Danger Signaling Protein HMGB1 Induces a Distinct Form of Cell Death Accompanied by Formation of Giant Mitochondria

High-mobility group box 1 (HMGB1) is an architectural transcription factor present in most eukaryotic cells (1), but it can also act as a cytokine when released from necrotic cells into the extracellular space (2). HMGB1 exerts pleiotropic biological effects by binding to several members of the Toll-like family such as receptor for advanced glycation end products (RAGE) or Toll-like receptor 2 (TLR2) and TLR4 (1), thereby promoting inflammatory processes as shown in a mouse model of sepsis (3). Interestingly, HMGB1 can also function in an antimicrobial manner. Phosphorylated recombinant HMGB1 and native HMGB1 extracted from tumor cells inhibit eukaryotic closed circular DNA replication in vitro (4). Similarly, recombinant and human adenoid-derived HMGB1 revealed antimicrobial activity (5, 6).

In addition to the established role of HMGB1 in inflammation, there is growing evidence for its contribution to cancer (7, 8). Breast cancer patients with a TLR4 single-nucleotide polymorphism had a worse clinical outcome due to the lack of immunoadjuvant effects of HMGB1 released from dying tumor cells (9). Interaction of glioblastoma-derived HMGB1 with TLR2 of dendritic cells was crucial for initiating an effective antitumor immune response in an intracranial glioma xenograft mouse model (10). Studies in an orthotopic syngeneic intracranial glioma rat model showed that the efficacy of an immunotherapy with adenoviruses expressing the cytokine Flt3L and thymidine kinase was dependent on circulating HMGB1 (11). In contrast, HMGB1 may also elicit direct antitumor activity. Endogenous HMGB1 influences the responsiveness of cancer cells to cisplatin by shielding the drug-induced DNA adducts from excision repair, thus promoting apoptosis specifically in tumor cells (12, 13).

In this study, we examined the potential of exogenous HMGB1 to induce cell death in cancer cells in vitro and in vivo. We describe a recombinant human HMGB1 (rhHMGB1)–dependent novel form of cell death, which is accompanied by the formation of vacuolated giant mitochondria and a rapid depletion of mitochondrial DNA (mtDNA). Thus, the HMGB1-induced signaling pathway might be a distinct mechanism of how tumor cell death evolves in vivo, thereby contributing in the formation and expansion of nonvital tissue areas classically viewed as tumor “necrosis.”

Human cancer cell lines were purchased from the American Type Culture Collection, expanded, and frozen in aliquots within 4 weeks of purchase. For the experiments described here, the cells were thawed and cultured for no more than eight passages. Nonimmortalized human astrocytic cells were purchased from ScienCell and immediately used for the experiments described. For detailed information, see Supplementary Materials and Methods.

Six-week-old male athymic CD1 nude mice (n = 8; Charles River) were injected s.c. with 7.5 × 10⁶ U251MG cells in 100 μL PBS in the right flank using a 30-gauge needle. After 3 weeks, treatment of the xenograft tumors was initiated with 10 μg rhHMGB1 in 500 μL PBS or PBS only (control group) by daily i.p. injections at the contralateral side for 6 weeks. For detailed information, see Supplementary Materials and Methods.

For large-format two-dimensional (2-D) electrophoresis and 2-D Western blots, isoelectric focusing was performed as described previously using dry polyacrylamide gel strips (IPG, 24 cm) with an immobilized pH gradient (pH 3–10) and the IPGphor1 Isolectric Focusing System (Amersham Pharmacia Biotech). For the second dimension, the ETTAN DALTsix electrophoresis system was used. Western blotting was performed as described previously. For detailed information, see Supplementary Materials and Methods.

U251MG cells (10,000) were seeded in drops of DMEM (10% FCS, 1% penicillin/streptomycin) on round coverslips and incubated for 24 hours at 37°C and 5% CO₂. The coverslips were put upside down in a 15-mL Falcon tube containing growth medium complemented with 10 μg/mL cytochalasin B. After 1-hour incubation at 37°C, the 15-mL tubes were centrifuged for 1.5 to 3 hours at 37°C with 4,000 rpm. Then, the coverslips were removed and put into 6-cm plates containing growth medium. After a recovering period of 8 hours, the cytoplasts were treated with rhHMGB1 for 24 hours. Successful enucleation was confirmed by immunofluorescence with 1 μg/mL 4′,6-diamidino-2-phenylindole.

