Interferon-induced RIP1/RIP3-mediated necrosis requires PKR and is licensed by FADD and caspases

Significance

The interferons are small secreted proteins with powerful antiviral and cytotoxic properties. Here, we outline a signaling pathway activated by interferons that results in the precipitous necrotic death of susceptible cells. Interferon-induced necrosis proceeds via a novel, progressive mechanism that requires RNA transcription, as well as the sequential activity of three serine-threonine kinases: PKR, RIP1, and RIP3. This pronecrotic kinase cascade is normally held in check by FADD and caspases. As FADD can be disabled by phosphorylation during mitosis, our findings suggest the existence of a putative cell cycle-dependent checkpoint that licenses interferon-induced necrosis.

Abstract

Interferons (IFNs) are cytokines with powerful immunomodulatory and antiviral properties, but less is known about how they induce cell death. Here, we show that both type I (α/β) and type II (γ) IFNs induce precipitous receptor-interacting protein (RIP)1/RIP3 kinase-mediated necrosis when the adaptor protein Fas-associated death domain (FADD) is lost or disabled by phosphorylation, or when caspases (e.g., caspase 8) are inactivated. IFN-induced necrosis proceeds via progressive assembly of a RIP1–RIP3 “necrosome” complex that requires Jak1/STAT1-dependent transcription, but does not need the kinase activity of RIP1. Instead, IFNs transcriptionally activate the RNA-responsive protein kinase PKR, which then interacts with RIP1 to initiate necrosome formation and trigger necrosis. Although IFNs are powerful activators of necrosis when FADD is absent, these cytokines are likely not the dominant inducers of RIP kinase-driven embryonic lethality in FADD-deficient mice. We also identify phosphorylation on serine 191 as a mechanism that disables FADD and collaborates with caspase inactivation to allow IFN-activated necrosis. Collectively, these findings outline a mechanism of IFN-induced RIP kinase-dependent necrotic cell death and identify FADD and caspases as negative regulators of this process.

Interferons (IFNs) are pleiotropic cytokines classified into two primary groups, type I (predominantly α/β) and type II (γ). Both classes of IFNs exert their effects via similar Janus kinase (JAK)-signal transducers and activators of transcription (STAT)-dependent signaling cascades to induce the expression of over 500 genes (1). Such IFN-stimulated genes (ISGs) have been reasonably well characterized in the context of antiviral or immune-modulatory signaling, but less is known about how they collaborate to mediate the cytotoxic and antiproliferative effects of IFNs.

Recent studies have shed light on a new form of regulated cell death that is activated when caspase-dependent apoptotic pathways are inhibited. This mode of necrotic cell death, sometimes called “necroptosis,” requires the serine-threonine kinases receptor-interacting protein 1 (RIP1) and RIP3, and results from overproduction of reactive oxygen species (ROS) and eventual mitochondrial dysfunction (2, 3). Strict negative control of the pronecrotic kinases RIP1 and RIP3 are essential for several aspects of mammalian development and homeostasis, including immune cell proliferation and progression through embryogenesis (4). The proteins FADD, caspase 8, and c-FLIP represent three such negative regulators; in the absence of any of these molecules, the RIP kinases trigger inopportune necrosis, often with severe consequences for the host (4). The core necrosis machinery is thus carefully regulated to execute cell death only in specific contexts, but how this regulation is achieved and which other upstream stimuli exploit RIP kinases to activate necrosis are still relatively poorly described.

In the present study, we show that both IFN-γ and IFN-α/β trigger a powerful necrotic program when the adaptor protein FADD is absent or disabled by phosphorylation, or when caspases are inactivated. Using fadd^−/− murine embryo fibroblasts (MEFs), we demonstrate that IFNs activate RIP1 by a progressive transcription-dependent mechanism requiring Jak/STAT signaling and initiated by the RNA-responsive kinase PKR. IFN-induced necrosis also requires the kinase RIP3, but is likely not responsible for the RIP1/3-dependent embryonic lethality of fadd^−/− mice. Finally, we identify phosphorylation of FADD as a putative mechanism that collaborates with caspase inactivation to license IFN-induced necrosis, and show that arresting cells in G₂/M—an event that triggers phosphorylation of FADD—also sensitizes cells to IFN-induced necrotic death. Collectively, these findings provide a molecular framework for IFN-triggered necrosis and propose mechanisms licensing its execution.

Results

FADD Protects Cells from IFN-γ–Activated Necrosis by Preventing Formation of the RIP1–RIP3 Necrosome.

