|Home | About | Journals | Submit | Contact Us | Français|
Interferons (IFNs) are cytokines with well-described immunomodulatory and antiviral properties, but less is known about the mechanisms by which they promote cell survival or cell death. Here, we show that IFN-γ induces RIP1 kinase-dependent necroptosis in mammalian cells deficient in NF-κB signaling. Induction of necroptosis by IFN-γ was found to depend on Jak1 and partially on STAT1. We also demonstrate that IFN-γ activates IκB kinase β (IKKβ)-dependent NF-κB to regulate a transcriptional program that protects cells from necroptosis. IFN-γ induced progressive accumulation of reactive oxygen species (ROS) and eventual loss of mitochondrial membrane potential in cells lacking the NF-κB subunit RelA. Whole-genome microarray analyses identified sod2, encoding the antioxidant enzyme manganese superoxide dismutase (MnSOD), as a RelA target and potential antinecroptotic gene. Overexpression of MnSOD inhibited IFN-γ-mediated ROS accumulation and partially rescued RelA-deficient cells from necroptosis, while RNA interference (RNAi)-mediated silencing of sod2 expression increased susceptibility to IFN-γ-induced cell death. Together, these studies demonstrate that NF-κB protects cells from IFN-γ-mediated necroptosis by transcriptionally activating a survival response that quenches ROS to preserve mitochondrial integrity.
Interferons (IFNs) are a family of powerful immunomodulatory and antiviral cytokines categorized into two major classes, type I and type II. Type I interferons, including several IFN-α subtypes and a single IFN-β species, are produced by most cell types in response to virus infections, while type II interferon (IFN-γ) is made primarily by T cells and NK cells and is not virus inducible (55). IFN-γ signals through the IFN-γ receptor (IFNGR) complex comprising IFNGR-1 and IFNGR-2 dimers (1, 55). Upon IFN-γ binding to its receptor, the Janus family tyrosine kinases Jak1 and Jak2 are activated and phosphorylate the latent cytoplasmic transcription factor signal transducer and activator of transcription 1 (STAT1). Phosphorylated STAT1 homodimerizes to form IFN-γ-activated factor (GAF) (15), which then translocates to the nucleus to transactivate genes containing IFN-γ-activated sequence (GAS) elements in their promoters (15, 38). Over 200 such genes, referred to as interferon-stimulated genes (ISGs), are regulated in this manner (18, 23).
While STAT1-dependent gene activation represents the predominant and most well-studied transcriptional response to IFN-γ, recent evidence suggests that additional pathways (such as those involving other STAT family members, mitogen-activated protein [MAP] kinases, and NF-κB) are also engaged downstream of IFNGR to modulate gene expression, but their functional significance remains unclear (46, 58). NF-κB refers to a family of transcription factors composed of combinations of RelA, P50, P52, c-Rel, and RelB homo- and heterodimers (26, 27). All NF-κB subunits share a Rel homology domain in their N termini while RelA, RelB, and c-Rel also possess C-terminal transactivation domains. Canonical NF-κB signaling proceeds via activation of RelA-P50 heterodimers, which are normally retained in the cytoplasm via binding to the I-κB class of inhibitory proteins (26, 27). In response to upstream signals, I-κBs are phosphorylated by IκB kinases (IKKs) and rapidly degraded by the proteasome. Degradation of IκB results in translocation of NF-κB to the nucleus, where it activates transcription of genes containing κB sites in their promoters (24, 26, 27).
Key roles of NF-κB include regulation of immune and inflammatory responses (26, 27). In addition, NF-κB activates an important cell survival program in response to certain stimuli, including tumor necrosis factor alpha (TNF-α) (5, 19, 26, 27). During TNF-α signaling, two primary pathways are activated downstream of TNF-α receptor 1 (TNFR1) (9). In a simplistic model, one of these cascades results in NF-κB activation and consequent transcription of cytoprotective genes while the other pathway triggers apoptotic cell death when survival signals are absent (9). Cells defective in NF-κB signaling are thus often susceptible to TNF-α-induced apoptosis (4). Among the survival genes activated by NF-κB are those encoding well-recognized antiapoptotic molecules such as c-FLIP and Bcl-xL (2, 34). In addition, NF-κB transcriptionally upregulates genes encoding antioxidant enzymes such as manganese superoxide dismutase (MnSOD) and ferritin heavy chain (FHC) (43, 47). MnSOD catalyzes the dismutation of superoxide anion into hydrogen peroxide and oxygen, promoting eventual reactive oxygen species (ROS) elimination (30, 32). In parallel, FHC oxidizes Fe2+ to Fe3+ to reduce the availability of free intracellular Fe2+ that can participate in the generation of free radicals through the Fenton reaction (12, 25, 56). Deficiency in either MnSOD or FHC renders cells susceptible to TNF-α-induced apoptosis, underscoring the importance of scavenging free radicals to cell survival during TNF-α signaling (42, 47).
