TNF-Induced Activation of the Nox1 NADPH Oxidase and Its Role in the Induction of Necrotic Cell Death

Molecular Cell

Article TNF-Induced Activation of the Nox1 NADPH Oxidase and Its Role in the Induction of Necrotic Cell Death You-Sun Kim,1,2 Michael J. Morgan,1,2 Swati Choksi,1 and Zheng-gang Liu1,* 1 Cell and Cancer Biology Branch, Center for Cancer Research, National Cancer Institute, National Institutes of Health, 37 Convent Drive, Bethesda, MD 20892, USA 2 These authors contributed equally to this work. *Correspondence: [email protected] DOI 10.1016/j.molcel.2007.04.021

SUMMARY

Tumor necrosis factor (TNF) is an important cytokine in immunity and inflammation and induces many cellular responses, including apoptosis and necrosis. TNF signaling enables the generation of superoxide in phagocytic and vascular cells through the activation of the NADPH oxidase Nox2/gp91. Here we show that TNF also activates the Nox1 NADPH oxidase in mouse fibroblasts when cells undergo necrosis. TNF treatment induces the formation of a signaling complex containing TRADD, RIP1, Nox1, and the small GTPase Rac1. TNFtreated RIP1-deficient fibroblasts fail to form such a complex, indicating that RIP1 is essential for Nox1 recruitment. Moreover, the prevention of TNF-induced superoxide generation with dominant-negative mutants of TRADD or Rac1, as well as knockdown of Nox1 using siRNA, inhibits necrosis. Thus our study suggests that activation of Nox1 through forming a complex with TNF signaling components plays a key role in TNF-induced necrotic cell death. INTRODUCTION Tumor necrosis factor (TNF) is a pleiotropic inflammatory cytokine and plays a critical role in diverse cellular events, including cell proliferation, differentiation, apoptosis, and necrosis (Chen and Goeddel, 2002; Wajant et al., 2003). Through binding to its two receptors, TNF-R1 (p55) and TNF-R2 (p75), TNF is a major mediator of both inflammation and immunity and has thus been implicated in a variety of pathological inflammatory conditions and autoimmune diseases. Since the discovery of its tumoricidal activity, the TNF pathway has become one of the most studied signaling pathways, resulting in the characterization of a vast superfamily of receptors, the TNF receptor superfamily, and their ligands. In response to TNF treatment, the transcription factor NF-kB and MAP kinases, such as c-Jun N-terminal kinase (JNK), are activated in most types of

cells and, in some cases, apoptosis or necrosis can also be induced (Fiers et al., 1999; Karin and Lin, 2002). Much is known about the molecular mechanisms of TNF signaling. Engagement of TNF-R1 by the TNF homotrimer initiates the binding of the adaptor protein, TNF-R1-associated death domain protein (TRADD), which then recruits other effector proteins, such as receptor interacting protein 1 (RIP1) and TNFR-associated factor 2 (TRAF2) to form a TNF-R1 signaling complex leading to the activation of several pathways, including NF-kB and MAP kinases (Chen and Goeddel, 2002; Wajant et al., 2003). Both TRAF2 and RIP1 play essential roles in the activation of NF-kB and MAP kinase pathways through recruitment and activation of IKK (IkB kinase) and MAP3Ks (Chen and Goeddel, 2002; Wajant et al., 2003). Under certain conditions, it is thought that the TRADD, RIP1, and TRAF2 proteins dissociate from the receptor and form a secondary protein complex containing other proteins (Micheau and Tschopp, 2003). Apoptosis is primarily initiated through the recruitment of Fas-associated death domain protein (FADD) to this secondary complex. FADD then mediates the activation of the initiator cysteine proteases caspase-8 and caspase-10, which drives apoptosis. Although caspase cleavage is the primary means of initiating apoptotic cell death by TNF, cell death still occurs (or is even enhanced) in some types of cells in the absence of caspase activity (Fiers et al., 1999). While the mechanism of TNF-induced apoptotic cell death is well elucidated, the signaling events that lead to TNF-initiated caspase-independent death are largely unknown. Caspase-independent necrotic cell death has been proposed to involve reactive oxygen species (ROS), which some studies suggest are derived from the mitochondria (Fiers et al., 1999). ROS can inactivate MPK phosphatases, leading to sustained JNK activation and resulting in eventual cell death (Kamata et al., 2005). RIP1 is necessary for the generation of ROS by TNF and is required for the initiation of necrotic cell death (Lin et al., 2004). During apoptotic cell death, RIP1 is cleaved by caspase-8 (Lin et al., 1999), limiting its ability to activate the pathway(s) of ROS generation. NADPH oxidases are enzymes specifically dedicated to the production of ROS (Lambeth, 2004). In macrophages and neutrophils, TNF is known to stimulate the activity of

Molecular Cell 26, 675–687, June 8, 2007 ª2007 Elsevier Inc. 675

Molecular Cell TNF-Induced Activation of Nox1 in Necrosis

one such oxidase, Nox2/gp91phox, resulting in the generation of superoxide (O2 ) that is important in the ability of these cells to kill invasive microorganisms. Nox2, which is a membrane glycoprotein that exists as a heterodimer with a 22 kDa subunit (p22phox), is activated by the p47phox and p67phox proteins. Mutations in any of the four oxidase subunits can result in chronic granulomatous disease (CGD), characterized by a susceptibility to severe and recurrent bacterial and fungal infections derived from the inability of phagocytic cells to destroy pathogens (Heyworth et al., 2003). When phosphorylated, the p47phox subunit binds to membrane phospholipids, interacts with p22phox, and recruits the p67phox subunit to the complex. The p67phox activator binds and stabilizes an interaction of the complex with the small GTPase Rac, and the fully formed complex is able to generate O2 in the presence of NADPH. TNF is a potent activator of Nox2 through a little-understood mechanism that may involve phosphorylation of p47phox (Frey et al., 2002; Dewas et al., 2003; Dang et al., 2006). NADPH oxidases have been characterized in nonphagocytic cell types, along with other regulatory adaptor proteins (Lambeth, 2004). The regulatory p41NOXO1 and p51NOXA1 subunits may function in some of these oxidase complexes similarly to p47phox and p67phox, respectively. Unlike p47phox, NOXO1 lacks an autoinhibitory region, binds to different lipids, and does not require phosphorylation for membrane translocation (Lambeth, 2004). Nox1 and Nox4 appear to be more ubiquitous than other NADPH oxidases in nonphagocytic cells. However, Nox4 is constitutively active when complexed with p22phox, does not appear to require additional regulatory subunits (Martyn et al., 2006), and may be regulated solely by its expression level. Thus, Nox1 may be the most likely source of immediate NADPH oxidase activity in response to TNF in nonphagocytic cell types where Nox2 is not expressed. In this study, we show that the Nox1 NADPH oxidase is activated during TNF-induced necrotic cell death by forming a complex with TRADD, RIP1, and Rac1. Interactions of NOXO1 with TRADD and RIP1 are critical for the activation of Nox1 by TNF, and RIP1 is essential for recruiting Nox1 to the complex in MEF cells when necrosis is triggered. Importantly, Nox1-mediated production of O2 is critical for TNF-induced necrotic cell death because such death is inhibited by ROS scavengers, downregulation of the Nox1 protein by siRNA, or expression of dominant-negative versions of TRADD or Rac1. Our data also suggest that Nox1 may mediate the induction of necrosis through regulating sustained JNK activation. RESULTS TNF-Induced O2 Generation and Necrotic Cell Death Are Inhibited by BHA Many experiments have implicated ROS in TNF-induced necrotic cell death, some of which have suggested mitochondrial involvement (Fiers et al., 1999). To address whether the ROS generated by NADPH oxidase also plays