During the course of our previous work (5, 6), we unexpectedly observed that the both IFN-α/β and IFN-γ were extraordinarily toxic to subconfluent, early-passage fadd^−/− MEFs, while only minimally affecting the growth of wild-type cells. Notably, this phenotype was significantly attenuated by growing fadd^−/− MEF monolayers to confluency—allowing our earlier studies (5, 6) to be performed—or by the serial passaging, immortalization, or transformation of these cells. As IFN-γ was a more potent inducer of cell death in fadd^−/− MEFs than any of the type I IFNs tested, we primarily focused on IFN-γ for the rest of these studies. Doses of IFN-γ in the 1–10 ng/mL range effectively killed ∼60–90% of fadd^−/− MEFs within 24 h (Fig. 1 A and B) and virtually 100% of these cells by 48 h posttreatment. Reexpressing FADD in fadd^−/− MEFs fully protected these cells from IFN-γ challenge (Fig. 1C), and acute ablation of FADD expression in HeLa cells by RNAi accelerated IFN-γ–mediated death (Fig. S1). Together, these experiments demonstrate that FADD protects mammalian cells from IFNs.

Fig. 1.

FADD protects cells from IFN-γ–induced necrosis by inhibiting formation of a RIP1–RIP3 necrosome. (A) Photomicrographs of fadd^+/+ and fadd^−/− MEFs treated with mIFN-γ (5 ng/mL) for 48 h. (B) Fadd^+/+ and fadd^−/− MEFs were treated with IFN-γ for 48 h, after which cell viability was determined by Trypan blue exclusion analysis. Error bars represent mean ± SD; n = 3. *P < 0.05, **P < 0.005. (C) Retrovirally transduced populations of fadd^−/− MEFs expressing either wild-type FADD or empty vector (Vec) were treated with IFN-γ (5 ng/mL), and viability was determined 48 h posttreatment. Restoration of FADD expression in fadd^−/− MEFs was confirmed by immunoblotting (Inset). Error bars represent mean ± SD; n = 3. **P < 0.005. (D) Fadd^−/− MEFs were pretreated with either Nec-1 or the pancaspase inhibitor z-VAD for 1 h before exposure to IFN-γ (5 ng/mL). Viability was determined 48 h after IFN-γ treatment. Error bars represent mean ± SD; n = 3. *P < 0.05, **P < 0.005. (E and F) Populations of fadd^−/− MEFs expressing two distinct shRNAs (Sigma) to either RIP1 (E) or RIP3 (F) were treated with IFN-γ (5 ng/mL) and examined for survival 48 h posttreatment. Nontargeting shRNAs were used as controls. Knockdown of RIP1 and RIP3 was confirmed by immunoblotting (Inset). Error bars represent mean ± SD; n = 3. **P < 0.005. (G) Immunoblot analysis of anti-RIP3 and anti-RIP1 IPs shows that slower migrating forms of RIP1 and RIP3 are preassociated with each other and increase in abundance over 6 h of IFN-γ (5 ng/mL) treatment only in the absence of FADD. The position of relevant molecular weight markers in this and subsequent figures are shown in kilodaltons to the Left of the blots. (H) λ phosphatase (PPtase) treatment of anti-RIP3 and anti-RIP1 IPs demonstrates that the modified forms of RIP1 and RIP3 in fadd^−/− MEFs are phosphorylated versions of these kinases. Gaps between lanes indicate where irrelevant lanes have been removed.

FADD typically functions as an adaptor protein in signaling cascades. Reasoning that FADD prevented cell death by physically interacting with a putative death effector molecule(s) to inhibit its activity, we carried out a yeast two-hybrid screen with full-length murine FADD, or its various deletion mutants, as bait. From this screen, we identified a total of nine interacting clones, seven of which encoded fragments of the well-described FADD-interacting adaptor protein TRADD. The two remaining clones, both isolated with the death domain of FADD (amino acids 90–204) as bait, encoded polypeptides corresponding to the C-terminal region of the pronecrotic kinase RIP1 (Fig. S2). Given this finding, we asked whether FADD mediated its survival function against IFNs by preventing RIP1-mediated necrosis. We therefore treated fadd^−/− MEFs with Necrostatin-1 (Nec-1), a selective RIP1 kinase inhibitor (7), before challenging them with IFN-γ. Nec-1 pretreatment increased survival of IFN-γ–treated fadd^−/− MEFs (from ∼20% to over 60% in a dose-dependent manner), whereas bioactive concentrations of the caspase inhibitor z-VAD had no protective effect (Fig. 1D). Two populations of fadd^−/− MEFs, each expressing a distinct shRNA to RIP1, also demonstrated significantly increased resistance to IFN-γ, compared with control cells expressing nonsilencing shRNAs (Fig. 1E). Similarly, shRNAs targeting the RIP family kinase RIP3 protected >80% of fadd^−/− MEFs from IFN-γ–induced cell death (Fig. 1F). These results indicate that IFN-γ employs the kinases RIP1 and RIP3 to execute a necrotic death program. Notably, HeLa cells do not express RIP3 (8) but still succumb to IFN-γ–induced RIP1-dependent necrotic death (ref. 9 and Fig. S1), indicating that IFN-γ, in a cell type-specific manner, can also activate RIP1-mediated necrosis without need for RIP3.

