Caspase Inhibition Sensitizes Inhibitor of NF-κB Kinase β-deficient Fibroblasts to Caspase-independent Cell Death via the Generation of Reactive Oxygen Species*

Cells lacking functional NF-κB die after ligation of some tumor necrosis factor (TNF) receptor family members through failure to express NF-κB-dependent anti-apoptotic genes. NF-κB activation requires the IκB kinase (IKK) complex containing two catalytic subunits named IKKα and IKKβ that regulate distinct NF-κB pathways. IKKβ is critical for classical signaling that induces pro-inflammatory and anti-apoptotic gene profiles, whereas IKKα regulates the non-canonical pathway involved in lymphoid organogenesis and B-cell development. To determine whether IKKα and IKKβ differentially function in rescuing cells from death induced by activators of the classical and non-canonical pathways, we analyzed death after ligation of the TNF and lymphotoxin-β receptors, respectively. Using murine embryonic fibroblasts (MEFs) lacking each of the IKKs, the caspase inhibitor benzyloxycarbonyl-Val-Ala-Asp-fluoromethyl ketone, and dominant negative Fas-associated death domain protein, we found that deletion of these kinases sensitized MEFs to distinct cell death pathways. MEFs lacking IKKα were sensitized to death in response to both cytokines that was entirely caspase-dependent, demonstrating that IKKα functions in this process. Surprisingly, death of IKKβ–/– MEFs was not blocked by caspase inhibition, demonstrating that IKKβ negatively regulates caspase-independent cell death (CICD). CICD was strongly activated by both TNF and lymphotoxin-β receptor ligation in IKKβ–/– MEFs and was accompanied by loss of mitochondrial membrane potential and the generation of reactive oxygen species. CICD was inhibited by the anti-oxidant butylated hydroxyanosole and overexpression of Bcl-2, neither of which blocked caspase-dependent apoptosis. Our findings, therefore, demonstrate that both IKKα and IKKβ regulate cytokine-induced apoptosis, and IKKβ additionally represses reactive oxygen species- and mitochondrial-dependent CICD.

Cytokines of the TNF 2 superfamily induce pro-inflammatory responses in many cell types via activation of the NF-B family of transcription factors (1). In addition to inducing pro-inflammatory gene transcription, NF-B also regulates the transcriptional activation of many anti-apoptotic genes including the caspase-8 inhibitor c-FLIP, the caspase inhibitors c-IAP1 and -2, the Bcl-2 homologue A1, and the signal inhibitor A20 (2)(3)(4). Induction of this anti-apoptotic gene profile is pivotal for the prevention of programmed cell death initiated in response to TNF and ligation of other death receptors.
Ligation of the TNF receptor 1 (TNFR1) by TNF initiates its ligand-induced trimerization leading to the recruitment of adapter proteins named TNFR-1-associated death domain protein and Fas-associated death domain protein (FADD). FADD initiates the recruitment and autocatalytic activation of procaspase-8 via its death effector domain, or alternatively, it may bind the caspase-8 inhibitor c-FLIP, thereby preventing the activation of this protease (5,6). In the absence of NF-B activation and induction of c-FLIP, caspase-8 activation results in proteolytic activation of effector caspases including caspase-3. This process occurs either directly or via a mitochondrial amplification process dependent upon the cleavage of Bcl-2 homology domain (BH)3-only-containing proteins such as Bid that bind and oligomerize the "multidomain" proteins Bax or Bak (7,8). This binding results in mitochondrial permeability, release of cytochrome c, and activation of the apoptosome.
The mechanisms through which the TNF family member lymphotoxin-␤ (LT␤) activates NF-B and initiates apoptosis are less well defined. The LT␤ receptor does not contain a death domain; however, the death process is dependent upon selfassociation of the receptor, TRAF3, TRAF2, c-IAP1, and SMAC (9 -12). It is clear, however, that for both TNF receptor-and LT␤ receptor-induced signaling, the precise series of biochemical events upstream of NF-B activation is critical for both the control of inflammation and the regulation of apoptosis.
