Reactive Oxygen Species- and Dimerization-induced Activation of Apoptosis Signal-regulating Kinase 1 in Tumor Necrosis Factor-α Signal Transduction*

Reactive oxygen species (ROS) have been implicated in the induction of apoptosis by tumor necrosis factor-α (TNFα) and other cytotoxic insults, although the molecule(s) regulated by ROS in TNFα signaling have not been identified. Apoptosis signal-regulating kinase 1 (ASK1) is a member of the mitogen-activated protein kinase kinase kinase (MAPKKK) superfamily that has been shown to be activated during TNFα-induced apoptosis. ASK1 increases apoptosis when overexpressed, but the mechanism of ASK1 activation and the mechanisms of ASK1-induced apoptosis are unclear. We now report that hydrogen peroxide induces the activation of ASK1 in 293 cells. TNFα-induced activation of ASK1 was inhibited by antioxidants. Hydrogen peroxide-induced apoptosis was markedly enhanced by the expression of ASK1. These results suggest that TNFα-induced activation of ASK1 is mediated by ROS. We also examined how ASK1 activity is regulated by ROS. We found that ASK1 formed dimers or higher order oligomers in 293 cells. TNFα or hydrogen peroxide treatment increased the dimeric form of ASK1, and pretreatment withN-acetylcysteine decreased it. Furthermore, synthetic dimerization of an ASK1-gyrase B fusion protein by coumermycin resulted in substantial activation of ASK1, suggesting that dimerization of ASK1 is sufficient for its activation. These results taken together suggest that TNFα causes ASK1 activation via ROS-mediated dimerization of ASK1.

Reactive oxygen species (ROS) have been implicated in the induction of apoptosis by tumor necrosis factor-␣ (TNF␣) and other cytotoxic insults, although the molecule(s) regulated by ROS in TNF␣ signaling have not been identified. Apoptosis signal-regulating kinase 1 (ASK1) is a member of the mitogen-activated protein kinase kinase kinase (MAPKKK) superfamily that has been shown to be activated during TNF␣-induced apoptosis. ASK1 increases apoptosis when overexpressed, but the mechanism of ASK1 activation and the mechanisms of ASK1-induced apoptosis are unclear. We now report that hydrogen peroxide induces the activation of ASK1 in 293 cells. TNF␣-induced activation of ASK1 was inhibited by antioxidants. Hydrogen peroxide-induced apoptosis was markedly enhanced by the expression of ASK1. These results suggest that TNF␣-induced activation of ASK1 is mediated by ROS. We also examined how ASK1 activity is regulated by ROS. We found that ASK1 formed dimers or higher order oligomers in 293 cells. TNF␣ or hydrogen peroxide treatment increased the dimeric form of ASK1, and pretreatment with N-acetylcysteine decreased it. Furthermore, synthetic dimerization of an ASK1-gyrase B fusion protein by coumermycin resulted in substantial activation of ASK1, suggesting that dimerization of ASK1 is sufficient for its activation. These results taken together suggest that TNF␣ causes ASK1 activation via ROS-mediated dimerization of ASK1.
The mitogen-activated protein kinases (MAPKs) 1 participate in intracellular signaling pathways that are activated by extracellular stimuli in a variety of cell types (1)(2)(3). In mammalian cells, there are at least three distinct members of the MAPK family: MAPK (also known as extracellular signal-regulated kinase), stress-activated protein kinase (SAPK) (also known as c-Jun NH 2 -terminal kinase (JNK)), and p38. The "classical" MAPK is known to be activated and involved in growth factor-mediated cell proliferation and differentiation (1,3). In contrast, SAPK and p38 are activated in response to proinflammatory cytokines and environmental stresses such as UV and osmotic shock and are implicated in inflammatory responses, cell cycle arrest, DNA repair, and apoptosis (3)(4)(5).