Given that RIP1 associates with RIP3 during TNF-α–induced necrosis (10), we asked whether FADD, by directly interacting with RIP1, prevented formation of a putative IFN-γ–activated RIP1–RIP3 complex. We treated wild-type (fadd^+/+) and fadd^−/− MEFs with IFN-γ for up to 6 h, immunoprecipitated RIP3 from these cells, and examined precipitates for RIP1. In wild-type MEFs, low amounts of RIP1 were found constitutively associated with RIP3 (Fig. 1G). In fadd^−/− MEFs, however, we noticed the appearance of a modified, slower migrating form of RIP1 in anti-RIP3 immunoprecipitates (IPs). This modified form of RIP1 was found basally associated with RIP3 and gradually increased in abundance over a 6 h time course (Fig. 1G). Similarly, reciprocal coimmunoprecipitation experiments with anti-RIP1 antibodies revealed that a slower migrating form of RIP3 was prebound with RIP1 in fadd^−/−MEFs, and also increased in association with RIP1 over a time course of 6 h after IFN-γ treatment (Fig. 1G). Exposure of RIP3 and RIP1 IPs to λ phosphatase resulted in loss of the slower migrating forms of RIP1 and RIP3, respectively, demonstrating that these forms represent phosphorylated versions of the RIP kinases (Fig. 1H).

IFN-γ–Activated Necrosome Formation and Necrosis Requires Jak1, STAT1, and Transcription.

Classical IFN-γ–induced signaling is mediated by Jak1/2-driven phosphorylation of STAT1 (1). To determine whether this Jak–STAT axis is necessary for RIP1–RIP3 association and consequent necrosis, we used RNAi to stably ablate expression of either Jak1 or STAT1 in fadd^−/− MEFs (Fig. 2A and Fig. S3A). Knockdown of Jak1 or STAT1—or pharmacological inhibition of Jak1—rescued viability of IFN-γ–treated fadd^−/− MEFs from ∼20% in control cells to ∼90% (for Jak1) or ∼70% (for STAT1; Fig. 2A and Fig. S3B). Consistent with these results, ablation of either Jak1 or STAT1 expression almost completely abolished IFN-γ–driven RIP1–RIP3 complex assembly (Fig. 2B and Fig. S3C). Collectively, these data demonstrate that IFN-γ uses classical Jak–STAT signaling to promote formation of the RIP–RIP3 pronecrotic kinase complex and trigger necrosis.

Fig. 2.

IFN-γ activates RIP1–RIP3 necrosome formation by a mechanism requiring Jak/STAT-dependent transcription. (A) shRNAs to Jak1 or STAT1 protect fadd^−/− MEFs from IFN-γ. Viability was determined by Trypan blue exclusion analysis. Knockdown of Jak1 and STAT1 was verified by immunoblotting (Inset). Error bars represent mean ± SD; n = 3. **P < 0.005. (B) Immunoblot analysis of anti-RIP3 IPs from control- or STAT1 shRNA-expressing fadd^−/− MEFs shows that STAT1 is required for IFN-γ–driven RIP1–RIP3 necrosome formation. (C and D) Inhibiting transcription with Act D (100 ng/mL) prevents IFN-γ–driven RIP1–RIP3 necrosome formation (C), but does not significantly affect TNF-α–triggered necrosome assembly (D). TCZ, TNF-α (50 ng/mL) plus cycloheximide (500 ng/mL) plus z-VAD (50 μM). Note that TCZ-induced necrosome formation was examined in fadd^+/+ MEFs, as TNF-α, unlike IFN-γ, requires FADD to assemble the necrosome in MEFs (30). (E and F) Nec-1 (50 μM) pretreatment does not prevent IFN-γ (5 ng/mL)–driven necrosome formation (E) but efficiently prevents TNF-α–induced necrosome formation in fadd^+/+ MEFs (F). TCZ, TNF-α (50 ng/mL) plus cycloheximide (500 ng/mL) plus z-VAD (50 μM). Gaps between lanes indicate where irrelevant lanes have been removed.

As formation of the IFN-γ–induced RIP1–RIP3 complex required the transcription factor STAT1, and as this complex formed in a gradual, accumulative manner, we sought to determine whether IFN-γ used a transcriptional mechanism to stimulate association between RIP1 and RIP3. Inhibiting transcription with the RNA polymerase II (Pol II) inhibitor Actinomycin D (Act D) prevented IFN-γ–driven RIP1–RIP3 complex formation (Fig. 2C), but had no effect on generation of the TNF-α–stimulated RIP1–RIP3 complex (Fig. 2D). Thus, although IFN-γ and TNF-α both use RIP1 and RIP3 to effect necrosis, only IFN-γ requires ongoing Pol II-dependent transcription to activate these kinases. Of note, we were unable to demonstrate that Act D could rescue the viability of IFN-treated fadd^−/− MEFs, as a dose of Act D as low as 5 ng/mL was by itself very toxic (<50% viability) to these cells over 24 h (Fig. S3D).

The Kinase PKR Is Required for IFN-γ–Activated Necrosis.