676 Molecular Cell 26, 675–687, June 8, 2007 ª2007 Elsevier Inc.

a role, we first examined whether O2 was generated under such conditions. Lucigenin, which reacts specifically with O2 and produces chemiluminescence (Teixeira et al., 1999), was used to measure O2 production in murine fibrosarcoma L929 cells, which undergo necrotic cell death in response to TNF treatment. The amount of chemiluminescence increased dramatically 30–40 min after addition of TNF to the cells (Figure 1A), indicating that the production of O2 followed exponential kinetics at this time period, while steady levels of luminescence were achieved 75 min after TNF treatment (see Figure 1E). Cells treated with another inflammatory cytokine, IL-1, failed to produce a detectable amount of O2 , indicating that the O2 generation in this case is specific to TNF (see Figure S1A in the Supplemental Data available with this article online). Notably, treatment with a pancaspase inhibitor, z-VAD-fmk, which has been shown to enhance TNF-induced cell death in L929 cells, significantly increased TNF-stimulated O2 generation (Figure S1B), suggesting that the O2 production was correlated with cell death. To test whether O2 is produced in other cell types under necrotic conditions, we used MEF cells deficient in the RelA subunit of NF-kB (p65 / ). These cells preferentially undergo apoptotic cell death in response to treatment with TNF alone but undergo necrosis when caspase inhibitor is present (Sakon et al., 2003). As shown in Figure 1B, TNF treatment induced the formation of O2 only under necrotic cell death conditions. This suggests that the generation of O2 is correlated with caspaseindependent necrotic cell death in response to TNF. As previously reported, the antioxidant BHA significantly blocked TNF-induced cell death in both L929 and p65 / cells (Figures 1C and 1D and Figures S1C and S1D). While BHA did have some protection in TNF-treated p65 / MEFs, it was far less efficient than in L929 cells (compare Figure 1C and Figure S1C with Figure 1D and Figure S1D). However, BHA prevented TNF-induced cell death more efficiently in the presence of zVAD (Figure 1D and Figure S1D), indicating that BHA more efficiently protects these cells from necrosis than from apoptosis. Pretreatment with the ROS scavengers NAC and TEMPOL gave similar results (Figure S2A). We then tested whether BHA had an effect on O2 generation. As shown in Figure 1E, BHA pretreatment abolished TNF-induced O2 formation in L929 cells. Similarly, the addition of superoxide dismutase (SOD), NAC, or TEMPOL reduced O2 to background levels (Figures S2B–S2D). Pretreatment for 30 min with cycloheximide did not block O2 generation in L929 or p65 / cells (data not shown), indicating that protein synthesis was not required. Taken together, these results indicated that antioxidants block both TNFinduced necrotic cell death and O2 accumulation. Nox1 NADPH Oxidase Is Responsible for TNF-Induced O2 Generation The production of O2 in response to TNF suggests the activation of an NADPH oxidase. Of the known NADPH oxidases, only Nox2/gp91 has been previously shown to

Molecular Cell TNF-Induced Activation of Nox1 in Necrosis

Figure 1. BHA Prevents TNF-Induced O2 Generation and Necrotic Cell Death (A) NADPH oxidase assays in TNF-treated and untreated L929 cells. Time of TNF treatment is indicated. (B) NADPH oxidase assays in p65 / MEFs. (C and D) Cell death assays (MTT). Cells were untreated, or pretreated for 30 min with BHA and/or zVAD and then treated with TNF for the indicated times (error bars, ±SEM). (E) NADPH oxidase assays in TNF-treated and untreated L929 cells in the presence or absence of BHA.

be activated in response to TNF (Frey et al., 2002). While Nox2 is found primarily in phagocytic cells, it is also expressed in several other cell types, including endothelial cells and vascular fibroblasts (Gorlach et al., 2000; Chamseddine and Miller, 2003). We therefore sought to determine whether Nox2 is expressed in L929 and p65 / MEF cells. As shown in Figure 2A, Nox2 was not detected by western blot in either of these two types of cells, though it was readily detected in the J774 and RAW264.7 macrophage cell lines. We then tested whether Nox1 is expressed in L929 and p65 / MEF cells because it has been reported to be expressed in several types of nonphagocytic cells (Suh et al., 1999; Takeya et al., 2003; Geiszt et al., 2003). We probed the same lysates with a Nox1

antibody and detected a 46 kDa band in several cell types, including L929 and p65 / MEFs (Figure 2A). This band ran below the expected Nox1 size, which is predicted to be 65 kDa. However, Nox2, before it is glycosylated, also runs lower than its actual molecular weight (Yu et al., 1999; Nguyen and Tidball, 2003), as is not uncommon for an integrated membrane protein. Nevertheless, we sought to more clearly establish the identity of this band as Nox1. Treatment of HL-60 cells with PMA or DMSO induced expression of this band (Figure 2A) in a manner similar to several other oxidase components, including p47phox and p67phox (Figure S3). In order to further verify the identity of the band, we performed in vitro translation of an untagged Nox1 cDNA with S35

Molecular Cell 26, 675–687, June 8, 2007 ª2007 Elsevier Inc. 677

Molecular Cell TNF-Induced Activation of Nox1 in Necrosis

Figure 2. The NADPH Oxidase Nox1 Is Expressed in L929 Cells and Is Responsible for TNF-Induced O2 Generation (A) Cell lysates from various cell lines blotted with anti-gp91 (Nox2), anti-Nox1, and anti-actin antibodies. (B) S-35 autoradiography of in vitro translated untagged (top) or Xpress-tagged (bottom) plasmid constructs of Nox1. In vitro translated p51NOXA1 and p47phox are shown for size comparison in addition to the indicated molecular markers. (C) Myc-Nox1 immunoprecipitated with an anti-mouse Myc antibody from transfected 293 cells and blotted with anti-Nox1 antibody. (D) Western blot of L929 cells transfected with Nox1-specific siRNA (#4 and #2) or control siRNA (Lamin A/C) 72 hr posttransfection. (E) NADPH oxidase assays in cells from (D). For clarity purposes luminescence data are now shown as the RLU ratio between the TNF-treated and untreated samples and indicate the relative fold activation of NADPH oxidase by TNF. (F) NADPH oxidase assays from cells cotransfected with Nox1 #4 siRNA oligo along with either a human Nox1 expression plasmid, an inactive point mutant (T341K), or empty vector.

