The p53-cathepsin axis cooperates with ROS to activate programmed necrotic death upon DNA damage

Abstract

Three forms of cell death have been described: apoptosis, autophagic cell death, and necrosis. Although genetic and biochemical studies have formulated a detailed blueprint concerning the apoptotic network, necrosis is generally perceived as a passive cellular demise resulted from unmanageable physical damages. Here, we conclude an active de novo genetic program underlying DNA damage-induced necrosis, thus assigning necrotic cell death as a form of “programmed cell death.” Cells deficient of the essential mitochondrial apoptotic effectors, BAX and BAK, ultimately succumbed to DNA damage, exhibiting signature necrotic characteristics. Importantly, this genotoxic stress-triggered necrosis was abrogated when either transcription or translation was inhibited. We pinpointed the p53-cathepsin axis as the quintessential framework underlying necrotic cell death. p53 induces cathepsin Q that cooperates with reactive oxygen species (ROS) to execute necrosis. Moreover, we presented the in vivo evidence of p53-activated necrosis in tumor allografts. Current study lays the foundation for future experimental and therapeutic discoveries aimed at “programmed necrotic death.”

Cells constantly interpret environmental cues before making a correct life versus death decision to assure proper operation of all biological processes (1). Impaired or overactive death results in human illness including cancer, autoimmune disorders, neurodegenerative diseases, stroke, and myocardial infarction (2). Research focused on the cell death/survival control provides a vantage point for the development of therapeutic agents that specifically trigger or prevent cell death (3, 4). Based on the morphological features, 3 forms of cell death have been described: apoptosis, autophagic cell death, and necrosis (5–7). Among the 3 distinct forms of cell death, apoptosis is best studied. The evolutionarily conserved signaling cascade, consisting of the BCL-2 family, the adaptor protein Apaf-1, and the caspase family, outlines the essential apoptotic network (8). The “activator” BH3-only molecules, BID, BIM, and PUMA, convey apoptotic signals to trigger homo-oligomerization of multidomain proapoptotic BAX and BAK, which in turn permeabilize mitochondria, leading to the efflux of cytochrome c, the assembly of apoptosome, and the ultimate activation of caspases (9–11). Autophagy serves as either a survival or a death mechanism depending upon the context of the signaling events (12). Although necrosis is generally perceived as an unavoidable consequence of extreme physicochemical stress, recent studies have indicated a molecular control of tumor necrosis factor (TNF)/Fas and DNA alkylation induced necrosis (13–18). However, whether necrosis triggered by other intrinsic death signals is similarly regulated and more importantly, whether necrotic cell death recruits de novo genetic programs in analogy to apoptosis remain undetermined.

Results

DNA Damage Activates a de Novo Genetic Program to Execute Necrotic Death in Bax and Bak DKO Cells.