Under certain conditions, such as when caspases are inhibited, TNF-α induces an alternative death pathway, termed programmed necrosis, or necroptosis (59). The study of necroptosis has been greatly facilitated by the recent discovery of necrostatins, small-molecule allosteric inhibitors of the kinase receptor-interacting protein 1 (RIP1) (16, 29). Indeed, RIP1 has emerged as a pivotal molecule in necroptosis and, together with RIP3, forms a kinase complex that impinges on mitochondria to alter metabolic rates (11, 28, 59, 66). As a consequence, excessive production of ROS appears to initiate mitochondrial dysfunction and consequent necroptotic death (11, 66). Interestingly, a recent whole-genome RNA interference (RNAi) screen for effectors of necroptosis identified several immune system genes, including those encoding modulators of type I IFN, TNF-α, and Toll-like receptor (TLR) signaling (29). These findings suggest that necroptosis may represent the outcome of a host antimicrobial mechanism that utilizes ROS and other effector mechanisms to eliminate cells during innate immune responses.
In this study, we show that mammalian cells deficient in NF-κB signaling are susceptible to IFN-γ-induced RIP1-dependent necroptosis. We demonstrate that IFN-γ, like TNF-α, induces an NF-κB-dependent transcriptional response that is cytoprotective. In the absence of NF-κB, IFN-γ promotes mitochondrial ROS accumulation and eventual loss of mitochondrial membrane potential (Δψm). Using microarray technology, we show that a subset of ISGs are dependent on the NF-κB subunit RelA, and we have identified sod2, encoding MnSOD, as a potential RelA-regulated survival target. Overexpression of MnSOD partially rescues RelA-deficient cells from IFN-γ-triggered necroptosis, whereas silencing MnSOD expression increases susceptibility to IFN-γ-induced cell death. Collectively, our results suggest that IFN-γ activates an NF-κB-mediated transcriptional program to control ROS output during immune responses and triggers RIP1-dependent necroptosis when ROS levels are unchecked.
Early passage rela+/+, rela−/−, ikkβ+/+, and ikkβ−/− mouse embryonic fibroblasts (MEFs) were prepared as previously described (60, 61). myd88/trif double-knockout MEFs were obtained from S. Akira (Osaka University, Japan), stat1+/+ and stat1−/− MEFs were provided by L. Sigal (Fox Chase Cancer Center), HeLa and 293T cells were purchased from the ATCC (Manassas, VA), and 293FT cells were from Invitrogen (Carlsbad, CA). Cytokines and chemicals were from the following sources: human and murine IFN-γ, Pestka Biomedical Laboratories; human and murine TNF-α, R&D Systems; necrostatin-1, Enzo Life Sciences; z-VAD-fluoromethyl ketone (FMK), Calbiochem; 5-(and-6)-chloromethyl-2′,7′-dichlorodihydrofluorescein diacetate acetyl ester (CM-H2DCFDA) and tetramethylrhodamine ethyl ester (TMRE), Molecular Probes. Antibodies were purchased from Millipore (STAT1 and phospho-STAT1), BD Biosciences (RIP1), ProSci, Inc. (RIP3), and Santa Cruz Biotechnology (RelA, P50, P52, c-Rel, RelB, IRF-1, and MnSOD). Mouse and rabbit IgG were obtained from Jackson ImmunoResearch (Bar Harbor, ME). All other reagents were from Sigma-Aldrich, unless otherwise mentioned. All cells were cultured in high-glucose Dulbecco's modified Eagle medium (DMEM) containing 10% fetal bovine serum (FBS) and antibiotics.
HeLa cells (6 × 104/well) seeded into six-well dishes were transfected with pools of four distinct proprietary small interfering RNAs ([siRNAs] SMARTpool; Dharmacon) to each target mRNA at 20 nM using Oligofectamine (Invitrogen) as a transfection reagent. As controls, nontargeting siRNA duplexes (Dharmacon) were employed. Cells were used in experiments at 72 h posttransfection.
Cells (1 × 106 cells/condition) were washed twice with cold phosphate-buffered saline (PBS) and suspended in 400 μl of hypotonic buffer A (10 mM HEPES, pH 7.9, 10 mM KCl, 1.5 mM MgCl2, 1 mM dithiothreitol [DTT], 0.5 mM phenylmethylsulfonyl fluoride, and protease inhibitor cocktail [Roche]). After incubation for 15 min on ice, 25 μl of 10% Nonidet P-40 was added for a further 10 min, and nuclei were collected by centrifugation at 3,000 × g for 5 min at 4°C. Nuclear pellets were subsequently lysed in 30 μl of buffer B (20 mM HEPES, pH 7.9, 400 mM NaCl, 1.5 mM MgCl2, 0.2 mM EDTA, 1 mM DTT, 5% glycerol, and protease inhibitor cocktail) and incubated for 1 h at 4°C with brief intermittent mixing. Nuclear lysates were clarified by centrifugation at 10,000 × g for 5 min at 4°C. The NF-κB consensus oligonucleotide 5′-GGGGACTTTCCC-3′ (Santa Cruz Biotechnology) or the GAS oligonucleotide 5′-GATCGATTTCCCCGAAAT-3′ (from the irf1 gene, custom synthesized at Fox Chase DNA Core Facility) was end labeled with [γ-32P]ATP using T4 kinase (Promega). Ten micrograms of nuclear protein was incubated with radiolabeled oligonucleotides in electrophoretic mobility shift assay (EMSA) reaction buffer [2 μg of poly(dI-dC), 12% glycerol, 20 mM HEPES, pH 7.0, 1 mM EDTA, 50 mM NaCl] for 30 min at room temperature and subjected to 5% nondenaturing PAGE in 0.5× Tris-borate-EDTA (TBE) buffer. For antibody supershift experiments, antibodies (1 μg) were added to nuclear extracts 15 min prior to addition of radiolabeled oligonucleotides. Gels were run at 80 V to 100 V, vacuum dried, and subjected to autoradiography.