To identify how an IFN-γ–driven transcriptional mechanism might activate RIP1–RIP3 complex formation, we first ruled out the possibility that IFN-γ induced the expression of a member of the TNF superfamily, which then activated a necrotic program in fadd^−/− MEFs (Fig. S4). We then took advantage of the surprising observation that Nec-1 did not affect the abundance of either basal or IFN-γ–driven necrosome formation (Fig. 2E), although it was capable of rescuing fadd^−/− MEFs from IFN-induced necrosis (Fig. 1D). This result was unexpected, as Nec-1 functions in TNF-α–induced necrosis to prevent phosphorylation-driven formation of the RIP1–RIP3 complex, as shown in Fig. 2F (10). Thus, the step in IFN-γ–triggered necrosis inhibited by Nec-1 is not RIP1–RIP3 complex formation, and the kinase that drives assembly of the phospho-RIP1–RIP3 complex is probably not RIP1 itself.

Hypothesizing that the initiating kinase was activated by transcriptional up-regulation and thus likely encoded by an ISG, we used DNA microarray analysis to identify all kinases whose expression was transcriptionally induced by both IFN-γ and IFN-β. A total of nine genes encoding kinases and kinase-like proteins, including PKR (encoded by the eif2ak2 gene) and the necrosome adaptor pseudokinase MLKL, were found induced at least twofold by both IFNs (Fig. 3A). Of these, the dsRNA-responsive antiviral kinase PKR was also isolated as a RIP1-interacting partner in a parallel yeast two-hybrid screen. We thus focused our attention on PKR.

Fig. 3.

The kinase PKR is necessary for IFN-γ–activated RIP1–RIP3 complex formation and necrosis. (A) Kinases induced at least twofold by both IFN-γ and IFN-β within 6 h of treatment in fadd^+/+ MEFs. Heat bar = log₂ scale. Signal in untreated MEFs was normalized to 1 (yellow). The arrow shows eif2ak2, the gene encoding PKR. (B) Fadd^−/− MEFs are rescued from IFN-γ–induced cell death by the PKR inhibitors C16 and 2-aminopurine (2AP) and by a dominant-negative mutant of PKR (hPKR Δ6, missing amino acids 361–366, LFIQME), but not an empty vector control. Viability was determined by Trypan blue exclusion analysis. Error bars represent mean ± SD; n = 3. *P < 0.05, **P < 0.005. (C) Two distinct shRNAs to PKR protect fadd^−/− MEFs from IFN-γ (5 ng/mL). Knockdown of PKR was verified by immunoblotting (Inset). Error bars represent mean ± SD; n = 3. **P < 0.005. (D) IFN-γ (5 ng/mL) treatment potently activates PKR within 2 h, without significantly affecting total PKR levels. Pretreatment of fadd^−/− MEFs with Act D inhibited IFN-γ–induced PKR activation, whereas C16 abolished all detectable PKR activity, as measured using an anti–phospho-PKR antibody. For phospho-PKR immunoblots, cells were treated with the phosphatase inhibitor calyculin A (200 nM, 1 h) before lysis. (E) PKR is preassociated with a phosphorylated RIP1 in fadd^−/− MEFs, and this association is robustly stimulated by IFN-γ (5 ng/mL) treatment. (F) IFN-γ (5 ng/mL)–driven PKR–RIP1 association is inhibited by Act D pretreatment, indicating dependence on mRNA transcription. (G) IFN-γ (5 ng/mL)–driven PKR–RIP1 association is disrupted by pretreatment with C16, demonstrating a requirement for PKR kinase activity for this interaction. (H) Recombinant active PKR can phosphorylate recombinant wild-type RIP1 (Left) or kinase-dead (RIP1 K45A) mutant RIP1 (Right) proteins in vitro. (Left) Nec-1 (50 μM) was used to prevent RIP1’s autophosphorylation activity from confounding interpretation of results. Note that Nec-1 pretreatment completely blocks RIP1 autophosphorylation (lane 4), but does not significantly affect PKR kinase activity (lanes 2 and 5). Gaps between lanes indicate where irrelevant lanes have been removed. (Right) PKR inhibitor 2AP (lanes 6 and 7) inhibits PKR autophosphorylation and blocks phosphorylation of RIP1 K45A. The use of kinase-dead RIP1 in this experiment was to examine whether RIP1 can serve as a substrate for PKR, without interference from RIP1’s intrinsic kinase activity.

To determine whether PKR was involved in IFN-γ–induced necrosis, we first asked whether a selective small-molecule inhibitor of PKR [called C16 (11)] can protect fadd^−/− MEFs from IFN-γ. C16 efficiently rescued fadd^−/− MEFs from IFN-γ–induced cell death and, indeed, afforded greater protection against IFN-γ than even Nec-1 (Fig. 3B). A second small-molecule PKR inhibitor 2-aminopurine (2AP) also protected fadd^−/− MEFs from IFN-γ (Fig. 3B), and populations of fadd^−/− MEFs expressing distinct shRNAs to PKR, or a dominant-negative mutant of this kinase [PKRΔ6 (12)], were largely resistant to the toxic effects of IFN-γ (Fig. 3 B and C). Extending these observations, we found that two other cell types susceptible to IFN-γ–induced necrosis—rela^−/− MEFs, and murine macrophages (9, 13)—were also rescued from IFN-γ by pretreatment with PKR inhibitors (Fig. S5). These findings identify a universal role for PKR in IFN-γ–triggered necrosis.

