. Thereafter, equivalent amounts of protein (15 µg/lane) were subjected to immunoblot detection of caspase-2 (A) and -9 (B). A representative immunoelectron micrograph of mitochondria labeled with a specific caspase-9 antibody is also shown (C).

View larger version (40K): [in this window] [in a new window]   Fig. 7.   Neutralization of the Z-VAD.afc-cleaving activity by immunodepletion of caspase-2 and -9. Soluble intermembrane proteins were depleted of caspase-2 and/or -9 using specific antisera, and the AIF activity (A) or Z-VAD.afc-cleaving activity (B) was assessed. Typical results out of three experiments are shown.

In conclusion, procaspase-2 and -9 are present in mitochondria of different organs and are released and activated upon induction of PT.

As expected from the fact that the Z-VADase does not cause isolated nuclei to become hypoploid in vitro (Figs. 2-4), recombinant caspase-2 and -9 have no direct apoptogenic effect in a cell-free system involving purified nuclei cultured in the absence of additional cytoplasmic factors (Fig. 7 A). However, caspase-2 and -9 do induce nuclear apoptosis, reduction of cell size, and phosphatidylserine exposure on the outer leaflet of the plasma membrane when microinjected into cells (Fig. 8), indicating that they can activate cellular factors capable of inducing apoptosis. The purified Z-VAD. fmk-resistant fraction of AIF also causes nuclear apoptosis upon microinjection, and this effect is not inhibited by Z-VAD.fmk (not shown). These data indicate that several mitochondrial factors (AIF, caspase-2, and caspase-9) may have a primary apoptogenic effect in vivo.

View larger version (29K): [in this window] [in a new window]   Fig. 8.   Apoptosis induction by microinjection of recombinant caspases. Rat-1 fibroblasts were left untreated or cultured with etoposide (1 µM, 6 h), in the absence or presence of Z-VAD.fmk (100 µM), to obtain a positive or negative control of apoptotic morphology, respectively. Alternatively, cells pretreated with Z-VAD.fmk (100 µM, 60 min) or left untreated were microinjected with buffer only or with recombinant caspase-2 or -9 (3 U/µl). Microinjected viable cells (100-200 per session, two to five independent sessions of injection) could be identified because the injectate contained FITC-dextran (0.25% [wt/vol], green fluorescence; not shown). Microphotographs representing the dominant (>70%) phenotype of microinjected cells stained either with Hoechst 33342 (blue fluorescence, A) or with Annexin V (red fluorescence, B) are shown after 90 min of culture. Z-VAD. fmk pretreatment prevented phosphatidylserine exposure induced by etoposide, caspase-2, or caspase-9, as measured with Annexin V (not shown).

To further investigate the in vivo impact of caspase-2 and -9, we analyzed their subcellular distribution during the process of apoptosis. Immunofluorescence detection revealed a granular pattern of distribution of both caspases, similar to that of cytochrome c in Rat-1 fibroblasts (Fig. 9), U937 myelomonocytic cells, SHEP neuroblastoma cells, and 2B4.11 T cell hybridoma cells (not shown). Upon induction of apoptosis with staurosporine A (Fig. 9), etoposide, or ceramide (not shown), the granular pattern of fluorescence is replaced by a diffuse pattern, indicating the abolition of subcellular compartmentalization of these molecules. This interpretation was confirmed by subcellular fractionation. Mitochondria and cytosols from T cell hybridoma cells induced to undergo apoptosis in response to the synthetic glucocorticoid receptor agonist DEX were purified. As shown in Fig. 10, in this cell line both procaspase-2 and -9 are localized in mitochondria. Some caspase-9 but no caspase-2 were found in the cytosolic fraction of untreated cells. Upon stimulation with DEX, procaspase-2 and -9 redistribute to the cytosol where they are at least partially processed, as revealed by the presence of bands corresponding to 33 kD (caspase-2) and 33 and 18 kD (caspase-9). Both the redistribution of these caspases and their activation are inhibited in cells overexpressing Bcl-2 (Fig. 10), thus confirming the inhibitory effect of Bcl-2 on caspase release observed in isolated mitochondria (Fig. 1).

View larger version (78K): [in this window] [in a new window]   Fig. 9.   Loss of subcellular compartmentalization of cytochrome c, caspase-2, and caspase-9. Rat-1 cells were left untreated or were cultured in the presence of staurosporine (4 h, 1 µM), fixed, and stained to determine the subcellular distribution of the indicated molecules by indirect immunofluorescence analysis.