MEFs (5 × 106/condition) were stimulated in duplicate with murine IFN-γ (250 U/ml). Total RNA was isolated in TRIzol reagent (Invitrogen, Carlsbad, CA) and purified using an RNeasy kit (Qiagen). Five hundred nanograms of total RNA was amplified and labeled using a Low RNA Input linear amplification kit (Agilent). Labeled cRNA targets were hybridized onto Agilent 4×44k (4 arrays/chip and ~44,000 transcripts/array) mouse whole-genome arrays. Microarray images were processed using Agilent Feature Extraction software (version 9.5). Identification of differentially expressed genes was performed with the LIMMA package (53) implemented in the R/Bioconductor platform (22).
Wild-type MEFs (4 × 106/condition) were processed for chromatin immunoprecipitation (ChIP) using an EZ ChIP Kit (Millipore), according to the manufacturer's instructions. The following primers were used to amplify the cytokine-responsive NF-κB site from the murine sod2 gene (33) by real-time PCR following immunoprecipitation with anti-RelA antibody (5 μg/condition; Santa Cruz): Forward, 5′-GGATTTTTGGAAATTGCAGATCTGGG; Reverse, 5′-GGCTCCACAGAAGGATTACAATCAAGA.
MEFs were plated in six-well plates containing 0.2% gelatin-coated glass coverslips prior to stimulation with IFN-γ (25 ng/ml) for the indicated time periods. After stimulation, the cells were loaded simultaneously with CM-H2DCFDA (ROS indicator; 5 μM) and TMRE (Δψm indicator; 100 nm) in the dark for 30 min at 37°C. Cells were then washed and resuspended in Hank's buffered salt solution (HBSS), pH 7.2. Images were acquired using a Carl Zeiss 510 Meta confocal system with excitation at 488 nm and 561 nm for ROS and Δψm, respectively. Five images were collected randomly for each sample, and fluorescence change was quantified using Image J software (13, 39).
Ad.Luc and Ad.MnSOD were obtained from the University of Iowa Gene Transfer Vector Core (67). MEFs (1 × 105/well) were plated in six-well plates and allowed to adhere for 24 h before infection. Adenoviral infections were performed in serum-free medium at multiplicities of infection (MOI) of 20. Recombinant protein expression was verified by immunoblot analysis, and experiments were performed 48 h postinfection.
A Student's t test was used for comparison between two groups, and P values of <0.05 were considered significant.
During the course of our work on the role of NF-κB in antiviral responses (3), we noticed that cells lacking the NF-κB subunit RelA succumbed to treatment with recombinant IFN-γ in the absence of any virus. To examine this phenotype further, we treated early-passage littermate-derived rela+/+ and rela−/− MEFs with IFN-γ and monitored cell survival over a period of 72 h. As a positive control, we used TNF-α, which is known to trigger apoptosis in the absence of RelA (5). As expected, TNF-α induced robust cell death of rela−/− MEFs within 12 to 24 h of challenge (Fig. 1A and B). IFN-γ, however, triggered a more progressive form of cell death in rela−/− MEFs, first manifesting only after ~24 h. By 72 h, however, ~80% of rela−/− MEFs had succumbed to IFN-γ (Fig. 1B). Of note, we found that late-passage, immortalized, or transformed rela−/− MEFs showed significantly greater resistance to IFN-γ-induced cell death (data not shown). Our subsequent dose-response analyses revealed that IFN-γ manifests cytotoxic activity on rela−/− MEFs at doses between 1 to 5 ng/ml while higher doses (10 to 100 ng/ml) did not significantly increase the magnitude of cell death (Fig. 1C).
To eliminate issues arising from chronic loss of RelA, we next performed RNAi experiments to acutely knock down RelA levels and asked whether such cells showed increased susceptibility to IFN-γ. We reduced RelA expression in HeLa cells via transfection of pooled specific siRNAs by >95% and confirmed functional ablation by monitoring susceptibility to TNF-α at 3 days posttransfection (Fig. 1D). As previously reported, we found that IFN-γ alone was toxic to HeLa cells and reduced their viability by ~40% after mock or control siRNA treatment within 3 days (17). Despite this, acute loss of RelA increased susceptibility to IFN-γ by a further ~40% (Fig. 1D). Together, these results demonstrate that RelA functions as a survival factor against IFN-γ in mammalian cells.
Our results thus far demonstrate that IFN-γ triggers a slow, nonsynchronous form of cell death in rela−/− MEFs that appears phenotypically distinct from TNF-α-induced cytotoxicity in the same cells. In agreement with these observations, the pan-caspase inhibitor z-VAD was unable to rescue IFN-γ-induced cell death in rela−/− MEFs at concentrations up to 50 μM (Fig. 2A and B).
