Requirement for Aspartate-cleaved Bid in Apoptosis Signaling by DNA-damaging Anti-cancer Regimens*

Lymphoid malignancies can escape from DNA-damaging anti-cancer drugs and γ-radiation by blocking apoptosis-signaling pathways. How these regimens induce apoptosis is incompletely defined, especially in cells with nonfunctional p53. We report here that the BH3-only Bcl-2 family member Bid is required for mitochondrial permeabilization and apoptosis induction by etoposide and γ-radiation in p53 mutant T leukemic cells. Bid is not transcriptionally up-regulated in response to these stimuli but is activated by cleavage on aspartate residues 60 and/or 75, which are the targets of caspase-8 and granzyme B. Bid activity is not inhibitable by c-FlipL, CrmA, or dominant negative caspase-9 and therefore is independent of inducer caspase activation by death receptors or the mitochondria. Caspase-2, which has been implicated as inducer caspase in DNA damage pathways, appeared to be processed in response to etoposide and γ-radiation but downstream of caspase-9. Knock down of caspase-2 by short interfering RNA further excluded its role in Bid activation by DNA damage. Caspase-2 was implicated in the death receptor pathway however, where it contributed to effector caspase processing downstream of inducer caspases. Granzyme B-specific serpins could not block DNA damage-induced apoptosis, excluding a role for granzyme B in the generation of active Bid. We conclude that Bid, cleaved by an undefined aspartate-specific protease, can be a key mediator of the apoptotic response to DNA-damaging anticancer regimens.

Execution of apoptosis requires the activation of effector caspases, which can be accomplished via the death receptor or the mitochondrial pathway (1). Death receptor-induced apoptosis does not require de novo protein synthesis but the mere recruitment of preformed molecules to the receptor, including the inducer caspases-8 and/or -10. Most cellular stresses, however, rely on the primordial mitochondrial pathway, which requires more elaborate intracellular events to become activated. Depending on the stimulus, its activation may involve new synthesis of apoptosis signaling molecules and/or their post-translational modification. The ultimate induction of mitochondrial permeability in this pathway allows the release of cytochrome c (Cyt c), 1 Smac/Diablo, and Omi/HtrA2, which are essential for activation of the inducer caspase-9 (1).
The response of the mitochondrion to upstream stimuli is a critical control point in apoptosis signaling. It is crucial, therefore, to understand how death signals are transmitted to this organelle. In mammals, multiple pro-apoptotic BH3-only members of the Bcl-2 family are specialized in sensing partially distinct stimuli. They integrate apoptotic stimuli into one single pathway by triggering mitochondrial permeabilization by the Bcl-2 family members Bax and/or Bak (2,3). There are two prominent models for mode of the action of BH3-only proteins. In one, they transiently interact with Bax and/or Bak, inducing them to form pores in the mitochondrial membrane. In this scenario, inhibitory Bcl-2 family members sequester BH3-only proteins to prevent their pro-apoptotic action (2,3). In the other model, BH3-only proteins interact with inhibitory Bcl-2 family members and thereby neutralize their function, which results in Bax/Bak activation (4).
To prevent inappropriate cell death, BH3-only family members are kept in check by mechanisms, which are tailored to each individual member (4). For example, Noxa and Puma are under transcriptional control, in particular by p53 in the DNA damage response (5)(6)(7). Other BH3-only proteins are regulated by various modes of post-translational modification. Bim and Bmf are sequestered to the cytoskeleton and released upon phosphorylation (8). Bad is sequestered in the cytosol and released upon dephosphorylation (9). Bid requires proteolytic cleavage to produce an active carboxyl-terminal fragment, which is achieved by caspase-8 in the death receptor pathway (10) and by granzyme B in apoptosis triggered by cytolytic T cells (11).
It is important to elucidate the molecular events involved in apoptosis signaling by anti-cancer therapies. The required active participation of cancer cells in bringing about their own demise implies that they can also put up resistance. Whether resistance to apoptosis generally confers therapy resistance is debated, because cells may well die a nonapoptotic death or be irreversibly arrested in the cell cycle in response to anti-cancer regimens (12). However, for lymphoid malignancies, convincing evidence has been provided that loss of p53 or Bcl-2 overex-pression and resulting apoptosis resistance promotes therapy resistance in vivo (13,14).
Although ␥-radiation and DNA-damaging anti-cancer drugs are complex inputs, there is a great deal of molecular evidence that the apoptosis signaling pathways initiated by these stimuli emerge from the DNA damage response. DNA damage is sensed by ATM and Chk2 kinases, which phosphorylate and stabilize p53, allowing it to function as a transcription factor (15,16). In this way and possibly also via transcription-independent mechanisms, p53 can bring about a cell cycle arrest and apoptosis (17). Cell cycle arrest usually precedes apoptosis, to allow for repair of DNA damage. Whether apoptosis ensues is negotiated between the DNA repair machinery and the apoptotic machinery (18). p53 mediates transcription of pro-apoptotic Bcl-2 family members (such as Bax (19), Puma (6), Noxa (7), and Bid (20)), death receptors like CD95 (21) and TRAIL receptor-2 (22), the caspase-9 activator Apaf-1 (23), and HtrA2/ Omi (24). In this scenario, one can envision that cells are sensitized for apoptosis induction via the mitochondrial pathway as well as the death receptor pathway. Whether they undergo apoptosis may depend on expression levels of antiapoptotic Bcl-2 family members and/or availability of the death receptor ligand.
