Granzyme B Induces BID-mediated Cytochrome c Release and Mitochondrial Permeability Transition*

Many cell death pathways converge at the mitochondria to induce release of apoptogenic proteins and permeability transition, resulting in the activation of effector caspases responsible for the biochemical and morphological alterations of apoptosis. The death receptor pathway has been described as a triphasic process initiated by the activation of apical caspases, a mitochondrial phase, and then the final phase of effector caspase activation. Granzyme B (GrB) activates apical and effector caspases as well as promotes cytochrome c (cyt c) release and loss of mitochondrial membrane potential. We investigated how GrB affects mitochondria utilizing an in vitro cell-free system and determined that cyt c release and permeability transition are initiated by distinct mechanisms. The cleavage of cytosolic BID by GrB results in truncated BID, initiating mitochondrial cytc release. BID is the sole cytosolic protein responsible for this phenomenon in vitro, yet caspases were found to participate in cyt c release in some cells. On the other hand, GrB acts directly on mitochondria in the absence of cytosolic S100 proteins to open the permeability transition pore and to disrupt the proton electrochemical gradient. We suggest that GrB acts by two distinct mechanisms on mitochondria that ultimately lead to mitochondrial dysfunction and cellular demise.

Granzyme B (GrB) 1 is an aspartyl serine protease located in the granules of cytotoxic T lymphocytes and natural killer cells that induces apoptosis (1,2). Upon cytotoxic T lymphocyte degranulation at the interface with a target cell, GrB crosses the plasma membrane and enters the cytoplasm aided by perforin (3,4). Execution of apoptosis by GrB requires the participation of the caspase family of proteases (5,6), which are central regulators of the apoptotic pathway for multiple cell death signals (7,8). Recent work suggests that mitochondria play an important if not key role in the regulation of cell death in mammalian cells (9 -13). Apoptotic signals induce mitochondria to release cytochrome c (cyt c), which interacts with APAF-1 in an ATP-dependent manner, and caspase 9 (14,15) to form an active holoenzyme that processes and activates downstream caspase 3 (16). Cells deficient in APAF-1 or caspase 9 are resistant to some apoptotic signals that are thought to require mitochondrial cyt c release (17,18).
The signals that induce cyt c release are often propagated through members of the BCL-2 family of pro-apoptotic proteins. Translocation from the cytoplasm to mitochondria during induction of apoptosis has been reported for BID (19,20), BAX (21), BAK (22), BAD (23,24), BIM (25,26), and NOXA (27). These molecules can be regulated by phosphorylation, dimerization, or proteolytic cleavage (10,24,28,29). For example, BAX, BAK, and BIM are held inactive in the cytoplasm and are translocated to the mitochondria after a cell death signal. Following Fas ligation and recruitment of FADD to the trimeric receptor complex, caspase-8 is activated and then cleaves cytoplasmic BID, producing a truncated fragment (tBID) that is able to translocate to mitochondria to induce cyt c release (19,20). BID is also cleaved by GrB, although the function of the cleaved form has not yet been defined (19,30).
In response to stress signals such as Ca 2ϩ or reactive oxygen species as well as most apoptosis signals, the mitochondrial membrane opens to allow solutes and water to enter the matrix in a phenomenon termed permeability transition (PT) (12,31). It has been proposed that this is a result of the opening of a multiprotein complex spanning the inner and outer membranes of mitochondria known as the PT pore complex (12,13,31). The most abundant components of the PT pore are the voltage-dependent anion channel/porin and adenine nucleotide translocator, plus proteins such as cyclophilin D, creatine kinase, and others. Upon PT pore opening, mitochondria lose their mitochondrial membrane potential (⌬⌿ m ) across the inner membrane, which is reported to be an early response to many apoptotic signals (32). BCL-2 family pro-apoptotic proteins that translocate to the outer mitochondrial membrane such as BAX are reported in some models to physically interact with the voltage-dependent anion channel and adenine nucleotide translocator, and it has been proposed that this results in PT pore opening and cyt c release (33)(34)(35). However, the exact mechanism of cyt c release remains controversial and unresolved.
Although one model of GrB action suggests that it directly cleaves and activates the downstream caspase-3 (6,36) and would thus mediate apoptosis directly without the need for mitochondria, in earlier work, we found that GrB and perforin rapidly induce cyt c release and decrease ⌬⌿ m before or coincident with apoptosis (37). Furthermore, in an in vitro cell-free system, the addition of mitochondria to S100 increases GrBinduced nuclear chromatin condensation 15-fold, indicating that mitochondria have a powerful amplification effect on GrB function (37). This activity requires proteolytically active GrB. These observations suggest a role for mitochondria in the GrB apoptotic response.