PKR Is Activated by IFN-γ and Associates with RIP1.

In agreement with a function for PKR in initiating IFN-driven necrosis, exposure of fadd^−/− MEFs to IFN-γ increased PKR activity in a manner requiring ongoing transcription (Fig. 3D). Notably, and despite the fact that PKR is encoded by an ISG, its protein levels remained relatively unchanged during this period (Discussion). As we identified a direct PKR–RIP1 union by yeast two-hybrid screening, and as these two kinases have previously been shown to interact with each other (14), we next examined whether PKR associates with RIP1 during IFN-γ–triggered necrosis. A slower migrating form of RIP1 was precomplexed with PKR (as it was with RIP3), and this association was robustly stimulated by IFN-γ treatment (Fig. 3E). IFN-γ–induced PKR–RIP1 association required both ongoing transcription (Fig. 3F), as well as PKR’s kinase activity (Fig. 3G). We also found that recombinant active PKR could phosphorylate recombinant wild-type RIP1 (Fig. 3H, Left; RIP1’s intrinsic kinase activity was nullified by preincubation with Nec-1) as well as a kinase-dead K45A mutant of RIP1 (Fig. 3H, Right) in in vitro kinase assays. These results are consistent with a model in which IFN-γ transcriptionally activates PKR to promote its association with—and phosphorylation of—RIP1.

PKR Initiates RIP1–RIP3 Complex Formation After IFN-γ Stimulation.

To determine whether activation of PKR was upstream of RIP1–RIP3 necrosome assembly, we analyzed RIP1–RIP3 complex formation in fadd^−/− MEFs expressing PKR shRNA, and found that both basal and IFN-γ–driven RIP1–RIP3 association were dependent on PKR (Fig. 4A). Importantly, the PKR inhibitor C16—unlike Nec-1—also blocked RIP1–RIP3 necrosome formation (Fig. 4B) but had no effect on either TNF-α–induced RIP1–RIP3 association or consequent necrosis (Fig. 4 C and D), indicating that PKR promotes RIP1–RIP3 association specifically in response to IFN.

Fig. 4.

PKR lies upstream of RIP1–RIP3 necrosome formation. (A) RNAi ablation of PKR expression attenuates IFN-γ (5 ng/mL)–driven RIP1–RIP3 necrosome formation in fadd^−/− MEFs. (B) PKR inhibitor C16 abrogates IFN-γ (5 ng/mL)–driven RIP1–RIP3 necrosome formation in fadd^−/−MEFs. (C and D) PKR is not essential for TNF-α–induced necrosis. The PKR kinase inhibitor C16 does not inhibit TNF-α–driven RIP1–RIP3 necrosome formation in fadd^−/− MEFs (C), nor does it rescue fadd^−/− MEFs from TNF-α–induced necrotic death (D). Cell viability was determined by Trypan blue exclusion analysis. Error bars represent mean ± SD; n = 3. **P < 0.005. (E) Directly activating PKR induces RIP1–RIP3 necrosome formation without need for transcription. Fadd^−/− MEFs pretreated with Act D (100 ng/mL) for 1 h were transfected with poly(I:C) (1 μg/mL in 3 μg/mL Lipofectamine 2000) and examined for RIP1–RIP3 necrosome formation in the presence or absence of C16. Activation of PKR by poly(I:C) was confirmed by immunoblotting with an anti–phospho-PKR antibody. For phospho-PKR immunoblots, cells were treated with the phosphatase inhibitor calyculin A (200 nM, 1 h) before lysis. (F) Directly activating PKR induces necrosis without need for autocrine type I IFN. Fadd^−/− MEFs were pretreated with Jak kinase inhibitor I (0.5 μM) for 1 h before transfection with poly(I:C) (1 μg/mL in 3 μg/mL Lipofectamine 2000) in the presence or absence of C16, 2AP, and Nec-1. Cell viability was measured 48 h posttransfection. Blockade of type I IFN signaling was confirmed by immunoblotting for phospho-STAT1 after exposure of fadd^−/− MEFs to IFN-β (Inset). Error bars represent mean ± SD; n = 3. *P < 0.05, **P < 0.005.