To interrogate nonapoptotic cell death, we used Bax and Bak double-knockout (DKO) cells that are deprived of the apoptotic gateway to mediate cytochrome c release for caspase activation (Fig. S1) (9–11, 19, 20). Despite the lack of caspase activation (20), DKO cells eventually succumb to various death signals manifesting a much slower death kinetics compared with wild-type cells (Fig. 1A, Fig. S2, and data not shown). To investigate the mechanism(s) underlying BAX/BAK-independent cell death, we first examined the morphological features of the dying DKO cells. Electron microscopy uncovered signature characteristics of necrosis in DKO cells after DNA damage, including the loss of plasma membrane integrity, the spillage of intracellular contents, the swelling of organelles, the appearance of translucent cytosolic compartment, and the formation of intracellular vacuoles (Fig. 1B). We next determined one of the biochemical hallmarks of necrosis-the release of high mobility group box I (HMGB1) protein, a proinflammatory cytokine (21). HMGB1 functions in the nucleus as an architectural chromatin-binding factor that bends DNA, stabilizes nucleosomes, and facilitates transcription, whereas outside the cell it is a potent mediator of inflammation, cell migration and metastasis (22). Indeed, HMGB1 was released into the culture medium when DKO cells died upon DNA insults (Fig. 1C). Analyses revealed a 2-step release of HMGB1-it first translocated from the nucleus to the cytosol ≈8 h after genotoxic stress (Fig. 1D), followed by the diffusion to the extracellular milieu once the plasma membrane integrity was lost (Fig. 1C). Double staining of necrotic cells with Annexin-V and PI have been shown in dsRNA- and TNF-induced necrosis (ref. 23 and T. Vanden Berghe and P. Vandenabeele, personal communications). By analogy, the dying Bax and Bak DKO cells were stained positively with both Annexin-V and propidium iodide (Fig. S2). Although some cells displayed increased double-membrane autophagosomes and conversion of LC3-I to LC3-II was observed (Fig. S3), knockdown of Beclin1 and ATG5 provided minimal protection at 3 days after DNA damage (Fig. S4). Furthermore, over-expression of BCL-2 or BCL-X_L had no effects, supporting that DNA damage-induced death in DKO cells is not caused by mitochondrion-dependent apoptosis (Fig. S5A). Remarkably, DNA damage-induced necrosis in DKO cells was completely blocked by either actinomycin D (ActD) or cycloheximide (CHX) that inhibits transcription and translation, respectively (Fig. 2 A and B). Moreover, transient cycloheximide treatment enhanced the clonogenic survival of DKO cells in response to DNA damage (Fig. 2C). Of note, cycloheximide did not prevent necrotic death triggered by hydrogen peroxide (Fig. S5 B and C). These findings indicate the essence of de novo gene expression in executing necrotic signaling cascade upon DNA damage, which contrasts the general perception of necrosis as passive, unorganized cellular demise.

Fig. 1.

DNA damage activates necrotic death in Bax and Bak DKO cells. (A) Bax and Bak DKO cells eventually succumb to DNA damage in a much slower death kinetics compared with wild-type cells. Wild-type or Bax and Bak DKO cells were treated with etoposide (10 μg/ml) for the indicated times. Cell death was quantified by Annexin-V and propidium iodide staining. Data are mean ± SD from 3 independent experiments. (B) Electron microscopy demonstrates necrotic features of Bax and Bak DKO cells after treatment with etoposide (10 μg/ml) for 2 days. (C) HMGB1 is released extracellularly during DNA damage-induced necrosis. Anti-HMGB1 immunoblot was performed on cell lysates (C) or culture medium (M) after treatment with etoposide for indicated time. (D) HMGB1 translocates from nucleus to cytosol upon DNA damage. Fluorescence microscopy of Bax and Bak DKO cells treated with 10 μg/ml etoposide for 8 h. Green, HMGB1 immunostaining; blue, Hoechst staining of DNA.

Fig. 2.

DNA damage activates a de novo genetic program that is orchestrated by p53 to execute necrotic death in Bax and Bak DKO cells. (A) DNA damage-induced cell death in Bax and Bak DKO cells requires de novo gene expression. SV40-transformed Bax and Bak DKO cells were treated with etoposide or etoposide plus actinomycin D (ActD, 1 μM) or etoposide plus cycloheximide (CHX, 2 μg/ml). Cell death was quantified by Annexin-V and propidium iodide staining the at indicated times. (B) E1A/Ras transformed Bax and Bak DKO cells were treated with etoposide (5 μg/ml) or etoposide plus cycloheximide (2 μg/ml). Cell death was quantified by Annexin-V staining after 4 days. (C) Cycloheximide enhances clonogenic survival of DKO cells in response to DNA damage. Three independent clones of SV40-transformed Bax and Bak DKO cells were treated with etoposide or etoposide plus cycloheximide for 24 h. Colonies were stained with crystal violet after 12 days. (D) A dominant negative mutant of p53 or knockdown of p53 protects DKO cells from DNA damage-induced necrotic death. SV40-transformed DKO cells transduced with control retrovirus (MIG or pSuper-Retro) or retrovirus expressing either a dominant negative mutant of p53 or shRNA against p53, were treated with etoposide for 3 days. Cell death was quantified by Annexin-V. (E) E1A/Ras transformed Bax and Bak DKO cells transduced with control retrovirus (MIG) or retrovirus expressing a dominant negative mutant of p53, were treated with etoposide (5 μg/ml) for the indicated times. Cell death was quantified by Annexin-V. Data in A, B, D, and E are mean ± SD from 3 independent experiments. *, P < 0.01.

p53 Transactivates Cathepsin Q to Trigger Necrotic Death in Bax and Bak DKO Cells.