Considerable progress has been made in identifying the upstream kinases that regulate each member of the MAPK family. The central core of each MAPK pathway is a conserved cascade of three protein kinases, which are commonly referred to as MAPK, MAPK kinase (MAPKK), and MAPKK kinase (MAPKKK) (1,2 (3,6). MAPKs are activated by dual phosphorylation on Thr and Tyr residues catalyzed by MAPKKs, which in turn are activated by dual phosphorylation on Ser and Ser/Thr residues catalyzed by MAPKKKs. The activation of the MAPKKKs in the classical MAPK pathway has been studied extensively. The activation of Raf-1 has been shown to involve Ras-mediated membrane recruitment, phosphorylation at critical sites (7), and possibly dimerization (8,9). However, the activation mechanism of MAPKKKs in the "stress-activated" MAPK pathways remains largely unknown and is one of the critical unanswered questions in signal transduction. ASK1 (also called MAPKKK5) is a member of the MAPKKK family, which can activate MKK4 and MKK7 (SAPK pathway) and MKK3 and MKK6 (p38 pathway) by direct phosphorylation (10,11). Several lines of evidence suggest that ASK1 is involved in TNF␣-induced apoptosis signaling. First, ASK1 is activated by TNF␣ treatment in many cells; second, the overexpression of ASK1 induces apoptosis in stably transfected Mv1Lu cells; and third, the expression of a dominant-negative ASK1 inhibits TNF␣-induced apoptosis in Jurkat cells (11). However, it remains to be determined how ASK1 is regulated by TNF␣ signaling.
TNF␣ is a multifunctional proinflammatory cytokine, which induces apoptosis or other responses depending on the cellular context. The effects of TNF␣ are mediated by cell surface receptors (p55 TNF-R1 and p75 TNF-R2), their adaptor proteins, and intracellular messengers (12,13). Among them, ROS have been proposed to act as signaling intermediates for TNF␣induced apoptosis, because they are produced in response to TNF␣ stimulation (14 -17) and because antioxidants inhibit TNF␣-induced cell death (18 -28). Furthermore, sensitivity to TNF␣ is inversely correlated with the levels of superoxide dismutase (29,30) and glutathione (28,31), the most abundant antioxidants in cells. These studies prompted us to examine whether ASK1 is regulated by ROS in TNF␣ signaling. We have found that ROS can activate ASK1. We further examined the molecular mechanism of ASK1 activation induced by ROS and found that dimeric forms of ASK1 exist in ROS-stimulated cells and that in vivo dimerization of ASK1 is sufficient for its activation. Thus, we provide the first evidence that dimerization is important for activating an enzyme that regulates the stress-activated MAPK pathways.
Expression Vectors and Transfections-pcDNA3-ASK1, pME18-HA-MKK6, and pSR␣-HA1-SAPK were described previously (11). pME18-HA-ASK1 was constructed by ligating a SalI-NotI fragment containing ASK1 into pME18-HA. pME18-Myc-ASK1 was constructed by ligating ASK1 cDNA into the SalI site of pME18-Myc. For constructing this plasmid, SalI sites were created at both ends of ASK1 by polymerase chain reaction-amplifying it with the primers 5Ј-CCGCGGTCGAC-GACTGTGAAAGTGACTTGCT-3Ј and 5Ј-GGCCGTCGACCCAGTCT-GTTTGTTTCGAAAGTC-3Ј. The modified gyrase B vector, which has a translation initiation site, gyrase B, and a linker for fusion protein with Bluescript KS backbone, was generously provided by Drs. Michael A. Farrar and Roger M. Perlmutter (University of Washington) (8). SpeI sites were added to both ends of ASK1 by polymerase chain reactionamplifying the full-length ASK1 cDNA with the primers 5Ј-GGGGAC-TAGTGCGGCCGCTCAAGTCTGTTTGTTTCGAAA-3Ј and 5Ј-GGG-GACTAGTATGAGCACGGAGGCGGACGAGGGCATCAC-3Ј. The amplified fragment was inserted into the SpeI site of the modified gyrase B vector. The gyrase B-linker-ASK1 fragment was cut out with NotI and inserted into NotI site of pcDNA3 (Invitrogen). For control experiments, gyrase B without ASK1 was also introduced into pcDNA3. For transfections, 7 ϫ 10 5 cells in 60-mm dishes were transfected with 1-5 g of DNA in total by using lipofectamine (Life Technologies, Inc.) according to the manufacturer's instructions.