methionine and detected a band at about 49 kDa (Figure 2B, top), which includes the 3 kDa signal peptide cleaved from the endogenous protein. We also expressed the full-length Nox1 and several potential Nox1 variants with an N-terminal Xpress tag (Xp), (3.5 kDa) for comparison (Figure 2B, bottom). In vitro translated Xp-Nox1 proteins gave a single band at about 51 kDa, while an Xpress-tagged naturally occurring C-terminal splice vari-

678 Molecular Cell 26, 675–687, June 8, 2007 ª2007 Elsevier Inc.

ant of Nox1 (Nox1-lv) had a smaller size than this (49.5 kDa). Cloning from the next in-frame start codon (Nox1trunc) gave an even smaller in vitro translated product of just over 47 kDa. These results indicated that the higher mobility of the endogenous band detected by the Nox1 antibody was not likely due to alternative translation or splice variation, as adjustment for the signal peptide and tag removal makes the full-length Nox1 in cells likely to

Molecular Cell TNF-Induced Activation of Nox1 in Necrosis

be observed at 46 kDa, which is the size we observed. Transfection of a C-terminal Myc-tagged Nox1 construct in several cell types resulted in little expression as detected by anti-myc antibody for unknown reasons (data not shown). However, we were able to immunoprecipitate the ectopically expressed Myc-tagged Nox1 protein from a large number of cells with an anti-Myc antibody and detected the precipitated protein with the anti-Nox1 antibody at just under 50 kDa (Figure 2C). Immunoprecipitation of in vitro translated Nox1-Myc with an anti-Myc antibody also gave a protein recognized by the Nox1 antibody at the expected size (data not shown). These data suggest that the Nox1 antibody properly recognizes the Nox1 protein at an observed weight of 46 kDa. We then tested whether Nox1 is responsible for TNF-induced O2 generation. Among the four Nox1 siRNA oligos tested, oligos #1 and #4 reduced the amount of Nox1 protein when transfected into L929 cells (Figure 2D and Figure S4A). Oligos #2 and #3 had little or no effect on Nox1 protein levels (Figure 2D and data not shown). The Lamin A/C control oligo caused no reduction in the Nox1 protein but efficiently reduced both the A and C forms of Lamin in these cells (Figure 2D and Figure S4A). Knockdown of Nox1 by oligos #1 and #4 led to a dramatic decrease in the production of O2 in these cells (Figure 2E and Figure S4B), but neither of the other siRNA oligos diminished the amount of O2 generated in response to TNF. In a rescue experiment, we cotransfected an expression plasmid encoding human Nox1 and Nox1 siRNA oligo #4. Because the target sequence of oligo #4 in mouse Nox1 is quite different from the sequence of human Nox1, oligo #4 cannot reduce human Nox1 expression. Cotransfection of the human Nox1 plasmid restored the TNF-induced O2 activation in siRNA transfected cells, while cotransfection of a catalytically inactive point mutant of Nox1, T341K, which is based on a homologous inactive mutation of Nox2 found in some patients with CGD, failed to restore the TNF-induced activation (Figure 2F). Based on these findings, Nox1 is likely the major NADPH oxidase responsible for TNF-induced O2 generation in L929 cells. Nox1 Forms a Complex with TRADD, RIP1, and Rac1 Following TNF Treatment Nox1 is localized to caveolin-enriched lipid rafts on the cell membranes of some cell types (Hilenski et al., 2004). Several TNF receptor machinery components have also been shown to relocalize to similar lipid rafts upon TNF simulation (Legler et al., 2003), suggesting Nox1 activation may be regulated by an interaction with TNF-R1 or its adaptor molecules. We therefore immunoprecipitated the endogenous TRADD protein in the presence or absence of TNF. As has been previously shown, the protein RIP1 coimmunoprecipitated with the TRADD protein upon TNF stimulation, indicating that a signaling complex had been formed upon TNF treatment (Figure 3A). When probed with the Nox1 antibody, a 46 kDa band was detected in the TNFtreated precipitants, but not in the untreated ones, indicating that Nox1 coprecipitated with TRADD. We also sought

to determine whether the complex contained Rac1, which is required for an active Nox1 NADPH oxidase complex. Rac1 protein was also found in the TRADD coimmunoprecipitants from the treated cells, but not the untreated cells (Figure 3A, bottom). This suggests that not only is Nox1 found in a complex with TRADD, but also that Nox1 is likely activated in the complex. We next sought to identify the means by which the Nox1 protein is recruited to the TRADD complex following TNF treatment since we did not observe a direct Nox1 interaction with TRADD or RIP1 (data not shown). Although little is known about Nox1 activation, it is thought to involve the NOXO1 and NOXA1 subunits. We therefore tested whether TRADD and/or RIP1 interacts with these two proteins upon overexpression in HEK293 cells. As shown in Figure 3B, left panel, RIP1 shows a strong interaction with NOXO1, but little interaction with NOXA1. Similarly, a specific interaction was also found between TRADD and NOXO1 (Figure 3B, right). None of the other known NADPH oxidase subunits including p22phox, p47phox, or p67phox interacted significantly with RIP1 or TRADD (data not shown). Therefore, the interaction between NOXO1 and RIP1 or TRADD provides a potential mechanism for recruiting Nox1 to the TNF-induced signaling complex. To further confirm the involvement of NOXO1 in the formation of the Nox1/TRADD/RIP1 complex, we immunoprecipitated NOXO1 in L929 cells in the presence or absence of TNF. We detected a constitutive interaction between the NOXO1 and Nox1 in untreated cells (Figure 3C), consistent with previous data (Cheng and Lambeth, 2004). In contrast, RIP1 and Rac1 were not coprecipitated with NOXO1 in untreated cells. RIP1 and Rac1 were observed in the complex 30 min after TNF treatment, suggesting that TNF induced the formation of a Nox1/TRADD/RIP1 signaling complex that also contained NOXO1 (Figure 3C). To determine whether endocytosis of the receptor was responsible for bringing the TNF signaling components into contact with the Nox1 machinery, we treated cells with Cytochalasin D or Latrunculin A, which inhibit actin polymerization by different mechanisms. Neither drug had any effect on TNF-induced JNK activation or IkBa degradation, nor prevented recruitment of Nox1 to TRADD (Figures S5A and S5B). The levels of O2 were actually increased upon drug treatment (Figure S5C), suggesting that inhibition of endocytosis may increase the amount of detectable extracellular O2 to some extent by keeping the complex in the cell membrane. Thus endocytosis is not responsible for the Nox1 recruitment. RIP1 Is Required for the Recruitment of Endogenous Nox1 in TNF-Treated Cells Because RIP1 showed a strong interaction with NOXO1, we then tested whether RIP1 was required for the formation of the endogenous Nox1/TRADD/RIP1 complex. Detection of the Nox1/TRADD/RIP1 complex in wild-type MEFs was seen under necrotic conditions (with TNF,