Wild-type p53 is not absolutely required for apoptosis induction by DNA-damaging regimens, as shown in cycling peripheral T lymphocytes and lymphoma cells of p53 Ϫ/Ϫ mice (25) and in various p53 mutant cancer cell lines. For instance, the Jurkat T leukemic cells used in this study are highly sensitive to DNA-damaging anti-cancer drugs, such as the topoisomerase II inhibitor etoposide, as well as to ␥-radiation (26,27), whereas they harbor mutant p53, which is transcriptionally inactive (28,29). In cycling murine T lymphocytes and lymphoma cells (25), as well as in Jurkat cells (26,27), Bcl-2 overexpression inhibits apoptosis induction by DNA-damaging anti-cancer drugs and ␥-radiation, indicating that in these cells the p53independent DNA damage response relies on the mitochondrial pathway for caspase activation. However, which signaling molecules activate the mitochondria in this pathway is unclear.
In the work reported here, we have addressed this question and found that Bid is the key mediator of mitochondrial activation in the p53-independent DNA damage response in Jurkat T cells. We provide evidence that Bid needs to be proteolytically activated by an aspartate-specific protease to perform its function in this pathway. However, we have excluded granzyme B and all known inducer caspases from playing a role in Bid processing. This includes caspase-2, which has been featured recently as an activator of the mitochondrial pathway in response to DNA-damaging anti-cancer drugs (30,31).

EXPERIMENTAL PROCEDURES
Reagents-Mouse anti-human CD95 monoclonal antibody (mAb) 7C11 was obtained from Immunotech (Marseille, France), soluble human recombinant TRAIL and enhancer from Alexis (Laufelfingen, Switzerland), and etoposide from Sigma. Polyclonal rabbit anti-Bid antibody was raised in our laboratory against a fusion protein of glutathione S-transferase and full-length Bid. It detects full-length and truncated (t)Bid (32). Anti-actin mAb C4 was obtained from Roche Applied Science; anti-Cyt c mAbs 7H8.2C12 and 6H2.B4, anti-caspase-2/ICH-1 L mAb (clone 35), and anti-active caspase-3 mAb were from BD Biosciences. Anti-HA mAb 12CA5 and anti-VSV mAb P5D4 were used as purified Ig prepared from available hybridomas. Rabbit anti-caspase-9 polyclonal antibody was purchased from New England Biolabs. Horseradish peroxidase-conjugated rabbit anti-mouse Ig and swine anti-rabbit Ig were obtained from Dako A/S (Glostrup, Denmark). Protein A-Sepharose beads and the enhanced chemiluminescence (ECL) kit were purchased from Amersham Biosciences.
Cells and Stimulation-The J16 clone was derived from the human T acute lymphoblastic leukemia cell line Jurkat by limiting dilution and was selected for CD95 sensitivity (26,27,32). CD95-resistant JA variant clones were derived by limiting dilution from the parental Jurkat line after 5 weeks of selection for cells resistant to anti-CD95 mAb (26). The functional status of p53 was analyzed in J16 and JA clones by monitoring the transactivation activity of separated alleles in yeast according to Ishioka et al. (33). All cells expressed one functional and one nonfunctional allele. The JS/J mixed lymphocyte culture was generated by co-culturing peripheral blood mononuclear cells of donor JS with irradiated Jurkat cells. Allogeneic responder T cells were propagated by weekly stimulation with feeder cell mixture and cloned by limiting dilution at 1 cell/well in the same feeder mixture (34). T cell clones JS/J7 and JS/J29 were selected for cytotoxicity toward J16 cells as displayed in a conventional 51 Cr release assay. Jurkat cells were cultured in Iscove's modified Dulbecco's medium and Phoenix-Ampho cells in Dulbecco's modified medium, both supplemented with 8% fetal calf serum, 2 mM glutamine, and antibiotics. Prior to stimulation, Jurkat clones were suspended in serum-free Yssel's medium (26) and seeded at 1 ϫ 10 6 per ml, 200 l per well in round-bottom 96-well plates for apoptosis assays and at 5-10 ϫ 10 6 per ml in 24-well culture plates for Cyt c release assays. Prior to irradiation or treatment with etoposide, cells were cultured overnight in Yssel's medium. After addition of stimulus, cells were incubated for the indicated times at 37°C, 5% CO 2 . Routinely, cells were stimulated by incubation with 50 ng/ml anti-CD95 mAb for 4 -6 h, with 100 ng/ml recombinant TRAIL plus 10-fold excess of enhancer for 6 -8 h, or with 10 g/ml etoposide for 16 -18 h. Cells were irradiated with 30 gray using a 137 Cs source (415 Ci; Von Gahlen Nederland, B.V.) and incubated for 16 -18 h.
Apoptosis Assays-To measure nuclear fragmentation (subdiploid DNA content), cells were lysed in 0.1% Triton X-100, 0.1% sodium citrate, 50 g/ml propidium iodide (PI) (35). Fluorescence intensity of PI-stained nuclei was determined on a FACScan (BD Biosciences), and data were analyzed using Cellquest software. Levels of active caspase-3 were determined as follows. After stimulation, cells were fixed in 4% paraformaldehyde, washed in phosphate-buffered saline (PBS) with 0.1% saponin and 0.5% bovine serum albumin (BSA), and stained for 45 min on ice with phycoerythrin-conjugated anti-active caspase-3 mAb in PBS with 0.1% saponin and 0.5% BSA. Subsequently, cells were washed and analyzed by flow cytometry. For determination of apoptosis inflicted by the cytotoxic JS/J clones, effector T cells and J16 target cells were co-cultured in 24-well plates at the indicated ratios and subsequently washed in PBS, fixed for 2 h at Ϫ20°C with 70% ethanol, washed in PBS, incubated for 1 h at 4°C with 100 g/ml RNase and 20 g/ml PI, and analyzed by flow cytometry on a FACScan. Percentage of subdiploid PI-positive particles was scored as percentage apoptosis. Background apoptosis in effector and target cells that had been cultured alone was minimal but subtracted from the percentage apoptosis determined in all co-cultures.