Hitomi and colleagues have recently shown that IFN-γ potentiates a novel form of cell death, called necroptosis, dependent on the kinase RIP1 (29). Given that aspects of IFN-γ induced cytotoxicity in rela−/− MEFs (such as the delayed kinetics and caspase independence) are suggestive of RIP1-dependent necroptosis (6), we asked whether IFN-γ activates necroptotic cell death in the absence of RelA. Accordingly, we preincubated rela−/− MEFs with necrostatin-1 (Nec-1), a small-molecule allosteric inhibitor of RIP1 kinase activity, and monitored cell survival over 72 h of IFN-γ treatment. In contrast to z-VAD, Nec-1 almost completely protected rela−/− MEFs from IFN-γ-induced cytotoxicity (Fig. 2A and B). HeLa cells are naturally susceptible to IFN-γ-induced cytotoxicity and succumb over a period of several days to chronic stimulation by this cytokine (17). We found that IFN-γ-induced cell death in HeLa cells was inhibited by Nec-1, indicating that it was necroptotic in nature (Fig. 2C). As confirmation, we determined that HeLa cells in which RelA expression was abolished by RNAi before treatment with IFN-γ were also protected by Nec-1 (Fig. 2D). Finally, ablation of RIP1 itself rescued HeLa cells from IFN-γ-triggered cytotoxicity (Fig. 2E). Together, these data strongly suggest that IFN-γ activates RIP1-dependent necroptosis in the absence of RelA.
The best-described signaling pathway activated by IFN-γ proceeds via Jak1/Jak2 (Jak1/2)-mediated phosphorylation and activation of the latent transcription factor STAT1. To determine if this pathway was required for IFN-γ-triggered necroptosis, we ablated expression of Jak1 and STAT1 in HeLa cells by RNAi and confirmed ablation by immunoblot analysis (Fig. 3A). STAT3 can also be activated by IFN-γ in certain contexts (58) and was included as a control in these experiments. EMSAs confirmed that HeLa cells rendered deficient in the expression of Jak1 and STAT1 (but not STAT3) by RNAi were unable to activate a GAS-binding complex, demonstrating functional loss of classical Jak/STAT signaling downstream of IFNGR in these cells (Fig. 3B). To test if Jak1, STAT1, or STAT3 was required for IFN-γ-activated necroptosis, we treated HeLa cells in which expression of Jak1, STAT1, or STAT3 was ablated by RNAi and monitored their viability over 3 days. While cells transfected with control siRNA exhibited significant (~45%) loss of viability in the presence of IFN-γ, cells lacking Jak1 were almost completely resistant and cells lacking STAT1 were partially resistant to IFN-γ-triggered cell death 3 days posttreatment (Fig. 3C and D). Knockdown of STAT3 did not affect viability of IFN-γ-treated HeLa cells to any significant extent compared to controls (Fig. 3C and D). Together, these studies show that IFN-γ activates necroptosis by a mechanism requiring Jak1 and at least partially dependent on STAT1.
TNF-α signaling has been shown to bifurcate downstream of its receptor into at least two arms, one of which activates an RelA-containing NF-κB complex to transcriptionally induce a robust cell survival program (9, 27, 31). If this axis is inactivated (e.g., in rela−/− MEFs), then another arm of TNF-α signaling triggers apoptosis (4). We next asked whether IFN-γ, like TNF-α, stimulates dual cell survival and cell death pathways and whether RelA mediates its survival functions by participating in an NF-κB dependent transcriptional response. We stimulated early-passage wild-type MEFs with two different doses of IFN-γ and examined nuclear lysates from these cells for DNA binding activity to a radiolabeled NF-κB element by EMSA. We found that both doses of IFN-γ reproducibly activate a single NF-κB DNA-binding complex, albeit to a lesser degree than TNF-α (Fig. 4A, upper panel). As expected, IFN-γ induced robust binding to a GAS probe, while TNF-α did not (Fig. 4A lower panel). In subsequent time course experiments, we observed that IFN-γ-induced NF-κB activity appeared with somewhat delayed kinetics compared to TNF-α and peaked ~1 h poststimulation (Fig. 4B, upper panel). In contrast, GAS activation was seen by 15 min and inactivated within 2 h after IFN-γstimulation (Fig. 3B, lower panel). We next carried out antibody supershift experiments to determine the specific composition of the IFN-γ-induced NF-κB complex. We found that, as with TNF-α, the IFN-γ-induced NF-κB complex primarily comprises RelA-p50 heterodimers (Fig. 4C).
Given that STAT1 was required for IFN-γ-activated necroptosis (Fig. 3), we next asked whether NF-κB activation by IFN-γ was also dependent on STAT1. We therefore treated stat1+/+ and stat1−/− MEFs with IFN-γ and evaluated NF-κB activity in nuclei from these cells by EMSA. As shown in Fig. 4D, we found that STAT1 was dispensable for NF-κB activation by IFN-γ. In fact, both basal and IFN-γ-stimulated NF-κB activity levels were noticeably elevated in cells lacking STAT1, suggesting that STAT1 may function as a negative regulator of NF-κB signaling, as previously noted (62). Expectedly, IFN-γ-stimulated NF-κB was abolished in rela−/− MEFs (Fig. 4E).
Although NF-κB activation by IFN-γ was reproducibly detectable, we along with others have observed that IFN-γ stimulated NF-κB to a significantly lesser extent than other well-recognized activators such as interleukin-1β (IL-1β) and TNF-α (14). We therefore next carried out experiments to rule out the possibility of lipopolysaccharide (LPS)/endotoxin contamination in the IFN-γ preparations used in our study. Since LPS activates NF-κB through TLRs, we treated MEFs doubly deficient for the adaptor molecules MyD88 and TRIF (and consequently lacking all known TLR signaling ) with IFN-γ and monitored NF-κB activation. These cells were completely unresponsive to exogenous LPS but continued to support NF-κB activation by IFN-γ (data not shown). We also found that IFN-γ activated NF-κB in the presence of the transcriptional inhibitor actinomycin D, arguing against the possibility that NF-κB activation was a secondary bystander effect mediated by ISGs (data not shown). Together, these data strongly suggest that IFN-γ directly activates RelA-containing NF-κB in a STAT1-independent manner.
