Recruitment of Tumor Necrosis Factor Receptor-associated Factor Family Proteins to Apoptosis Signal-regulating Kinase 1 Signalosome Is Essential for Oxidative Stress-induced Cell Death*

Apoptosis signal-regulating kinase 1 (ASK1) plays a pivotal role in oxidative stress-induced cell death. Reactive oxygen species disrupt the interaction of ASK1 with its cellular inhibitor thioredoxin and thereby activates ASK1. However, the precise mechanism by which ASK1 freed from thioredoxin undergoes oligomerization-dependent activation has not been fully elucidated. Here we show that endogenous ASK1 constitutively forms a high molecular mass complex including Trx (∼1,500-2,000 kDa), which we designate ASK1 signalosome. Upon H2O2 treatment, the ASK1 signalosome forms a higher molecular mass complex at least in part because of the recruitment of tumor necrosis factor receptor-associated factor 2 (TRAF2) and TRAF6. Consistent with our previous findings that TRAF2 and TRAF6 activate ASK1, H2O2-induced ASK1 activation and cell death were strongly reduced in the cells derived from Traf2-/- and Traf6-/- mice. A novel signaling complex including TRAF2, TRAF6, and ASK1 may thus be the key component in oxidative stress-induced cell death.

Among various stimuli tested, oxidative stress is one of the most potent activators of ASK1 (3). Thioredoxin (Trx), a reduction/oxidation (redox) regulatory protein, inhibits the kinase activity of ASK1 by its direct binding to the N-terminal noncatalytic region of ASK1 (6,9). Reactive oxygen species (ROS) such as H 2 O 2 dissociate Trx from ASK1 and thereby activate ASK1 (6). Following the dissociation of Trx, autophosphorylation-dependent activation of ASK1 occurs most likely through mechanisms involving homo-oligomerization (10,11); however, precise mechanisms of the oligomerization-dependent activation of ASK1 remained unknown. TNF receptor-associated factor 2 (TRAF2) has been shown to bind to and thereby activate ASK1 (5). TRAF2 is not only a pivotal intermediate in TNF␣-induced NF-B activation (12)(13)(14)(15)(16), but it also activates JNK in TNF␣ signaling through the association with MAPKKKs such as ASK1 and MEKK1 (5,17,18). A recent report has shown that AIP1/DAB2IP, a novel member of Ras-GAP family, regulates TRAF2-dependent activation of ASK1 in TNF␣-treated endothelial cells (19). In accordance with the findings that ROS function as second messengers in TNF␣ signaling (20,21), TNF␣-induced association of ASK1 with TRAF2 and subsequent activation of ASK1 have been shown to depend largely on intracellular ROS production by TNF␣ (6,17).
In addition to TRAF2, we have previously shown that among extensively characterized six members of the TRAF family (TRAF1-TRAF6) (22)(23)(24), TRAF5 and TRAF6 also interact with and activate ASK1 upon overexpression (5). TRAF5 has been implicated in NF-B and JNK activation in signaling through lymphotoxin-␤ receptor, CD40, and CD27 (25)(26)(27). Analyses of TRAF5-deficient mice indicated that TRAF5 is required for CD40-and CD27-mediated lymphocyte activation (28). TRAF6 is a critical regulator of NF-B and MAP kinases in the signaling pathways through the TNF receptor superfamily and Toll/interleukin-1 receptor family, as has been demonstrated by many studies using TRAF6-deficient mice (29 -33). Activation of NF-B and MAP kinase pathways by TRAF6 has been reported to be induced through the association with MAPKKKs such as TAK1 and MEKK3 (34,35).
In this study, we show that ASK1 constitutively forms a high molecular mass complex (ϳ1,500 -2,000 kDa). Upon H 2 O 2 treatment, ASK1 was dissociated from Trx but reciprocally associated with TRAF2 and TRAF6. The recruitment of TRAF2 and TRAF6 but not TRAF5 to ASK1 signaling complex was required for H 2 O 2 -induced activation of ASK1 and cell death. These results propose a novel signaling mechanism involving TRAF family proteins and ASK1 in oxidative stress-induced cell death.