Immunoblotting-For monitoring Cyt c release, stimulated cells were washed twice with ice-cold PBS, suspended in 100 l of extraction buffer (50 mM PIPES-KOH, pH 7.4, 220 mM mannitol, 68 mM sucrose, 50 mM KCl, 5 mM EGTA, 2 mM MgCl 2 , 1 mM dithiothreitol, and protease inhibitors), and allowed to swell on ice for 30 min. Cells were homogenized by passing the suspension through a 25-gauge needle (10 strokes). Homogenates were centrifuged in a Beckman Airfuge at 100,000 ϫ g for 15 min at 4°C, and supernatants were harvested as cytosolic fractions. Fifteen g of cytosolic protein, as determined by the Bio-Rad protein assay (Bio-Rad), were loaded per lane onto 15% SDS-polyacrylamide gels and blotted onto nitrocellulose. For monitoring protein expression levels or caspase processing, cells were washed with PBS and lysed in 10 mM triethanolamine HCl, pH 7.6, 1% Nonidet P-40, 150 mM NaCl, 5 mM EDTA, and protease inhibitors. Lysates were cleared by centrifugation and analyzed for protein content by Bio-Rad protein assay. Equal amounts of protein (50 -75 g per lane) were separated on 12% SDSpolyacrylamide gels and transferred to nitrocellulose sheets. After transfer, membranes were blocked for 1 h in PBS, 0.05% Tween with 5% not-fat dry milk and probed in PBS, 0.05% Tween with 7H8.2C12 anti-Cyt c mAb (1:1000) and anti-actin mAb (1:10,000), or with anticaspase-2 (1:1000), anti-Bid (1:1000), or anti-caspase-9 antibodies (1: 2000). After incubation with a 1:7500 dilution of the appropriate horseradish peroxidase-conjugated secondary antibody, immunostained proteins were visualized by ECL.
Isolation of Mitochondria-Mouse liver cells were lysed by Dounce homogenization in mitochondrion incubation buffer (MIB): 5 mM HEPES-KOH, pH 7.2, 250 mM mannitol, 0.5 mM EGTA, 0.1% (w/v) BSA, 1 g/ml leupeptin, and 0.1 mM phenylmethylsulfonyl fluoride (36). Nuclei and debris were removed by centrifugation at 600 ϫ g for 5 min at 4°C, and a pellet containing mitochondria was obtained by a spin at 10,000 ϫ g for 10 min at 4°C. The pellet was suspended in MIB and layered on a gradient consisting of layers of 10, 18, 30, and 70% Percoll in 25 mM HEPES-KOH, pH 7.2, 225 mM mannitol, 0.5 mM EGTA, and 0.1% (w/v) BSA. Purified mitochondria were collected at the 30/70% Percoll interphase after centrifugation in an SW-41 rotor for 35 min at 13,500 ϫ g at 4°C. The harvested fraction was diluted at least 5-fold in MIB and centrifuged for 10 min at 6,300 ϫ g at 4°C. After two more washes in MIB, mitochondria were suspended in Wang buffer B, which is 20 mM HEPES-KOH, pH 7.5, 220 mM mannitol, 68 mM sucrose, 100 mM KCl, 1.5 mM MgCl 2 , 1 mM EDTA, 1 mM EGTA, 1 mM dithiothreitol, and 0.1 mM phenylmethylsulfonyl fluoride (10) to a protein concentration of 5 mg/ml.
In Vitro Mitochondrion Assay-Cytosols were prepared by hypotonic lysis as described above in Wang buffer B with 10 mM KCl and subsequently incubated with anti-Bid or normal rabbit serum and protein A-Sepharose beads five times for 1 h at 4°C. Purified mitochondria (25 g/sample) were incubated with 100 g of cytosolic protein in a final volume of 30 l of Wang buffer B at 30°C for 1 h and subsequently centrifuged for 5 min at 10,000 ϫ g at 4°C. Mitochondrial pellets were solubilized in SDS sample buffer and separated by 15% SDS-PAGE at 12.5 g of protein/sample. Immunoblotting for Cyt c content was performed as described above.
Flow Cytometric Analysis of Cyt c Release-Cells were analyzed for Cyt c content by intracellular fluorescence of digitonin-permeabilized and subsequently fixed cells, according to a protocol published previously (37). This method relies on the fact that digitonin permeabilizes the plasma membrane but not the mitochondrial membrane. Prior to permeabilization, cells were washed twice in PBS. Anti-Cyt c mAb 6H2.B4 was used at a 1:100 dilution and a Cy5-conjugated goat antimouse Ig antibody (Molecular Probes) at a 1:100 dilution. Cells were analyzed using a FACSCalibur and Cellquest software (BD Biosciences).
Stable Knock Down of Human Caspase-2 by Short Interfering (si)RNA-pSUPER-GFP (a gift from Dr. H. Spits and R. Schotte, Amsterdam Medical Center, Amsterdam, The Netherlands) is identical to the pSUPER vector described previously (38), except that it expresses GFP under a phosphoglycerate kinase promoter. Complementary sense and antisense oligonucleotides (caspase-2 si II sense oligo 5Ј-acagctgttgttgagcgaa-3Ј, caspase-2 si III sense oligo 5Ј-tgcaagagaaactgcagaa-3Ј) were annealed and cloned into pSUPER-GFP. The siRNA GFP cassette was cloned into pSin, an LZRS-based retroviral vector that allows expression of the siRNA under a polymerase III promoter (kindly provided by Dr. H. Spits and R. Schotte). Packaging cell supernatants containing ecotropic retrovirus encoding caspase-2 si II and caspase-2 si III were used to transduce J16 cells expressing the ecotropic retroviral receptor (39). (This receptor was introduced by stable transduction of the cDNA cloned in pBabe-puro and selection by puromycin.) Transduced cells expressing caspase-2 si constructs were selected for GFP expression using a MoFlo high speed cell sorter.