In this study, we demonstrate that both APAF-1 and caspase 9, and therefore, a mitochondrial cyt c pathway, are required for efficient GrB-induced apoptosis of murine embryonic fibroblast (MEF) cells. GrB mediates cyt c release from isolated mitochondria in vitro through an S100 protein that we identify as the BCL-2 family member BID. Furthermore, GrB in the absence of BID or other S100 proteins can open the mitochondrial PT pore with the loss of ⌬⌿ m . Thus, we have identified two mechanisms by which GrB disrupts mitochondrial function and promotes cell death.

EXPERIMENTAL PROCEDURES
Cell Lines and Reagents-HeLa cells were cultured in Dulbecco's minimal essential medium (Life Technologies, Inc.) supplemented with 10% fetal bovine serum (Cansera). MEFs deficient in APAF-1 or caspase 9 were cultured as described previously (17). Rat GrB and perforin were isolated from RNK-1 cells as described previously (2). Recombinant GrB (rGrB) and mutant inactive GrB (S203A) were a gift of Dr. Tim Ley (Washington University Medical School, St. Louis, MO). Yeast-derived rGrB has been shown to have the same proteolytic activity and specificity and response to inhibitors as purified wild-type GrB (38). In addition, rGrB can replace the lost apoptotic activity of cells deficient in GrA Ϫ/Ϫ GrB Ϫ/Ϫ (39). Purified recombinant BID (rBID) protein and rat anti-BID antiserum were a gift of Dr. Junying Yuan and have been described previously (19).
Mitochondrial Isolation and S100 Preparation-For each experiment, fresh mitochondria were isolated from mouse liver with six strokes of a Dounce homogenizer before purifying on a Percoll gradient as described previously (40). The mitochondria were kept on ice in HB (300 mM sucrose, 5 mM TES, and 0.2 mM EGTA, pH 7.2) and used within 4 h of isolation. Murine liver S100 was prepared by collecting the supernatant at the 8740 ϫ g step and centrifuging it twice at 14,400 ϫ g for 10 min before a final centrifugation at 100,000 ϫ g for 1 h at 4°C. S100 proteins from HeLa cells were prepared by harvesting cells at the log phase of growth and washing four times with Hanks' balanced salt solution and 10 mM HEPES, pH 7.2, before resuspension at 10 8 cells/ml in HB plus 0.2% bovine serum albumin. The cells were lysed by 2-4-s bursts of a Polytron homogenizer (setting 6), and the cell debris was removed by centrifugation at 792 ϫ g for 10 min, followed by centrifugation of the supernatant twice at 14,400 ϫ g for 15 min and finally at 100,000 ϫ g for 1 h at 4°C.
Cytochrome c and AIF Release Assay-Freshly isolated mouse liver mitochondria (1 mg/ml) were incubated with GrB and mouse liver S100 (100 g/ml), partially purified BID, or rBID in a V-bottom 96-well plate in a total volume of 60 l of Mitochondrial (M) buffer (220 mM sucrose, 68 mM mannitol, 10 mM KCl, 5 mM KH 2 PO 4 , 2 mM MgCl 2 , 0.5 mM EGTA, 5 mM succinate, 2 mM rotenone, and 10 mM HEPES, pH 7.2) at 37°C for 2 h. After centrifugation at 760 ϫ g for 5 min, 30 l of the supernatant was resolved on a 15% SDS-polyacrylamide gel as previously described (37) and Western-blotted for cyt c, AIF, and BID. The lanes labeled M in the figures represent the maximum cyt c release and were prepared using an aliquot of mitochondria in 60 l of M buffer. Mouse monoclonal anti-cyt c antibody (Pharmingen, Mississauga, Ontario, Canada) and rat anti-BID antibody (19) were used with goat anti-mouse and goat anti-rat peroxidase conjugates (Sigma), respectively, to develop the blots.
FPLC Purification of Cytochrome c-releasing Factor-All chromatography steps were carried out using an FPLC system (Amersham Pharmacia Biotech), and samples were maintained on ice throughout. Mouse liver S100 (30 ml from three livers) was applied to a self-packed hydroxyapatite column (10-ml bed volume; HA-Ultrogel, BioSepra Inc., Marlborough, MA) equilibrated with buffer A (2 mM phosphate, pH 7.0, with protease inhibitors (0.5 g/ml aprotinin, 0.5 M leupeptin, and 100 g/ml phenylmethylsulfonyl fluoride)). The column was washed with 5 column volumes of buffer A and eluted with a linear gradient of buffer B (buffer A plus 2 M NaCl). The fractions with GrB-mediated cyt c-releasing activity were pooled. Solid (NH 4 ) 2 SO 4 was directly added to the active fraction (0.5 M final concentration). After centrifugation at 10,000 ϫ g for 10 min, the pool was applied to a phenyl-Sepharose 6 column (HiPrep 16/10 phenyl (high sub), Amersham Pharmacia Biotech). The column was equilibrated with buffer C (buffer B plus 0.5 M (NH 4 ) 2 SO 4 ) and eluted with a linear gradient of buffer D (10 mM phosphate, pH 7.5, with the protease inhibitors listed above). The pooled active fractions were concentrated to 2 ml with a Centricon-10 microconcentrator (Millipore Corp., Bedford, MA) and diluted 10 times with buffer D before application to a Mono Q column (HR 5/5, Amersham Pharmacia Biotechnology), followed by elution with a linear gradient of buffer E (buffer D plus 2 M NaCl). The highest activity fractions were diluted 10 times with buffer D and applied to a second Mono Q column, again eluting with 0 -50% buffer E.