To pinpoint the responsible transcription factor(s) in orchestrating this necrotic program, we investigated several candidates known to control cell death and identified p53 as the key necrotic programmer. Overexpression of a dominant negative mutant of p53 or retrovirus-mediated stable knockdown of p53 protected DKO cells from DNA damage-induced necrosis (Fig. 2 D and E and Fig. S6). To identify the molecular signatures underlying the observed necrosis, we performed gene expressing profiling on DKO cells upon DNA damage and recognized up-regulation of several lysosomal cysteine proteases including cathepsin F, H, and Q, but not cathepsin B and L (data not shown). The observed induction of cathepsin F, H, and Q was confirmed by quantitative RT-PCR analyses (Fig. 3A). Importantly, the up-regulation of these cathepsins was abrogated when p53 was knocked down, positioning p53 in orchestrating the cathepsin program (Fig. 3A). The importance of individual cathepsins in necrosis was investigated with shRNA-mediated stable knockdown of respective cathepsins (Fig. S7). Knockdown of cathepsin Q, but not F or H, protected DKO cells from DNA damage-induced necrosis activated by etoposide (Fig. 3B). Furthermore, deficiency of cathepsin Q also protected DKO cells from UV- and camptothecin-induced necrosis (Fig. 3C and data not shown). Although cathepsin Q was barely detected by anti-cathepsin Q Western blot, DNA damage induced up-regulation of cathepsin Q protein that was readily detectable (Fig. S7D and S8A). Real time RT-PCR revealed a ubiquitous, mostly low-level expression of cathepsin Q in all mouse tissues except lung where no message was detected (Fig. S8B), which contrasts the prior reported placenta-restricted expression (24). Indeed, the expression of cathepsin Q protein could be detected in bone marrow (Fig. S8C). To directly link p53 to the transactivation of cathepsin Q, we searched for consensus p53 binding sites based on the p53MH algorithm (25) and identified one candidate p53-binding site located at the intron 1 of cathepsin Q. Indeed, intron 1 but not the promoter of cathepsin Q responded to p53-mediated transactivation (Fig. 3D). Moreover, mutation of the identified p53-binding site abolished its responsiveness to p53 (Fig. 3D). Taken together, our data uncovered an active genetic program, the p53-cathepsin Q axis, which initiates “programmed necrotic death” in response to DNA damage.

Fig. 3.

p53 transactivates cathepsin Q to trigger necrotic death in Bax and Bak DKO cells. (A) p53-dependent transactivation of cathepsin expression upon DNA damage. DKO cells with or without p53 knockdown were treated with etoposide for 6 h. Levels of various cathepsins were analyzed by qRT-PCR and presented as mean ± SD. (B) DKO cells transduced with control retrovirus (pSuper-Retro) or retrovirus expressing scrambled shRNA or shRNA against indicated cathepsins were treated with etoposide for 3 days. (C) DKO cells transduced with retrovirus expressing scrambled shRNA or shRNA against cathepsin Q were irradiated with UV and cell death was assessed at day 4. (D) Intron 1 of cathepsin Q is responsive to p53. Luciferase reporter constructs containing the intron 1 of cathepsin Q (CtsQ I1-Luc), the intron 1 of cathepsin Q with a mutated p53 binding site (CtsQ I1m-Luc), or the promoter of cathepsin Q (CtsQ prom-Luc) were assayed for transactivation by wild-type p53 in Saos-2 cells 24 h after transfection with the indicated reporters and either p53 or vector control. Data are presented as mean fold activation ± SD from 3 independent experiments. (E) Necrotic death in Bax and Bak DKO cells can be blocked by a cathepsin L inhibitor. Bax and Bak DKO cells were treated with etoposide or etoposide plus z-FY-CHO (100 μM) for 3 days. (F) Bax and Bak DKO cells were treated with etoposide, followed by addition of z-FY-CHO (100 μM) at 0, 6, 12, or 24 h. Cell death was assessed after 3 days. Cell death was quantified by Annexin-V. Data are mean ± SD from 3 independent experiments. *, P < 0.01.