NF-B proteins are retained inactive in the cytosol through interaction with members of the IB family of proteins typified by IB␣ (1). Activation of NF-B requires the signal-induced phosphorylation, ubiquitination, and degradation of IBs, thereby allowing NF-B to translocate to the nucleus. IB phosphorylation is facilitated by the IB kinase (IKK) complex that contains two catalytic subunits named IKK␣ and IKK␤ and a regulatory subunit named NEMO (NF-B essential modulator) (13). Genetic studies manipulating components of the IKK complex revealed that NF-B activation induced by the rapid and transient degradation of IB family members in response to pro-inflammatory cytokines is dependent upon IKK␤ and NEMO (13). This "classical" NF-B pathway is largely independent of IKK␣ (14). In contrast, IKK␣ is critical for activation of the "non-canonical" NF-B pathway that involves IKK␣-dependent processing of NF-B2/p100 to generate p52 (15,16). Activation of the non-canonical pathway occurs in response to a limited number of ligands including lymphotoxin ␣1␤2 (LT␣1␤2), CD40L, and BAFF and functions primarily in the maturation of B cells and the development of organized secondary lymphoid tissue (14). In addition to its role in the noncanonical pathway, IKK␣ functions in the nucleus to regulate NF-B-dependent target genes by a number of mechanisms including promoter-associated histone H3 phosphorylation (17,18). A kinase-independent role for IKK␣ in epidermal differentiation (19) and a negative-regulatory role in controlling the function of IKK␤ (20) have also been described.
Current evidence, therefore, indicates that signals induced by distinct cytokines utilize separate components of the IKK complex. In this regard TNF strongly activates the classical pathway but does not induce non-canonical signaling, suggesting that this cytokine predominately activates IKK␤. However, LT␣1␤2 only moderately activates the classical pathway but strongly stimulates the non-canonical pathway, suggesting that it predominantly signals via IKK␣. We, therefore, sought to determine whether IKK␣ and IKK␤ play separate or overlapping roles in regulating apoptosis in response to TNF and LT␣1␤2. Using IKK␣ Ϫ/Ϫ and IKK␤ Ϫ/Ϫ mouse embryonic fibroblasts (MEFs), the broad range caspase inhibitor zVADfmk, and transduction with dominant negative FADD to block receptor-induced caspase activation, we found that deletion of these kinases sensitized MEFs to distinct cell death pathways. Thus, MEFs lacking IKK␣ were sensitized to death that was entirely caspase-mediated, whereas the death of IKK␤ Ϫ/Ϫ cells was not blocked by caspase inhibition. Interestingly, we did not observe similar caspase-independent cell death (CICD) in NEMO-deficient cells, demonstrating that IKK␤ normally represses CICD in a NEMO-independent manner. We further found that CICD in IKK␤ Ϫ/Ϫ MEFs was dependent upon the generation of reactive oxygen species (ROS) and could be blocked by overexpression of Bcl-2. Our findings, therefore, demonstrate that both IKK␣ and IKK␤ regulate cytokine-induced apoptosis, and IKK␤ additionally represses ROS and mitochondrial-dependent CICD.

EXPERIMENTAL PROCEDURES
Materials-Alexa Fluor 488-conjugated annexin V, propidium iodide, JC-1, and H2DCFDA were all purchased from Molecular Probes (Eugene, OR). Butylated hydroxyanosole (BHA) was purchased from Sigma. zVADfmk and zVDVADfmk were purchased from Calbiochem. Polyclonal anti-Mn-SOD and anti-Bax were purchased from Upstate (Temecula, CA), anti-catalase was purchased from Abcam (Cambridge, MA), and anti-Bcl-2 was from BD Bioscience. Phoenix cells and the LZRS expression construct were gifts from Dr. Gary Nolan (Stanford University, CA). WT, IKK␣ Ϫ/Ϫ , and IKK␤ Ϫ/Ϫ cells were provided by Dr. Inder Verma, and NEMO Ϫ/Ϫ cells were from Dr. Michael Karin (both of the University of California, San Diego, CA).
Retroviral Transduction-Ecotropic phoenix cells were transfected with retroviral expression constructs LZRS-EGFP, LZRS-FADD DN , LZRS-IKK␤, and LZRS-Bcl-2 using FuGENE 6 (Invitrogen). The FADD DN construct encodes a protein lacking the N-terminal 80 amino acids encompassing the death effector domain and does not associate with caspases (21). Twenty-four hours after transfection, phoenix cells were selected using growth media supplemented with puromycin (1 g/ml). After expansion of the cell population, media supplemented with puromycin was removed, and cells were washed then returned to normal growth media for a period of 12-24 h. This media was collected, supplemented with Polybrene (8 g/ml), filtered, and used immediately or stored at Ϫ80°C. MEFs were retrovirally transduced by 2-3 rounds of incubation with retroviral-conditioned media for 6 -12 h at 37°C. This procedure resulted in infection of at least 95% of the population.