Reagents and Antibodies-Transfection was performed with FuGENE 6 (Roche Applied Science) according to the manufacturer's instructions. Human recombinant TNF␣ was obtained from WAKO. The antibody to HA (clone 3F10) was from Roche Applied Science. Phospho-JNK (Thr 183 /Tyr 185 ) and phospho-p38 (Thr 180 /Tyr 182 ) antibodies were from Cell Signaling Technology. Anti-ASK1, anti-Trx, anti-TRAF2, anti-TRAF5, and anti-TRAF6 antibodies were from Santa Cruz Biotechnology, Inc. Affinity-purified rabbit polyclonal antibody raised against phospho-ASK1 (Thr 838 ) was described previously (11). To generate anti-TRAF6 antibody, the synthetic peptide RKEGFQPRSTD was used as immunogen, and the antisera obtained from immunized rabbit was affinity-purified by the synthetic peptide (36).
Gel Filtration Chromatography-S-100 fraction from HEK293 cells was loaded onto a Superose6 10/300 GL column pre-equilibrated with buffer A plus 0.1% Chaps and 0.01% Brij35. The proteins were eluted at 0.3 ml/min. Each fraction (0.5 ml/fraction) was precipitated using acetone/ethanol (1:4). Samples from one or four column runs were pooled and analyzed by Western blotting for the indicated antibodies.
Immunoblotting-The cells were lysed in buffer A plus 0.2% Nonidet P-40, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, and 5 g/ml aprotinin. The cell extracts were resolved by SDS-PAGE and transferred to polyvinylidene difluoride membranes. After blocking with 5% skim milk in TBS-T (150 mM NaCl, 50 mM Tris-HCl, pH 8.0, and 0.05% Tween 20), the membranes were probed with the indicated antibodies. The antibody-antigen complexes were detected using the ECL system (Amersham Biosciences).
Co-immunoprecipitation Analysis-The cells were lysed in buffer A plus 0.2% Nonidet P-40, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, and 5 g/ml aprotinin. The cell extracts were immunoprecipitated with 20 l of protein A-or protein G-Sepharose beads (Amersham Biosciences) for overnight at 4°C with the indicated antibodies. The beads were washed four times with the same buffer before analysis by SDS-PAGE.
Preparation of Splenocyte-derived Macrophage-Because the number of bone marrow cells was reduced significantly in Traf6Ϫ/Ϫ mice because of severe osteopetrosis, macrophages were generated from splenocytes of Traf6ϩ/Ϫ and Traf6Ϫ/Ϫ mice (30). After 6 days of culture with macrophage colony-stimulating factor (M-CSF), almost equal numbers of adherent cells, which displayed typical macrophage morphology, were obtained from both Traf6ϩ/Ϫ and Traf6Ϫ/Ϫ splenocytes. More than 95% of the adherent cells from both Traf6ϩ/Ϫ and Traf6Ϫ/Ϫ mice expressed CD11b and F4/80 (data not shown), indicating that TRAF6 deficiency does not affect macrophage differentiation.
Trypan Blue Dye Exclusion Assay-For the trypan blue dye exclusion assay, cells (1 ϫ 10 4 cells/well) were plated in 12-well plates. Twentyfour h after the plating, the cells were treated with 0.1 mM H 2 O 2 for 10 h. The cells were then directly stained with 0.4% trypan blue dye or harvested with trypsin followed by staining with 0.4% trypan blue dye. The trypan blue-positive cells were counted by using a hemocytometer under phase contrast microscopy.