View larger version (48K): [in this window] [in a new window]   Fig. 10.   Redistribution of caspase-2 and -9 from the mitochondrion. T cell hybridoma cells transfected with a Neo control vector or with Bcl-2 were cultured in the absence or presence of 1 µM DEX, followed by subcellular fractionation, as described in Materials and Methods. Equivalent amounts of proteins were subjected to immunoblot analysis in order to determine the subcellular localization and activation of cytochrome c (A), caspase-2 (B), or caspase-9 (C). Results are representative of three independent determinations.

In summary, upon induction of apoptosis, the pro-forms of caspase-2 and -9 redistribute from mitochondria to the cytosol in a Bcl-2-inhibitable fashion. These caspases become activated after or during the mitochondrial release and may actively participate in the apoptotic process.

	Discussion

Top Abstract Introduction Materials & Methods Results Discussion References

Although some investigators have tacitly implied that cytochrome c would be the only relevant apoptogenic activity released from mitochondria (13, 32, 33, 53), the data presented in this paper suggest that mitochondria release several apoptogenic factors. These factors include AIF (7), a DNAse, and at least two caspase zymogens, namely caspase-2 and -9. Caspase-2 and -9 are found in the intermembrane space of mitochondria from five different organs (liver, kidney, heart, brain, and spleen), as well as in several cell lines (including different lymphoid and neuroblastoma cell lines), and are released in a Bcl-2-inhibitable fashion upon induction of PT in isolated mitochondria and upon apoptosis induction in cells. It has been reported very recently that a variable portion (10- 90%) of caspase-3 zymogen is located in the intermembrane space of mitochondria from different cell types (54). However, in mouse liver, we have failed to detect caspase-3, in accord with the very minor portion of caspase-3 zymogen detected in human liver mitochondria (54). Moreover, we failed to detect caspase-6, -7, or -8 in mouse liver mitochondria (not shown). As demonstrated here, during or after the release from mitochondria, the caspase-2 and -9 zymogens become proteolytically processed and enzymatically active. At present, the mechanisms responsible for this phenomenon are not clear. As a possibility, the release of caspase zymogens from mitochondria alters the equilibrium between caspase activators and local caspase inhibitors, thereby favoring their activation. Alternatively, procaspases could become activated by extramitochondrial factors (in cells) and/or by mitochondrial surface proteins with whom they normally cannot interact (in isolated mitochondria). Irrespective of the exact mode of activation, cytochrome c is not a rate-limiting factor for caspase-2 and -9 activation upon mitochondrial release, as suggested by experiments in which cytochrome c has been immunodepleted from mitochondrial supernatants. However, these results do not rule out the possibility that a rather low level of cytochrome c and/or early interactions between caspase zymogens and cytochrome c, within the intermembrane space, may participate in caspase activation. In addition to caspases, mitochondria also release AIF. Whereas one portion of AIF activity is well inhibited by caspase inhibitors (Z-VAD.fmk), another, biochemically distinct portion of AIF activity is Z-VAD.fmk resistant. Since this latter AIF activity disappears when mitochondria are pretreated with Z-VAD.fmk, it appears plausible that caspase activation is required for its generation and/or action. Using a different protocol of AIF purification (7, 8), in which both AIF species copurify, we have previously overlooked this phenomenon. In any case, the inhibitory profile of AIF, whose molecular characterization is still in progress, clarifies that it is not identical to caspase-2, -9, or -3 (which are all inhibited by Z-VAD. fmk) nor with DFF/CAD (which is inhibited by ATA and acts on naked DNA) (35, 47). Altogether, these data indicate that mitochondria contain several potentially apoptogenic proteins in addition to cytochrome c.

The data reported here add to the intricate relation between mitochondria and the caspase activation cascade. First, in apoptosis-induction pathways directly coupled to caspase activation, namely those involving ligation of death domain-containing surface receptors (prototype: CD95), caspases are activated at a premitochondrial stage, before the mitochondrial membrane permeability is perturbed (8, 55, 56). Thus, certain caspases (e.g., caspase-8 in the CD95 pathway) can be activated upstream of mitochondria and can, as shown by addition of recombinant caspase to isolated mitochondria, directly disrupt the barrier function of mitochondrial membranes (10, 57). Second, mitochondria contain caspase zymogens, in particular procaspase-2, -3, and -9 (54; and this paper), which are released upon permeabilization of the outer mitochondrial membrane. Third, mitochondrial intermembrane proteins, including cytochrome c and caspase-2 and -9, can activate postmitochondrial caspases, including caspase-3, -6, and -7 (8, 55), which in turn can directly act on mitochondrial membranes (8, 10, 57) and/or redistribute into mitochondria, as has been recently shown for caspase-7 (58). Thus, they would engage in a self-amplifying feedback loop in which mitochondrial membrane permeabilization causes caspase activation and vice versa.