For measurement of reactive oxygen species (ROS), 1.5 × 10⁶ U251MG cells were seeded in 6-cm plates, treated with 40 nmol/L rhHMGB1 for 24 hours, and incubated with the fluorescent H2DCF-DA (dichlorodihydrofluorescein diacetate; 10 μmol/L; Invitrogen) for 1 hour at 37°C. ROS was measured immediately at FL-1. For detailed information, see Supplementary Materials and Methods.

For iodination of rhHMGB1, 100 μg protein was incubated at room temperature for 15 minutes in 50 mmol/L phosphate (pH 7.0) with 74 MBq carrier-free Na¹²⁵I (Hartmann Analytic) in the presence of one Iodobead (Pierce). U251MG cells were seeded on 10-cm culture dishes. After treatment of cells with 80 nmol/L of ¹²⁵I-coupled rhHMGB1 for 0, 4, 16, and 24 hours, the cells were detached carefully by incubation with cell dissociation buffer (30 minutes). The radioactivity of each subcellular fraction was measured in a liquid scintillation counter (Packard TriCarb 2900). For detailed information, see Supplementary Materials and Methods.

Enrichment of mitochondrial fractions was performed using the ApoAlert Cell Fractionation kit (Clontech) as described previously (14). Briefly, after several centrifugation steps, U251MG cells were homogenized with a Dounce tissue grinder, and the homogenate was centrifuged at 700 × g for 10 minutes at 4°C and the resulting supernatant at 10,000 × g for 25 minutes at 4°C.

U251MG cells were plated onto 8-well chamber slides (Nunc) or grown confluently in 6-cm culture dishes containing round glass cover plates, permeabilized as described (14), and incubated with monoclonal rabbit fluoro-conjugated anti–COX IV antibody (1:10, 3E11, Alexa Fluor 488 conjugate; Cell Signaling) or monoclonal mouse anti–poly-his-tag antibody (1:1,000; R&D Systems) for 30 minutes at 4°C. For detailed information, see Supplementary Materials and Methods.

All data are means ± SD. The significance of differences between groups was determined by Student's t test. P values of <0.05 were considered statistically significant.

To study the cellular effects of exogenous rhHMGB1, we treated human glioblastoma cells with increasing concentrations of rhHMGB1 for several days (Fig. 1A). Treatment with rhHMGB1 resulted in a strong inhibition of tumor cell proliferation and induction of cell death even at nanomolar concentrations. Cytotoxicity assays revealed that glioblastoma cells, but not human nonimmortalized astrocytes, were susceptible to rhHMGB1-induced cell death (Fig. 1B). Similarly, rhHMGB1 induced a strong inhibition of clonogenic cell survival of glioma cells, but not of astrocytes (Supplementary Fig. S1). Except U251MG glioblastoma cells, the cytotoxic effect of rhHMGB1 was also observed in other long-term (T98G and LN18) and primary (NCH89 and NCH82) glioblastoma cell lines. In these cell lines, the following rhHMGB1 concentrations were required for half-maximal cell killing (72 hours): T98G, 15 nmol/L; LN18, 15 nmol/L; NCH89, 30 nmol/L; and NCH82, 30 nmol/L. Tumor cell lines of other tissue origins were also susceptible to rhHMGB1-induced cytotoxicity, although to a somewhat lower extent (IC₅₀ concentrations of rhHMGB1: HeLa cells, 35 nmol/L; MCF7 cells, 60 nmol/L; HCT116 cells, 65 nmol/L; SW480 and RKO cells, >160 nmol/L; Supplementary Fig. S2). Further, combined treatment of glioma cells with rhHMGB1 and the death ligand TRAIL [tumor necrosis factor (TNF)–related apoptosis-inducing ligand] or the anticancer drug temozolomide, which is the first-line chemotherapy for glioblastomas, resulted in a strong synergistic induction of cell death (Fig. 1C). In contrast, astrocytes were not susceptible to the synergistic interaction of rhHMGB1 and temozolomide (Supplementary Fig. S3). Rat recombinant HMGB1 protein, which exhibits a high homology to human HMGB1 (>98%; ref. 15), showed similar cytotoxic effects as rhHMGB1 (Supplementary Fig. S4). Digestion of rhHMGB1 with 100 μg/mL proteinase K before treatment completely abrogated the cell death (data not shown). To test the anticancer activity of rhHMGB1 in vivo, we systemically treated CD1 athymic nude mice bearing subcutaneous glioma xenograft tumors with rhHMGB1. After 6 weeks, the rhHMGB1-treated animals showed significantly smaller tumors than the control group (Fig. 1D).