Cell Viability Assays-In all assays floating cells were collected and pooled with remaining attached cells that were harvested using trypsin/EDTA.
Annexin V Staining-Cells were washed in PBS, and each sample was re-suspended in 100 l of annexin V staining buffer (10 mM HEPES, 140 mM NaCl, 2.5 mM CaCl 2 , pH 7.4) containing Alexa Fluor 488-conjugated annexin V (5 l/sample). Samples were incubated at room temperature for 15-20 min before the addition of a further 400 l of annexin V staining buffer and immediate analysis by FACS using Cell Quest software (FACSort, BD Biosciences).
Propidium Iodide (PI) Permeabilization Assay-Cells were washed once in PBS before the addition of PBS containing PI (25 g/ml). After incubation for 20 min at 37°C, the cells were analyzed by FACS.
Cell Cycle and Hypodiploid DNA Analysis-Harvested cells were washed in PBS and fixed for 30 -60 min in 70% ethanol at 4°C. After fixation, samples were washed once more before incubation in PBS containing PI (50 g/ml) and RNase A (1 g/ml) for at least 1 h. DNA analysis was performed by FACS.
Mitochondrial Membrane Potential (⌬⌿) Analysis-Floating and attached cells were harvested following the procedures described above. Cells were washed once in PBS containing 1% bovine serum albumin before re-suspension in 200 l of PBS/ bovine serum albumin containing JC-1 (10 g/ml). After 15 min of incubation at 37°C MEFs were washed, re-suspended in PBS, and analyzed by FACS.
Western Blotting-For all immunoblots, each well of a 12-well plate was treated as described. After experimental manipulation, any detached cells were collected, washed in PBS, and pooled with the lysate from the same well prepared by washing the adherent cells twice in ice-cold PBS before the addition of 100 -200 l of TNT lysis buffer (50 mM Tris-Cl, pH 7.5, 150 mM NaCl, and 1% Triton X-100) containing complete protease inhibitor mixture (Roche Applied Science), NaF (2 mM), and ␤-glycerophosphate (2 mM). Protein content was determined using Coomassie Plus Reagent (Pierce), and for each sample an equal amount of protein (10 -20 g) was fractionated by SDS-PAGE then transferred electrophoretically to polyvinylidene difluoride membranes (Immobilon P, Millipore, Milford, MA) and immunoblotted with the appropriate primary and species-specific horseradish peroxidase-conjugated secondary antibodies (Jackson ImmunoResearch, Westgrove, PA). Detection of the bound antibody by chemiluminescence was performed using Luminol reagent according to the manufacturer's instructions (Santa Cruz).
ROS Measurement-After experimental manipulation, floating and attached cells were harvested and pooled, then washed once in PBS before incubation in prewarmed PBS containing H2DCFDA (10 M) for 20 min at room temperature. Samples were then washed once in PBS before re-suspension in PBS. After incubation at room temperature for a further 10 -15 min, samples were analyzed by FACS.

TNF and LT␣1␤2 Induce Cell Death in IKK␣
Ϫ/Ϫ and IKK␤ Ϫ/Ϫ MEFs-To compare the mechanisms of cell death induced by TNF and LT␣1␤2, we utilized MEFs lacking either IKK␣ or IKK␤. Although IKK␤ critically regulates the classical NF-B pathway and induction of anti-apoptotic genes, we found that deletion of either kinase sensitized MEFs to cell death in response to both TNF and LT␣1␤2. Death was detected by detachment of MEFs from the culture dish (not shown), an increase in annexin V staining (Figs. 1, A and B), and an increase in membrane permeability determined by uptake of propidium iodide (Fig. 1, C and D). Interestingly, MEFs lacking IKK␤ were less sensitive to death in response to LT␣1␤2 than TNF. In contrast, MEFs lacking IKK␣ were equally sensitive to either TNF or LT␣1␤2.