Chemical Cross-linking-Cells (60-mm dish) were washed with phosphate-buffered saline and homogenized with 150 l of phosphate-buffered saline containing 1 mM dithiothreitol and 1 mM phenylmethylsulfonyl fluoride. After centrifugation at 10,000 ϫ g for 3 min, 50 l of the supernatants were incubated with 10 mM EDC for 10 min at 30°C. To stop the reaction, EDC was removed by spin column (ProbeQuant G-50 MicroColumns; Amersham Pharmacia Biotech), and the eluates were boiled with Laemmli sample buffer.
Apoptosis Assay-The terminal deoxynucleotidyltransferase end labeling assay was carried out using the In Situ Cell Death Detection Kit (Boehringer Mannheim) according to the manufacturer's instructions.
ROS Assay-Intracellular ROS production was measured by the method of Bass et al. (34) as modified by Bae et al. (35). Briefly, dishes treated with or without TNF␣ for 10 min were washed with Earle's balanced salt solution without phenol red and incubated in the dark for 5 min in the presence of 5 M dichlorodihydrofluorescein diacetate. When dichlorodihydrofluorescein diacetate is oxidized within the cell, it becomes a highly fluorescent 2Ј7Ј-dichlorofluorescein and can be detected (excitation, 488 nm; emission, 515-540 nm).

RESULTS
To study the activation mechanism of ASK1 in response to TNF␣ stimulation, we first examined whether ceramide can activate ASK1. Ceramide is known to be generated in response to TNF␣ and can induce apoptosis in U937, lymphoblastoid, and endothelial cells (36 -38). Human embryonic kidney 293 cells were transfected with HA-tagged ASK1 and treated with a membrane-permeable ceramide analogue, C 2 -ceramide. HA-ASK1 was immunoprecipitated, and its activity was determined by a coupled kinase assay using recombinant MKK6 and a kinase-negative p38. As shown in Fig. 1A, the C 2 -ceramide treatment had little effect on ASK1 activity (ϳ1.3-fold activation). However, the same concentration of C 2 -ceramide induced apoptosis in 293 cells (data not shown), suggesting that ceramide may not be a major mediator of ASK1 activation.
TNF␣ treatment increases ROS, such as H 2 O 2 , that are thought to be essential for TNF␣-induced apoptosis (see Introduction). In fact, the ROS level measured by dichlorodihydrofluorescein diacetate was increased after 15 min of TNF␣ treatment in Mv1Lu cells (5.8-fold on average) in which ASK1 activation is highest at 30 min after TNF␣ treatment (11). Therefore, it is possible that ROS is involved in ASK1 activation. H 2 O 2 treatment induced apoptosis in a dose-dependent manner (EC 50 ϳ 50 M) as determined by the appearance of apoptotic bodies and nuclear condensation in 293 cells (data not shown). Importantly, H 2 O 2 treatment significantly increased ASK1 activity in 293 cells transiently transfected with HA-ASK1 (Fig. 1A). Activation of ASK1 by H 2 O 2 was detected not only by a coupled kinase assay but also by an autophosphorylation assay for ASK1 (Fig. 1B). The doses of H 2 O 2 required for the activation of ASK1 (EC 50 ϳ30 M) were approximately the same as those required for inducing apoptosis (Fig. 1B). The activation of ASK1 by H 2 O 2 treatment was sustained for at least 2 h (Fig. 1C). Endogenous ASK1 was also activated by H 2 O 2 treatment as determined by immunoprecipitation with an anti-ASK1 antibody (Fig. 1D). In contrast, treatment with 10 g/ml anisomycin did not increase ASK1 activity (Fig. 1A), although anisomycin is a strong activator of the SAPK and p38 pathways. This implies that ASK1 is only activated by a subset of the stimuli that activate p38 and SAPK. In fact, it is not clear that ASK1 is a major MAPKKK for p38 and SAPK in TNF signaling.