Next, we examined the intracellular signaling events leading to rhHMGB1-induced cell death. rhHMGB1 treatment resulted in a strong generation of ROS, which was partially prevented by the ROS scavenger N-acetylcysteine (NAC; Fig. 2A, left). Cotreatment of the cells with NAC substantially inhibited rhHMGB1-induced cell death (Fig. 2A, right). Similarly, both the rhHMGB1-induced increase in ROS and the rhHMGB1-dependent cell death were significantly inhibited by PEG-superoxide dismutase (SOD) and PEG-catalase as well as by overexpression of mitochondrially targeted manganese-SOD (MnSOD) and (to a lesser extent) catalase (Supplementary Fig. S5). The rhHMGB1-induced ROS generation occurred as early as 6 hours after treatment began. In contrast, loss of plasma membrane integrity as assessed by lactate dehydrogenase release was a late event observed after ∼36 hours of rhHMGB1 treatment (Supplementary Fig. S6). Moreover, rhHMGB1 caused a continuous and dose-dependent phosphorylation of c-Jun NH₂-terminal kinase 1/2 (JNK; Fig. 2B, left). The activation of JNK contributes to rhHMGB1-dependent cytotoxicity because coincubation with the JNK inhibitor SP600125 not only reduced the amount of phosphorylated JNK (Supplementary Fig. S7) but also partially inhibited cell death induced by rhHMGB1 (Fig. 2B, right). The NAC-dependent inhibition of ROS prevented the activation of JNK after rhHMGB1 treatment (Fig. 2C, left), indicating that rhHMGB1-induced ROS generation occurs upstream of JNK activation. To investigate the functional importance of intact mitochondria in rhHMGB1-induced cytotoxicity, we generated cells devoid of intact mtDNA, so-called pseudo rho zero cells (ρ⁰ cells; ref. 16). Importantly, ρ⁰ cells were almost completely resistant to rhHMGB1-induced cell death (Fig. 2C, right), whereas the apoptotic cell death signaling cascade was still intact in these cells (Supplementary Fig. S8). These results suggested that mitochondria represent a main target for cytotoxic actions of rhHMGB1. This hypothesis was further confirmed by the observation that rhHMGB1-induced cell death in U251MG glioma cells was substantially inhibited by adding uridine to the culture medium (Fig. 2C, right). Whereas ρ⁰ cells devoid of intact mtDNA are pyrimidine auxotrophs (16, 17), normal cells essentially depend on mitochondrial-derived uridine. Therefore, the uridine-mediated protection from rhHMGB1-induced cell death indicates that the disturbance of mitochondrial metabolic pathways is essential for the cytotoxic activity of rhHMGB1.

Given the mitochondria-dependent effects of rhHMGB1, we investigated whether rhHMGB1 induces morphologic changes in mitochondria. Because treatment of glioma cells with rhHMGB1 resulted in the formation of large vacuolated organelles as observed by phase-contrast microscopy (data not shown), we performed electron microscopy of cells treated with rhHMGB1. The ultrastructure of the vacuoles revealed a double-membrane and rudimentary cristae (Fig. 3), confirming the mitochondrial nature of the vacuolated organelles. The size of these giant mitochondria increased in a time- and concentration-dependent manner after incubation with rhHMGB1 (Supplementary Fig. S9). To further characterize the rhHMGB1-dependent mitochondrial changes, we performed immunofluorescence studies using the mitochondrial marker enzyme COX IV. These experiments unambiguously identified the vacuolated organelles as extremely enlarged spherical mitochondria in 100% of cells treated with rhHMGB1 (Fig. 4A). In contrast, tumor cells devoid of intact mtDNA (ρ⁰ cells) that were highly resistant to the cytotoxic actions of rhHMGB1 (Fig. 2C, right) did not form giant mitochondria (Supplementary Fig. S10). Given the chromatin-binding function of rhHMGB1, we next investigated whether the nuclear localization of rhHMGB1 is required for the formation of giant mitochondria (e.g., by nucleomitochondrial translocation). Therefore, we generated U251MG cells devoid of nuclei, the so-called cytoplasts. Treatment of cytoplasts with rhHMGB1 resulted in the formation of giant mitochondria to a similar extent and in a similar time frame compared with nucleus-containing control cells (Fig. 4B), showing that nuclear events are not required for the generation of giant mitochondria. To address the question of how exogenous HMGB1 can induce such marked mitochondrial effects, we investigated the subcellular localization of rhHMGB1. When U251MG glioma cells were incubated with ¹²⁵I-labeled rhHMGB1, a time-dependent increase of radioactivity was observed in the mitochondrial fraction (Fig. 4C), suggesting that after penetration of the cell membrane, a substantial amount of rhHMGB1 translocated to the mitochondria. Translocation of rhHMGB1 to the mitochondria was also confirmed by confocal immunofluorescence microscopy after costaining with a mitochondrial marker (Fig. 4D).