Immunofluorescent Microscopy-Cytochrome c was detected in HeLa cells with the mouse monoclonal anti-cyt c antibody and a Cy3-conjugated goat anti-mouse secondary antibody (Chemicon International, Inc., Don Mills, Ontario) as previously described (37). When required, cells were pretreated for 15 min with 20 M Z-VAD-fmk before GrB and perforin treatment. Image analysis was performed with Northern Eclipse Version 5.0 software (Empix Inc., Toronto, Ontario) as described (41).
Transmembrane Potential of Isolated Mitochondria Determined by Fluorometry-We used an in vitro cell-free assay to examine the GrB effect on mitochondria (37,42). Freshly isolated mitochondria were added to a 96-well plate in M buffer in the presence or absence of the S100 fraction and GrB and then incubated at 37°C for the indicated times. Finally, 5 M rhodamine 123 (excitation at 485 nm and emission at 538 nm) or 150 M TMRM (excitation at 544 nm and emission at 590 nm) was added, and the samples were read on a Titertek Fluoroskan II fluorometer (MTX Labsystems, Vienna, VA).
PT Pore Opening and ⌬⌿ m Determined by Confocal Laser Imaging in Cells and Purified Mitochondria-Freshly isolated mitochondria were centrifuged onto coverslips at 1475 ϫ g for 5 min at 4°C, washed, resuspended in M buffer, and then stained for 15 min with 100 nM tetramethylrhodamine and 8 M calceinAM. After four washes, the mitochondria were resuspended in M buffer, and GrB was added. When required, mitochondria were pretreated with CsA 30 min before GrB treatment. The mitochondria were imaged, and total fluorescence was determined on an Olympus IX70 inverted confocal laser microscope using Fluoview 2.0 software (Carson Group Inc., Markham, Ontario).
Whole cells were grown on coverslips for 2 days and then washed and incubated in Hanks' balanced salt solution, 10 mM HEPES, and 4 mM NaH 2 CO 3 , pH 7.4, for 90 min with GrB and sublytic amounts of perforin. When required, cells were pretreated with 20 M CsA for 30 min. The cells were stained with 1 M calceinAM and 5 mM CoCl 2 for 15 min at 23°C. CoCl 2 quenches the cytoplasmic calcein fluorescence and allows visualization of mitochondria (43). Cells were washed four times and resuspended in Hanks' balanced salt solution and 10 mM HEPES, and calcein images were captured on an Olympus IX70 inverted confocal laser microscope using Fluoview 2.0 software and a band bypass filter of 488 nm, with Nomarski optics for transmitted light images of the same cells. The fluorescence of individual cells was determined using Northern Eclipse Version 5.0 software.

RESULTS
Enhanced Resistance of apaf-1 Ϫ/Ϫ and caspase 9 Ϫ/Ϫ MEFs to Granzyme B-induced Apoptosis-GrB can induce mitochondrial cyt c release and loss of ⌬⌿ m (30,37,44), and mitochondria can amplify nuclear apoptotic responses to GrB in vitro (37). Although this suggests that mitochondria may be important regulators of GrB-induced apoptosis, it is possible that the observed mitochondrial dysregulation is secondary to caspase activation by GrB and is not responsible for the induction of apoptosis (6). To test this hypothesis, we examined apaf-1 Ϫ/Ϫ or caspase 9 Ϫ/Ϫ MEF cells, which are highly resistant to many apoptotic death signals except Fas/CD95, which can bypass mitochondria and directly activate downstream caspases (17,18). If GrB were able to bypass mitochondria, then a deficiency of either protein would have no effect on GrB apoptosis. We therefore treated apaf-1 Ϫ/Ϫ or caspase 9 Ϫ/Ϫ MEFs with GrB and perforin and measured apoptosis by Hoechst staining of condensed chromatin. Fig. 1 (A and B) illustrates that both apaf-1 Ϫ/Ϫ and caspase 9 Ϫ/Ϫ MEFs were more resistant to GrB compared with normal MEFs over a wide dose range. Adriamycin was also unable to initiate apoptosis in the deficient cells (data not shown) (18), thus confirming an earlier report (17). The MEFs were not completely resistant to higher doses of GrB, indicating that part of the GrB apoptotic activity did not require either APAF-1 or caspase 9. When examining the kinetics of GrB-induced nuclear changes ( Fig. 1B), nuclear condensation of the wild-type MEFs occurred rapidly, as almost 60% were apoptotic at 2 h, and a plateau of 80% apoptosis was reached by 4 h of treatment. In contrast, apoptosis of the apaf-1 Ϫ/Ϫ and caspase 9 Ϫ/Ϫ MEFs was 15% of the wild-type levels in the first hours of the assay and then slowly increased, so by 24 h, it was ϳ40% of the wild-type MEF levels. Clearly, GrB-induced nuclear apoptosis is most efficient when utilizing APAF-1 and caspase 9 on the mitochondrial-dependent pathway.