Activation of ASK1 by either TNF␣ or H 2 O 2 was 3-5-fold compared with basal level (Fig. 1). The -fold activation of MAP-KKKs is generally low compared with that of MAPKs. A model proposed by Ferrell might explain why a small change in activity of a MAPKKK can trigger robust activation of a MAPK; the requirement for dual phosphorylation to activate the MAPKK and MAPK, the three-kinase cascade (MAPKKK-MAPKK-MAPK), and near saturation of the enzymes produce multistep cooperativity or "ultrasensitivity" and thus create a switch-like response from a graded stimulus (39).
Since H 2 O 2 activated ASK1 in the cell, we next examined if ROS are required for TNF␣-induced activation of ASK1. We pretreated 293 cells with an antioxidant, NAC, and then treated the cells with TNF␣. In the presence of 20 mM NAC, the activation of ASK1 by TNF␣ treatment was markedly inhibited, as shown in Fig. 2. The basal activity of ASK1 was also reduced by NAC treatment (ϳ30% reduction) (Fig. 2). In addition to NAC, catalase (3000 units/ml) (a peroxide-scavenging enzyme) and pyrrolidine dithiocarbamate (100 M) reduced the TNF␣-induced activation of ASK1 (data not shown; Fig. 2). These results suggest that the production of ROS is essential for ASK1 activation in TNF␣ signaling.
Since ASK1 was activated by H 2 O 2 and ASK1 has been implicated in apoptosis signaling, we tested whether ASK1 expression enhances H 2 O 2 -induced apoptosis. ASK1 was inducibly expressed in Mv1Lu cells stably transfected with an ASK1 construct driven by the metallothionein promoter (11). Only a small fraction of cells grown in serum undergo apoptosis by ASK1 expression (Fig. 3). 2   ASK1 was immunoprecipitated with anti-HA, and its activity was measured either by a coupled kinase assay using His-MKK6 and Hiskinase-negative p38 (KN-p38) (A and C) or by an autophosphorylation assay (B). D, after 293 cells were treated with or without 100 M H 2 O 2 for 10 min, endogenous ASK1 was immunoprecipitated with a rabbit anti-ASK1 antiserum (11), and its kinase activity was measured by a coupled kinase assay. The activities were shown in arbitrary units. The results are representative of at least three independent experiments. In B, the bottom is based on five independent experiments, and the top is a representative result. went apoptosis in the Mv1Lu cells transfected with an empty vector (Fig. 3). More than 200 M H 2 O 2 was required to induce apoptosis in the latter cells. Therefore, ASK1 synergized with H 2 O 2 in inducing apoptosis, which is consistent with the hypothesis that ASK1 plays a role in ROS-mediated apoptosis signaling. A kinase-dead mutant (K709R) ASK1 did not inhibit H 2 O 2 -induced apoptosis in 293 cells (data not shown), suggesting that ASK1 is not the only target for H 2 O 2 in inducing apoptosis in 293 cells.
We then investigated the molecular mechanism of ASK1 activation induced by ROS. Because Raf1, another member of the MAPKKK superfamily, has recently been reported to be activated by dimerization, we examined whether ASK1 also forms a dimer. First we carried out chemical cross-linking experiments using EDC, which cross-links between carboxyl groups and amine groups. After treatment with or without EDC, the cell lysates were subjected to immunoblotting with anti-HA to determine the mobility of HA-ASK1. In addition to the monomeric HA-ASK1 (ϳ190 kDa), some bands at the position around 350 kDa were detected only in the lysate incubated with EDC (Fig. 4). The mobility of this band appeared to correspond to the molecular weight of a dimer of ASK1. To confirm that ASK1 forms a dimer, Myc-tagged ASK1 and HA-tagged ASK1 were either co-expressed or expressed separately in 293 cells to see if they would coprecipitate. Myc-tagged ASK1 was detected in anti-HA immunoprecipitates from the cells expressing both HA-and Myc-tagged ASK1 but not from the cells expressing only either Myc-or HA-ASK1 (Fig. 5). In the converse experiment, HA-tagged ASK1 was detected in anti-Myc immunoprecipitates only from the cells expressing both HAand Myc-ASK1 (Fig. 5). These results suggest that ASK1 either forms dimers or higher order oligomers.