Molecular Cell 26, 675–687, June 8, 2007 ª2007 Elsevier Inc. 679

Molecular Cell TNF-Induced Activation of Nox1 in Necrosis

Figure 3. TRADD, RIP1, and Rac1 Form a Complex with Nox1/NOXO1 Following TNF Treatment (A) Western blots of endogenous L929 cell protein coimmunoprecipitated by anti-TRADD antibody from TNF-treated cells for the indicated times. (B) Western blots showing immunoprecipitation (IP) experiments in 293 cells overexpressing proteins from Xp-His, Xp-His-NOXO1, or Xp-His-NOXA1 plasmids in combination with either Flag-RIP1 (left) or Flag-TRADD (right). Due to the differences in binding affinity, the RIP1 immunoprecipitation was carried out for 5 hr after lysis, while the TRADD immunoprecipitation was done overnight. (C) Western blots of anti-NOXO1 antibody precipitates of endogenous protein from untreated L929 cells or after TNF treatment. Lanes were cropped where indicated to simplify the figure. (D) Immunoprecipitation of endogenous protein from wild-type, RIP / , and TRAF2 / MEFs using the anti-TRADD antibody. Cells were pretreated with CHX and zVAD for 30 min before TNF treatment or left untreated for the control.

CHX, and zVAD), but not under apoptotic conditions (TNF and CHX) or in cells treated with TNF alone (data not shown). Therefore, coimmunoprecipitation experiments were performed in wild-type, RIP1 / , and TRAF2 / MEF cells under necrotic conditions. While the antiTRADD antibody brought down Nox1 in the wild-type and TRAF2 / cells, this interaction was not seen in RIP1 / MEFs, though Nox1 was present in the cell lysates (Figure 3D). This indicates that RIP1 was required for recruitment of endogenous Nox1 to the complex in response to TNF and implies that the activation of Nox1 requires the RIP1 protein. TRADD Association with NOXO1 Requires a Polyproline Region that Is a Consensus SH3 Ligand Motif While the N terminus of TRADD contains a TRAF binding region and the C terminus contains a death domain that interacts with TNF-R1, RIP1, and FADD, the central portion of the molecule has no ascribed function, except that it contains a nuclear export sequence. This uncharacterized region of TRADD contains a consensus SH3 ligand motif with seven prolines flanked by an N-terminal lysine. This region could be responsible for the interaction with NOXO1, since NOXO1 contains two SH3 domains. There-

680 Molecular Cell 26, 675–687, June 8, 2007 ª2007 Elsevier Inc.

fore, we hypothesized that the interface between TRADD and NOXO1 might involve this polyproline region. We created a TRADD protein with alanine substitutions in the proline region (P-A). These mutations caused a small shift in the mobility of the protein, possibly due to the increased flexibility (Figure 4). When coexpressed with NOXO1, this mutant protein failed to interact with NOXO1 as the wild-type protein did (Figure 4A). Because the mobility of the mutant TRADD protein had changed, we checked whether the overall structure of the mutant was altered by examining its interaction with RIP1, FADD, and TRAF2. As shown in Figure 4B, the P-A mutant retained the interaction with these proteins, and coimmunoprecipitated with them as efficiently as the nonmutant TRADD protein did. These results suggest that the TRADD P-A mutant protein loses its interaction with NOXO1 specifically. To test whether the SH3 domains of NOXO1 were involved in binding to TRADD, we mutated the conserved tryptophan in the binding pockets of NOXO1. The resulting W196K and W268K mutations would not be expected to bind to proline-rich ligand sequences any longer but should retain the structure of the SH3 domains. As shown in Figure 4C, mutation of the first SH3 domain (W196K), but not the second domain (W268K), prevented the binding of NOXO1 to TRADD. The mutations on both SH3

Molecular Cell TNF-Induced Activation of Nox1 in Necrosis

Figure 4. Interaction of NOXO1 with TRADD Requires Its Polyproline Region (A and B) Western blots of immunoprecipitation (IP) experiments performed in 293 cells transfected with the Flag, Flag-TRADD, or FlagTRADD P-A mutant (TRADD prolines 192–198 changed to AAAAAGT) plasmids and (A) Xpress-His Vector or Xp-His-NOXO1 plasmids, (B [top]) Myc or Myc-RIP1 plasmids, (B [bottom left]) GFP-FADD plasmid, or (B [bottom right]) HA-TRAF2 plasmid. (C) Immunoprecipitation of His-tagged proteins from 293 cells cotransfected with Flag or Flag-TRADD along with Xpress-His Vector, Xp-His-NOXO1, or Xp-His-NOXO1 SH3 mutant plasmids.

domains also abolished the interaction between TRADD and NOXO1. These data suggest that the first SH3 domain is critical for binding to TRADD. The TRADD P-A Mutant Blocks TNF-Induced Nox1 Activation and Necrotic Cell Death To test the role of TRADD in Nox1 activation, several L929 cell lines were created in which the Flag vector, FlagTRADD, and Flag-TRADD P-A mutant were stably expressed (Figure 5A). We used these cell lines to examine the effect of the TRADD P-A mutant on different TNF signaling pathways. As shown in Figure 5B, no deficiencies in immediate JNK activation and IkBa degradation were found in these cells, but less Nox1 coimmunoprecipitated with TRADD upon TNF treatment (Figure 5C), suggesting a specific dominant-negative effect of this TRADD mutant on Nox1 recruitment. We then measured the TNF-induced O2 production in these cell lines. The O2 production was dramatically impaired in the cells with ectopic expression of TRADD P-A mutant protein, but not in the other stable cell lines (Figure 5D). To further confirm these results, we produced a second set of stable cell lines and obtained

similar data (Figure S6A). These results indicated that the TRADD P-A mutant selectively blocks Nox1 activation. The above data suggest that the TRADD P-A mutant acts as a dominant-negative molecule with respect to the activation of NADPH oxidase by TNF but has no effect on TNF-induced activation of NF-kB and JNK. We then tested whether the specific reduction in O2 generation by Nox1 affected TNF-induced necrotic cell death. We treated stable cell lines for 10 hr with TNF and observed the resulting cell death. A significant increase of cell survival was seen in the cells expressing the TRADD P-A mutant protein compared to cells having the empty vector or the nonmutant TRADD, as measured by a tetrazolium dye colorimetric test (MTT) (Figure 5E, left). Phase contrast microscopy verified these results (Figure 5E, right), and a decrease in TNFinduced cytotoxicity was also seen in the other independently derived stable cell lines when compared to controls (Figure S6B). In the presence of zVAD, where NADPH oxidase activity was much higher and cell death occurred quickly, a difference in cell viability between the FlagTRADD P-A mutant and the other cell lines could be easily detected as early as 2 hr (Figure S6C). These data suggest

Molecular Cell 26, 675–687, June 8, 2007 ª2007 Elsevier Inc. 681

Molecular Cell TNF-Induced Activation of Nox1 in Necrosis

Figure 5. Effect of Stable Expression of TRADD P-A Mutant on TNF-Induced O2 Generation and Cell Death (A) Expression levels of TRADD and TRADD P-A mutant in L929 stable cell lines. Similar protein amounts were added. (B) Western blots showing intact JNK and IkBa degradation pathways in Flag-vector, Flag-TRADD, and Flag-TRADD P-A mutant stable cell lines from (A) in response to TNF. (C) TRADD immunoprecipitation experiments in TRADD and TRADD P-A mutant stable cell lines. (D) NADPH oxidase assays in cells from (A). The luminescence ratio is shown indicating the relative fold activation of NADPH oxidase by TNF. (E) Cell viability data of stable cell lines from (A) treated for TNF for 10 hr and analyzed by MTT assay data (error bars, ±SEM) (left) or phase-contrast microscopy (right).

that a specific reduction in Nox1 activation by TNF results in an increase in cell viability and implicates the essential role of O2 in TNF-induced necrotic cell death. The Dominant-Negative Rac1 Mutant, N17Rac1, Reduces TNF-Induced Nox1 Activity and Necrotic Cell Death To further verify that a reduction in O2 production in response to TNF leads to a decrease in cell death, we sought to reduce the Nox1 activation in other ways.