ROS Cooperates with p53-Cathepsin Q to Execute DNA Damage-Induced Necrotic Death.

Because cathepsin Q is homologous to cathepsin L and no specific cathepsin Q inhibitors are available, we used a cathepsin L inhibitor to chemically probe the cathepsin Q executed necrosis. Cathepsin L inhibitor, z-FY-CHO, significantly inhibited necrotic death in DKO cells despite the minor toxicity associated with the chemical (Fig. 3E). Of note, other cathepsin inhibitors including E64, CA-074 Me (cathepsin B inhibitor), or pepstatin A (cathepsin D inhibitor) did not protect DKO cells (Fig. S9 A–C). Furthermore, knockdown of cathepsin L or cathepsin B provided minimal protection (Fig. S9 D–F). With this inhibitor in hand, we investigated at which time point when inhibition of cathepsin Q no longer provides protection. We added this inhibitor at 6, 12, or 24 h after etoposide treatment. Z-FY-CHO no longer prevented necrotic death when added at 12 h after etoposide treatment (Fig. 3F), which correlated with the peak of cathepsin Q induction by DNA damage (Fig. S8A). These data suggest that cathepsin Q needs to be induced to execute necrotic death. However, the ≈24-h lag between cathepsin Q induction and the detectable onset of cell death (Fig. 1A), and the inability of cathepsin Q overexpression alone to trigger cell death indicate the existence of additional death effector(s) that cooperates with cathepsin Q to induce death. Because ROS is one of the most important secondary messengers implicated in necrotic death (6, 13, 14, 26), we investigated whether ROS is such a candidate. Indeed, both the antioxidant N-acetyl cysteine (NAC) and flavin-dependent oxidoreductase inhibitor, diphenylene iodonium (DPI), protected DKO cells from etoposide-induced necrosis (Fig. 4A). More importantly, DPI further protected DKO cells with stable knockdown of cathepsin Q from necrotic death (Fig. 4B), supporting the notion that cathepsin Q and ROS are the necrotic death effectors. Although p53 was reported to up-regulate a series of redox genes to produce ROS (27), expression of p53 DN or knockdown of p53 did not prevent the accumulation of ROS in DKO cells upon genotoxic stress (Fig. S10 and data not shown). By contrast, both cycloheximide and DPI significantly reduced the ROS production (Fig. S10B). Accordingly, DPI further protected cells expressing p53 DN from necrotic death (Fig. 4C). DPI also protected Bax, Bak and p53 triple knockout MEFs (data not shown). In summary, our data are consistent with a model in which p53 coordinates the induction of cathepsin Q that cooperates with ROS to execute “programmed necrotic death” (Fig. 5D).

Fig. 4.