Granzyme B-induced Cytochrome c Release in Vitro-As APAF-1/caspase 9-directed apoptosis is dependent on cyt c (15), we established an in vitro cell-free system using purified murine liver mitochondria and S100 to further examine GrB regulation of cyt c release (see "Experimental Procedures"). We found that GrB in the presence of S100 initiated cyt c release from isolated mitochondria within 30 min ( Fig. 2A) and in a dose-dependent manner (Fig. 2B), whereas GrB or S100 alone had no effect ( Fig. 2A). Thus, we assume that GrB must act through a factor(s) in S100 that exerts its effect on mitochondria. To determine whether caspases are important for the release of cyt c, the peptide caspase inhibitors DEVD-fmk, YVAD-fmk, Z-VAD-fmk, and control Z-FA-fmk at concentrations previously shown to inhibit caspase activity (37) were mixed with S100 before the addition of GrB. None of the inhibitors was able to block cyt c release (Fig. 2C). We also investigated whether another mitochondrial intermembrane protein, AIF, was released along with cyt c and determined that it was not (Fig. 2D), suggesting that cyt c was released in a restricted manner and was not part of the general permeabilization of the outer mitochondrial membrane.
Previously, we had demonstrated in HeLa cells that GrBmediated cyt c release was significantly inhibited by the pan caspase inhibitor Z-VAD-fmk (37), whereas our current in vitro data suggested that GrB was acting by a caspase-independent mechanism. To determine whether the requirement for caspases differs between cell types or between intact cells and the in vitro cell-free assay, we compared the caspase dependence of GrB-induced cyt c release in whole HeLa cells with the in vitro assay using liver mitochondria and HeLa cell S100 over a dose range and time course. We found that Z-VAD-fmk significantly blocked cyt c release from mitochondria in intact HeLa cells following GrB and perforin treatment over a wide dose range and up to 8 h of treatment (Fig. 2, E and F). The most pronounced effects were seen at 2 h with GrB doses between 400 and 800 ng/ml, but inhibition at other doses and times was always Ͼ50% of that of the untreated cells. In contrast, Z-VAD-fmk had little effect on cyt c release in vitro either in a HeLa S100 dose response or over an extended time course (Fig. 2, G and H).
Granzyme B Cleaves BID Protein-We then started to purify the protein in mouse liver S100 that mediates cyt c release by passing S100 through sequential FPLC columns, hydroxyapatite, phenyl-Sepharose, and Mono Q (see "Experimental Procedures"). Positive fractions from the hydroxyapatite column were pooled and loaded on the phenyl-Sepharose column, and then the positive eluted fractions were loaded on the Mono Q column (Fig. 3A). We analyzed the Mono Q fractions for GrBinduced cyt c release and BID, a pro-apoptotic protein and GrB substrate known to mediate Fas signaling to the mitochondria through caspase 8. The peak activity fractions coincided exactly with the elution of BID (Fig. 3A).
It has been reported that GrB can cleave BID (19,30). We confirmed these data by showing that BID in S100 was rapidly cleaved to a p15 truncated form (tBID) by rat GrB or mouse rGrB in a time-and dose-dependent manner (Fig. 3B). rGrB cleaved BID and induced cyt c release at similar doses using either BID-containing Mono Q fractions or whole S100 (Fig.  3C). GrB proteolytic activity was required for BID cleavage, as rGrB with an inactivating mutation (S203A) failed to cleave BID and to induce cyt c release (Fig. 3C).
Granzyme B Uses BID to Induce Cytochrome c Release in Isolated Mitochondria-To determine whether BID protein in S100 is required for GrB-induced cyt c release, we immunoprecipitated BID with rat antiserum. BID was highly depleted from S100 by the rat antiserum, but not by normal rat serum (Fig. 4A). The treated sera were then serially diluted and added to mitochondria with a constant dose of GrB, and cyt c release was measured. The BID-depleted serum was unable to support GrB-induced cyt c release compared with controls (Fig. 4B).