Caspase Inhibition Does Not Prevent Death in IKK␤ Ϫ/Ϫ MEFs-To determine whether the cell death we observed in response to cytokine treatment of IKK␣-or IKK␤-deficient MEFs was caspase-dependent apoptosis, we employed the broad range caspase inhibitor zVADfmk. To confirm its activity, we determined the effects of zVADfmk on cell death in HUVEC induced by TNF plus cycloheximide as this death is well characterized as apoptosis (22). Treatment of HUVEC with TNF plus cycloheximide resulted in apoptosis measured by increased uptake of PI that was completely blocked by co-treatment with zVADfmk (25 M) (Figs. 2, A and B). zVADfmk also effectively blocked DNA fragmentation in HUVEC (Fig. 2C). MEFs lacking IKK␣ behaved in a similar manner to HUVEC. Thus, LT␣1␤2 induced death characterized by PI uptake that was completely blocked by incubation with zVADfmk (Figs. 2, D and E). Similar effects were observed when IKK␣ Ϫ/Ϫ MEFs were treated with TNF (not  shown).
Surprisingly, however, the death of IKK␤-deficient cells in response to TNF was unaffected by co-treatment with zVADfmk. We did not observe any inhibition of cell detachment from the culture plate (not shown) or uptake of PI (Fig. 3, A and open bars in B) at any concentration of zVADfmk tested (up to 50 M). Remarkably, the extent of LT␣1␤2-induced cell death in IKK␤ Ϫ/Ϫ MEFs actually increased in the pres- ence of zVADfmk (Fig. 3, A and gray bars in B). Despite failing to inhibit cell death, zVADfmk prevented the generation of hypodiploid DNA in IKK␤ Ϫ/Ϫ MEFs (Fig. 3C), indicating that the death mechanism was not apoptosis. Because zVADfmk is a relatively poor inhibitor of caspase-2, we also used the caspase-2 inhibitor zVDVADfmk. Treatment of IKK␤ Ϫ/Ϫ MEFs with TNF in the presence of zVDVADfmk (25 M) either alone or in combination with zVADfmk (25 M) did not prevent cell death (data not shown). Hence, a lack of caspase-2 inhibition and escape from protease inhibition was unlikely to account for the failure of zVADfmk to prevent cell death.

IKK␤ Represses Death in zVADfmk-treated MEFs in a NEMO-independent
Manner-To determine whether the TNF-induced death of zVADfmk-treated IKK␤ Ϫ/Ϫ cells could be repressed by exogenous IKK␤, we reconstituted IKK␤-deficient MEFs by retroviral transduction with LZRS-IKK␤. As shown in Figs. 4, A and B, transduction with IKK␤ but not LZRS alone markedly reduced the TNF-induced cell death observed in the presence of zVADfmk. The activity of IKK␤ in regulating the classical NF-B pathway is critically dependent upon its interaction with NEMO. We, therefore, questioned whether the death observed in zVADfmk-treated IKK␤ Ϫ/Ϫ MEFs also occurred in cells lacking NEMO. Inter- estingly, TNF-induced death of NEMO Ϫ/Ϫ MEFs was completely blocked by ZVADfmk (Fig. 4, C and D), demonstrating that IKK␤ functions independently of NEMO to repress cell death in zVADfmk-treated cells.
Dominant Negative FADD Recapitulates the Effects of zVADfmk in IKK␤ Ϫ/Ϫ MEFs-Our findings strongly suggest that the cell death observed in zVADfmk-treated cells lacking IKK␤ is caspase-independent. However, to ensure that the effects were not an artifact of the inhibitor and could be recapitulated by non-pharmacological caspase inhibition, we transduced IKK␣ Ϫ/Ϫ and IKK␤ Ϫ/Ϫ cells with a dominant negative version of FADD (FADD DN ) to prevent receptor-mediated caspase activation. Similar to zVADfmk, FADD DN prevented TNF-induced death of IKK␣ Ϫ/Ϫ MEFs but exacerbated the death observed in IKK␤ Ϫ/Ϫ cells (Figs. 5, A and B). Taken together, the findings in Figs. 3-5 lead us to conclude that cell death induced in IKK␤ Ϫ/Ϫ cells by TNF and LT␣1␤2 in the presence of ZVADfmk is caspase-independent cell death (CICD).