ROS cooperates with p53-Cathepsin Q to execute DNA damage-induced necrotic death and DNA damage-induced necrotic program is present in wild-type cells. (A) Bax and Bak DKO cells were treated with etoposide or etoposide plus N-acetylcysteine (NAC, 20 mM) or etoposide plus diphenylene iodonium (DPI, 100 μM) for indicated time. *, P < 0.01. (B) DKO cells transduced with retrovirus expressing scrambled shRNA or shRNA against cathepsin Q were treated with etoposide or etoposide plus diphenylene iodonium for 4 days. *, P < 0.01 between CtsQ KD and CtsQ KD plus DPI or between DPI and DPI plus CtsQ KD. (C) DKO cells transduced with control retrovirus (MIG or pSuper-Retro) or retrovirus expressing a dominant negative mutant of p53, were treated with etoposide or etoposide plus diphenylene iodonium for 4 days. *, P < 0.01. (D) Knockdown of cathepsin Q protects wild-type cells from DNA damage-induced cell death. Wild-type MEFs with or without cathepsin Q knockdown were treated with etoposide for 18 h. *, P < 0.01. (E) A549 cells were irradiated with UV plus the indicated caspase (z-VAD-FMK, 50 μM) or cathepsin inhibitors (z-FY-CHO, 100 μM) and cell death was assessed after 4 days. *, P < 0.01. (F) U031 cells were irradiated with UV plus the indicated caspase (z-VAD-FMK, 50 μM) or cathepsin inhibitors (z-FY-CHO, 50 μM) and cell death was assessed after 3 days. *, P < 0.01. Cell death was quantified by Annexin-V. Data are mean ± SD from 3 independent experiments.

Fig. 5.

In vivo evidence of p53-activated necrotic death upon DNA damage. (A) Electron microscopy demonstrates necrotic features in the tumor allografts derived from Bax and Bak DKO cells expressing GFP but not p53 DN after treatment with etoposide. (B) Tumor weights from groups of 7 mice injected with Bax and Bak DKO cells expressing GFP or p53 DN after etoposide treatment. *, P = 0.0011 (n = 7 vs. 7). (C) Tumor volume from groups of 7 mice injected with Bax and Bak DKO cells expressing GFP or a dominant negative mutant of p53 before etoposide treatment. (D) Model depicts that DNA damage activates the p53-cathepsin axis and ROS to execute “programmed necrotic death.”

DNA Damage-Induced Necrotic Program Is Present in Wild-Type Cells.

As apoptotic machinery efficiently executes cellular demise, the necrotic component is easily masked in cells with functional BAX or BAK. To investigate DNA damage-induced necrosis in the presence of apoptosis, cathepsin Q was knockdowned in wild-type cells. As expected, knockdown of cathepsin Q provided minor yet significant protection of wild-type cells against DNA damage-induced cell death (Fig. 4D). These findings are consistent with the fact that apoptosis dominates over necrosis with its high efficiency and fast kinetics. It was reported that HMGB1 binds tightly and irreversibly to the condensed chromatin of apoptotic cells such that its extracellular release is only observed in necrotic but not apoptotic cells (22). Our observation of early cytoplasmic translocation of HMGB1 in DKO cells after DNA damage raises a possibility that HMGB1 might translocate to the cytosol before chromatin condensation in wild-type cells. Similar to DKO cells, wild-type cells also exhibited early cytoplasmic translocation of HMGB1 before any discernable features of apoptosis were evident (Fig. S11). On the contrary, HMGB1 remained in the nucleus upon apoptosis triggered by staurosporine (Fig. S11). These data suggest that wild-type cells are equipped to mount a full-blown necrotic event once extreme damages compromise the integrity of the plasma membrane. Since there is no direct human orthologues of cathepsin Q been recognized so far, we used protease inhibitors to probe whether the same de novo necrotic pathway operates in human cells. Indeed, the combination of caspase inhibitor (z-VAD-FMK) and cathepsin L inhibitor (Z-FY-CHO) provided a synergistic protection of DNA damage induced cell death in human cancer cell lines, including A549 (lung carcinoma) and U031 (renal cell carcinoma) (Fig. 4 E and F and Fig. S12A). The combination of caspase inhibitor (z-VAD-FMK) and cathepsin L inhibitor (Z-FY-CHO) also provided a synergistic protection of DNA damage induced cell death in E1A/Ras transformed wild-type MEFs (Fig. S12B). These data indicate that apoptotic and necrotic programs work together to execute cellular demise.

Chemotherapy Induces p53-Dependent Necrotic Death in Tumor Allografts.