Multiplex Ligation-dependent Probe Amplification (MLPA)-The principle of MLPA has been described previously (40). This method allows amplification and quantitative analysis of transcripts from multiple genes in a one-tube assay. Total RNA was prepared from nonstimulated or 3-, 6-, or 9-h etoposide-stimulated or ␥-irradiated J16 cells by the Trizol method and reverse-transcribed by using a gene-specific oligonucleotide probe mix. The resulting cDNA was annealed overnight to the MLPA probe mix. Probes used in this study were designed to hybridize with cDNA of 36 human apoptosis-related genes. Detailed reaction conditions for MLPA with these particular probes and their performance have been described (41). The annealed oligonucleotides and cDNAs were ligated, and products were amplified by PCR, with one unlabeled and one fluorescently labeled primer. The resulting DNA mixture was analyzed on an Applied Biosystems 3100 capillary sequencer with GeneScan and Genotyper software packages (Applied Biosystems, Warrington, UK). As expression of housekeeping genes varied minimally throughout the time courses, the sum of the fluorescence intensities of all products amplified in one reaction was set at 100% to correct for fluctuations in total signal between samples, and individual peaks were calculated relative to the 100% value.

Common Signaling Requirements for Apoptosis Induction by DNA-damaging Anti-cancer Regimens and Death
Receptors-In wild-type J16 Jurkat T cells, apoptosis induction by the topoisomerase II inhibitor etoposide and ␥-radiation shares aspects with apoptosis induction by the death receptor CD95. This was first discovered when variant Jurkat cells, selected for resistance to CD95-mediated apoptosis, proved cross-resistant to these DNA-damaging anti-cancer regimens (26). In variant JA cells, etoposide and ␥-radiation bring about a cell cycle arrest in G 2 , as they do in wild-type J16 cells (Fig. 1A). This arrest is followed by an apoptotic response in the case of J16, but JA cells do not undergo apoptosis. Upon prolonged incubation, they remain arrested in G 2 and ultimately die by necrosis within a week (results not shown). Apparently, these cells have retained the capacity to sense the detrimental con-FIG. 1. Apoptosis signaling by DNA-damaging inputs in Jurkat cells shares aspects with death receptor signaling but is independent of death receptor activation. A, the wild-type clone J16 and the variant clone JA1.2, which was selected for CD95 resistance, were stimulated with anti-CD95 mAb or etoposide (E) or ␥-irradiated (IR). The medium control (M) was incubated under the same conditions. Apoptosis was read out as nuclear fragmentation by PI staining of nuclei and fluorescence-activated cell sorter analysis. B, J16 cells, stably transduced with empty vector, c-Flip L , or dn FADD cDNA were treated with etoposide, ␥-radiation, or anti-CD95 mAb, and apoptosis was read out as indicated in A. C, J16 cells transduced with empty vector or Bcl-2 cDNA were treated with etoposide, ␥-radiation or anti-CD95 mAb, and apoptosis was read out as indicated in A. All data are representative of multiple independent experiments and show means Ϯ S.D. from triplicate samples in one experiment. sequences of treatment with etoposide and ␥-radiation and can translate these in a cell cycle arrest but have lost the ability to activate the apoptotic machinery.
It has been reported previously that in Jurkat T cells, DNAdamaging anti-cancer drugs and ␥-radiation induce apoptosis in a death receptor-independent manner (42). This was confirmed in this study by retroviral transduction of J16 cells with dn FADD and c-Flip L . dn FADD, which has a death domain but lacks the death effector domain, abrogates death receptor signaling by disallowing recruitment of caspase-8/10 to the death receptor. c-Flip L blocks activation of caspase-8/10 by displacing it from FADD at the cytoplasmic tail of death receptors. Although dn FADD and c-Flip L effectively blocked CD95-induced apoptosis, they did not interfere with apoptosis induction by etoposide and ␥-radiation (Fig. 1B), indicating that these regimens do not require death receptor engagement and/or FADD and caspase-8/10 function to convey the apoptotic signal.
An aspect of apoptosis signaling that is shared by DNAdamaging anti-cancer regimens and death receptors in Jurkat T cells is the requirement of a mitochondrial contribution to activation of the apoptotic execution machinery. This is shown for J16 cells in Fig. 1C, in which retrovirus-mediated overexpression of Bcl-2 not only blocks the apoptotic response to etoposide and ␥-radiation but also strongly inhibits apoptosis induction via CD95 and TRAIL receptor. In JA cells, the capacity of etoposide, ␥-radiation, and CD95 to bring about Cyt c release is abrogated, confirming that the response to these stimuli is under common mitochondrial control (26). These data argue that etoposide and ␥-radiation induce apoptosis in J16 cells via a mitochondrial pathway that shares aspects with the death receptor pathway downstream from caspase-8/10.