The experiments to this point supported the hypothesis that the S100 protein BID mediates cyt c release. If this is correct, then we would predict that BID protein alone can replace S100. Thus, we next examined the effect of purified rBID protein on cyt c release in the presence and absence of GrB. GrB induced mitochondria to release cyt c in the presence of BID protein, and the effect was dependent on both GrB and BID concentration (Fig. 4, C and D). We found that BID protein at higher levels induced cyt c release in the absence of GrB; however, at BID doses that were ineffective alone, a GrB-dependent BIDmediated cyt c release was observed. The cyt c release in the absence of GrB is not due to contaminating caspases, as it has been shown previously that higher doses of full-length BID can also cause cyt c release by an unknown mechanism (19,20).  1. Granzyme B-induced apo ptosis of apaf-1 ؊/؊ and caspase 9 ؊/؊ MEFs. A, MEF cells from apaf-1 Ϫ/Ϫ (f), caspase 9 Ϫ/Ϫ (OE), and parental (wild-type; q) mice were incubated with GrB at the indicated doses plus perforin (0.12 g/ml) or with perforin (Pf) alone for 2 h and then stained with Hoechst dye, and apoptotic cells were enumerated. The data represent means Ϯ S.E. from three experiments. B, MEFs were treated and analyzed as described for A using GrB (2 g/ ml) plus perforin (0.12 g/ml) over a 24-h time course. The closed symbols are the same as described for A, and the open symbols represent MEFs treated with perforin only. The data represent means Ϯ S.E. from three experiments. FIG. 2. Granzyme B-induced cytochrome c release in an in vitro cell-free system. A, purified mouse liver mitochondria were incubated with GrB (0.7 g/ml) and mouse liver S100 (100 g/ml) for the indicated times. The supernatant was recovered after centrifugation and Western-blotted for cyt c. M is a control from untreated mitochondria (in A, C, G, and H). B, GrB at the indicated doses was incubated with purified mitochondria and S100 for 2 h and then analyzed as described for A. C, peptide caspase inhibitors (50 M) were preincubated with S100 for 15 min before adding GrB and mitochondria for 2 h, and then cyt c release was analyzed as described for A. D, mitochondria were treated as described for B and Western-blotted for AIF. E, HeLa cells were preincubated with (E) and without (q) 50 M Z-VAD-fmk for 30 min before treatment with perforin (0.2 g/ml) and GrB at various doses. Cyt c release was detected by immunostaining cells with anti-cyt c antibody plus the Cy3-conjugated secondary antibody. Cells were imaged by fluorescent microscopy, and the number of cells staining positively for cyt c were counted. The data represent means Ϯ S.E. from three experiments. F, HeLa cells were treated and analyzed as described for E using GrB (1 g/ml) plus perforin (0.2 g/ml) (q) or perforin only () over an 8-h time course. The open symbols represent cells that receive a 50 M Z-VAD-fmk treatment for 30 min. The data represent means Ϯ S.E. from three experiments. G, purified mouse liver mitochondria and serially diluted HeLa S100 were preincubated, separately, with 50 M Z-VAD-fmk for 15 min before adding GrB (0.7 g/ml) for 2 h. Cyt c release was analyzed as described for A. H, purified mouse liver mitochondria and HeLa S100 were preincubated, separately, with 50 M Z-VAD-fmk for 15 min before adding GrB (0.7 g/ml) over a 3-h time course. Cyt c release was analyzed as described for A. A Western blot for actin was performed as a loading control.
However, tBID is 100 times more efficient at inducing the release of cyt c compared with full-length BID (19).
Granzyme B Induces Permeability Transition and ⌬⌿ m Suppression-PT and suppression of ⌬⌿ m occur as a result of GrB and perforin treatment of cells (37,44). To determine the proteins required for ⌬⌿ m loss, we utilized a fluorescent in vitro cell-free assay in which we could combine mitochondria with GrB in the presence or absence of the murine liver S100 fraction. We observed a significant suppression of ⌬⌿ m whether or not S100 was present, indicating that no cytosolic protein was required (Fig. 5A). The response was dependent on the dose of GrB, with a maximum effect at 3-4 g/ml (Fig. 5B). A minor but inconsistent inhibitory effect of S100 on ⌬⌿ m was observed at some concentrations. The earliest detectable ⌬⌿ m dissipation was observed at 30 min after the start of GrB (3 g/ml) treatment and was complete by 90 -120 min (Fig. 5C).