ASK1 Forms a High Molecular Mass Signaling Complex in the Cell
-We previously demonstrated that phosphorylation of a threonine residue within the activation loop of ASK1 kinase domain (Thr 838 and Thr 845 of human and mouse ASK1, respectively) is required for activation of ASK1 in response to various stresses (11). We also showed that the activation state of ASK1 can be monitored by the phospho-ASK1 antibody that detects activating phosphorylation of this threonine residue (11). Immunoblotting analysis using the phospho-ASK1 antibody revealed that basal activity of a mutant form of ASK1 that lacked C-terminal coiled-coil domain (ASK1⌬coil) was much lower than that of wild-type ASK1 (ASK1WT) when each construct was expressed in HEK293 cells (Fig. 1, A and B). Overexpression of ASK1⌬coil was not able to induce the activation of JNK and p38 nor to inhibit the activity of co-transfected ASK1WT (data not shown). These results confirm our previous finding that homo-oligomerization through the C-terminal coiled-coil domain is required for the activation of ASK1 (11); however, regulatory mechanisms of ASK1 oligomer have been largely unknown.
To investigate how homo-oligomerization of ASK1 is involved in its activation, we first examined the potential complex formation of ASK1 in the cells using a gel filtration column chromatography by an approach similar to that used to characterize the apoptosome (37). The expected molecular mass of monomeric ASK1 is ϳ160 kDa. When HA-tagged ASK1WT was expressed in HEK293 cells, we found that ASK1WT formed a high molecular mass (HMM) complex in fractions 1-8 (Ͼ1,500 kDa; Fig. 1C, top panel, lanes 1-8) with relatively high basal activity detected by the phospho-ASK1 antibody (Fig. 1C, second panel). A kinase-negative mutant of ASK1 (ASK1K709R) also formed an HMM complex (data not shown), suggesting that the kinase activity of ASK1 is not a prerequisite for the HMM complex formation. On the other hand, ASK1⌬coil formed a smaller complex (ϳ500 kDa; but still larger than monomeric ASK1) with no detectable activity (Fig. 1C, bottom two panels, lanes [11][12][13][14][15]. It is noteworthy that a minor part of transfected ASK1⌬coil also formed an HMM complex with certain kinase activity (Fig. 1C, bottom two panels, lanes 5-7). These results suggest that oligomerization-dependent formation of the HMM complex is required for the kinase activity of ASK1.
ROS Induce a Higher Molecular Mass Complex Formation of ASK1 -We next examined the states of endogenous ASK1 by gel filtration analysis. Fractionation of the extracts from HEK293 cells through a gel filtration column revealed that although confined to fractions 5-8 (ϳ1,500 -2,000 kDa), endogenous ASK1 formed a similar HMM complex as was formed by exogenously expressed ASK1WT ( Fig. 2A, top  panel). A similar pattern of HMM complex was detected in all of the cultured cells we tested, e.g. HeLa cells, HT1080 cells, and MEFs (data not shown), suggesting that the HMM complex formation is a general feature of endogenous ASK1. Under unstimulated conditions, the kinase activity of ASK1 is negatively regulated by its association with Trx (6). To examine whether Trx is a component of the unstimulated HMM complex, ASK1-containing fractions (corresponding to lanes 5-8 in Fig. 2A, top panel) were collected and subjected to immunoprecipitation with anti-ASK1 antibody. Trx was found to be co-immunoprecipitated with ASK1 ( Fig. 2B), indicating that endogenous ASK1 forms a static HMM complex containing Trx, which may function as a signalosome competent to ROS-dependent activation of ASK1.