If IFN-γ functions to initiate necrosis primarily by transcriptional activation of PKR, then directly activating PKR in fadd^−/− MEFs will be expected to trigger RIP1–RIP3 complex formation and consequent necrosis without need for IFN signaling. To test this prediction, we stimulated PKR by transfecting cells with the well-established PKR activator poly(I:C) in the presence of inhibitors of transcription or IFN signaling. Transfection of poly(I:C) into fadd^−/− MEFs activated PKR and triggered PKR-dependent RIP1–RIP3 necrosome formation without requirement for transcription (Fig. 4E). Poly(I:C), in the presence of a Jak inhibitor—added to rule out secondary effects that may arise from autocrine type I IFN signaling—also killed ∼50% of fadd^−/− MEFs within 48 h of transfection; this death was inhibited by both Nec-1 and C16 (Fig. 4F). These findings implicate PKR as a critical IFN-γ–induced pronecrotic kinase upstream of RIP1 and RIP3.

Type I IFNs Activate Necrosis by a Mechanism Similar to IFN-γ.

Recombinant IFN-β (as well as IFN-α, but not the proinflammatory cytokines IL-1β, IL-6, and TNF-α) also efficiently killed fadd^−/−MEFs (Fig. 5A). IFN-β–triggered cell death was rescued by pretreatment of fadd^−/− MEFs with Jak inhibitor I, C16, and Nec-1 (Fig. 5B), demonstrating that, like IFN-γ, IFN-β also activates Jak-, PKR-, and RIP1 kinase-dependent necrosis in the absence of FADD. In agreement, fadd^−/− MEFs expressing shRNAs to Jak1, STAT1, PKR, RIP1, and RIP3 were all largely resistant IFN-β–triggered necrotic cell death (Fig. 5B). Notably, RIP1 shRNAs or Nec-1 consistently afforded less protection against IFN-β than they did against IFN-γ, indicating that RIP1 may not be as necessary for type I IFN-driven necrosis as it is for IFN-γ. Nevertheless, the kinetics of IFN-β–induced necrosome formation parallel those of IFN-γ (Fig. 5C), and the molecular steps involved also appear to be largely similar, in that (i) IFN-β also activates PKR in a transcription-dependent manner, and (ii) PKR—but not RIP1—activity is required for necrosome formation. Thus, both type I and type II IFNs use similar Jak–STAT-dependent transcriptional mechanisms to trigger necrosis when FADD is absent.

Fig. 5.

Type I IFNs also induce PKR–RIP1–RIP3-dependent necrosis, but IFNs are likely not responsible for the embryonic lethality of fadd^−/− mice. (A) The type I IFNs IFN-α and IFN-β, but not other proinflammatory cytokines, are toxic to fadd^−/− MEFs. Viability was measured 48 h posttreatment by Trypan blue exclusion analysis. Error bars represent mean ± SD; n = 3. **P < 0.005. (B) Type I IFNs induce Jak/STAT-dependent necrosis requiring the kinases PKR, RIP1, and RIP3. Fadd^−/− MEFs treated with the indicated kinase inhibitors, or stably expressing the indicated shRNAs, were treated with IFN-β (5 ng/mL), and viability was determined 48 h posttreatment. Error bars represent mean ± SD; n = 3.*P < 0.05, **P < 0.005. (C) IFN-β—like IFN-γ—induces progressive RIP1–RIP3 and PKR–RIP1 complex formation in the absence of FADD. (D) IFNs are not solely responsible for embryonic lethality of fadd^−/− mice. Results from intercrosses of Fadd^+/− heterozygotes on a stat1^−/− background are shown. None of 60 live births recorded was fadd^−/− (Upper). By contrast, a similar cross of fadd^+/− mice on a ripk3^−/− background produced live fadd^−/− pups at roughly Mendelian frequency (Lower).

IFNs Are Likely Not Responsible for fadd^−/− Embryonic Lethality.

Fadd^−/− embryos die prenatally from RIP1/RIP3-dependent necrosis, but the upstream stimulus responsible for activating RIP kinases during embryogenesis in the absence of FADD remains unknown (4, 15). To test whether IFN signaling was responsible for the embryonic lethality of fadd^−/− mice, we intercrossed fadd^+/− heterozygotes on an IFN signaling-deficient stat1^−/− background. None of the 60 live births tested from this cross were homozygously null for fadd (Fig. 5D). By contrast, a parallel intercross using fadd^+/− mice on a ripk3^−/− background produced fadd^−/− pups at roughly Mendelian frequency [9 of 40 live births, in agreement with a previous report (16)], indicating that IFN signaling is likely not the trigger for the premature RIP kinase-dependent death of fadd^−/− embryos (Fig. 5D).

Phosphorylation of FADD Licenses IFN-Induced Necrosis.

FADD appears to prevent IFN-activated necrosis by physically binding RIP1 and preventing its interaction with either PKR or RIP3, indicating that signaling mechanism(s) that license IFN necrosis likely function by destabilizing the FADD–RIP1 interaction. Phosphorylation of FADD on S191 (equivalent to S194 in humans) is the best-characterized covalent modification that regulates FADD function (17–19). Hypothesizing that phosphorylation of FADD on S191/S194 may represent one mechanism by which the FADD–RIP1 association can be destabilized, we used the published human FADD–Fas complex (20) as template to model the FADD–RIP1 death domain association. This model predicts that FADD S194 is in close enough proximity to RIP1 to permit hydrogen bonding between S194 and RIP1 K625 (Fig. 6A), and suggests that phosphorylation on S191/194 might disrupt the FADD–RIP1 association to permit IFN-driven necrosome formation.