GrB could induce the loss of ⌬⌿ m by opening the PT pore. To examine the state of the PT pore, we used the fluorescent dye calceinAM (45). CalceinAM is freely permeable across cellular membranes, but becomes fluorescent and impermeable upon cleavage by intracellular esterases, which prevents its exit from mitochondria until the PT pore is opened. Isolated mitochondria were loaded with calceinAM and simultaneously stained with the potentiometric dye TMRM to follow ⌬⌿ m . The mitochondria were then treated with GrB and analyzed by confocal laser microscopy at increasing time periods (Fig. 6, A  and B). Within 20 min, calcein and TMRM fluorescence was reduced to ϳ50% of control levels, eventually falling to almost zero fluorescence by 40 min. To verify that the loss of fluorescence was due to opening of the PT pore, the potent pore inhibitor cyclosporin A was added under the same treatment conditions. CsA maintained calcein and TMRM fluorescence near the control maximum levels well beyond 60 min of incubation (Fig. 6, A and B).
We next determined whether GrB-induced opening of the PT pore in cells was the same as that observed in isolated mitochondria (Fig. 6C). GrB plus perforin treatment of HeLa cells induced a decrease in mitochondrial calcein fluorescence in FIG. 3. Identification of BID in S100 protein as the mediator of granzyme B-induced cytochrome c release. A, BID in mouse liver S100 copurifies with granzyme B-induced cyt c releasing activity. Mouse liver S100 was sequentially passed through hydroxyapatite, phenyl-Sepharose and two Mono Q FPLC columns (see "Experimental Procedures"). Fractions able to support cyt c release in the presence of GrB (heavy bars) were pooled and loaded on the next column. Mono Q fractions were analyzed for BID protein and GrB-induced mitochondrial cyt c release by Western blotting. B is the sample before loading, and C is the sample not binding to the column. B, shown are the results from GrB hydrolysis of BID protein in S100. GrB (0.7 g/ml) was incubated with mouse liver S100 and then Western-blotted for BID. C, the partially purified Mono Q fractions shown in A or murine liver S100 was added to mitochondria and rGrB or mutant inactive rGrB (S203A) at the indicated doses for 2 h, and then the supernatant was Western-blotted for cyt c or BID. M is a control from untreated mitochondria.

FIG. 4. BID is necessary and sufficient for granzyme B-induced cytochrome c release from isolated mitochondria.
A, mouse liver S100 (Control) was incubated with rat anti-BID antiserum or normal rat serum and immunoprecipitated. S100 was then Westernblotted for BID. B, GrB (0.7 g/ml) was added to mitochondria and serially diluted S100 fractions treated as described for A for 2 h, and then supernatants were Western-blotted for cyt c. M is a control from untreated mitochondria (in B and D). C, rBID was added to serially diluted GrB plus mitochondria for 2 h, and the supernatants were Western-blotted for cyt c. D, serially diluted rBID was added to mitochondria in the presence or absence of GrB and incubated for 2 h before Western blot analysis of supernatants for cyt c.
comparison with either of the untreated controls (Fig. 6D), GrB alone, or perforin alone (data not shown). The addition of CsA blocked the GrB-and perforin-induced calcein release from mitochondria (Fig. 6D). Using image analysis software, we were able to determine the relative fluorescent intensity of individual cells and the number of cells with high, medium, and low mitochondrial fluorescence. The percentage of cells with low calcein fluorescence (calcein lo ) increased from 9 to 71% following GrB and perforin treatment and was decreased to 18% by CsA. Corresponding shifts in populations with medium (calcein med ) and high (calcein hi ) calcein fluorescence levels were observed with these treatment conditions. This experiment was repeated three times with similar results, and one example is shown in Fig. 6D.
We observed that GrB acts on mitochondria directly to cause ⌬⌿ m loss in the absence of any other cytosolic protein; however, it has been reported that the addition of BID can produce the same effect (19,30). We examined the effect of rBID on the relative level of ⌬⌿ m in isolated mitochondria treated with GrB using TMRM fluorescence. We found that both GrB and rBID suppressed ⌬⌿ m in a dose-dependent manner and that the combination of rBID and GrB reduced ⌬⌿ m in an additive manner (Fig. 7, A and B). DISCUSSION We have examined how GrB mediates apoptosis by disrupting mitochondrial function and have determined that GrB acts through two distinct mechanisms: (i) cleaving the cytosolic protein BID, resulting in the release of cyt c, and (ii) acting directly on mitochondria to cause PT and loss of ⌬⌿ m . Since mitochondria appear to be dysregulated in GrB-mediated apoptosis, we examined whether GrB acts directly on the mitochondria or indirectly through other proteins to perturb mitochondrial function. One regulatory mechanism that is important for apoptosis is the activation of the downstream effector caspase cascade via activation of caspase 9 during integration into the holoenzyme (15,9,46). Caspase 3 is generally activated by caspase 9 in a mitochondrial-dependent pathway; however, it has been suggested that GrB can directly activate caspase 3 (36,47,48). In this study, we show that the GrB mitochondrial-dependent pathway is the dominant apoptotic pathway using apaf-1 Ϫ/Ϫ and caspase 9 Ϫ/Ϫ MEF cells. Other death signals delivered to apaf-1 Ϫ/Ϫ and caspase 9 Ϫ/Ϫ MEFs have been shown to reduce caspase 3 activation and to increase survival (18,49). As APAF-1 and caspase 9 require cyt c to form an active holoenzyme for caspase 3 processing (50, 51), this suggests that cyt c release from mitochondria may be a key event in GrB-induced apoptosis. We were able to deter-mine from the kinetics of the nuclear changes induced by GrB in apaf-1 Ϫ/Ϫ and caspase 9 Ϫ/Ϫ MEFs that only ϳ15-20% of the apoptotic signal of wild-type cells is seen in the first few hours of GrB and perforin treatment. Apoptosis occurs much more slowly; and by 24 h, apaf-1 Ϫ/Ϫ and caspase 9 Ϫ/Ϫ MEF cell apoptosis increases to ϳ40% of that seen in wild-type cells. These observations led us to conclude that mitochondria are the major regulators of GrB-induced apoptosis in MEFs. Whether the mitochondrial pathway will be as important in other cell types remains to be determined.