In an effort to identify other NF-κB pathway members that phenocopy RelA in protecting cells from IFN-γ-induced necroptosis, we screened MEFs singly deficient in various genes implicated in NF-κB signaling (e.g., traf2, tak1, tbk1, and ikkβ) for susceptibility to IFN-γ. From these experiments, MEFs lacking IKKβ showed the strongest phenotype, and we therefore studied further. At doses of between 50 ng/ml and 100 ng/ml IFN-γ, ikkβ−/− MEFs showed ~50% reduction in viability after 72 h, whereas control ikkβ+/+ MEFs were mostly viable at this time (Fig. 5A). Notably, ikkβ−/− MEFs were more resistant to IFN-γ than rela−/− MEFs, which succumbed with faster kinetics to significantly lower doses of this cytokine over the same period (compare Fig. 5A and and1B).1B). Since stimulus-specific redundancy exists among IKKs, these results may reflect possible compensation by other IKK family members (e.g., IKKα or Tank-binding kinase 1 [TBK-1]). We subsequently confirmed that the mode of death triggered by IFN-γ in ikkβ−/− MEFs was RIP1-dependent necroptosis by almost completely rescuing these cells from IFN-γ-induced cytotoxicity with Nec-1 pretreatment (Fig. 5B).
IKKβ is the primary kinase involved in canonical NF-κB (RelA-p50) activation, and previous work has shown a role for IKKβ in induction of a subset of ISGs by IFN-γ. To determine whether IKKβ is directly required for NF-κB activation following stimulation with IFN-γ, we treated early-passage ikkβ−/− and ikkβ+/+ MEFs with IFN-γ and examined NF-κB activation by EMSA. We found that IFN-γ-induced NF-κB complex formation was mostly abolished in the absence of IKKβ, indicating that NF-κB activation by IFN-γ is largely IKKβ dependent (Fig. 5C). Nevertheless, residual IFN-γ-induced NF-κB DNA binding activity was still seen in ikkβ−/− MEFs, suggesting some redundancy at the level of IKKs in NF-κB regulation by IFN-γ (Fig. 5C).
Our results shown thus far are consistent with a model in which IFN-γ stimulates NF-κB to activate a cytoprotective transcriptional response. We therefore next sought to identify direct transcriptional targets of NF-κB that mediate cell survival during IFN-γ signaling. We treated rela+/+ and rela−/− MEFs with IFN-γ for either 3 or 6 h and subjected mRNA from these cells to whole-genome DNA microarray analyses. For the purposes of these studies, genes induced at least 2-fold after 6 h of IFN-γ treatment were considered ISGs. By this criterion, we selected a total of 262 genes for further analysis. Most of these genes (212/262) were induced to approximately the same extent in both rela+/+ and rela−/− MEFs and are most likely direct STAT1 targets (Fig. 6A). In agreement with this, conventional IFN-γ-induced Jak/STAT signaling appears to be relatively normal in the absence of RelA (data not shown). Of the differentially induced genes, those that were induced 1.5-fold or less at either 3 or 6 h in rela−/− MEFs were considered partially RelA dependent, and those induced 1.5-fold or less at both 3 and 6 h were categorized as fully RelA dependent and clustered accordingly (Fig. 6A). A total of 22 ISGs were found to be critically dependent on RelA for their expression (Fig. 6B); a further 28 were partially dependent on RelA (data not shown). From these analyses, we identified immune response gene 1 (irg1) as the primary RelA target (Fig. 6B). Among the other obligate RelA-dependent genes identified was sod2, encoding MnSOD (Fig. 6B). Expression of sod2 was induced ~2 to 3-fold in rela+/+ MEFs but was not activated to any significant extent in rela−/− MEFs (Fig. 6B, arrow).
Normally, ROS generated in mitochondria are rapidly eliminated by antioxidant enzymes. In particular, superoxide is converted to hydrogen peroxide (H2O2) by superoxide dismutases (SODs), and H2O2 is subsequently eliminated by catalase and glutathione peroxidases (GPXs) (Fig. 6C, top). Of the genes encoding these enzymes, only sod2 was found to be a RelA-dependent ISG (Fig. 6C, bottom). In agreement, chromatin immunoprecipitation (ChIP) analysis shows that RelA robustly associates with a previously identified NF-κB-responsive site in the murine sod2 gene (33) in an IFN-γ-dependent manner (Fig. 6D). We next examined MnSOD protein expression by immunoblot analysis of whole-cell extracts derived from rela+/+ and rela−/− MEFs stimulated with IFN-γ. In congruence with our microarray and ChIP data, we found that MnSOD protein levels increased ~2 to 3-fold within 12 to 24 h of IFN-γ treatment in rela+/+ but not rela−/− MEFs (Fig. 6E), consistent with the ~2-fold activation by IFN-γ of an NF-κB element in luciferase reporter assays (data not shown).