To further characterize the rhHMGB1-mediated cytotoxicity, we examined whether rhHMGB1-induced cell death shares the phenotypic characteristics of apoptosis, autophagy, senescence, or classic necrosis. First, after rhHMGB1 treatment, the typical morphologic features of apoptosis, such as membrane blebbing, formation of apoptotic bodies, or chromatin condensation, were absent. In addition, activation of caspase-3 or release of cytochrome c was not observed after rhHMGB1 treatment (Fig. 5A), and rhHMGB1-induced cell death was not blocked by the caspase inhibitor zVAD-fmk (Supplementary Fig. S11A). Electron microscopy showed neither autophagosomes characteristic for autophagic cell death nor disintegration and rupture of the outer cell membrane indicative of necrotic cell death (Fig. 3). Moreover, autophagy was excluded by a lack of formation of LC3-GFP dots (Supplementary Fig. S11B) and by a lack of LC3I/II conversion (Supplementary Fig. S11C) after incubation with rhHMGB1. The rhHMGB1-induced cell death was not blocked by the autophagy inhibitor 3-methyladenine or the necroptosis inhibitor necrostatin (Supplementary Fig. 11A). Next, we examined whether the senescence-associated DNA double-strand break repair enzyme H2A.X (18) was activated by rhHMGB1. However, H2A.X phosphorylation was not observed on rhHMGB1 treatment (Fig. 5B). Senescence was also excluded by analyzing β-galactosidase activity after treatment with rhHMGB1 (Supplementary Fig. S11D). Similarly, the necrosis-associated BCL2/adenovirus E1B 19-kDa interacting protein 3 (BNIP3) was not upregulated after treatment of cells with rhHMGB1 (Fig. 5B). Thus, the rhHMGB1-mediated cell death lacks the typical features of apoptosis, autophagy, or classic necrosis.

To show that the rhHMGB1-induced formation of giant mitochondria does not only occur in cultured tumor cells but also in vivo, we examined the tumor tissue of xenografted nude mice treated systemically with rhHMGB1 (Fig. 1D). After immunohistochemical staining of COX IV, numerous giant mitochondria were observed in the xenografted tumor cells of rhHMGB1-treated animals (42% of cells; Fig. 5C, arrows). However, giant mitochondria were to a lesser extent also visible in perinecrotic areas in untreated tumors (16% of perinecrotic cells; Fig. 5C, asterisk). Therefore, we hypothesized that giant mitochondria–associated cell death might represent a mechanism commonly occurring in the setting of tumor necrosis. To test this hypothesis, we examined a panel of 25 human malignant tumors of diverse tissue origin containing necrotic areas by COX IV immunohistochemistry to detect giant mitochondria. In fact, in the majority of cases (80%), tumor cells with giant mitochondria surrounding nonvital tissue areas were observed, especially in glioblastomas and squamous cell carcinomas (Fig. 5D).

To further characterize the nature of the rhHMGB1-mediated mitochondrial damage, we examined the mtDNA content of cells after rhHMGB1 treatment. Dot blot analysis showed a rhHMGB1-dependent substantial and rapid depletion of mtDNA (Fig. 6A). The rapid mtDNA depletion after rhHMGB1 treatment was accompanied by a severe depletion of the isoelectric-neutral proteins and an increase of predominantly alkaline proteins (Fig. 6B). Moreover, 2-D Western blot analysis of the mitochondrial proteome revealed a substantial increase in covalently bound malondialdehyde in the alkaline range after treatment with rhHMGB1 (Fig. 6C). Similarly, rhHMGB1 induced the formation of hydroxynonenal (19) adducts (Supplementary Fig. S12A). HMGB1 also altered the ratio of glutathione (GSH) to glutathione disulfide (GSSG) shown by a significant increase in GSSG and a decrease in GSH after 48 hours of rhHMGB1 treatment (Supplementary Fig. S12B). Taken together, these results show that rhHMGB1 induces lipid peroxidation, severe oxidative stress, and inhibition of mtDNA replication.