Importantly, both TNF␣ and H 2 O 2 treatment moderately increased the amount of ASK1 at the dimer position (Fig. 4). Moreover, pretreatment with NAC markedly reduced the amount of ASK1 at the dimer position (Fig. 4). Therefore, TNF␣ and its downstream mediator H 2 O 2 may promote dimerization of ASK1 within the cell.
The next important question is whether ASK1 can be activated by dimerization. To address this, we produced a fusion protein between gyrase B and ASK1 to determine if dimerization of ASK1 is sufficient for its activation. The application of coumermycin to cells induces the dimerization of gyrase B, but novobiocin, a monovalent analogue of coumermycin, does not ( Ref. 8; Fig. 6A). Activation of ASK1 within the cell was monitored by the activity of a co-transfected substrate, MKK6. Treatment of the gyrase B-ASK1-transfected 293 cells with coumermycin induced activation of ASK1 efficiently (Fig. 6B). In contrast, coumermycin had little effect on MKK6 activity in cells transfected with gyrase B without ASK1 (Fig. 6B). In addition, novobiocin did not induce gyrase B-ASK1 activation (Fig. 6B). The activation of gyrase B-ASK1 by coumermycin was also detected when the activity of ASK1 was monitored by the activity of another co-transfected substrate, SAPK (JNK) (Fig. 6C). These results strongly suggest that dimerization of ASK1 is sufficient for its activation. DISCUSSION Very little is known about the activation mechanism of the MAPKKKs in the stress-activated MAPK pathways, although the activation and the functions of those pathways have been studied extensively (4). In this report, we have investigated the activation mechanism of ASK1, a MAPKKK that is involved in TNF␣-induced apoptosis. Since both ROS and ASK1 are thought to be important for propagating cytotoxic effects of TNF␣, we examined the relationship between ROS and ASK1. We propose that ROS, especially H 2 O 2 , are involved in TNF␣induced ASK1 activation based on the following evidence: 1) TNF␣ induces both H 2 O 2 production and ASK1 activation (11,16,17), 2) H 2 O 2 treatment increases ASK1 activity (this study), 3) antioxidant treatment antagonizes TNF␣-induced ASK1 activation (this study), and 4) H 2 O 2 treatment and ASK1 expression synergize in inducing apoptosis (this study). These results are consistent with the idea that ASK1 is activated downstream of H 2 O 2 in TNF␣ signaling. Therefore, ASK1 as well as other potential ROS targets (direct or indirect; cf. NF-B, AP-1, MAPK, phospholipase A2, protein kinase C, SAPK, p38, Lck, and protein-tyrosine phosphatases (40 -48)) might participate in mediating ROS-induced cellular effects.
We further examined the molecular mechanism of ASK1 activation. The results presented here demonstrate that ASK1 dimerizes in vivo. In addition, we have shown here that 1) treatment of the cells with either H 2 O 2 or TNF␣ increased the dimeric form of ASK1, 2) NAC treatment decreased the dimeric form, and 3) synthetic dimerization of ASK1 by a gyrase Bcoumermycin system was sufficient for ASK1 activation in the cells. These data strongly suggest that dimerization of ASK1 plays a role in its activation. This hypothesis is consistent with the fact that overexpression of ASK1 leads to a high level of basal activity in the absence of stimulation (11). Also, it is possible that immunoprecipitation artificially increases ASK1 activity by aiding dimer formation. Therefore, it would be good to develop a method to assay ASK1 without immunoprecipitation. It still remains to be addressed whether mechanisms other than dimerization are involved in ASK1 activation and whether the dimerization of ASK1 is required for its activation.