CICD in IKK␤ Ϫ/Ϫ MEFs Is Dependent on the Mitochondria-To characterize the CICD induced in the absence of IKK␤, we explored the effects of TNF on mitochondrial membrane potential (⌬⌿). Using the ⌬⌿-sensitive dye JC-1, we found that treatment of IKK␤ Ϫ/Ϫ MEFs with TNF caused a loss of ⌬⌿ that could not be prevented by co-incubation with zVADfmk (Fig.  6A). Treatment with LT␣1␤2 and zVADfmk led to a similar loss of ⌬⌿ (not shown). We next questioned whether TNF and/or zVADfmk treatment affected the levels of the mitochondrial regulatory proteins Bcl-2, Bax, and Bcl-2-binding protein beclin-1. The expression levels of each of these proteins were unaffected by treatment of either WT or IKK␤ Ϫ/Ϫ MEFs with TNF in the absence or presence of zVADfmk (data not shown).
Overexpression of Bcl-2 prevents mitochondria-dependent cell death in response to a wide range of stimuli and conditions including treatment with the DNA-damaging chemotherapeutic agent etoposide (23,24). We confirmed that etoposide induced death in IKK␤ Ϫ/Ϫ MEFs and that Bcl-2 overexpression effectively inhibited this process (Fig. 6B). We, therefore, investigated the effect of Bcl-2 overexpression on the death of IKK␤ Ϫ/Ϫ MEFs in response to TNF and LT␣1␤2 plus ZVADfmk. As shown in Fig. 6, C and D, Bcl-2 overexpression did not affect the death of IKK␤ Ϫ/Ϫ MEFs treated with LT␣1␤2 or TNF alone; however, it did prevent death in response to these cytokines in the presence of zVADfmk (Figs. 6, C and D). This, therefore, suggests that the death pathway in IKK␤ Ϫ/Ϫ MEFs treated with cytokines changes in the presence of zVADfmk to one that is dependent upon the mitochondria.

CICD in IKK␤ Ϫ/Ϫ MEFs Is Dependent on the Generation of ROS-
Because mitochondrial function is closely linked to the generation of ROS, we questioned whether TNF treatment induced ROS in IKK␤ Ϫ/Ϫ MEFs and whether scavenging such species could inhibit death. Using the dye H2DCFDA, we found that TNF treatment of WT MEFs did not lead to ROS generation, whereas a significant increase was observed in MEFs lacking IKK␤ (Figs. 7, A and B). In WT MEFs, treatment with zVADfmk caused an increase in the generation of ROS in that it was reduced by co-treatment with TNF (Figs. 7, A and B). In contrast, treatment of IKK␤ Ϫ/Ϫ MEFs with zVADfmk alone had no effect on ROS, and co-treatment with zVADfmk plus TNF still resulted in increased ROS levels (Figs. 7, A and B). To determine whether the generation of ROS contributed to TNFinduced cell death in IKK␤ Ϫ/Ϫ MEFs, we utilized the ROS scavenger BHA. Pretreatment of IKK␤ Ϫ/Ϫ MEFs with BHA significantly diminished the death of IKK␤ Ϫ/Ϫ cells treated with TNF plus ZVADfmk but did not affect the death induced by TNF treatment alone (Figs. 7, C and D). Similar effects were observed when cells were treated with LT␣1␤2 and zVADfmk (not shown). Together with the data in Fig. 6, C and D, these findings further support a model in which the death pathway activated in IKK␤ Ϫ/Ϫ MEFs by TNF or LT␣1␤2 alters the phenotype after the addition of zVADfmk to become dependent on the mitochondria and the generation of ROS.
ROS levels can be regulated by the expression of scavenging enzymes. We, therefore, questioned whether treatment of IKK␤ Ϫ/Ϫ MEFs with TNF and zVADfmk affected the expression of Mn-SOD and catalase that have been implicated recently in the regulation of cell death in response to both of these mediators (25,26). Mn-SOD expression was induced by TNF in WT MEFs, and this was not affected by the presence of zVADfmk ( Fig. 8A; lanes 1-4). In MEFs lacking IKK␤, basal Mn-SOD expression was significantly reduced and was not modulated by TNF either alone or in combination with zVADfmk ( Fig. 8A; lanes 5-8). Despite the recent report that zVADfmk decreases catalase levels by a posttranscriptional mechanism (26), we did not observe any clear effect of TNF, zVADfmk, or TNF plus zVADfmk on its expression in either WT or IKK␤ Ϫ/Ϫ MEFs (Fig. 8B). We did observe modification of catalase resulting in the appearance on immunoblots of bands both above and below the catalase band (lanes 3 and 4). The nature of these modifications remains unclear; however, no such modified catalase was detected in IKK␤ Ϫ/Ϫ cells treated with zVADfmk (lanes 7 and 8), suggesting that this is dependent upon functional IKK␤.