Identification of the ROS scavenger sod2 as a direct RelA target during IFN-γ signaling prompted us to ask whether IFN-γ triggered ROS accumulation in rela−/− MEFs. For these studies, we used CM-H2DCFDA, a cell-permeable substrate that is converted to a fluorescent product upon oxidation by ROS. Since prolonged ROS accumulation can trigger loss of mitochondrial membrane potential (Δψm), we also used TMRE as an indicator of Δψm integrity. TMRE is a fluorescent probe that accumulates in mitochondria in a Δψm-dependent manner. As Δψm dissipates (for example, following ROS accumulation), TMRE escapes from mitochondria into the cytosol and nucleus. Thus, loss of Δψm can be visualized by a change in TMRE localization from a punctate mitochondrial localization to a diffuse cytosolic/nuclear distribution. rela+/+ and rela−/− MEFs were loaded with CM-H2DCFDA and TMRE prior to stimulation with IFN-γ and analyzed by confocal imaging. Both rela+/+ and rela−/− cells showed punctate TMRE staining and no significant ROS production before treatment with IFN-γ (Fig. 7A). No substantial increase in ROS levels was observed in rela+/+ MEFs up to 72 h after IFN-γ stimulation, suggesting that either IFN-γ does not induce much ROS accumulation in wild-type MEFs or that ROS is efficiently scavenged in these cells (Fig. 7A and B). In contrast, in rela−/− MEFs, ROS levels increased gradually starting from 12 h and peaked between 24 to 48 h after IFN-γ treatment to amounts ~15-fold higher than those found in untreated cells (Fig. 7A and B). Parallel examination for TMRE fluorescence revealed continued punctate staining in rela−/− MEFs for up to 48 h, followed by dramatic leakage into the cytosol and nucleus between 48 and 72 h (Fig. 7A and C). The differences in kinetics between ROS accumulation and loss of Δψm in IFN-γ-treated rela−/− MEFs suggest that ROS accrual to critical levels over ~48 h is necessary to precipitate Δψm collapse. Collectively, these findings demonstrate that rela−/− MEFs are defective in their ability to eliminate ROS following IFN-γ stimulation. As a consequence, ROS buildup results in catastrophic loss of Δψm.
To establish a causal link between IFN-γ-induced RIP1-dependent necroptosis and ROS accumulation in rela−/− MEFs, we asked whether RIP1 activity was responsible for increased ROS levels seen in these cells. Accordingly, we pretreated rela−/− MEFs with Nec-1 at a dose (25 μM) that effectively inhibits cell death in these cells (Fig. 2A) and measured ROS production 24 and 48 h after IFN-γ treatment. As before, we found that IFN-γ increased ROS levels in untreated rela−/− MEFs by ~15-fold within 24 h. In rela−/− MEFs pretreated with Nec-1, however, ROS production following IFN-γ treatment was almost completely inhibited (Fig. 8A and B). These results indicate that RIP1 lies upstream of ROS generation during IFN-γ-triggered necroptosis.
Our results thus far implicate unchecked ROS accrual in the absence of sod2 induction in rela−/− MEFs as a probable cause of death of these cells following IFN-γ stimulation. We therefore tested whether overexpression of MnSOD could rescue rela−/− MEFs from necroptosis. As a start to these studies, we first examined the ability of ectopic MnSOD to quench ROS production in rela−/− MEFs. We infected rela−/− MEFs with recombinant adenovirus expressing MnSOD (Ad.MnSOD) 2 days prior to treatment with IFN-γ. As a control, we also infected parallel populations of rela−/− MEFs with an adenovirus expressing luciferase (Ad.Luc). We were careful to choose a multiplicity of infection (MOI; 20) that resulted in MnSOD levels similar to what is seen after IFN-γ treatment of rela+/+ MEFs (i.e., 2- to 3-fold over basal levels) (Fig. 9A, inset). Cells were subsequently treated with IFN-γ, and ROS levels were measured by quantifying CM-H2DCFDA fluorescence at 24 h posttreatment. Overexpressed MnSOD effectively reduced IFN-γ-stimulated ROS levels to that seen in unstimulated cells (Fig. 9A).
We next asked whether restoring MnSOD to wild-type levels in rela−/− MEFs can protect these cells from IFN-γ-triggered necroptosis. We infected rela−/− MEFs with either Ad.Luc or Ad.MnSOD (MOI of 20) 2 days before treatment with IFN-γ. As a positive control, we included N-acetyl cysteine (NAC), a well-established small-molecule antioxidant, in these experiments. NAC serves as a precursor for the synthesis of glutathione, which in turn is a substrate for antioxidant GPXs (65). As shown in Fig. 9B, IFN-γ treatment reduced viability of Ad.Luc-infected rela−/− MEFs to ~40% of untreated controls, comparable to the effect of IFN-γ on uninfected rela−/− MEFs at the same time point. Consistent with a role for ROS in mediating IFN-γ-induced necroptosis in the absence of NF-κB, NAC was able to rescue viability of IFN-γ-treated rela−/− MEFs to ~70% of untreated controls. Importantly, overexpression of MnSOD also rescued IFN-γ-treated rela−/− MEFs to the same extent (Fig. 9B). Supplementing Ad.MnSOD-infected cells with NAC did not further augment MnSOD's protective ability, in agreement with our finding that MnSOD by itself is sufficient to quench most of the ROS produced following IFN-γ stimulation. Although these experiments identify sod2 as a primary target of the NF-κB cytoprotective response to IFN-γ, note that overexpression of MnSOD did not fully rescue rela−/− MEFs from IFN-γ-induced necroptosis even though it could efficiently scavenge most of the ROS being generated. These data indicate that other NF-κB target genes are necessary to complement sod2 in maintaining cell survival during cytokine responses.