In this study, we describe that rhHMGB1 protein induces a novel form of cell death that is characterized by the formation of giant mitochondria and a severe damage to mtDNA and proteins. It has been described that both endogenous HMGB1 that is released from dying tumor cells and recombinant HMGB1 act as immune adjuvants, enabling dendritic cells and cytotoxic T cells to attack specifically tumor cells (7, 10). Moreover, HMGB1 signaling through TLR2 and TLR4 is crucial for an effective immune response mediating tumor regression in humans (9) and in rodents (10, 11). However, the giant mitochondria–associated rhHMGB1-induced cell death we describe is independent of TLR2, TLR4, or RAGE signaling, because the TLR4, TLR2, and RAGE agonists lipopolysaccharide (LPS), PAM3CSK4, and AGE did not induce formation of giant mitochondria or cell death in glioma cells (Supplementary Fig. S13). Moreover, overexpression of the receptors TLR2, TLR4, RAGE, or their intracellular transducers MyD88 and TRIF, or the small interfering RNA–mediated downregulation of TLR4 or RAGE did not alter the susceptibility of glioma cells to the rhHMGB1-induced cell death (Supplementary Fig. S14; data not shown).

Recently, it was shown that endogenous HMGB1 can translocate from the nucleus to the mitochondria, resulting in reorganization of the mitochondria in endothelial cells (20). In our study, we show that rhHMGB1 enters the mitochondria, which is followed by the formation of giant mitochondria. However, the mechanisms that enable HMGB1 to pass the cell membrane and the mitochondrial membranes still have to be clarified. Interestingly, HMGB1 has been used to potentiate the transfection of cells with compacted DNA, yielding greater transfection rates compared with conventional techniques (21). This might be due to the polybasic penetratin-like domain of HMGB1, common to proteins such as Antennapedia, lactoferrin, and the HIV protein Tat, which can transport DNA or RNA across cell and organelle membranes (1). In this context, it is interesting that the uptake of rhHMGB1 in glioma cells was not inhibited by the endocytosis inhibitor brefeldin A and that several endocytosis inhibitors did not alter the extent of HMGB1-induced cell death (Supplementary Fig. S15), suggesting that endocytosis does not play an essential role for the internalization of rhHMGB1.

Our results suggest that rhHMGB1 is the first known mammalian protein that specifically induces the generation of giant mitochondria. The giant mitochondria observed in our study are morphologically similar to megamitochondria, which can be induced by several chemical compounds, including ethanol, chloramphenicol, hydrazine, and ethidium bromide (22). However, in the glioblastoma cell model we used, the giant mitochondria–associated cell death was not observed after treatment with compounds that are reportedly known to induce megamitochondria and depletion of mtDNA in other cell systems (22), such as high doses of ethidium bromide or ethanol (data not shown). Moreover, we show that even very low concentrations of rhHMGB1 are capable of inducing giant mitochondria. Once translocated to mitochondria, rhHMGB1 leads to a rapid depletion of mtDNA. A similarly rapid but reversible and less effective depletion of mtDNA is described for ethanol in the liver cells of an alcoholic binge mouse model (23). Importantly, the rhHMGB1-dependent effects on mitochondria are absent in pseudo ρ0 cells harboring mutated and intercalated mtDNA instead of intact, wild-type mtDNA. In this context, it is of interest that the human HMGB1 protein can inhibit closed circular plasmid replication in vitro (24) and that the yeast homologue of HMGB1, the mitochondrial HM protein, is required for the maintenance and expression of the yeast mitochondrial genome. In contrast, a deficiency of HM under certain circumstances leads to loss of mtDNA, which can be prevented by overexpression of NHP6A, the yeast nuclear variant of HM (25, 26). Therefore, HMGB1 or its yeast homologues are important regulatory proteins that can maintain or inhibit the expression of the mitochondrial genome. We postulate that the interaction of rhHMGB1 with mtDNA mediates the deleterious effects of rhHMGB1 on the mitochondrial genome and proteome as well as the giant mitochondria–associated cell death.