When cells were treated with H 2 O 2 , the ASK1 signalosome was eluted in higher molecular mass fractions, i.e. fractions 1-4 (Ͼ3,000 kDa; Fig. 2A, second and third panels). This elution profile was similar to that of exogenously expressed ASK1WT (Fig. 1C), suggesting that a certain population of exogenous ASK1 comprises an activated complex without any stimuli. The shift of the ASK1-containing fractions was inhibited by the pretreatment of cells with anti-oxidants such as propyl gallate ( Fig. 2A, bottom panel) and N-acetyl cysteine (data not shown). The changes in the elution pattern of the ASK1 complex correlated well with the activation states of ASK1 and its downstream targets, JNK and p38 (Fig. 2C), as monitored by immunoblotting analysis using the aliquots of each cell extract prepared for the gel filtration. These results suggest that recruitment of additional component(s) to the HMM signaling complex of ASK1 may be involved in H 2 O 2 -induced activation of ASK1. To examine the binding states between Trx and ASK1 after ROS exposure, nontransfected HEK293 cells lysate was subjected to the coimmunoprecipitation analysis. As previously demonstrated in ASK1transfected HEK293 cells (6), we confirmed that H 2 O 2 treatment disrupted the endogenous interaction between Trx and ASK1 in nontransfected HEK293 cells (Fig. 2D). When nontransfected HEK293  The cell extracts were prepared as described under "Materials and Methods " and fractionated through a Superose6 10/300 GL column. Each fraction (0.5 ml/fraction) was subjected to protein precipitation using acetone/ethanol (1:4). The samples were analyzed by immunoblotting with the indicated antibodies. Apparent molecular mass is shown as described above. The elution positions of standard proteins are indicated at the top of the figure. B, co-immunoprecipitation of Trx with ASK1. The cell extracts from HEK293 cells (2.0 ϫ 10 8 cells) were prepared as described under "Materials and Methods." After fractionation on a Superose6 10/300 GL column, ASK1 containing fractions ( Fig. 2A, lanes [5][6][7][8] were collected and immunoprecipitated with ASK1 antibody or control mouse IgG. The samples were then analyzed by immunoblotting with anti-ASK1 and anti-Trx antibodies. The lower two panels (Input) show the presence of ASK1 and Trx by using 5% of the total cell extracts that were used to perform the immunoprecipitation. C, H 2 O 2 -induced activation of endogenous ASK1. Subconfluent HEK293 cells were pretreated with or without 2 mM N-acetyl cysteine (NAC) or 20 M propyl gallate (PG) for 20 min and then treated with 0.5 mM H 2 O 2 for the indicated periods. The cell extracts were subjected to immunoblotting with the indicated antibodies. D, H 2 O 2dependent dissociation of Trx from ASK1. An aliquot of each cell extract that was prepared for the gel filtration in E was immunoprecipitated with anti-Trx antibody or control rabbit IgG. The samples were then analyzed by immunoblotting with the indicated antibodies. The lower three panels (Input) show the presence of ASK1 and Trx by using 1.5% of the total cell extracts that were used to perform the immunoprecipitation. E, co-immunoprecipitation of ASK1with Trx. The cell extracts from HEK293 cells were treated with or without 0.5 mM H 2 O 2 for 15 min and then prepared as described under "Materials and Methods." After fractionation on a Superose6 10/300 GL column, each fraction was immunoprecipitated with anti-Trx antibody. The samples were then analyzed by immunoblotting with the indicated antibodies. Apparent molecular mass is shown as described above. The elution positions of standard proteins are indicated at the top of the figure. WB, Western blot; IP, immunoprecipitation. cells lysate was analyzed by gel filtration, endogenous Trx was found to exist in all fractions regardless of the presence of ASK1 (Fig. 2E, second  and fourth panels). Co-immunoprecipitation analyses of individual fractions showed that the interaction of ASK1 and Trx in unstimulated cells was clearly detectable in the HMM fractions almost in parallel with the amount of ASK1 in each fractions (Fig. 2E, top panel, lanes 5-10; compare with Fig. 2A, top panel); however, Trx-ASK1 interaction was strongly reduced in H 2 O 2 -treated cells (Fig. 2E, third panel). This indicates that the H 2 O 2 -induced higher molecular mass complex of ASK1 (corresponding to lanes 1-4 of the second and third panels in Fig. 2A) does not contain Trx. These results suggest that dissociation of Trx from and reciprocal association of additional component(s) to the preexisting HMM signaling complex of ASK1 may occur during ROS-induced activation of ASK1.