Fig. 6.

Phosphorylation of FADD licenses IFN-activated necrosis. (A) The putative complex of RIP1 death domain (blue) and FADD death domain (green), modeled from the template of FAS–FADD complex (PDB ID code 3EZQ) is portrayed in ribbon representation. Also shown in sphere representation is the key residue S194 of FADD, and RIP1 residue K625, which is immediately proximal and capable of hydrogen bonding interactions. (B) Fadd^−/− MEFs were stably reconstituted with either wild-type FADD, FADD-S191A, or FADD-S191D and, after confirmation of expression by anti-FADD immunoblot analysis (Inset), examined for survival by Trypan blue exclusion analysis 48 h after IFN-β or IFN-γ treatment. TRAIL plus Act D (50 ng/mL) was used to demonstrate that FADD-dependent apoptosis (measured 12 h posttreatment) was unaffected by mutating S191. Error bars represent mean ± SD; n = 3. *P < 0.05, **P < 0.005, compared with fadd^−/− MEFs reconstituted with WT FADD. (C) Fadd^−/− MEFs stably expressing FADD-S191D were pretreated with the indicated inhibitors (0.5 μM Jak Inh. and C16; 50 μM Nec-1) before exposure to IFN-γ (5 ng/mL). Viability was quantified 48 h posttreatment. Error bars represent mean ± SD; n = 3. *P < 0.05, **P < 0.005. (D) Fadd^+/+ MEFs were treated with Noc for 18 h and examined for phosphorylation of FADD using a FADD-S191P–specific antibody (Inset; also see Fig. S7B) or subsequently treated with IFNs (5 ng/mL each) for a further 30 h, at which time viability was measured. For phospho-FADD immunoblots, cells were treated with the phosphatase inhibitor calyculin A (200 nM, 1 h) before lysis. Error bars represent mean ± SD; n = 3. *P < 0.05. (E) Fadd^+/+ MEFs were pretreated treated with Noc (40 ng/mL, 18 h) followed by IFN-γ (5 ng/mL) in the presence of the indicated inhibitors (0.5 μM Jak Inh. and C16; 50 μM Nec-1) for a further 30 h before quantification of viability. Error bars represent mean ± SD; n = 3. *P < 0.05. (F) Fadd^−/− MEFs reconstituted with either wild-type FADD or FADD-S191A were treated with Noc (40 ng/mL, 18 h) followed by IFN-γ (5 ng/mL) for a further 30 h before quantification of viability. Error bars represent mean ± SD; n = 3. *P < 0.05. (G) Fadd^+/+ MEFs were pretreated treated with Noc (40 ng/mL, 18 h) followed by IFN-γ (5 ng/mL), and necrosome formation was examined in lysates from these cells at the indicated times posttreatment. One hour before addition of IFN-γ, all cells were treated with a low dose (5 μM) of z-VAD to facilitate necrosome formation.

To test this possibility, we reconstituted fadd^−/− MEFs with either wild-type FADD, nonphosphorylatable FADD-S191A, or phosphomimetic FADD-S191D point mutants, and exposed these MEF populations to IFN-β or IFN-γ. As shown in Fig. 6B, both wild-type FADD and FADD-S191A were largely able to rescue fadd^−/− MEFs from the cytotoxic effects of IFN-β and IFN-γ. FADD-S191D–expressing cells, however, were unable to survive in the continued presence of either IFN-β or IFN-γ, and about one-half of these cells died within 48 h of treatment with either IFN (Fig. 6B, dark gray bars). IFN-driven cell death in FADD-S191D cells depended on Jak1, PKR, and RIP1 (Fig. 6C and Fig. S6), but the degree of necrosis manifested in these cells was significantly less than that seen in FADD-deficient cells (Fig. 6 B and C). Notably, FADD-S191D completely restored responsiveness to FADD-dependent apoptotic death induced by the cytokine TRAIL, demonstrating that its inability to prevent necrosis was not the result of some gross defect in FADD expression or function (Fig. 6B).