Etoposide and ␥-Radiation Convey the Apoptotic Signal via Presynthesized and Aspartate-cleaved Bid-To identify the BH3-only protein that may be instrumental in the DNA damage pathway in Jurkat cells, we first determined which apoptosis-regulatory proteins were transcriptionally regulated by etoposide and ␥-radiation. J16 cells were stimulated for 3, 6, and 9 h, a time frame in which Cyt c release increments to maximal levels, and RNA was isolated. The transcript levels of most known BH3-only proteins, Bak, Bax, inhibitory Bcl-2 family members, and some other apoptosis regulators, were determined by MLPA, a PCR-based assay that allows the quantitative comparison of multiple transcripts amplified in one reaction (40,41). No or only marginal up-or down-regulation was observed for Bcl-2 family members, including Bid, Noxa, Bax, Bak, Bcl-2, and Bcl-x L (see supplemental figure). The exception was Puma, for which transcripts were not detectable in unstimulated cells but modestly up-regulated in stimulated cells. However, expression of Puma siRNA construct, which had been validated in transiently transfected HeLa and 293T cells, had no effect on apoptosis induction by etoposide or ␥-radiation (results not shown). Therefore, analysis of new transcription in response to DNA damage did not clarify which BH3-only protein(s) may convey the apoptotic signal in the pathways activated by etoposide and ␥-radiation in Jurkat cells.
We have demonstrated previously that in JA cells, tBid cannot signal to mitochondria. Recombinant tBid, when incubated with exogenous mouse liver mitochondria in the presence of cytosol from JA cells, was unable to bring about Cyt c release, although it functioned normally in the context of J16 cytosol (32). Therefore, common resistance to CD95-, etoposide-, and ␥-radiation-induced apoptosis in JA cells can possibly be attributed to the inability of tBid to induce mitochondrial permeabilization in all these pathways. To explore whether Bid might be the common mediator involved in signaling to the mitochondria in the DNA damage and death receptor pathways, we first used an in vitro mitochondrion assay. Cytosols were derived from wild-type J16 cells, which had been stimulated with etoposide or ␥-radiation or left untreated for various times. These cytosols activated in vivo were either mock-depleted or depleted of Bid by repeated immunoprecipitation and incubated with mouse liver mitochondria. Subsequently, mitochondria were assayed for the presence of Cyt c. In case mitochondria were incubated with mock-depleted etoposide-or radiation-activated cytosols, Cyt c was effectively released (Fig. 2A). However, etoposide-and radiation-activated cytosols depleted of Bid lost the capacity to release Cyt c from mouse liver mitochondria, indicating that Bid is required to permeabilize mitochondria in response to these stimuli.
In the death receptor pathway, Bid needs to be cleaved to become active and to convey the apoptotic signal to mitochondria. Therefore, we investigated whether overexpression of a noncleavable Bid mutant could inhibit the apoptotic response to DNA damage. J16 cells were retrovirally transduced with a Bid cDNA, containing point mutations that alter the caspase-FIG. 2. Aspartate-cleaved Bid is required for Cyt c release and apoptosis in response to etoposide and ␥-radiation. A, in vitro mitochondrion assay. Mitochondrion-free cytosols were derived from J16 cells, which had been incubated without stimulus (medium) or treated with etoposide (E) or ␥-radiation (IR) for the indicated times. Cytosols were immunodepleted for Bid (⌬Bid) or mock-depleted with normal rabbit serum (control) and incubated with mouse liver mitochondria for 60 min at 30°C. Cyt c content of the mitochondrial fraction was assayed by immunoblotting (IB). B, J16 cells were transduced with empty vector or the Bid D60E/D75E mutant cDNA, in which caspase-8 and granzyme B cleavage sites were mutated. Cells were stimulated with the indicated concentrations of etoposide or doses of ␥-radiation. Apoptosis was read out as nuclear fragmentation. Data represent means Ϯ S.D. of duplicate samples from three independent experiments. C, in vivo Cyt c release assay. The same cells as described for B were stimulated with etoposide (E) or ␥-radiation (IR), or left untreated (M) for the indicated times. Mitochondrion-free cytosols were prepared and assayed for the presence of Cyt c and actin (as a loading control). 8/10 (D60E) and granzyme B (D75E) cleavage sites by substituting aspartate for glutamate (10). This noncleavable Bid mutant has been shown previously to block signaling to the mitochondria by CD95 and TRAIL receptor in J16 cells (32). We demonstrate here that it also fully inhibits etoposide-and ␥-radiation-induced apoptosis in these cells (Fig. 2B). Immunoblotting of mitochondrion-free cytosols, derived from J16 cells treated with etoposide or ␥-radiation, showed that the BidD60E/D75E mutant impeded Cyt c release in both cases (Fig. 2C). We conclude that, surprisingly, the DNA-damaging anti-cancer regimens etoposide and ␥-radiation require aspartate-cleaved Bid to induce Cyt c release and apoptosis in Jurkat T cells.
Bid Acts Upstream of the Mitochondria in Response to DNA Damage-It is known that caspase-3 can cleave Bid at the same aspartate residue as caspase-8/10 and with that amplify Cyt c release in a mitochondrial feedback loop (43). Therefore, we assessed whether in the DNA damage pathway cleaved Bid was required to convey the initial apoptotic signal to the mitochondria or acted downstream of the mitochondria in a feedback loop. To this end, J16 cells were retrovirally transduced to express an active site mutant of caspase-9, which blocks the apoptotic pathway directly downstream of the mitochondria (1). This dn caspase-9 strongly inhibited the apoptotic response to etoposide and ␥-radiation, as expected from the complete reliance on the mitochondrial pathway for effector caspase activation (Fig. 3A). Immunoblotting of mitochondrion-free cytosols, prepared from J16 cells stimulated with etoposide or ␥-radiation, showed that Cyt c release was not detectably reduced by expression of dn caspase-9 (Fig. 3B). From this, it can be concluded that the Cyt c release observed in response to etoposide and ␥-radiation does not rely on a caspase-9-dependent feedback loop. Because overexpression of the Bid mutant completely prevented the release of Cyt c (Fig. 2C), we can conclude that cleaved Bid is an obligatory signaling intermedi-ate in the DNA damage pathway upstream of the mitochondria.