TNF␣ Also Induces Higher Molecular Mass Complex Formation of ASK1-We have previously demonstrated that TRAF2 interacts with and activates ASK1 in TNF␣ signaling (5). We have also reported that ROS-dependent prior dissociation of Trx is required for the TNF␣induced TRAF2-ASK1 interaction and activation of ASK1 (6, 17). Fig.   3A shows that TNF␣ activated endogenous ASK1 in HEK293 cells within 15 min and peaked at 30 min after the treatment with TNF␣. The interaction between exogenously expressed ASK1 (ASK1-HA), and endogenous TRAF2 was also found to peak at 30 min after the treatment with TNF␣ in HEK293 cells, and ASK1 co-immunoprecipitated with TRAF2 was found to be phosphorylated as determined by immunoblotting with phospho-ASK1 antibody (Fig. 3B). When TNF␣-treated cell extract was analyzed by the gel filtration column, a shift of the ASK1containing HMM fractions toward the higher molecular mass fractions was clearly observed in a TNF␣-dependent manner (Fig. 3C). These results suggested that TNF␣ activates the static HMM ASK1 complex by recruiting additional components including TRAF2.
ASK1 Interacts with TRAF2 in a ROS-dependent Manner-The similarity between H 2 O 2 -and TNF␣-induced changes in the ASK1 signaling complex ( Figs. 2A and 3C), together with the finding that TNF␣induced activation of ASK1 depends largely on ROS (6, 17) prompted us to examine whether TRAF2 may also be involved in the mechanism of ROS-induced activation of ASK1. In support of this idea, it has recently been reported that TRAF2 is required for H 2 O 2 -induced JNK activation FIGURE 3. TNF␣ also induces higher molecular mass complex formation of the ASK1 signalosome. A, TNF␣-induced activation of endogenous ASK1. Subconfluent HEK293 cells were treated with 100 ng/ml TNF␣ for the indicated periods. The cell extracts were subjected to immunoblotting with anti-ASK1 antibody and phospho-ASK1 antibody. B, TNF␣-induced interaction between TRAF2 and ASK1. HEK293 cells were transiently transfected with ASK1WT-HA in 6-well plates. After 36 h, the cells were treated with 100 ng/ml TNF␣ for the indicated periods. The cell extracts were immunoprecipitated with anti-TRAF2 antibody followed by immunoblotting with the indicated antibodies. The lower two panels (Input) show that appropriate expression of proteins by using 2.5% of the cell extracts that was used to perform the immunoprecipitation. C, gel filtration analysis of endogenous ASK1. HEK 293 cells (1.0ϫ10 8 cells) were treated with 100 ng/ml TNF␣ for 30 min. Gel filtration analysis of endogenous ASK1 was performed as described for Fig. 2A. D, the kinetics of the ASK1 activation by H 2 O 2 . Subconfluent HEK293 cells were treated with 0.5 mM H 2 O 2 for the indicated periods. The cell extracts were subjected to immunoblotting with anti-ASK1 antibody and phospho-ASK1 antibody. E, H 2 O 2 -induced interaction of TRAF2 and ASK1. HEK293 cells were transiently transfected with ASK1WT-HA in 6-well plates. After 36 h, the cells were treated with 0.5 mM H 2 O 2 for the indicated periods. The cell extracts were immunoprecipitated with anti-TRAF2 antibody followed by immunoblotting with the indicated antibodies. The lower two panels (Input) show the appropriate expression of proteins by using 2.5% of the cell extracts that were used to perform the immunoprecipitation. WB, Western blot; IP, immunoprecipitation. and cell death (38). We therefore examined whether TRAF2 is involved in H 2 O 2 -induced activation of ASK1. In parallel with the time course of H 2 O 2 -induced activation of ASK1 (Fig. 3D), ASK1 was co-immunoprecipitated with TRAF2 in an H 2 O 2 -dependent manner, and co-immunoprecipitated ASK1 was phosphorylated and thus activated (Fig. 3E). These results strongly suggested that the molecular components responsible for H 2 O 2 -induced larger complex formation of the ASK1 signalosome include TRAF2.