682 Molecular Cell 26, 675–687, June 8, 2007 ª2007 Elsevier Inc.

Rac1 has been recently shown by several groups to be essential for Nox1 activity (Cheng et al., 2006; Miyano et al., 2006; Ueyama et al., 2006) and was detected in the Nox1/ TRADD/RIP1 complex (Figure 3A). Therefore, we stably expressed a Rac1 mutant (T17N) in L929 cells (Figure 6A). This mutant, which acts as a dominant-negative molecule by preventing GDP-GTP exchange, had no detectable effect on early JNK activation and activation of NF-kB by TNF (Figure 6B). Expression of N17Rac1 significantly reduced the amount of O2 produced in response to TNF

Molecular Cell TNF-Induced Activation of Nox1 in Necrosis

Figure 6. Effect of Dominant-Negative Rac1 in TNF-Induced O2 Generation and Cell Death (A) Expression levels of the N17-Rac1 mutant in L929 stable cell lines. (B) Western blots showing intact JNK and IkBa degradation pathways in Flag-vector or Flag-N17-Rac1 mutant stable cells from (A) in response to TNF at the indicated times. (C) NADPH oxidase assays in cells from (A). The luminescence ratio is shown indicating the relative fold activation of NADPH oxidase by TNF. (D) MTT assay of stable cell lines from (A)–(C) treated with TNF for 10 hr (error bars, ±SEM). (E) Viability assays (MTT) of L929 cells transfected with Nox1-specific siRNA (#4 and #2) or control siRNA (Lamin A/C) and treated with TNF. As shown in Figure 2D, little Nox1 reduction was seen in cells transfected with Nox1 oligo #2. Representative phase-contrast microscopy images (top) and MTT assay data at 24 hr (error bars, ±SEM) (bottom) are shown. (F) MTT assay of L929 cells cotransfected with Nox1 #4 siRNA oligo and either a human Nox1 expression plasmid, an inactive point mutant (T341K), or empty vector and treated with TNF for 24 hr (error bars, ±SEM).

when compared with the empty vector transfected cells (Figure 6C). We next examined the effect of N17Rac1 on TNF-induced necrotic cell death. As shown in Figure 6D,

a significant reduction in the amount of cell death was seen in cells stably expressing N17Rac1 compared to control cells. Therefore, like the TRADD P-A mutant,

Molecular Cell 26, 675–687, June 8, 2007 ª2007 Elsevier Inc. 683

Molecular Cell TNF-Induced Activation of Nox1 in Necrosis

Figure 7. Prolonged JNK Activation Is Involved in TNFInduced O2 Generation and Cell Death (A and B) Time course of JNK phosphorylation in (A) Flag-TRADD and Flag-TRADD P-A mutant L929 stable cell lines or (B) Flag or Flag-N17Rac1 mutant L929 stable cells. (C) Cell viability after TNF treatment (8 hr) as determined by MTT assay (error bars, ±SEM) in L929 cells pretreated (30 min) with BHA or SP600125, an inhibitor of JNK. (D) Schematic diagram of proposed pathway.

N17Rac1 also specifically inhibits TNF-induced Nox1 activation and necrotic cell death.

Prolonged JNK Activation Is Restrained in Cells in which Nox1 Activation Is Reduced JNK is activated within 15 min after TNF treatment, but in cell types that are not sensitive to TNF-induced cell death, this activation is transient due to JNK phosphatases and is barely detected after 60 min. It has been previously shown that prolonged JNK activation after this time contributes to TNF-induced cell death and that increased ROS levels play a role in this process (Tang et al., 2001; Sakon et al., 2003). It was proposed that this prolonged JNK activation occurs through activation of the MAP3K ASK1 (Liu et al., 2000; Tobiume et al., 2001) and also through the inactivation of JNK phosphatases by ROS (Kamata et al., 2005). We therefore examined whether prolonged JNK activation was reduced in our stable cell lines expressing the dominant-negative TRADD or Rac1 mutants. Notably, TNF-induced activation of JNK at 15 and 30 min occurs normally in these cells (Figures 5B, 6B, 7A, and 7B). But at extended time points when O2 was produced in L929 cells (>45 min), we saw a significant reduction in the amount of JNK phosphorylation in cells expressing the TRADD P-A mutant (Figure 7A) or the N17Rac1 mutant (Figure 7B), in which Nox1 activation is reduced, compared to the control cells. This reduction in sustained JNK activation was especially obvious 3 hr after treatment. To determine whether prolonged JNK activation was involved in cell death in L929 cells, we treated the cells with a pharmacological inhibitor of JNK, SP600125, and measured TNF-induced cell death. The JNK inhibitor reduced the amount of cell death similarly to treatment with BHA after 8 hr of TNF treatment (Figure 7C), suggesting that JNK activation is an important step in TNF-induced necrotic cell death (Figure 7C). SP600125 also significantly reduced the amount of cell death at early time points (<6 hr) in p65 / cells treated with TNF and zVAD (data not shown) in a manner similar to BHA. These data imply that sustained JNK activation occurs in response to the production of O2 generated downstream of TNF and that prolonged JNK activation contributes to caspaseindependent necrotic cell death induced by TNF. DISCUSSION

Knockdown of Nox1 Reduces TNF-Induced Necrotic Cell Death To eliminate the possibility that unexpected nonspecific effects of the TRADD and Rac1 mutants contribute to the decrease in TNF-induced necrotic cell death, we knocked down the expression of Nox1 protein with the Nox1 siRNA oligo #4 as described in Figure 2. Knocking down Nox1 protein level significantly protects the cells against TNF-induced necrotic cell death in L929 cells (Figure 6E). A similar reduction in death was seen using a second siRNA oligo (#1) to Nox1 (Figure S7), but not the Lamin A/C oligo, or the Nox1 oligo (#2) that fails to reduce Nox1 protein levels (Figures 6E and 2D). Cell death due to TNF could be rescued by cotransfection of the human Nox1 plasmid, but not its catalytically inactive point mutant T341K or empty vector (Figure 6F).