DISCUSSION
Cell death is a highly regulated process necessary for development. Dysregulation of the signals that control cell death can result in either failure to die or enhanced death that in turn supports tumor growth or leads to excessive tissue damage, respectively. To develop a clearer picture of the death signals induced by TNF family members, we utilized MEFs lacking either IKK␣ or IKK␤ to determine whether these kinases played distinct roles in regulating death induced by TNF or LT␣1␤2. We found that cells lacking either kinase were sensitized to death in response to both cytokines, although more death occurred in IKK␤ Ϫ/Ϫ MEFs incubated with TNF than those treated with LT␣1␤2. Conversely, LT␣1␤2 treatment caused  greater cell death in IKK␣ Ϫ/Ϫ MEFS than TNF. We, therefore, conclude that signaling through IKK␣ and IKK␤ differentially predominates to counteract the cell death signaling pathways induced by the separate cytokines.
Our finding that both TNF and LT␣1␤2 induce apoptosis in IKK␣ Ϫ/Ϫ cells was surprising in light of previous studies of mice lacking each of the IKK subunits. Animals lacking either NEMO or IKK␤ die during development from massive hepatocyte apoptosis (27,28), whereas mice lacking IKK␣ survive but die shortly after birth due to morphogenic defects (19,29). IKK␣ is, therefore, thought to play no role in the induction of classical NF-B-dependent anti-apoptotic genes (e.g. cIAPs, c-FLIP, A20) and protection from TNF-induced liver apoptosis. However, it has been reported recently that hepatocyte-specific deletion of NEMO but not IKK␤ sensitizes mice to TNF-induced liver apoptosis (30), suggesting that IKK␣ can substitute for IKK␤ and activate NF-B to protect hepatocytes from death. Separate studies have also linked IKK␣ to classical NF-B target gene expression (31,32), but we do not yet know whether IKK␣ controls anti-apoptotic genes by a direct regulatory function in the classical pathway or by another mechanism such as histone phosphorylation (17,18). Although further studies are required to determine its precise mechanism of action, our findings clearly demonstrate that IKK␣ plays a crucial role in protecting MEFs from apoptosis in response to TNF and LT␣1␤2.
Cell death occurs by various mechanisms including apoptosis, autophagy, and necrosis; however, only apoptosis is dependent upon caspase activation. To determine whether the death of IKK␣and IKK␤-deficient MEFs was caspase-dependent apoptosis or a separate non-apoptotic mechanism, we utilized the broad range caspase inhibitor zVADfmk. This inhibitor has been extensively characterized and shown to block the processes that characterize apoptosis including membrane blebbing, outer membrane inversion and exposure of phosphatidylserine, cell shrinkage, nuclear condensation, generation of nucleosomes, and DNA breakdown. Death of both IKK␣ Ϫ/Ϫ and IKK␤ Ϫ/Ϫ MEFs in response to TNF and LT␣1␤2 displayed several of these characteristics including nuclear condensation and DNA fragmentation. Consistent with this, caspase inhibition blocked the cell death induced by both TNF and LT␣1␤2 in IKK␣-deficient MEFs, demonstrating that this occurs via caspase-dependent apoptosis. Surprisingly however, caspase inhibition with either ZVADfmk or non-pharmacologically via transduction of FADD DN did not block death in IKK␤deficient MEFs, providing further evidence that IKK␣ and IKK␤ regulate separate cell death processes. Intriguingly, TNFinduced death of NEMO-deficient MEFs was blocked by caspase inhibition, demonstrating that the repression of CICD by IKK␤ occurs independent of its typical functional association with NEMO. Although further studies are required to determine the targets of IKK␤ in repressing CICD, our findings establish a novel NEMO-independent role for IKK␤ in regulating cell death.