It is not clear whether ASK1 is a direct or indirect target for ROS. In vitro incubation of an ASK1 immunoprecipitate with H 2 O 2 did not increase its activity. 3 However, the altered oligomerization of ASK1 induced by H 2 O 2 and TNF␣ may be due to changes in sulfhydryl oxidation in ASK1. When we carried out anti-HA immunoblotting of extracts prepared from HA-ASK1-expressing cells under nonreducing conditions, we observed extra bands that were not detected in extracts prepared under standard reducing conditions. 3 This might suggest that ASK1 is directly modulated by ROS. Interestingly, ASK1 has a cysteine-rich domain in its NH 2 terminus, which might be a target for the action of ROS. In future studies, it will be important to determine whether this domain is involved in the dimerization and activation of ASK1.
How is ASK1 activated by dimerization? The dimerization might result in a conformational change that either renders ASK1 active or allows ASK1 to be activated by upstream activator(s). Alternatively, the dimerization could result in intermolecular autophosphorylation that is involved in activation, as is the case for receptor-tyrosine kinases. However, it is not clear thus far whether phosphorylation is important for ASK1 activation.
Previous papers have shown that Raf1 is also activated by synthetic dimerization using either the gyrase B-coumermycin system or the FKBP-FK1012A system (8,9). Raf1 is also known to be activated by membrane recruitment by active Ras and subsequent phosphorylation by unknown kinase(s) (7). It is not clear if membrane recruitment and phosphorylation are required for Raf1 dimerization and vice versa. In contrast to Raf1, ASK1 is mostly localized in the cytoplasm even after stimulation with H 2 O 2 , 3 so membrane recruitment is unlikely to be a prerequisite for activation of ASK1.
It is possible that binding protein(s) and phosphorylation are also involved in the activation of MAPKKKs in the stressactivated MAPK pathways. STE20, a yeast Ser/Thr kinase, is genetically upstream of STE11, a yeast MAPKKK. There exist a number of mammalian STE20 homologues such as PAKs, MST1/2/3, HPK1, SOK1, GCK, NIK, LOK, and KHS (3). SOK1 is activated by H 2 O 2 (49), so it would be interesting to see if SOK1 can phosphorylate ASK1. Binding of the STE20-like kinases to MAPKKKs has been shown in some cases such as HPK1-MEKK1, NIK-MEKK1, and HPK1-MLK3 (50 -52), but there is no evidence that STE20-like kinases can directly activate MAPKKKs. It is plausible that some MAPKKKs are downstream of the STE20-like kinases, since dominant-negative mutants of MEKK1, MLK3, or TAK1 have been reported to inhibit HPK1-or GCK-induced SAPK activation (10,50,52). In addition, MAPKKKs may be activated without STE20-like kinases. Recent papers have reported direct binding of Rac1 and Cdc42 to MEKK1, MEKK4, and MLK3 (53,54), and this event may regulate activity. A newly identified protein, TAB1, binds 3 Y. Gotoh, unpublished observation. to and activates TAK1 (55), but the precise activation mechanisms have not been determined.
In the present study, we have provided the first evidence that a stress-activated MAPKKK can be activated by dimerization. In addition to ASK1, other MAPKKKs such as MEKK1, MLK3, and TAK1 are highly active without activation/mutation when they are overexpressed in mammalian cells (49,50), suggesting that they may also be stimulated by oligomerization. Furthermore, it is known that a yeast MAPK cascade, STE20-STE11-STE7-FUS3/KSS1, is activated by dimerization of a tethering protein STE5 and its activator STE4 (see Ref. 56 and references therein). Therefore, it is an intriguing possibility that dimerization/multimerization is a common mechanism for activation of some of the MAPKKK class of kinases.