684 Molecular Cell 26, 675–687, June 8, 2007 ª2007 Elsevier Inc.

The signaling events that mediate TNF-induced caspaseindependent necrotic cell death remain largely elusive, though previous studies indicate that the generation of ROS is essential for TNF-induced necrotic cell death (Fiers et al., 1999). ROS may mediate necrotic cell death by inactivating MPK phosphatases, leading to sustained JNK activation (Kamata et al., 2005). In our current study, we have shown that ROS generated through activation of Nox1 plays a critical role in TNF-induced necrotic cell death (Figure 7D). Our data suggest that Nox1 is the major NADPH oxidase responsible for TNF-induced generation of O2 in mouse fibroblasts and is activated through a signaling complex containing Nox1, NOXO1, Rac1, and the TNF effector proteins, TRADD and RIP1. TRADD and RIP1 bind

Molecular Cell TNF-Induced Activation of Nox1 in Necrosis

to the oxidase adaptor NOXO1 when overexpressed, and TNF treatment leads to the formation of an active Nox1 complex containing Rac1. Mutation of the proline region in TRADD, which is required for the interaction with the first SH3 domain of NOXO1, appears to diminish TNFinduced O2 generation. Thus, the mutant protein functions as a dominant-negative molecule with regard to Nox1 activation, without affecting transient JNK activation and IkBa degradation. While we cannot exclude the possibility that the mutation in TRADD also affects the binding of another protein that participates in Nox1 activation, the simplest explanation is that a TRADD-NOXO1 interaction at this polyproline region is necessary for the activation of NADPH oxidase by TNF. The protein RIP1 is necessary for the generation of ROS by TNF and required for the initiation of necrotic cell death (Lin et al., 2004; Holler et al., 2000). In this study, we found that Nox1 interacts with a TRADD complex in wild-type and p65 / MEFs treated to undergo necrotic cell death but that Nox1 fails to coimmunoprecipitate with TRADD in RIP1 / MEFs under similar conditions. This finding suggests that RIP1 functions in recruiting Nox1 to the signaling complex and that the interaction between RIP1 and NOXO1 may be necessary for this recruitment. Based on the relative affinities of RIP1 and TRADD for NOXO1, we would suggest a model in which the stronger RIP1 interaction with NOXO1 recruits it and Nox1 and/or NOXA1 to the complex, where an interaction between NOXO1 and TRADD is necessary for the stabilization of the complex and/or the activation of Nox1. A previous study proposed a role for the kinase activity of RIP1 in necrotic cell death (Holler et al., 2000). While we have not as yet excluded a role for the kinase activity of RIP1 in the activation of Nox1, our previous data suggest that the kinase activity of RIP1 is dispensable for JNK activation by TNF and necrotic cell death in fibroblasts (Lin et al., 2004). Additionally, the kinase domain of RIP1 appears not to be important for the interaction between RIP and NOXO1 (data not shown). While we have established that the nonmitochondrial oxidase Nox1 is involved in TNF-induced necrotic cell death, this does not exclude mitochondrial involvement in the process. We have also not excluded the possibility that the O2 produced by nonmitochondrial oxidases leads to mitochondrial-derived ROS. The potency by which BHA blocks TNF-induced cell death may therefore potentially be explained by an ability to block ROS generation from both sources. Some of the O2 generated by Nox1 appears to be extracellular in nature, as SOD and other ROS scavengers were able to immediately quench it. This infers that the O2 generated outside the cell must act via conversion to another species, such as H2O2, which is able to diffuse through the cell membrane and react inside the cell, or by an oxidation of extracellular (or endosomal) lipids and therefore the production of second messengers, as has been suggested previously (Saran, 2003). Alternatively, the endocytosis of TNF-R1 following TNF treatment may lead to production of O2 intracellularly.

TNF can induce biphasic activation of JNK: while the prolonged JNK activation contributes to TNF-induced cell death, the transient JNK activation protects cells against death (Karin and Lin, 2002; Ventura et al., 2006). Recent studies suggest that this prolonged JNK activation occurs through activation of the MAP3K ASK1 (Liu et al., 2000; Tobiume et al., 2001) or through the inactivation of JNK phosphatases by ROS reaction with their catalytic cysteine (Kamata et al., 2005). Our current work is consistent with the later observation, particularly, in that the kinetics of O2 generation by Nox1 correlates with the profile of prolonged JNK activation. TNF-R1 appears to be the receptor that activates Nox1, as TRADD is not recruited to the TNF-R2 complex. This is consistent with our previous results showing that necrotic cell death is induced only through TNF-R1 (Lin et al., 2004). Due to the antibodies involved and the proximity of both bands to the IgG heavy chain, we have not yet determined whether the TNF-R1 protein is physically involved in the Nox1 complex or whether this complex is downstream of the receptor. Nox1 is localized to lipid rafts in some cell types, and TNF-R1 has been reported to relocalize to lipid rafts after TNF stimulation, as has RIP1 and TRAF2 (Legler et al., 2003). Previous data showed the receptor components RIP1 and TRADD bind to TNFR1 within 5 min upon stimulation and begin to release from the receptor at about 20–30 min. We were able to detect Nox1 recruitment into the complex (with TRADD) within 15 min of TNF treatment, which may suggest TNF-R1 involvement in the complex. The maximal amount of Nox1 coprecipitated with TRADD at about 30 min after TNF treatment. This time point corresponds to the time at which the maximal amount of Rac1 is also recruited to the complex, and we are able to detect a rise in O2 levels shortly after this time. RIP1 is heavily modified by K63linked polyubiquitination at early time points after TNF treatment whereby it interacts with the NEMO subunit of IKK to activate NF-kB. It is possible that this ubiquitination prevents its interaction with NOXO1 until enough nonubiquitinated RIP1 is recruited to the receptor. TNF-R1 and Nox1 are expressed ubiquitously, so it is unclear at present under what conditions this necrotic pathway may be induced. In many systems, TNF induces inflammation rather than cell death due to the activation of NF-kB, which not only upregulates proinflammatory genes but also results in the upregulation of antioxidant proteins such as MnSOD and prosurvival proteins such as c-Flip, A20, and cIAPs1/2, which inhibit apoptosis. Many of these proteins may also inhibit necrosis. Nevertheless, there are likely some conditions under which necrotic death is initiated. In model cell culture systems such as ours, the addition of zVAD is sometimes required to initiate necrotic cell death, and, as mentioned before, in many situations the cleavage of RIP1 by caspases during apoptosis (Lin et al., 1999) probably prevents necrotic cell death from occurring. Given that many microbes encode antiapoptotic proteins such as CrmA, E1B19, and p35, many of which directly block caspase activation, the