Etoposide-and ␥-Radiation-induced Cyt c Release Cannot Be Inhibited by CrmA-The experiment with dominant negative caspase-9 indicated that DNA-damaging inputs rely on upstream Bid cleavage and Cyt c release on an aspartate-specific protease that can be activated independent of caspase-9. Therefore, the most likely candidates for this protease were other inducer caspases. The lack of effect of c-Flip L on DNA damageinduced apoptosis argued that caspase-8 and -10 were not involved, but we could not formally exclude that caspase-8 and/or -10 were activated in this pathway in a death effector domain-independent manner, which would not be inhibitable by c-Flip L . Alternatively, other potential inducer caspases might be involved. To test these possibilities, we determined the effect of the coxpox serpin CrmA on the DNA damage pathway. CrmA efficiently inhibits caspase-8, -9, and -10, as well as caspase-1, -4, and -5, a group that is primarily implicated in cytokine processing but may have a role in apoptosis as well (44). The structure of CrmA suggests that it acts like other serpins, which mimic substrates and trap proteases in a covalent complex (45). The effect of CrmA overexpression on Cyt c release was monitored by flow cytometry, a method that more easily allows quantification than immunoblotting (37). Clearly, CrmA was functional, as testified by the very efficient inhibition of Cyt c release upon CD95 stimulation in CrmA-expressing J16 cells (Fig. 4). In contrast, etoposide-and ␥-radiationinduced Cyt c release were not affected at all at the 7-h time point when Cyt c release is maximal (Fig. 4) and at earlier time points (not shown). This indicates that the CrmA-inhibitable caspase-1, -4, -5, -8, -9, and -10 are not involved in these pathways (Fig. 4). We have demonstrated previously that the caspase inhibitor Z-VAD-fmk does not block Cyt c release induced by etoposide and ␥-radiation in J16 cells (27), which excludes the involvement of the non-CrmA inhibitable caspase-3, -6, and -7 as well (44).
Granzyme B Is Not Involved in DNA Damage-induced Apoptosis-Apart from caspases, the serine protease granzyme B is aspartate-specific and known to cleave Bid at Asp-75 (10). Granzyme B is present in lysosomal compartments in cytolytic T lymphocytes. Although its intracellular localization is inconsistent with a role in cleavage of cytoplasmic proteins like Bid, it has been implicated recently (46) in the response to DNAdamaging anti-cancer regimens in myeloid malignancies. In these cells, granzyme B protein levels increased in response to etoposide and ␥-radiation. In addition, granzyme B was released from intracellular compartments into the cytosol. The lack of effect of CrmA on etoposide and ␥-radiation-induced Cyt c release already argued against an involvement of granzyme B in these apoptotic pathways, because CrmA can inhibit granzyme B activity (47). To investigate further a potential role of granzyme B in etoposide-and ␥-radiation-induced apoptosis, we used more specific inhibitors of granzyme B, which do not affect caspase activity. To this end, we transduced J16 cells with the human serpin PI-9, or its mouse homologue SPI-6 (48). Retrovirus-mediated expression of these serpins was confirmed by Western blotting (Fig. 5A). However, it did not affect the apoptotic response to etoposide or ␥-radiation (Fig. 5B). As a positive control, we tested the effect of serpin expression on the sensitivity of J16 cells to apoptosis induction by cytolytic T cell clones. Upon specific recognition of target cells by means of their T cell antigen receptor, cytolytic T cells exocytose lysosomal granules that contain granzyme B and deliver this into the target cells with the aid of perforin (11). Fig. 5C shows that PI-9 and SPI-6 effectively inhibited J16 apoptosis as induced by two specific cytolytic T cells clones, JS/J7 and JS/J29. We conclude that granzyme B is not involved in etoposide-and ␥-radiationinduced apoptosis in J16 cells and is therefore not responsible for Bid cleavage in this pathway.

Caspase-2 Processing in Response to DNA Damage and Death Receptor Triggering Occurs Downstream of Initiator Proteases-Recently
, it was found that in certain cell types caspase-2 participates upstream of the mitochondria in the response to DNA-damaging anti-cancer drugs (30). Because caspase-2 cannot effectively be inhibited by CrmA or Z-VADfmk (44), it was a candidate to be involved in upstream Bid processing and Cyt c release. To investigate this, we first determined whether caspase-2 was processed in response to etoposide and ␥-radiation and whether this occurred upstream of Bid. J16 cells overexpressing the Bid D60E/D75E mutant and vector control cells were stimulated with etoposide, ␥-radiation, anti-CD95 mAb, or recombinant TRAIL (Fig. 6A). Caspase-2 processing was detectable in response to all these stimuli, although the response to ␥-radiation was quite modest. The Bid mutant blocked caspase-2 processing induced by etoposide and ␥-radiation but left the death receptor response partially unaffected. In the presence of dn caspase-9, caspase-2 processing in response to etoposide and ␥-radiation was also abrogated, arguing that it was dependent on mitochondrial activation. Again, the response to death receptors was partially unaffected (Fig. 6B). However, both dn FADD and c-Flip L drastically inhibited the caspase-2 processing in response to CD95 and TRAIL receptor stimulation (Fig. 6C). This indicates that caspase-2 processing in the death receptor pathway is dependent on caspase-8/10 activation. The collective data allow us to conclude that detectable caspase-2 processing in response to etoposide and ␥-radiation or death receptor stimulation occurs downstream of inducer caspase activation by the mitochondria or the death receptor complex, respectively.