TRAF2 and TRAF6 but Not TRAF5 Are Required for ROS-induced Activation of the ASK1-JNK/p38 Pathways-Among the TRAF family members, not only TRAF2 but also TRAF5 and TRAF6 activate ASK1 (5). Furthermore, our recent report demonstrated that ROS-dependent interaction of TRAF6 and ASK1 is crucial for lipopolysaccharide-induced activation of the ASK1-p38 pathway (36). We thus explored the possibility of whether TRAF2, TRAF5, and TRAF6 are involved in ROSinduced ASK1 activation. To examine the requirement of TRAFs for ROS-induced ASK1 activation, we analyzed MEFs deficient for TRAF2 (39), TRAF5 (28) and TRAF6 (29,33). Although the expression of endogenous ASK1 was relatively low in MEFs, almost equivalent expression of endogenous ASK1 in Traf2Ϫ/Ϫ, Traf5Ϫ/Ϫ, Traf6Ϫ/Ϫ, and wild-type MEFs was confirmed by immunoblotting following immunoprecipitation (Fig. 4A). H 2 O 2 -induced activation of ASK1 was strongly suppressed in TRAF2-and TRAF6-deficient but not TRAF5deficient MEFs when compared with each of genotype-matched control MEFs (Fig. 4B). Consistently, H 2 O 2 -induced activations of JNK and p38 were also inhibited in TRAF2-and TRAF6-deficient MEFs (Fig. 4, C and  D). These results suggested that TRAF2 and TRAF6 but not TRAF5 are required for H 2 O 2 -induced activation of the ASK1-JNK/p38 pathways. Nevertheless, the relatively late (60 or 90 min) responses of JNK and p38 to H 2 O 2 in TRAF2-and TRAF6-deficient MEFs were only partially impaired (Fig. 4, C and D), suggesting that H 2 O 2 evoked secondary signals, which induce TRAF2-or TRAF6-independent activations of the JNK and p38 pathways in the late phase. Impairment of ASK1 activation in TRAF2-deficient cells treated with H 2 O 2 for 90 min was also limited probably for the same reason, whereas H 2 O 2 -induced ASK1 activation was abolished throughout the time course tested in TRAF6-deficient cells (Fig. 4B). This contrast in the extent of impairment of ASK1 activation suggests that TRAF6 is more critically required for H 2 O 2 -induced ASK1 activation than TRAF2.
We further examined whether TRAF6 is required for H 2 O 2 -induced ASK1 activation in primary cultured cells. In the macrophages independently prepared from two Traf6Ϫ/Ϫ mice (30), H 2 O 2 -induced acti- show the appropriate expression of proteins by using 2.5% of the cell extracts that was used to perform the immunoprecipitation. B, endogenous interaction between ASK1 and TRAF6. An aliquot of each cell extract that was prepared for the gel filtration in Fig. 2E was immunoprecipitated with anti-TRAF6 antibody or control rabbit IgG. The samples were then analyzed by immunoblotting with the indicated antibodies. The lower three panels (Input) show the presence of ASK1 and TRAF6 by using 0.5% of the total cell extracts that were used to perform the immunoprecipitation. C-F, gel filtration analysis of WT (C-E) and Traf6Ϫ/Ϫ MEFs (F). Both MEFs were infected with adenoviruses encoding HA-tagged ASK1 WT (6). After 36 h, the cell extracts were treated with or without 0.5 mM H 2 O 2 for 15 min and then prepared as described under "Materials and Methods," and each lysate was fractionated through a Superose6 10/300 GL column. Each fraction was analyzed by immunoblotting with anti-HA antibody (C and F) or phospho-ASK1 antibody (D) or was immunoprecipitated with anti-TRAF6 antibody followed by immunoblotting with anti-HA antibody (E). The apparent molecular mass is shown as described above. The elution positions of standard proteins are indicated at the top of the figure. WB, Western blot; IP, immunoprecipitation.
vation of ASK1-JNK/p38 pathways were strongly suppressed as was seen in TRAF6-deficient MEFs (Fig. 4E). On the other hand, activation of ASK1 by the treatment with thapsigargin, which triggers endoplasmic reticulum stress by depletion of lumenal calcium stores (4), was not impaired in Traf6Ϫ/Ϫ macrophages (Fig. 4F). Taken together with our previous finding that TRAF2 and TRAF6 activates ASK1 (5), TRAF2 and TRAF6 appear to be crucial components of the H 2 O 2 -induced activation of ASK1.