Candidate proteins mediating the regulation of cyt c release come from the BCL-2 family, which contains both anti-apoptotic (BCL-2 and BCL-X L ) and pro-apoptotic (BAX, BAK, BID, etc.) proteins that either directly or indirectly affect mitochondrial function through heterodimerization (for reviews, see Refs. [52][53][54][55]. In our search for the proteins involved in regulating GrB-induced cyt c release in vitro, we determined that BID was the only cytosolic protein required. BID has been shown to interact via its BH3 domain with the BCL-2 and BCL-X L domains and the BAX BH1 domain (56 -58). Precedence for BID in mitochondrial signaling comes from the Fas/ CD95 death pathway, in which caspase 8-cleaved BID is responsible for cyt c release as well as the subsequent apoptotic cellular changes resulting from holoenzyme formation and activation of downstream caspases (19,20,59,60). In vitro BID BH3 mutants and BID-depleted cell extracts have reduced capability to release cyt c (20,56). Our observations that GrBinduced cyt c release requires BID and that APAF-1 and caspase 9 are necessary for efficient apoptosis support the idea that GrB may use a mitochondrial pathway similar to Fas/ CD95 in which GrB cleavage of BID leads to the activation of downstream caspases.
In the cell-free assay, we demonstrated that caspases were not required, as GrB and rBID induced cyt c release in the absence of any other S100 proteins. Also, caspase inhibitors failed to block cyt c release when S100 was present, supporting a model in which BID is directly cleaved by GrB. This is in agreement with previously published data in Jurkat cells, where cyt c release following GrB and perforin treatment was found to be caspase-independent (44), and BID was cleaved within 60 min in the presence of Z-VAD-fmk (30). However, we have also demonstrated that Z-VAD-fmk, while efficiently inhibiting cyt c release in HeLa cells, fails to block release in vitro using S100 from HeLa cells and liver mitochondria (37). This indicates that in some cells, GrB operates primarily through a caspase-dependent mechanism, although in others such as Jurkat, it primarily targets BID. Why the caspase dependence FIG. 5. Granzyme B acts directly on isolated mitochondrial to suppress ⌬⌿ m . A, freshly isolated mitochondria were treated with GrB, and ⌬⌿ m was determined by measuring rhodamine 123 fluorescence. ⌬⌿ m suppression occurred in the presence (white bars) or absence (black bars) of the S100 fraction after a 1-h treatment with GrB. B, shows is the effect of GrB concentration on ⌬⌿ m after a 1-h incubation with mitochondria in the absence of S100. C, shown is a time course of GrB (3 g/ml)-mediated ⌬⌿ m suppression in the absence of S100.
of cyt c release in HeLa cells is not duplicated in the in vitro assay is not clear, but suggests that in a disrupted cell, GrB prefers BID as a substrate or that the caspase targeted by GrB is not present in the S100 fractions of the HeLa cells. We were also unable to determine whether the participation of caspases in cyt c release in HeLa cells is a result of direct processing of a caspase by GrB or is secondary to GrB-induced BID processing in a feed-forward amplification from the mitochondrial activation of downstream caspases. Whether in the presence or absence of active caspases, it appears that BID is a pivotal protein in GrB-mediated apoptotic cellular changes.