Caspase-2 Is Not Required for the Apoptotic Response to DNA Damage but Does Contribute to Death Receptor-induced Apoptosis-The observation that caspase-2 processing was only observed downstream of inducer caspases did not fully exclude the possibility that it might be involved in Bid processing upstream of the mitochondria in the DNA damage pathway. There is evidence that caspase-2 can be activated in the absence of its proteolytic processing (49). To examine directly whether caspase-2 was required for apoptosis induction by DNA-damaging regimens, we introduced a retroviral siRNA construct in J16 cells to stably down-regulate its expression. To enhance transduction efficiency, we used J16 cells expressing the ecotropic retroviral receptor. Two different siRNA constructs were tested, which had proven effective in transient transfection experiments in 293T cells. One of these constructs down-regulated endogenous caspase-2 to about 35% of its original level (Fig. 7A). All further experiments were done with this caspase-2 si II construct. The reduction in caspase-2 protein  5. Granzyme B does not participate in apoptosis induction by etoposide and ␥-radiation. J16 cells were stably transduced with empty vector or vector encoding the serpins PI-9 (HA-tagged) or SPI-6 (VSV-tagged), which inhibit the serine protease granzyme B. A, total cell lysates were separated by SDS-PAGE and subjected to immunoblotting to detect expression (anti-HA and anti-VSV) of the transduced serpins. B, J16 cells were treated with etoposide or ␥-irradiated and incubated for the indicated times. Apoptosis was read out as nuclear fragmentation. The experiment is representative of three. C, J16 cells were used as targets for the cytolytic T cell clones JS/J7 and JS/J29. Apoptosis was read out as nuclear fragmentation, as outlined under "Experimental Procedures." level had no impact at all on apoptosis outcome in response to DNA damage (Fig. 7B). However, we observed an ϳ50% reduction in apoptosis upon TRAIL stimulation when caspase-2 protein levels were down-regulated (Fig. 8A). This result demonstrated that the caspase-2 knock down was functional and that caspase-2 contributes to the apoptotic response to death receptor stimulation. However, it is apparently not essential for apoptosis induction by the DNA-damaging anti-cancer regimens in Jurkat cells.
To map further caspase-2 in the death receptor pathway, we examined the effect of its down-regulation on Cyt c release and caspase-3 activation. Mitochondrion-free cytosols derived from TRAIL-stimulated control and caspase-2 knock down cells were analyzed for the presence of Cyt c. As shown in Fig. 8B, caspase-2 down-regulation did not affect Cyt c release at all. In contrast, generation of active caspase-3 was consistently reduced in the caspase-2 knock down cells treated with TRAIL (Fig. 8C). The collective data indicate that caspase-2 does not contribute to effector caspase activation in the DNA damage pathway but does so in the death receptor pathway, downstream of inducer caspase activity. DISCUSSION The data collected in this study lead to the model depicted in Fig. 9. Two independent lines of evidence argue that Bid is the essential mediator of mitochondrial activation in the apoptotic response to etoposide and ␥-radiation in the T leukemic cells studied: depletion of Bid from cytosols of stimulated cells and use of a Bid mutant. In addition, the finding that tBid is the common intermediate in the death receptor and DNA damage pathways in Jurkat cells explains the cross-resistance to etoposide and ␥-radiation in Jurkat cells selected for CD95 resistance. Consistent with the present model, we have previously found that in the resistant cells a cytosolic factor prevents tBid from permeabilizing mitochondria (32).
This unambiguous functional implication of Bid in anti-cancer drug-and ␥-radiation-induced apoptosis is novel. Initial analysis of death pathways in Bid Ϫ/Ϫ mice only indicated a requirement for Bid in CD95-induced apoptosis in hepatocytes. Etoposide-induced apoptosis in Bid Ϫ/Ϫ mouse embryonic fibroblasts appeared normal (3,50). However, it was reported recently (20) that Bid Ϫ/Ϫ mouse embryonic fibroblasts have an impaired apoptotic response to adriamycin or 5-fluorouracil. In the same study, it was found that the Bid gene contains a p53 response element. In colon and spleen tissue sections derived from ␥-irradiated mice, Bid mRNA was induced in a p53-dependent manner (20). These findings suggest a role for Bid as a sensor of DNA damage in at least some tissues. Most inter- esting, Bid Ϫ/Ϫ mice are prone to develop a malignancy resembling chronic myelomonocytic leukemia, identifying Bid as a tumor suppressor (51). Whether this reflects a role for Bid in DNA damage responses or in feedback control of cell proliferation remains to be established.
The Jurkat cells used in our studies have one dysfunctional p53 allele as revealed in a yeast reporter assay (see "Experimental Procedures"). Functionally, they are expected to display a p53-deficient phenotype, which is consistent with our observations. The apoptosis gene expression profile before and after etoposide treatment and ␥-radiation revealed only Puma as a DNA damage response gene in J16 cells among the array of apoptosis regulators tested. We could not reliably detect Puma protein by immunoblotting or metabolic labeling with antibod-ies, validated by us and others (6). Because studies in Puma Ϫ/Ϫ mice recently revealed that Puma is important for etoposideand ␥-radiation-induced p53-dependent apoptosis in thymocytes (52, 53), we further tested its involvement by using a validated siRNA construct (41). However, we found no effect of Puma siRNA on etoposide-and ␥-radiation-induced apoptosis in J16 cells (results not shown).