To determine whether MnSOD was necessary for protection against IFN-γ, we took advantage of the fact that HeLa cells succumb to IFN-γ by necroptosis in a RelA-dependent manner (Fig. 1D and and2E).2E). We ablated MnSOD expression in HeLa cells by RNAi and challenged these cells with IFN-γ in the presence or absence of Nec-1 pretreatment. For comparison, we ablated RelA expression in parallel populations of cells, as described for Fig. 1D. As noted previously (Fig. 1C), we found that IFN-γ was toxic to HeLa cells even in the absence of specific knockdown and reduced their viability by ~40% (Fig. 10A and B). HelA cells in which MnSOD was ablated by RNAi, however, manifested ~80% loss of viability in the same time frame, results that were comparable to loss of RelA itself (Fig. 10A and B). Nec-1 pretreatment rescued both MnSOD- and RelA-deficient cells from IFN-γ-induced cell death (Fig. 10A and B). These results demonstrate that MnSOD is essential for protecting cells from IFN-γ-initiated necroptosis.
Alternate signaling pathways regulated by IFN-γ are areas of emerging interest, and several signaling cascades can be activated downstream of the IFN-γ receptor (46, 58). Many of these nonclassical pathways have transcriptional consequences, and microarray experiments have revealed that >100 genes are differentially regulated by IFN-γ in the absence of STAT1 (23, 45). In this study, we demonstrate that NF-κB represents one such pathway activated by IFN-γ. Using cells deficient in the NF-κB subunit RelA, we find that NF-κB may regulate ~10 to 20% of all ISGs in MEFs. Consistent with roles for IFN-γ and NF-κB in immune cell activation and proinflammatory responses, several RelA-dependent ISGs were found to encode modulators of inflammation (such as chemokines cxcl11, ccl3, and cxcl16, interleukins il15 and il18, the IL-1β-converting enzyme casp1, and tlr2).
A previous report showing activation of NF-κB by IFN-γ proposed a mechanism of NF-κB regulation dependent on the double-stranded RNA (dsRNA)-dependent protein kinase PKR (14). In contrast, other studies have demonstrated that IFN-γ stimulation led to IKK activation and induction of a subset of ISGs but suggest that IKKs regulate gene expression without NF-κB DNA binding (50–52). We found that IFN-γ can, indeed, activate canonical NF-κB via IKKβ but to a significantly lesser degree than TNF-α. As similar results have been reported with type I IFNs (64), these findings indicate that IKKs may regulate ISG expression by two different mechanisms, one relying on direct promoter activation by canonical NF-κB and another involving NF-κB signaling components but not requiring interaction with DNA. Indeed, our DNA microarray data reveal that only a subset of ISGs reported to be regulated by IKKs are dependent on the downstream NF-κB transcription factor subunit RelA, arguing for NF-κB-dependent and -independent gene regulation by IKKs (52). For example, cxcl10, previously identified as an IKKβ-dependent ISG (52), was found in our arrays to be RelA independent (data not shown).
It is currently unclear how IKKs are activated by IFN-γ, but the mechanism involved appears to be dependent on Jak1 and PKR (14) but independent of STAT1. Our time course analyses demonstrate that NF-κB complex formation reaches maximal levels approximately 1 h after addition of IFN-γ. While these kinetics are significantly slower than those of classical Jak1/2-mediated STAT1 activation, the fact that IFN-γ activates NF-κB in the presence of actinomycin D suggests a direct cytoplasmic mechanism of NF-κB stimulation without the need for gene transcription. In addition to PKR (14), a potential adaptor molecule linking IFN-γ to the NF-κB signaling apparatus is the TNF receptor-associated death domain-containing protein (TRADD). Although originally identified as a key NF-κB-activating component of TNF-α signaling, recent studies using tradd−/− mice have revealed that TRADD participates in NF-κB activation by additional innate immune pathways, including those involving TLRs and RIG-1-like receptors (RLRs) (10, 20, 44). Intriguingly, two reports have identified TRADD as an interacting partner for STAT1, suggesting possible roles for TRADD in IFN signaling (62, 63). Although preliminary analyses indicate that TRADD is not essential for STAT1 activation itself (10), an examination of its role in IFN-induced NF-κB signaling remains to be performed.
When the IKKβ–NF-κB signaling axis downstream of IFN-γ is compromised, our data show that IFN-γ utilizes a Jak1-dependent mechanism to trigger RIP1-mediated ROS production and eventual necroptosis. While the mechanism by which IFN-γ stimulates RIP1-dependent mitochondrial ROS production is currently unclear, our data demonstrate that ripk1 is a direct transcriptional target of IFN-γ (data not shown). Since overexpression of RIP1 can by itself trigger the molecular sequelae of cell death in susceptible cells (54), our results suggest that ripk1 induction by IFN-γ represents one mechanism by which IFN-γ triggers cytotoxicity in rela−/− and ikkβ−/− MEFs. However, STAT1 deficiency was only partially able to rescue cells from IFN-γ-triggered necroptosis (especially compared to Jak1 loss) (Fig. 3), indicating that IFN-γ very likely employs diverse transcriptional and nontranscriptional mechanisms to potentiate cytotoxicity. In this regard, an antisense RNA screen for effectors of IFN-γ-induced cell death in HeLa cells identified several proteins, collectively called death-associated proteins (DAPs), that were required for IFN-γ-triggered cell death in HeLa cells (17). Of the DAPs, the serine-threonine kinase DAPK1 is perhaps the best studied, and, like ripk1, is encoded by an ISG (21, 48, 49). Further, both RIP1 and DAPK1 have death domains that are commonly involved in homotypic interactions. Given that IFN-γ treatment of HeLa cells can be largely rescued by inhibition of RIP1 kinase activity, it is tempting to speculate that RIP1 and DAPK1 function in the same necroptotic death pathway.