This scenario helps to clarify the complex relationship between caspase activation and the (presumably) mitochondrial Bcl-2/Bcl-X_L checkpoint of the apoptotic process. Activation of upstream caspases (e.g., caspase-8) is not controlled by Bcl-2/Bcl-X_L (55, 56, 59). If caspase-8 (or secondarily, caspase-8-activated caspases) succeed in proteolytically inactivating Bcl-2/Bcl-X_L and/or bypassing its mitochondrial membrane-stabilizing effect, they can induce apoptosis (8, 10, 55, 60, 61). If not, Bcl-2 interrupts the apoptotic cascade by retaining caspase-activating agents such as cytochrome c (7, 15), AIF (7), and procaspase-2 and -9 (this paper) in the mitochondrial intermembrane space. This would explain how Bcl-2 abolishes the activation of downstream caspases, including caspase-3 and -6 (8, 55, 62, 63). In the nematode Caenorhabditis elegans, the death-regulatory machine is composed of three core interacting proteins: CED-4 (the equivalent to mammalian Apaf-1), the caspase CED-3, and the Bcl-2 homologue CED-9. Similarly, in mammalian cells Bcl-X_L, Apaf-1, and caspase-9 have been shown to interact (64). It is tempting to speculate that these complexes are formed at the mitochondrial outer membrane/intermembrane interface, where at least two of the three molecules reside. However, at present it is not clear whether the Bcl-X_L/Apaf-1/caspase-9 complex participates in the Bcl-2/Bcl-X_L-mediated regulation of mitochondrial membrane permeability or whether it is only formed after permeabilization of the outer mitochondrial membrane. If Apaf-1 is truly a cytosolic protein, as suggested previously (32, 33), then this latter possibility would apply.

Apoptosis-induction pathways directly coupled to caspase activation such as CD95 are rare. In most cases, apoptosis induction involves other second messengers such as ceramide, Ca²⁺, alterations in redox status, or changes in the expression level, subcellular distribution, and/or posttranslational modifications of Bcl-2 homologues. In all of these cases, mitochondria can respond by increasing the permeability of their membranes (4, 21, 22, 25), suggesting that they can integrate many different "private" proapoptotic pathways and unify them in one common pathway. Thus, the mitochondrion functions as an integrator of different death pathways.

In addition, mitochondria coordinate the cellular events of the apoptotic degradation at several levels. At the first level, mitochondria release factors capable of activating hydrolases (caspases and nucleases). As shown here, mitochondria release at least two caspases (caspase-2 and -9), which have been considered to be apical in a number of different apoptosis-induction pathways (34, 65). By virtue of these caspases, and caspase-activating factors such as cytochrome c, AIF, and a DNAse, mitochondria thus can trigger a series of catabolic reactions that participate in the acquisition of apoptotic morphology. At a second level, mitochondria generate reactive oxygen species due to interruption of electron transport on the respiratory chain, which lack cytochrome c, the electron shuttle between complexes III and IV (69). At a third level, PT pore opening causes a loss of mitochondrial antioxidant function and a cessation of ATP synthesis. These latter changes probably suffice to kill cells. Thus, overexpression of Bax (31, 36), Bak (70), or c-Myc (70), glucocorticoid receptor ligation (9, 71), DNA damage (9), or HIV-1 infection (72), all stimuli that usually induce apoptosis, can induce a nonapoptotic cell death in the presence of caspase inhibitors. Cells exposed to such proapoptotic stimuli in the context of caspase inhibitors undergo delayed cytolysis and fail to manifest several hallmarks of apoptotic death, including DNA fragmentation (9, 31, 36, 70). However, they do manifest an increased permeability of mitochondrial membranes with dissipation of the inner mitochondrial transmembrane potential (m) and/or the release of cytochrome c through the outer mitochondrial membrane (9, 16, 17, 31, 36, 71). Nonetheless, in normal conditions, in the absence of caspase inhibitors, the mitochondrial changes are closely tied to caspase activation, as can be deduced from the strong mutual relationship between mitochondrial membrane permeabilization and caspase activation (see above).

In summary, the mitochondrial release of several independent apoptogenic molecules, including caspase-2 and -9, reinforces the tight link existing between the disruption of mitochondrial membrane barrier function and the activation of the apoptotic degradation phase. Mitochondria fulfill a double role in apoptosis. In addition to integrating different proapoptotic initiation cascades, they coordinate the death response.