TRAF6 Is Required for ROS-induced Higher Molecular Mass Complex of ASK1-We next examined whether TRAF6-ASK1 interaction can be induced by H 2 O 2 . Transfected ASK1 was co-immunoprecipitated with TRAF6 in an H 2 O 2 -dependent manner, and co-immunoprecipitated ASK1 with TRAF6 was phosphorylated (Fig. 5A). H 2 O 2 -dependent endogenous interaction of TRAF6 and ASK1 was also readily detectable in HEK293 cells (Fig. 5B). These results suggested that TRAF2 and TRAF6 are likely candidates for the components of H 2 O 2 -induced higher molecular mass complex of ASK1. To confirm this possibility, MEFs transiently transfected with HA-tagged ASK1WT were treated or untreated with H 2 O 2 and subjected to the gel filtration analysis as determined using HEK293 cells in Fig. 1C. Fig. 5C shows that H 2 O 2 induced the shift of ASK1-containing fractions to the higher molecular mass fractions also in MEFs. Immunoblotting analysis of the same membrane with the phospho-ASK1 antibody revealed that H 2 O 2 -induced active form of ASK1 was more abundant in the higher molecular mass complex (Fig. 5D, lower panel, lanes 1-4). Furthermore, immunoprecipitation analysis of each fraction revealed that the H 2 O 2 -dependent complex formations of TRAF6 and ASK1 were more evident in the higher molecular fractions (fractions 1 and 2) (Fig. 5E). These results indicate that H 2 O 2 -dependent higher molecular mass complex is the active complex of ASK1 and that the recruitment of TRAF6 is responsible at least in part for H 2 O 2 -dependent formation of higher molecular mass complex. We have not been successful in detecting TRAF2 in the H 2 O 2dependent higher molecular mass complex, probably because of the low affinity of TRAF2 antibody or the lesser amount of endogenous TRAF2 compared with TRAF6 in MEFs (data not shown). To examine the requirement of TRAF6 for H 2 O 2 -dependent formation of the higher molecular mass complex, TRAF6-deficient MEFs were analyzed. We could not detect H 2 O 2 -dependent shift of the ASK1 signalosome in Traf6Ϫ/Ϫ MEFs (Fig. 5F). These results suggested that recruitment of TRAF6 is required for H 2 O 2 -induced higher molecular mass complex formation of the ASK1 signalosome.
Different Susceptibilities of Traf2Ϫ/Ϫ, Traf5Ϫ/Ϫ, and Traf6Ϫ/Ϫ MEFs to H 2 O 2 -induced Cell Death-Given that ASK1 is required for H 2 O 2 -induced cell death (8), it is formally possible that TRAF2 and TRAF6 are also involved in such cell death. We therefore investigated the susceptibility of Traf2Ϫ/Ϫ and Traf6Ϫ/Ϫ MEFs to H 2 O 2 -induced cell death by trypan blue dye exclusion assay (Fig. 6). Traf2Ϫ/Ϫ MEFs were apparently resistant to H 2 O 2 -induced cell death as recently reported (38), whereas wild-type and Traf5Ϫ/Ϫ MEFs were similarly vulnerable to H 2 O 2 treatment. Traf6Ϫ/Ϫ MEFs were even more resistant than Traf2Ϫ/Ϫ MEFs to H 2 O 2 -induced cell death, correlating well with the results that H 2 O 2 -induced activation of the ASK1-JNK/p38 pathways was more severely impaired in Traf6Ϫ/Ϫ than Traf2Ϫ/Ϫ MEFs (Fig. 4, C and D). These results demonstrated that TRAF2 and TRAF6 are novel components of the ROS-induced ASK1 signaling pathway and play an essential role in ROS-induced cell death.