To demonstrate the DNA damage-activated, p53-dependent necrotic death in vivo, we examined chemotherapy-induced cell death in tumor allografts using E1A/Ras transformed DKO MEFs. Consistent with our in vitro studies, etoposide induced necrosis in tumors derived from DKO cells expressing GFP control but not p53 DN (Fig. 5A). Accordingly, the tumor volume was significantly lower in tumors derived from control DKO cells than those from p53 DN-expressing DKO cells in response to etoposide treatment (Fig. 5B). All of the tumors were within the same size range before chemotherapy (Fig. 5C).

Discussion

Emerging evidence indicates that in response to a given death stimulus, there is often a continuum of apoptosis and necrosis (6, 26). Many insults induce apoptosis at lower doses and necrosis at higher doses. Certain death stimuli such as ischemia-reperfusion injury and excitotoxicity even concurrently induce both apoptotic and necrotic features in neurons (6). Here, we demonstrated that DNA damage activates p53 to trigger a necrotic program (Fig. 5D) in addition to the well-characterized mitochondrion-dependent apoptotic program (28, 29). Necrosis is clearly a fail-safe cell death mechanism for apoptosis. Together, they ensure the clearance of damaged cells to avoid accumulation of detrimental mutations. This further highlights the paramount importance of p53 as the guardian of genome stability and a tumor suppressor (30, 31). DNA damage triggered by topoisomerase inhibitors (etoposide and camptothecin) or UV apparently activates a necrotic signaling pathway that is different from necrosis induced by DNA alkylating agents. DNA alkylation-induced necrosis does not involve p53 nor recruits a de novo genetic program (18). Consistent with a previous report (18), PARP inhibitors protected DKO cell from necrosis induced by alkylating DNA damage but not by DNA double-strand break (Fig. S12C). Moreover, neither knockdown of cathepsin Q nor z-FY-CHO provided protection against alkylating DNA damage (data not shown). Interestingly, a recent study indicated a role of DRAM (damage-regulated autophagy modulator), a p53 downstream autophagy inducer, in assisting p53-initiated apoptosis (32). By contrast, deficiency of ATG5 or Beclin1 provided minimal protection against necrosis in DKO cells of which the mitochondrion-dependent apoptotic machinery is disabled (Fig. S4).

Lysosomes were described as “suicide bags” of the cells by the discoverer, De Duve, due the inherent danger of releasing hydrolytic enzymes through lysosomal membrane permeabilization (LMP) (33). Our observation that knockdown of cathepsin Q but not other lysosomal proteases such as cathepsin B and L blocked DNA damage-induced necrotic death in DKO cells suggests that LMP is unlikely an initiating factor of death. Indeed, we found that DKO cells did not lose Lysotracker staining until they became AV-positive, indicating that LMP is a consequence of dying (data not shown). In contrast, cathepsin Q appears to initiate the necrotic program before the appearance of discernible dying features (Fig. 3F). Although lysosomal cysteine cathepsins were believed to be involved primarily in intralysosomal protein degradation, genetic studies using knockout mouse models start to uncover their important roles in various biological processes (34, 35). Cathepsin B is important in TNF-induced apoptosis of hepatocytes, cathepsin K in bone remodeling, cathepsin S in MHC Class II antigen presentation, cathepsin C in the activation of granzyme A and B, and cathepsin L in positive selection for CD4⁺ T cells and terminal differentiation of keratinocytes (34, 35). Cathepsin Q is clearly not a placenta-restricted protease and is induced by DNA damage through p53. How cathepsin Q executes necrotic death and what is the human counterpart of mouse cathepsin Q require further experimentation.