Molecular Cell 26, 675–687, June 8, 2007 ª2007 Elsevier Inc. 685

Molecular Cell TNF-Induced Activation of Nox1 in Necrosis

Nox1-mediated necrotic death pathway may be a default cell death pathway that is activated by TNF following the inflammatory response in order to remove the remaining damaged or infected cells that are not removed by apoptosis. To our knowledge, there are no current reports in the literature where Nox1 / mice have been challenged by infection (the only described phenotype with regard to these mice is lower than average blood pressure), but such experiments would test this hypothesis. Interestingly, Nox1 did not appear to contribute to necrotic cell death induced by dsRNA and interferon in L929 cells (Figure S8) but did appear to contribute specifically to the TNF-induced death. Previously, the activation of Nox2 by TNF has been proposed to involve the activation of a tyrosine kinase or PKCzeta. We found that inhibitors of these pathways had no effect on the activation of Nox1 by TNF and did not protect from cell death (Figure S9). It is uncertain as to whether TNF may activate Nox2 by a similar mechanism in phagocytic or cardiovascular cell types, in which the p47phox protein is expressed instead of NOXO1. Though we did not detect an obvious interaction between p47phox and TRADD or RIP1, an interaction with these proteins is still a possibility in a physiological context. Further study is necessary to investigate whether Nox2 is recruited to TNF-R1 complex through RIP1/TRADD similarly as Nox1. EXPERIMENTAL PROCEDURES Reagents Murine rTNFa was from R&D Systems. Antibodies were from commercial sources: anti-Nox1, anti-TRADD, anti-IkBa, anti-Myc, anti-His, anti-HA, anti-GFP, and anti-Xpress from Santa Cruz; anti-RIP1 from BD Transduction Laboratories; anti-p-JNK from Biosource; anti-JNK from Pharmingen; anti-Rac1 from Upstate; and anti-NOXO1 from Rockland. SP600125, zVAD, and CHX were from Calbiochem. Lucigenin was from Invitrogen. Anti-actin and anti-Flag antibodies, NADPH, BHA, MTT, and puromycin were from Sigma. Treatment Conditions In all experiments, TNFa was used at 30 ng/ml, zVAD at 20 mM, CHX at 10 mg /ml, and BHA at 100 mM.

proteins were removed by boiling in SDS buffer and resolved in 4%–20% SDS-polyacrylamide gels for western blot analysis. Transfection Transfection experiments in 293 cells were performed with Lipofectamine PLUS reagent by following the instructions provided by the manufacturer (GIBCO/BRL). NADPH Oxidase Assays The NADPH oxidase activity was measured by O2 -dependent lucigenin chemiluminescence. Approximately 3 3 105 cells in 1 ml HBSS were incubated 20 min with 400 mM NADPH. Lucigenin (200 mM) was added and chemiluminescence was measured before and after addition of TNFa. Cells were maintained at 37 C during the experiment. Small Interfering RNA Knockdown of Nox1 L929 cells were plated in 6-well plates, and cells were transfected with 100 pmol of Nox1- or Lamin A/C- specific RNAi oligo (Dharmacon) using Lipofectamine 2000 reagent (Invitrogen). After 72 hr, knockdown was analyzed by western blotting and the remaining cells were used for the NADPH oxidase and cytotoxicity assays. Cytotoxicity Assay Cell death was determined using MTT. MTT absorbance was read at 570 nm. Representative images were taken by a phase-contrast microscope. In Vitro Translation In vitro translation was done using the T7 and SP6 directed TNTcoupled wheat germ and reticulocyte lysate systems (Promega). Supplemental Data Supplemental Data include nine figures and can be found with this article online at http://www.molecule.org/cgi/content/full/26/5/675/ DC1/. ACKNOWLEDGMENTS We would like to thank Drs. W.C. Yeh and T.W. Mak for TRAF2 / MEFs and Dr. M.A. Kelliher for RIP / MEFs. This research was supported by the Intramural Research Program of the Center for Cancer Research, NCI, NIH. Received: December 20, 2006 Revised: March 9, 2007 Accepted: April 27, 2007 Published: June 7, 2007 REFERENCES

Cell Culture L929, HEK293, and the various 3T3-like immortalized MEF cell lines were cultured in DMEM with 10% fetal bovine serum, 2 mM glutamine, and 100 U/ml penicillin/streptomycin. L929 stable cell lines were established by transfecting cells with the designated vector and then maintaining in media containing 20 mg/ml puromycin. Western Blot Analysis and Coimmunoprecipitation Upon treatment, cells were lysed in M2 buffer (20 mM Tris at pH 7, 0.5% NP-40, 250 mM NaCl, 3 mM EDTA, 3 mM EGTA, 2 mM DTT, 0.5 mM PMSF, 20 mM b-glycerol phosphate, 1 mM sodium vanadate, and 1 mg/ml leupeptin). For western blots, 50 mg protein was visualized by enhanced chemiluminescence (ECL, Amersham). For immunoprecipitation assays, cells were treated with TNF for 30 min and collected in M2 buffer. Lysates were mixed and precipitated with antibody and protein A-Sepharose or protein G-agarose beads by incubation at 4 C overnight, or for 5 hr (for RIP1 immunoprecipitation experiments). Beads were washed 4–6 times with 1 ml lysis buffer, and the bound

686 Molecular Cell 26, 675–687, June 8, 2007 ª2007 Elsevier Inc.

Chamseddine, A.H., and Miller, F.J., Jr. (2003). Gp91phox contributes to NADPH oxidase activity in aortic fibroblasts but not smooth muscle cells. Am. J. Physiol. Heart Circ. Physiol. 285, H2284–H2289. Chen, G., and Goeddel, D.V. (2002). TNF-R1 signaling: a beautiful pathway. Science 296, 1634–1635. Cheng, G., and Lambeth, J.D. (2004). NOXO1, regulation of lipid binding, localization, and activation of Nox1 by the Phox homology (PX) domain. J. Biol. Chem. 279, 4737–4742. Cheng, G., Diebold, B.A., Hughes, Y., and Lambeth, J.D. (2006). Nox1dependent reactive oxygen generation is regulated by Rac1. J. Biol. Chem. 281, 17718–17726. Dang, P.M., Stensballe, A., Boussetta, T., Raad, H., Dewas, C., Kroviarski, Y., Hayem, G., Jensen, O.N., Gougerot-Pocidalo, M.A., and El-Benna, J. (2006). A specific p47phox -serine phosphorylated by convergent MAPKs mediates neutrophil NADPH oxidase priming at inflammatory sites. J. Clin. Invest. 116, 2033–2043.