DISCUSSION
The importance of the so called signalosome has been emerging, in that cells may take advantage of the preformed multi-molecular com-plex to commence rapid as well as specific signaling events in response to the cellular input. For example, the IB kinase (IKKs) complex, a signalosome composed of IKK␣, IKK␤, and NEMO (IKK␥), forms an HMM complex and plays an essential role in NF-B activation (40 -45). Activations of certain caspases were also reported to require HMM complex formation, such as apoptosome for caspase-9 (37, 46) and inflammasome for caspase-1 and -5 (47)(48)(49). Apaf-1 and NALP1, the members of the nucleotide binding-site family, are the specific components to assemble large caspase complexes for apoptosome and inflammasome, respectively. Apoptosome, for example, which is formed by the recruitment of caspase-9 to Apaf-1, brings several caspase-9 molecules into close proximity, resulting in their efficient cross-activation (37,46). These findings validate the comprehensive analysis of ASK1 signalosome to understand the ASK1-dependent signaling mechanisms of oxidative stress-induced cell death.
In the present study, we found that ASK1 forms a static HMM complex composed of homo-oligomeric ASK1, Trx, and yet unidentified component(s) in unstimulated cells. We also found that oxidative stressdependent dissociation of Trx and reciprocal recruitment of TRAF2 and TARF6 to the ASK1 signaling complex are required for the activation of ASK1. Because H 2 O 2 -induced activation of ASK1 was almost completely abolished in Traf6Ϫ/Ϫ cells (Fig. 4, B and E), TRAF6 appears to be the most important among TRAF family proteins for the activation of ASK1. However, how TRAFs mechanistically contribute to the activation of ASK1 is yet to be elucidated. In this regard, two possibilities, physical and/or chemical modifications of ASK1, might apply. Because TRAF family proteins appear to form trimers in vivo (50 -52), physical association of TRAF6 may facilitate multimeric complex formation, resulting in close proximity of ASK1 molecules and thereby autophosphorylation-dependent activation. Although ASK1⌬coil, a deletion mutant of the C-terminal oligomerization domain, exhibited very low kinase activity (Fig. 1A), it was shown that forced homo-oligomerization of ASK1⌬coil by chemical oligomerizer induced kinase activation of this mutant (11), suggesting that a close proximity of ASK1 molecules is sufficient to activate ASK1. Nevertheless, the endogenous ASK1 signaling complex appears to be present as a homo-oligomerized but still inactive form (Fig. 2). These findings suggest that the homooligomerization of ASK1 through the C-terminal coiled-coil domain is not sufficient for ASK1 activation. Thus, the endogenously homo-oligomerized ASK1 is likely to be kept inactive by the presence of Trx, a "safety lock" for ASK1. Dissociation of Trx from ASK1 signaling complex appears to allow the association of TRAF2 and TRAF6 to ASK1 and thereby induce the close proximity-dependent conformational changes in the interface of the kinase domain of ASK1. The tight oligomerization assisted by the associated molecules may be physiologically necessary and sufficient for the autophosphorylation-dependent activation of ASK1. Activation status of the ASK1 signaling complex in oxidative stress response may thus be switched by reciprocal changes in the components of the signalosome; from ASK1-Trx signaling complex (inactive form) to the ASK1-TRAF2/TRAF6 signaling complex (active form) (Fig. 7).
Another possible mechanism by which TRAFs activate ASK1 may include chemical modifications, such as ubiquitination. TRAF6 is a RING domain-containing ubiquitin ligase that synthesizes Lys 63 -linked polyubiquitin chains (53). The targets of Lys 63 -linked polyubiquitination include NEMO, TRAF2, and TRAF6 itself, and the ubiquitinated TRAF6 is implicated in the activation of downstream effectors such as NEMO and TAK1 (54,55). Similar implications of TRAF2 are also reported (56,57). It was also reported that RIP1 is ubiquitinated by TRAF2 in a TNF␣-dependent manner, which induces the activation of NF-B and p38 MAP kinase (58). Although the role of Lys 63 -linked polyubiquitination remains unclear, ASK1 might require the Lys 63linked polyubiquitination by TRAF2 or TRAF6 for the kinase activation.
Thus, further studies are required for the elucidation of mechanisms by which the TRAF family proteins activate ASK1. Uncovering the unidentified components in the ASK1 signaling complex may reveal precise mechanisms of the ASK1-dependent signaling cascade that links oxidative stress and cell death.