During TNF-α-triggered necroptosis, RIP1 associates with the homologous serine/threonine kinase RIP3 to form a pronecrotic kinase complex (11, 28). Many of the bioenergetic events suggested to lie downstream of the RIP1-RIP3 complex produce ROS, either directly or inadvertently (59, 66). Mitochondrial ROS generation most likely represents an unavoidable consequence of increased mitochondrial metabolism downstream of RIP1/RIP3 (59, 66). For example, the putative RIP3 substrates glycogen phosphorylase, glutamate ammonia ligase, and glutamate dehydrogenase 1 all function upstream of the tricarboxylic acid cycle (TCA) to drive oxidative phosphorylation-mediated ATP biogenesis via the electron transport chain (66). Conceivably, RIP1-RIP3 activation by TNF-α and IFN-γ thus ramps up cellular metabolism by such mechanisms to stimulate ATP production during host innate immune and inflammatory responses. ATP generated in this way serves the twin purposes of (i) acting as a damage-associated molecular pattern (DAMP) when released from dying cells and (ii) fueling antimicrobial enzymes such as NADPH oxidase, myeloperoxidase, and nitric oxide synthase (7, 57). During oxidative phosphorylation, however, complex I and III of the electron transport chain can prematurely leak electrons to oxygen, producing superoxide (37). Normally, the unstable superoxide anions produced in this fashion are scavenged by MnSOD and do not accrue in mitochondria. If superoxide is allowed to accumulate, the resultant oxidative stress causes perturbation of mitochondrial integral membrane proteins and results in mitochondrial membrane permeabilization, Δψm loss, and eventual ATP depletion (37). NF-κB-mediated activation of sod2 expression by IFN-γ thus ensures mitochondrial integrity by scavenging by-product mitochondrial ROS during inflammatory responses (8, 37, 47).
In contrast to the process of mitochondrial ATP biogenesis (which generates ROS as a result of leaky electron transport), dedicated systems exist that produce ROS for the express purpose of killing intracellular pathogens. For example, the membrane-associated NADPH oxidase system is rapidly activated by TNF-α to generate ROS in microbe-containing phagosomes (36, 41). In this case, RIP1 participates in directly stimulating NADPH oxidase activity in response to TNF-α to promote pathogen clearance. Mutations in human NADPH oxidase subunits cause chronic granulomatous disease (CGD), characterized by susceptibility to recurrent bacterial and fungal infections, underscoring the importance of this mechanism in host innate immunity (40). Taken together, these findings suggest that the location of RIP1-RIP3-mediated ROS production is crucial to the timing of necroptotic outcome. In mitochondria, ROS buildup as a by-product of increased metabolic flux is toxic, and antioxidant enzymes are necessary to rapidly eliminate ROS as it is produced. Phagosomal ROS accrual, however, is necessary for microbicidal activity and is probably better tolerated than mitochondrial ROS. The fact that the NADPH oxidase subunit Nox1 is also required for TNF-α-mediated necroptosis nevertheless indicates that even nonmitochondrial ROS is eventually lethal (36). Necrotic cells destroyed in this manner, however, may catalyze the host defense by releasing DAMPs and other immune cell stimulators.
In summary, we show that IFN-γ utilizes IKKβ to activate canonical NF-κB in MEFs. When NF-κB signaling is compromised (for example, in rela−/− or ikkβ−/− MEFs), IFN-γ instead activates RIP1-dependent necroptotic cell death. Necroptosis in the absence of NF-κB proceeds via progressive accumulation of mitochondrial ROS, and we have identified sod2 as an NF-κB target gene that encodes MnSOD to quench ROS and prevent its buildup in mitochondria. Collectively, these data are consistent with a model in which IFN-γ activates NF-κB to buffer mitochondria during proinflammatory signaling and protect cells from RIP1-driven necroptotic death (depicted schematically in Fig. 11).
We thank Yutaro Kumagai, Taro Kawai, and Shizuo Akira for myd88−/−/trif−/− cells. We also thank Yuesheng Li for generating microarray data. We are grateful to Glenn Rall and Christoph Seeger for valuable comments.
This work was supported by an ACS Research Scholar Grant (RSG-09-195-01-MPC) to S.B. and National Institutes of Health grant (RO1-HL086699) to M.M. Additional funds were provided by the Fox Chase Cancer Center via institutional support of the Kidney Cancer Keystone Program.
R.J.T. and S.H.B. performed most of the experiments and participated in writing the manuscript. K.I. and K.M. carried out all experiments relating to mitochondrial ROS and Δψm measurements. S.N. performed chromatin immunoprecipitation experiments. M.J.S. analyzed microarray data. A.A.B. generated rela−/− and ikkβ−/− MEFs and littermate control MEFs and edited the manuscript. M.M. oversaw work performed by K.I. and K.M. and edited the manuscript. S.B. designed the experiments, interpreted data, and wrote the manuscript.
We declare that we have no conflicts of interest.
Published ahead of print on 16 May 2011.