Most interesting, tissue-specific roles for Puma and Noxa are emerging, as Noxa was not required for thymocyte apoptosis in response to ␥-radiation (53) but made a significant contribution to intestinal crypt cell apoptosis (52). Puma was also found to contribute to p53-independent pathways, such as serum deprivation-induced apoptosis in myeloid cells (52). Clearly, the division of labor between BH3-only proteins needs to be studied carefully for each apoptotic input in different tissues and their developmental stages. Systematic study of the responses in Bid Ϫ/Ϫ mice will hopefully reveal which primary cell lineage the pathway dominates as we have elucidated in this study. Because Jurkat T cells represent mature thymocytes, the pathway revealed here might be operational in cycling mature T cells, which die in response to etoposide and ␥-radiation in a p53-independent manner (25).
Proteolytic cleavage of Bid within the flexible loop formed between helix 2 and 3 reveals the hydrophobic residues in the BH3 domain helix 3 that trigger the apoptotic pathway. Substitution of aspartate by glutamate at residues 60 and/or 75 in this loop is not expected to alter its conformation (54,55). Therefore, we interpret the finding that the Bid mutant inhibits etoposide-and ␥-radiation-induced apoptosis to mean that aspartate-cleaved Bid is a signaling intermediate in these pathways. It is considered unlikely that mutated Bid represses the DNA damage pathway because of an acquired alternative conformation and resulting activity. Unfortunately, in Jurkat cells the amount of Bid that is proteolytically processed upstream of the mitochondria is below detection levels, even in the death receptor pathway. The Bid processing that can be detected is the result of a mitochondrial feedback loop, because In response to etoposide and ␥-radiation, pre-existing Bid protein conveys the apoptotic signal to the mitochondria, contingent upon its activation by an undefined protease, which is not granzyme B or any known inducer caspase, but cleaves at aspartates 60 and/or 75. Inhibition of Bid by Bcl-2 abrogates the apoptotic response. Caspase-2 (Casp-2) is activated downstream of caspase-9 either directly or indirectly via effector caspases and does not detectably contribute to apoptotic execution. Upon death receptor activation, caspase-8 is activated and cleaves Bid, which signals to the mitochondria. Inhibition of Bid by Bcl-2 strongly inhibits the apoptotic response but does not fully abrogate it. Caspase-8 also processes caspase-2, which contributes to caspase-3 activation and apoptotic execution but not to mitochondrial activation. Caspase-8 may also process caspase-3 directly. it is fully abrogated by Bcl-2 overexpression (26,56). Therefore, we cannot validate the effect of the Bid mutant on endogenous Bid processing in vivo. However, its dramatic effect on both death receptor-and DNA damage-induced Cyt c release strongly suggests that it acts in a dominant negative manner.
Several groups have reported links between caspase-2 and release of mitochondrial pro-apoptotic factors (30,31,57,58). The finding that caspase-2 resides in the nucleus and can signal from this site to the mitochondria (58) makes it an attractive candidate to control the DNA damage response. Indeed, Lassus et al. (30) have presented an elegant study with siRNA that indicates an important contribution of caspase-2 to etoposide-induced apoptosis in U2OS sarcoma cells, A549 carcinoma cells, and E1A-expressing IMR90 fibroblasts. Nevertheless, by using a similar siRNA approach to down-regulate caspase-2, we could not demonstrate such a contribution in Jurkat T cells. Lymphoblasts from caspase-2 Ϫ/Ϫ mice have a normal apoptotic response to etoposide (59), and thymocytes have a normal response to ␥-radiation (60), suggesting that the contribution of caspase-2 to the DNA damage pathway may be lineage-specific. At present, we cannot reconcile our findings with those of Robertson et al. (31), who found by using tetrapeptide inhibitors and transfection with a caspase-2 antisense construct that caspase-2 is involved in mitochondrial activation during etoposide-induced apoptosis in Jurkat cells.
The effect of siRNA for caspase-2 on TRAIL-induced apoptosis served as a positive control in our study. Our findings agree with those of Droin et al. (61) who down-regulated caspase-2 in Jurkat cells with an antisense construct. These authors also found a partial reduction in the apoptotic response to CD95 and TRAIL receptor stimulation. However, they concluded that caspase-2 was involved in caspase-8 activation and ensuing Bid cleavage, because these events were affected by caspase-2 down-regulation. However, as mentioned above, the caspase-8 and Bid processing that is detectable in Jurkat cells is the result of a mitochondrial feedback loop. Therefore, our interpretation that caspase-2 acts downstream from caspase-8, as based on c-FlipL overexpression, is most likely the correct one ( Fig. 9).
Caspase-8 (10), caspase-2 (57), and caspase-3 (62) are known to cleave Bid at Asp-60 (Asp-59 in murine Bid), whereas granzyme B cleaves Bid at Asp-75 (10). We have excluded caspase-8, -2 (and -9), as well as granzyme B from cleaving Bid in the DNA damage pathway. Therefore, the search is on for an aspartatespecific protease, which can act as inducer in this pathway. We have reported previously (27) that the caspase inhibitor Z-VAD could not block Cyt c release in response to etoposide and ␥-radiation in J16 cells. This finding argues against caspase-3 or other caspases playing a role in this pathway. Moreover, effector caspases are unlikely candidates because they require aspartate-specific cleavage to become activated themselves. To our knowledge, no other aspartate-specific proteases than caspases and granzyme B have been defined in mammalian cells. The cysteine protease calpain has been implicated in radiation-induced apoptosis in thymocytes (63), but calpain cleaves Bid at Gly-70 (62). Lysosomal extracts cleave Bid at Arg-65 (55). Because the mutant Bid protein used in this study seems to act as a dominant negative, it may bind with high affinity to the upstream protease involved in its cleavage, which should allow its identification.