The mechanism of cyt c release from mitochondria has not been determined; however, several theoretical models have been suggested such as the release through the PT pore (34) or a cyt c-specific channel (61) or a nonspecific release from rupturing of the outer mitochondrial membrane due to swelling from PT (62). GrB-induced cyt c release through the PT pore is unlikely, as the release is not blocked by PT pore inhibitors CsA and bongkrekic acid. 2 The mechanism is also not due to outer membrane rupture, as we did not see a coordinated release of cyt c and other intermembrane proteins such as AIF. This suggests that BID must be acting on some other protein associated with the mitochondria. BID is a cytosolic protein until a death signal results in its cleavage to the truncated form (tBID) that is translocated to the mitochondria. Using rBID, we found that it is cleaved directly by GrB as predicted since BID contains a single GrB cleavage site ( 72 IEPDS 76 for murine and 72 IEADS 76 for humans) and has been reported to be directly cleaved by GrB at this site (19,63) and during GrB-mediated apoptosis (30). The current model suggests that mitochondrial membrane-bound death antagonists BCL-2 and BCL-X L and agonist BAX can function independently of each other, but compete for tBID as a ligand to block or induce cyt c release ( change enabling its integration into the mitochondrial membrane, resulting in cyt c release (64). Furthermore, BAX-induced cyt c release has been shown to be independent of the PT pore (66), similar to our observations with GrB-induced cyt c release. Another possibility for the action of BID can be found in its tertiary structure, which resembles BAX and the poreforming domains found in the bacterial toxins diphtheria and colicin (63). It has been demonstrated that BID can form pores in membranes under low pH conditions (67), so perhaps BID is sufficient to induce cyt c release by itself under certain conditions.
It is noteworthy that cytochrome c release and PT, accompanied by loss of ⌬⌿ m , were independent of one another even though both were detected at similar times in response to GrB in the in vitro cell-free assay. Both PT and ⌬⌿ m suppression occur in the absence of any cytosolic S100 protein, indicating that caspases are not required and that GrB acts directly on the mitochondrial outer membrane. The caspase independence of the GrB-mediated ⌬⌿ m loss mechanism in a cell-free system is similar to our findings in intact HeLa cells (37), which was confirmed by Heibein et al. (44) in Jurkat cells. As CsA inhibits PT, we may assume that GrB acts on the PT pore or some other protein associated with the pore. The PT pore acts as a gated channel regulating mitochondrial matrix Ca 2ϩ , pH, volume, and ⌬⌿ m (33). Opening of the 1.5-kDa PT pore results in PT and the equilibration of solutes across the inner membrane. This can lead to uncoupling of the electron transport chain, loss of ⌬⌿ m , inhibition of ATP synthesis, mitochondrial swelling, and an increased reactive oxygen species production. The general serine protease inhibitor phenylmethylsulfonyl fluoride can block GrB-mediated PT and ⌬⌿ m loss; therefore, we determined whether the PT pore components voltage-dependent anion channel and adenine nucleotide translocator as well as some of the proteins shown to bind to or regulate the pore (BAX and BCL-2) were cleaved by GrB using Western analysis. However, we were unable to demonstrate that GrB cleaved any of these proteins (data not shown).
Although our in vitro assays indicate that S100 (and therefore, any BID it contains) is not required for GrB to suppress ⌬⌿ m , we found that rBID increases the loss of ⌬⌿ m . This suggests either that the concentration of BID in S100 is too low to act on the PT pore after GrB processing or that S100 contains another protein(s) that interferes with tBID-induced suppression of ⌬⌿ m . We are unable to determine whether or not rBID and GrB are acting at the same or distinct sites on the mitochondrial outer membrane.
We can now view GrB-induced apoptosis as a three-phase process similar to other death signals in which the pre-mitochondrial phase is followed by convergence at the mitochon-dria, resulting in the release of pro-apoptotic factors that activate the post-mitochondrial effector phase. Although cell death pathways exist that can bypass the mitochondria such as in the Fas/CD95 type I apoptotic program, disruption of mitochondrial function through an amplification feedback loop is still observed. Downstream caspase 3 can activate upstream caspase 8, which leads to efficient BID cleavage and mitochondrial dysfunction (68). We do not yet know if GrB can bypass mitochondria and then feedback in a similar manner.
GrB acts on many substrates in the cell, including several caspases, BID, ICAD, and PARP, suggesting that GrB could potentially act on all simultaneously, inducing death by multiple mechanisms to guarantee the cells' demise. However, there appear to be dominant pathways such as the caspaseand mitochondrial-dependent pathways; and if blocked, then cell death could proceed more slowly via secondary pathways. For example, if neither the mitochondrial nor caspase pathway is utilized, then GrB could inactivate ICAD directly to cause CAD activation, ensuring DNA fragmentation, albeit in a delayed and less efficient manner (39). The presence of the dominant mitochondrial pathways mediates rapid and efficient apoptosis; but clearly, there are other apoptosis mechanisms that come into play when the mitochondrial pathway is rendered inactive. In conclusion, we have demonstrated that GrB disrupts the mitochondria through two very distinct mechanisms: cleavage of the cytosolic protein BID, resulting in release of cyt c; and the induction of PT and loss of ⌬⌿ m , which ultimately lead to the disruption of mitochondrial function, assuring the death of the cell.