Modern biomedical research profoundly impacts the development of targeted therapeutics. Indeed, the studies on apoptosis have led to the development of drugs that modulate apoptosis in treating human illness (3, 4). Recently, a small molecule inhibitor of RIP1 kinase, necrostatin-1, was reported to provide protection in a murine ischemic brain injury model (36, 37). Of note, necrostatin-1 failed to prevent DNA damage-induced necrotic death in DKO cells (Fig. S12D), suggesting that DNA damage triggered necrotic program is different from “necroptosis” that is activated upon death receptor signaling (36). Further elucidation of the necrotic signaling cascades will certainly enable targeted therapeutics in eliminating cancer cells that evade apoptosis (38). Noticeably, the BAX/BAK-independent necrotic program appears to be more pronounced in transformed cells, suggesting that this pathway could be a promising cancer-specific target. Our study demonstrates a p53-initiated genetic program in executing “programmed necrotic death,” which assigns DNA damage-induced necrosis as a form of “programmed cell death” and opens a new avenue for rationally designed therapeutics aimed at the necrotic pathways.

Materials and Methods

Plasmid Construction and Retrovirus Production.

Retrovirus-mediated knockdown constructs were generated using pSuper-Retro-Puro or pSuperior-Retro-Puro according to the manufacture's instruction (Oligoengine). A dominant negative mutant of p53 (amino acids 1–14 and amino acids 303–390) was dually tagged with FLAG and HA at the N-terminus and cloned into MSCV-IRES-GFP (pMIG) or MSCV-Puro (Clontech). The production of retroviruses was described in ref. 10. The promoter (−2150 to + 51) and Intron 1 of cathepsin Q were cloned into pGL2-basic and pGL2-promoter (Promega), respectively. The mutation of the p53 binding site identified at the intron I of cathepsin Q was generate by PCR based site-directed mutagenesis. The sequence was changed from “agaCatGcccaccaCttGtta” to “agaTatTcccaccaTttTtta.” All of the constructs were confirmed by DNA sequencing. The target sequences of shRNA are listed in SI Materials and Methods.

Antibodies, Reagents, and Immunoblot Analysis.

Antibodies and reagents are listed in SI Materials and Methods. Cell lysates were resolved by 10% or 4–12% NuPAGE (Invitrogen) gels, transferred onto PVDF membrane (Immobilon-P; Millipore). Antibody detection was accomplished using enhanced chemiluminescence method (Western Lightning, PerkinElmer) and LAS-3000 Imaging system (FUJIFILM).

Cell Culture and Viability Assay.

Mouse embryonic fibroblasts (MEFs) were generated from E13.5 embryos. The 3 clones used in Fig. 2C were derived from 3 different Bax and Bak DKO embryos. SV40 transformation was performed as described in ref. 10. For E1A/Ras transformation, primary DKO MEFs were infected with retrovirus expressing E1A-IRES-Ras as described in ref. 39, followed by selection under 2 μg/ml puromycin. Pools of puromycin-resistant cells were used by the indicated experiments. Saos-2 osteosarcoma cells were maintained in McCoy's 5A medium supplemented with 15% FBS. A549 cells and U031 cells were maintained in RPMI-1640 medium supplemented with 10% FBS. For UV-induced cell death, DKO cells were irradiated with UV-C (1500 J/m²) at day 0 and day 1 and cell death was assessed at day 4; A549 cells were irradiated with UV-C (250 J/m²) at day 0 and day 1 and cell death was assessed at day 4; and U031 cells were irradiated with UV-C (500 J/m²), and cell death was assessed after 3 days. Cell death was quantified by Annexin-V (BioVision) or propidium iodide (Sigma) staining according to manufacturer's protocols, followed by flow cytometric analyses using a FACSCalibur (BD Biosciences) and CellQuest Pro Software. P values for statistical analyses were obtained using Student's t test.

Clonogenic Assays.

SV40-transformed DKO cells were treated with etoposide (10 μg/ml) or etoposide plus cycloheximide (2 μg/ml) for 24 h. A total of 10⁴ or 2 × 10⁴ cells were washed with PBS and plated in 24-well tissue culture plates. Colonies were fixed with methanol and stained with 0.5% crystal violet in 25% methanol after 12 days.

Measurement of ROS.