Previous studies have established that IKK␤ regulates classical NF-B-dependent expression of negative regulators of caspase-dependent apoptosis (2)(3)(4). Our finding that IKK␤ Ϫ/Ϫ MEFs treated with zVADfmk and, therefore, lacking functional caspase activity continue to die in response to TNF and LT␣1␤2 therefore demonstrates that IKK␤ also represses a CICD process. A potential model of the interplay between IKK␤ and caspases in regulating cell death is shown in Fig. 9. Remarkably, we consistently observed that cytokine-induced death in IKK␤-deficient MEFs was greater in the presence of zVADfmk than in untreated cells, and this was particularly notable in response to LT␣1␤2. We noted a similarly enhanced death response in IKK␤ Ϫ/Ϫ MEFs transduced with FADD DN that blocks receptor-induced caspase activation. This, therefore, suggests that CICD is normally limited by caspases in addition to being repressed by IKK␤. Previous studies have described CICD in the fibrosarcoma L929 cell line after TNF stimulation (33) and in caspase-8 Ϫ/Ϫ Jurkat cells (34), and CICD has been reported in vivo where zVADfmk exacerbates the lethal toxicity of TNF shock in mice (35). Our findings, therefore, extend these previous studies and demonstrate that IKK␤ is critical for repression of CICD induced by TNF and LT␤ receptor ligation. The cell death in cytokine-stimulated IKK␤ Ϫ/Ϫ MEFs treated with zVADfmk was typical of necrosis; we observed membrane leakiness, phosphatidylserine exposure, and loss of mitochondrial membrane potential that occurred in the absence of the DNA and nuclear breakdown seen in apoptosis. Autophagic death is hallmarked by accumulation of vacuoles that sequester and target cytoplasmic components for lysosomal degradation in a process dependent on Atg genes including Beclin-1 (36). We did not observe any changes in Beclin-1 levels upon TNF stimulation of either WT or IKK␤ Ϫ/Ϫ MEFs in the pres-   (Fig. 8A). B, in the absence of IKK␤ cytokines activate caspase-dependent apoptosis. ROS levels are increased possibly via a lack of induced Mn-SOD expression (Fig. 8A), and mitochondrial membrane potential is lost (Fig. 6A). However, in the presence of active caspases, CICD does not occur, and neither Bcl-2 (Figs. 6, C and D) nor BHA (Figs. 7, C and D) can block cell death. C, caspase-dependent apoptosis is blocked in IKK␤ Ϫ/Ϫ MEFs stimulated with cytokines in the presence of zVADfmk. These cells die by CICD that can be inhibited by mitochondrial regulation via Bcl-2 overexpression (Figs. 6, C and D) or by ROS scavenging via BHA treatment (Figs. 7, C and D). ence of zVADfmk (data not shown). Furthermore treatment with the autophagy inhibitor 3-MA (36, 37) did not block cytokine plus zVADfmk-induced death of IKK␤ Ϫ/Ϫ MEFs (data not shown). Consequently, we conclude that in contrast to the CICD previously reported in L929 cells (26,36), the death of IKK␤-deficient MEFs treated with cytokines and zVADfmk is not autophagic but is instead necrotic.
The continued loss of mitochondrial membrane potential in TNF-treated IKK␤-deficient MEFs after caspase inhibition led us to test whether exogenous expression of mitochondrial regulators might affect CICD. Overexpression of the anti-apoptotic mitochondrial protein Bcl-2 did not affect cytokine-induced death of IKK␤-deficient MEFs, indicating that death occurs by apoptosis via a mitochondrial-independent process. Activation of caspase-8 and the mitochondrial-independent activation of caspase-3 in a so-called "extrinsic pathway" after death receptor ligation have been described in various cell types (22). Bcl-2 has also been shown previously to inhibit CICD (22,38,39) most likely by preventing the loss of mitochondrial membrane potential, and consistent with this, we found that in the presence of zVADfmk, overexpressed Bcl-2 diminished death in cytokine-stimulated IKK␤ Ϫ/Ϫ MEFs. Our data, therefore, demonstrate that in cells lacking IKK␤, a switch in the cytokineinduced death phenotype from mitochondrial-independent apoptosis to caspase-independent, mitochondrial-dependent cell death occurs in the presence of zVADfmk.