Molecular Cell TNF-Induced Activation of Nox1 in Necrosis

Dewas, C., Dang, P.M., Gougerot-Pocidalo, M.A., and El-Benna, J. (2003). TNF-alpha induces phosphorylation of p47(phox) in human neutrophils: partial phosphorylation of p47phox is a common event of priming of human neutrophils by TNF-alpha and granulocyte-macrophage colony-stimulating factor. J. Immunol. 171, 4392–4398. Fiers, W., Beyaert, R., Declercq, W., and Vandenabeele, P. (1999). More than one way to die: apoptosis, necrosis and reactive oxygen damage. Oncogene 18, 7719–7730. Frey, R.S., Rahman, A., Kefer, J.C., Minshall, R.D., and Malik, A.B. (2002). PKCzeta regulates TNF-alpha-induced activation of NADPH oxidase in endothelial cells. Circ. Res. 90, 1012–1019. Geiszt, M., Lekstrom, K., Witta, J., and Leto, T.L. (2003). Proteins homologous to p47phox and p67phox support superoxide production by NAD(P)H oxidase 1 in colon epithelial cells. J. Biol. Chem. 278, 20006–20012. Gorlach, A., Brandes, R.P., Nguyen, K., Amidi, M., Dehghani, F., and Busse, R. (2000). A gp91phox containing NADPH oxidase selectively expressed in endothelial cells is a major source of oxygen radical generation in the arterial wall. Circ. Res. 87, 26–32. Heyworth, P.G., Cross, A.R., and Curnutte, J.T. (2003). Chronic granulomatous disease. Curr. Opin. Immunol. 15, 578–584.

Martyn, K.D., Frederick, L.M., von Loehneysen, K., Dinauer, M.C., and Knaus, U.G. (2006). Functional analysis of Nox4 reveals unique characteristics compared to other NADPH oxidases. Cell. Signal. 18, 69–82. Micheau, O., and Tschopp, J. (2003). Induction of TNF receptor Imediated apoptosis via two sequential signaling complexes. Cell 114, 181–190. Miyano, K., Ueno, N., Takeya, R., and Sumimoto, H. (2006). Direct involvement of the small GTPase Rac in activation of the superoxideproducing NADPH oxidase Nox1. J. Biol. Chem. 281, 21857–21868. Nguyen, H.X., and Tidball, J.G. (2003). Null mutation of gp91phox reduces muscle membrane lysis during muscle inflammation in mice. J. Physiol. 553, 833–841. Sakon, S., Xue, X., Takekawa, M., Sasazuki, T., Okazaki, T., Kojima, Y., Piao, J.H., Yagita, H., Okumura, K., Doi, T., and Nakano, H. (2003). NF-kappaB inhibits TNF-induced accumulation of ROS that mediate prolonged MAPK activation and necrotic cell death. EMBO J. 22, 3898–3909. Saran, M. (2003). To what end does nature produce superoxide? NADPH oxidase as an autocrine modifier of membrane phospholipids generating paracrine lipid messengers. Free Radic. Res. 37, 1045–1059.

Hilenski, L.L., Clempus, R.E., Quinn, M.T., Lambeth, J.D., and Griendling, K.K. (2004). Distinct subcellular localizations of Nox1 and Nox4 in vascular smooth muscle cells. Arterioscler. Thromb. Vasc. Biol. 24, 677–683.

Suh, Y.A., Arnold, R.S., Lassegue, B., Shi, J., Xu, X., Sorescu, D., Chung, A.B., Griendling, K.K., and Lambeth, J.D. (1999). Cell transformation by the superoxide-generating oxidase Mox1. Nature 401, 79–82.

Holler, N., Zaru, R., Micheau, O., Thome, M., Attinger, A., Valitutti, S., Bodmer, J.L., Schneider, P., Seed, B., and Tschopp, J. (2000). Fas triggers an alternative, caspase-8-independent cell death pathway using the kinase RIP as effector molecule. Nat. Immunol. 1, 489–495.

Takeya, R., Ueno, N., Kami, K., Taura, M., Kohjima, M., Izaki, T., Nunoi, H., and Sumimoto, H. (2003). Novel human homologues of p47phox and p67phox participate in activation of superoxide-producing NADPH oxidases. J. Biol. Chem. 278, 25234–25246.

Kamata, H., Honda, S., Maeda, S., Chang, L., Hirata, H., and Karin, M. (2005). Reactive oxygen species promote TNFalpha-induced death and sustained JNK activation by inhibiting MAP kinase phosphatases. Cell 120, 649–661.

Tang, G., Minemoto, Y., Dibling, B., Purcell, N.H., Li, Z., Karin, M., and Lin, A. (2001). Inhibition of JNK activation through NF-kappaB target genes. Nature 414, 313–317.

Karin, M., and Lin, A. (2002). NF-kappaB at the crossroads of life and death. Nat. Immunol. 3, 221–227. Lambeth, J.D. (2004). NOX enzymes and the biology of reactive oxygen. Nat. Rev. Immunol. 4, 181–189. Legler, D.F., Micheau, O., Doucey, M.A., Tschopp, J., and Bron, C. (2003). Recruitment of TNF receptor 1 to lipid rafts is essential for TNFalpha-mediated NF-kappaB activation. Immunity 18, 655–664. Lin, Y., Devin, A., Rodriguez, Y., and Liu, Z.G. (1999). Cleavage of the death domain kinase RIP by caspase-8 prompts TNF-induced apoptosis. Genes Dev. 13, 2514–2526. Lin, Y., Choksi, S., Shen, H.M., Yang, Q.F., Hur, G.M., Kim, Y.S., Tran, J.H., Nedospasov, S.A., and Liu, Z.G. (2004). Tumor necrosis factor-induced nonapoptotic cell death requires receptor-interacting protein-mediated cellular reactive oxygen species accumulation. J. Biol. Chem. 279, 10822–10828. Liu, H., Nishitoh, H., Ichijo, H., and Kyriakis, J.M. (2000). Activation of apoptosis signal-regulating kinase 1 (ASK1) by tumor necrosis factor receptor-associated factor 2 requires prior dissociation of the ASK1 inhibitor thioredoxin. Mol. Cell. Biol. 20, 2198–2208.

Teixeira, M.M., Cunha, F.Q., Noronha-Dutra, A., and Hothersall, J. (1999). Production of singlet oxygen by eosinophils activated in vitro by C5a and leukotriene B4. FEBS Lett. 453, 265–268. Tobiume, K., Matsuzawa, A., Takahashi, T., Nishitoh, H., Morita, K., Takeda, K., Minowa, O., Miyazono, K., Noda, T., and Ichijo, H. (2001). ASK1 is required for sustained activations of JNK/p38 MAP kinases and apoptosis. EMBO Rep. 2, 222–228. Ueyama, T., Geiszt, M., and Leto, T.L. (2006). Involvement of Rac1 in activation of multicomponent Nox1- and Nox3-based NADPH oxidases. Mol. Cell. Biol. 26, 2160–2174. Ventura, J.J., Hubner, A., Zhang, C., Flavell, R.A., Shokat, K.M., and Davis, R.J. (2006). Chemical genetic analysis of the time course of signal transduction by JNK. Mol. Cell 21, 701–710. Wajant, H., Pfizenmaier, K., and Scheurich, P. (2003). Tumor necrosis factor signaling. Cell Death Differ. 10, 45–65. Yu, L., DeLeo, F.R., Biberstine-Kinkade, K.J., Renee, J., Nauseef, W.M., and Dinauer, M.C. (1999). Biosynthesis of flavocytochrome b558. gp91(phox) is synthesized as a 65-kDa precursor (p65) in the endoplasmic reticulum. J. Biol. Chem. 274, 4364–4369.

Molecular Cell 26, 675–687, June 8, 2007 ª2007 Elsevier Inc. 687