Production of ROS was monitored by flow cytometric analyses using the redox-sensitive dye, 2′,7′-dichlorofluorescein diacetate (H₂DCF-DA, Invitrogen). Cells were incubated with 1 μM H₂DCF-DA at 37 °C for 30 min, followed by flow cytometric analyses using a FACSCalibur (BD Biosciences). Mean fluorescence detected by FL1 channel was assessed by FlowJo (Tree Star). Fold increase of ROS was determined by dividing the mean fluorescence after etoposide treatment for 2 days with the untreated controls.

Reverse-Transcription and Quantitative PCR.

Total RNA was extracted from cells using TRIZOL (Invitrogen) according to the manufacturer's instruction. Reverse transcription was performed with oligo(dT) plus random decamer primers (Ambion) using SuperScript II (Invitrogen). Quantitative PCR was performed with SYBR green master mix (Applied Biosystems) in duplicates using indicated gene specific primers (listed in SI Materials and Methods). Quantitative PCR was performed on an ABI Prism 7300 sequence detection system (Applied Biosystems). Data were analyzed as described previously by normalization against GAPDH (39). GAPDH was detected using a rodent-specific GAPDH Taqman probe (Applied Biosystems).

HMGB1 Release Assay.

Cells with or without etoposide (10 μg/ml) treatment grown in DMEM plus 1% FBS were pelleted and lysed in PBS containing 1% Triton X-100 plus complete protease inhibitors (Roche Applied Science). Medium was harvested and spun at 800 × g for 5 min of which the supernatant was then filtered through a 0.45-μm centrifugal filter device (Millipore) to remove cell debris. Proteins in the filtrates were precipitated with trichloroacetic acid (Sigma). Proteins from either cell lysates or cell-free-medium were analyzed by anti-HMGB1 Western blots.

Luciferase Assay.

Saos-2 cells, plated in a 6-well plate (3 × 10⁵ cells per well), were transfected with 1 μg of the indicated reporter constructs, 1 μg of lacZ-expressing CH110 (Amersham Pharmacia), or 50 ng of mouse p53 cloned into pCDNA3 (Invitrogen) using FUGENE6 (Roche) according to the manufacturer's protocol. Cells were harvested and lysed in cell lysis buffer (BD Biosciences) at 24 h after transfection. Cell lysates were assayed for luciferase activity using an Enhanced Luciferase Assay Kit (BD Biosciences). The data obtained by a luminometer were normalized for the transfection efficiency based on the beta-galactosidase activity.

Electron Microscopy.

Electron Microscopy was performed by the Electron Microscopy Facility at the Department of Cell Biology and Physiology, Washington University, St. Louis. Cells were thin sectioned on a Reichert-Jung Ultracut, poststained in uranyl acetate and lead citrate, viewed on a Zeiss 902 Electron Microscope, and recorded with Kodak E.M. film.

Indirect Immunofluorescence Microscopy.

Cells, fixed in 4% paraformaldehyde and permeabilized with 0.1% Triton X-100, were sequentially incubated with anti-HMGB1 antibody (BD Biosciences), Alexa Fluro488 conjugated goat anti-rabbit secondary antibody (Invitrogen), and Hoechst 33342 (Invitrogen). Images were acquired with a SPOT camera (Diagnostics Instruments) mounted on an Olympus EX51 microscope.

Tumor Allografts.

Tumors were induced in 5-week-old NOD/SCID/IL2Ry-null mice by s.c. injection of 10⁷ E1A/Ras transformed Bax and Bak DKO MEFs infected with control retrovirus (GFP) or retrovirus expressing a dominant negative mutant of p53. Before injection, cells were resuspended in DMEM with 33.3% Matrigel. After 13–17 days, tumors reached an average size of 80–150 mm³. Tumor size was measured in 2 dimensions by caliper and volume was calculated as (4π/3) × (width/2)² × (length/2). Etoposide was administered by i.p. injection at 15 mg/kg/day for 6 or 7 consecutive days (3 groups for 6 days and 4 groups for 7 days). Finally, the tumors were removed and subjected to histological (hematoxylin and eosin staining) or electron microscopic analyses. Electron Microscopy was performed by the Electron Microscopy Facility at the Department of Pathology and Immunology, Washington University in St. Louis.