What are the downstream effectors of rhHMGB1-induced cell death? Our data indicate an important role of ROS, which rapidly rise after treatment of cells with rhHMGB1, whereas the loss of plasma membrane integrity is a rather late event in rhHMGB1-induced cytotoxicity. Both the ROS increase and cell death on rhHMGB1 treatment were inhibited by NAC as well as by specific ROS scavengers such as PEG-catalase and PEG-SOD. Similarly, a ROS increase and higher intramitochondrial concentrations of its toxic lipid peroxidation product malondialdehyde were measured during the formation of megamitochondria in hepatocytes from rats fed with the megamitochondria-inducing compound hydrazine (22, 27), which resulted in the induction of apoptosis in this cell model (28). Accordingly, our experiments revealed high concentrations of intramitochondrial toxic malondialdehyde protein adducts after treatment with rhHMGB1, resulting in a severe damage of the mitochondrial proteome. The mitochondria seem to be the main mediators of the rhHMGB1-induced cell death, because nuclear signaling events are not required for giant mitochondria formation as shown in glioma cells devoid of nuclei. With respect to both morphology and signal transduction events, the rhHMGB1-induced cell death is different from apoptosis, autophagocytotic cell death, and senescence. It is also different from classic necrosis, which is characterized by cytoplasmic swelling, rupture of the plasma membrane, and mild swelling (but not fusion) of cytoplasmic organelles (29). Moreover, the giant mitochondria–associated cell death observed in our study clearly differs from the recently described “necroptosis,” a form of programmed necrosis that depends on the kinase RIP1 (30). Giant mitochondria have thus far not been reported to be associated with necroptosis, and necrostatin-1 had no effect on rhHMGB1-induced cell death (Supplementary Fig. S11A), which excludes an essential role for the RIP1 kinase activity. Because cells with giant mitochondria are present at perinecrotic areas in human tumor tissue and because the HMGB1 protein is released during necrosis, the HMGB1-induced, giant mitochondria–associated cell death might represent an important novel mechanism of specialized necrosis.

The molecular basis of giant mitochondria formation is still not understood. Similarly, further studies are needed to elucidate why exogenous HMGB1 protein is much more effective in inducing giant mitochondria than the endogenously expressed HMGB1 protein. Although giant mitochondria formation is observed on transfection and overexpression of hHMGB1 (Supplementary Fig. S6), the extent of giant mitochondria generation is larger after exogenous treatment with HMGB1. Thus, we hypothesize that exogenous HMGB1 undergoes a modification when it passes the cell membrane, for example, as described for serine-phosphorylated HMGB1 that is directed toward secretion (31). Tyrosine residue modification seems not to play a crucial role, as the covalent binding of ¹²⁵I to tyrosine phosphorylation sites did not interfere with rhHMGB1-induced giant mitochondria formation and cell death (data not shown).

Finally, the application of rhHMGB1 might be a promising new approach in the therapy of patients with malignant tumors. Because glioblastoma cells, but not normal astrocytic cell cultures, were highly susceptible to rhHMGB1-induced cell death, the rhHMGB1-mediated mitochondrial effects might be tumor specific. However, the potential toxicity of rhHMGB1 will need to be investigated thoroughly, because rHMGB1 can mediate endotoxemia in mice. Wang and colleagues showed that in mice between 20 and 50 μg, HMGB1 is released into the murine circulation after stimulation with LPS, whereas in our xenograft mouse model 10 μg of rhHMGB1 were administered i.p. (3). Future studies will have to examine in detail the possibilities and limitations of a rhHMGB1-based experimental therapy.

No potential conflicts of interest were disclosed.

We thank Angelika Bierhaus (University Hospital of Heidelberg, Heidelberg, Germany) for generously providing rat recombinant HMGB1 and AGEs; Heikki Rauvala (University of Helsinki, Helsinki, Finland) for plasmids encoding for RAGE, ΔRAGE, and p30 (HMGB1); Peter Krammer and Marcin Kaminski (German Cancer Research Center, Heidelberg, Germany) for plasmids encoding MnSOD and catalase and the corresponding antibodies; David Capper and David Reuss (Department of Neuropathology, University Hospital Heidelberg) for providing histologic sections of glioblastomas; Arne Warth (Institute of Pathology, Heidelberg) for providing histologic sections of lung cancer; and Thomas Hoffmann and Dirk Sombroek (German Cancer Research Center, Heidelberg, Germany) for generously providing the senescence assay kit.

Grant Support: Deutsche Krebshilfe (German Cancer Aid, Max Eder Program; W. Roth); and Heinrich FC Behr Foundation (G. Gdynia).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.