It has been proposed that mitochondria and the generation of ROS contribute to TNF-mediated CICD (33,35,40). We, therefore, questioned whether cytokine stimulation of IKK␤deficient cells led to increased ROS and whether the anti-oxidant BHA affected cell death. Similar to Kamata et al. (25), we found that TNF increased ROS in IKK␤ Ϫ/Ϫ but not WT MEFs, and this is also consistent with the effects observed in TRAF2/5 double knock-out, p65 Ϫ/Ϫ , and ⌬N-IB␣-transduced cells (41,42). It is, therefore, likely that induction of classical NF-B-dependent anti-oxidant enzymes such as Mn-SOD that we (Fig.  8A), and others (25,43) have observed contributes to the rescue by IKK␤ from TNF-induced ROS generation. Despite the increased ROS generation in TNF-stimulated IKK␤-deficient MEFs, BHA did not protect against cell death. Similarly, treatment of WT MEFs with zVADfmk increased ROS but did not cause death. We found that BHA only protected against death in IKK␤ Ϫ/Ϫ MEFs in response to cytokines in the presence of zVADfmk and conclude that ROS only cause death in cells lacking both IKK␤ and functional caspases. It is not clear why our findings conflict with those of with Kamata et al. (25) who reported that BHA alone blocked IKK␤-deficient cell death; however, our data strongly suggest that ROS-independent, caspase-dependent apoptosis remains functional and maintains cell death in the face of ROS inhibition. Consistent with this model we also found that Bcl-2 overexpression inhibited TNF-induced death in IKK␤-deficient MEFs only when caspases were blocked. It is, therefore, likely that Bcl-2 and BHA inhibit the same CICD pathway involving premitochondrial or mitochondrial generation of ROS (Fig. 9).
Post-translational down-regulation of catalase expression by zVADfmk has been reported to cause increased ROS in L929 cells (26). We did not detect any reduction in catalase levels in zVADfmk-treated WT or IKK␤ Ϫ/Ϫ MEFs, although we did observe its modification in WT cells. This, therefore, suggests that IKK␤ plays an important role in regulating catalase in cells lacking functional caspases. We do not yet know the nature of this modification, but it is possible that it renders catalase inactive, leading to increased ROS in WT MEFs. This modification was not present in the IKK␤-deficient cells, although the failure to induce Mn-SOD in response to TNF (Fig. 7A) may play a predominant role in maintaining elevated ROS levels in these cells. A further possibility is that zVADfmk actively induces ROS generation and potential mechanisms for this include enhanced leakiness of the mitochondrial electron transport chain (44), decreased mitochondrial turnover (45), or enhanced cPLA2 activation (35). Further work will be required to determine whether any or all of these mechanisms occur in IKK␤deficient cells.
Intriguingly, IKK␤ negatively regulates JNK (25,46), and we confirmed that the duration of JNK activity was increased in IKK␤ Ϫ/Ϫ MEFs (not shown). Some studies have implicated the generation of ROS in cell death mediated by sustained JNK activity (25,41,42,47). We did not observe any effect of zVADfmk on either the JNK response (data not shown) or generation of ROS in IKK␤ Ϫ/Ϫ cells. Because JNK is not required for cell death mediated by ROS (42), our data suggest that ROS are generated in IKK␤ Ϫ/Ϫ MEFs treated with zVADfmk by a mechanism that does not involve JNK. Consistent with this, a pharmacological inhibitor of JNK (JNK inhibitor II) failed to block death in TNF-and zVADfmk-treated IKK␤deficient MEFs (data not shown). Consequently, we conclude that JNK does not control the CICD pathway activated in the absence of IKK␤ and caspases.
In summary, we have shown that IKK␣ plays an anti-apoptotic role in death receptor signaling. The anti-apoptotic function of IKK␣ is most notable in response to LT␣1␤2, suggesting that distinct death-inducing cytokines preferentially utilize separate IKK subunits. We have also demonstrated that in addition to inhibiting caspase-dependent apoptosis, IKK␤ negatively regulates CICD in a NEMO-independent manner. The function of IKK␤ as a CICD inhibitor is revealed in the presence of zVADfmk that induces a change in the death phenotype from apoptosis to necrosis. Furthermore our data suggest that caspases normally inhibit CICD and that CICD depends on the mitochondria and the generation of ROS. Although further studies are required to determine the precise mechanism of ROS-induced cell death, our findings clearly identify an unanticipated interplay between IKK␤ and caspases in regulating CICD (see model Fig. 9).