Finally, we asked whether a stimulus capable of inducing phosphorylation of FADD on S191 sensitized cells to IFN-induced necrosis. Several groups have now shown that mammalian FADD is phosphorylated on this residue during the G₂/M phase of the cell cycle, and that artificially arresting cells in the G₂/M phase robustly stimulates FADD phosphorylation (17). Borrowing from these studies, we used relatively nontoxic doses of the microtubule-destabilizing agent nocodazole (Noc) to robustly trap wild-type MEFs in G₂/M (Fig. S7A) and induce FADD phosphorylation (Fig. 6D, Inset). When arrested cells were exposed to IFNs, both doses of Noc sensitized wild-type MEFs to IFN-induced cell death that was similar in magnitude to what was seen when FADD-S191D–expressing cells were treated with IFNs (Fig. 6D, compare with Fig. 6B). Noc-licensed IFN-driven cell death also depended on Jak1, PKR, and RIP1 activity (Fig. 6E), and—importantly—was significantly attenuated in cells carrying nonphosphorylatable FADD-S191A (Fig. 6F). Noc pretreatment also permitted IFN-γ–driven RIP1–RIP3 necrosome formation in FADD-containing cells (Fig. 6H). Notably, necrosome formation upon Noc treatment was only detected when cells were additionally exposed to a low dose (5 μM) of z-VAD (see below). Together, these results demonstrate that phosphorylation on FADD in its death domain is a putative mechanism for licensing a necrotic outcome, suggest collaborative roles for phosphorylation of FADD and inactivation of caspases in IFN-induced necrosome activation, and identify transition through G₂/M as a possible window of susceptibility to IFN-induced necrosis.

Caspases, Including Caspase 8, Are Critical Regulators of IFN-Induced Necrosis.

As caspases are negative regulators of necrosis in other contexts (2, 4) and as low doses of the caspase inhibitor z-VAD sensitized MEFs to Noc-induced necrosome formation after IFN treatment (Fig. 6H), we examined whether caspases—like FADD—regulated necrotic outcome in response to IFNs. When we exposed wild-type MEFs to the combination of IFN-γ and a fully bioactive dose of z-VAD (50 μM), we observed an ∼80% decrease in cell viability 48 h posttreatment, at which time neither IFN-γ nor z-VAD by themselves had exerted any appreciable toxicity on these cells (Fig. 7A). z-VAD treatment, like FADD ablation, induced the spontaneous assembly of a “basal” RIP1–RIP3 necrosome in wild-type MEFs (Fig. 7B, lane 2). Subsequent exposure of z-VAD–treated cells to IFN-γ increased the abundance of this complex over a 6-h timeframe (Fig. 7B, lanes 3–5). Cell death induced by the combination of IFN-γ and z-VAD was efficiently rescued by pretreatment with a Jak inhibitor, C16, Nec-1, and the antioxidant butyrated hydroxyanisole (BHA) (Fig. 7C), as well as by genetic ablation of STAT1, PKR, and RIP3 (Fig. 7D). IFN-γ or IFN-β also reduced the viability of caspase 8^−/− MEFs (Fig. 7E), albeit to a lesser extent than was seen upon IFN treatment of either fadd^−/− MEFs or z-VAD–exposed wild-type MEFs (compare Fig. 7E to Figs. 1B and 7A). Importantly, a low dose of z-VAD (5 μM) sensitized cells expressing phosphomimetic FADD-S191D—but not those expressing wild-type FADD—to precipitous death within 36 h of IFN-γ exposure, comparable in magnitude to that seen upon IFN treatment of fadd^−/− MEFs, or upon IFN-γplus high-dose z-VAD (50 μM) treatment of wild-type MEFs, for 48 h (Fig. 7F). Taken together, these results identify caspases (such as caspase 8, but likely others as well) as negative regulators of IFN-triggered necrotic death, and show that caspase inactivation can collaborate with FADD phosphorylation to license IFN-induced necrosis.

Fig. 7.

Caspases regulate IFN-induced necrosis. (A) IFN-γ (5 ng/mL), in the presence of the pancaspase inhibitor z-VAD (50 μM), is toxic to wild-type (Fadd^+/+) MEFs over 48 h. Error bars represent mean ± SD; n = 3. **P < 0.005. (B) IFN-γ, in the presence of z-VAD (50 μM), induces progressive RIP1–RIP3 necrosome formation in WT MEFs. (C) IFN-γ (5 ng/mL) plus z-VAD (50 μM)-induced cell death is prevented by pretreatment with Jak inhibitor (500 nM), C16 (500 nM), Nec-1 (50 μM), or the antioxidant butyrated hydroxyanisole (BHA) (50 μM). Cell viability was measured 48 h after IFN/z-VAD treatment. Error bars represent mean ± SD; n = 3. **P < 0.005. (D) MEFs genetically deficient in expression of STAT1, PKR, or RIP3 are largely resistant to IFN-γ (5 ng/mL) plus z-VAD (50 μM)-induced necrosis. The eif2ak2 gene encodes PKR. Error bars represent mean ± SD; n = 3. **P < 0.005. Immunoblots confirming gene knockouts are shown to the Right. (E) Caspase 8 deficiency partially sensitizes MEFs to IFN-induced cell death. Caspase 8^+/+ and caspase 8^−/− MEFs were treated with either IFN-γ (10 ng/mL) or IFN-β (10 ng/mL), and viability was determined 48 h posttreatment. Error bars represent mean ± SD; n = 3. **P < 0.005. (F) Caspase inactivation collaborates with FADD phosphorylation to permit IFN-induced necrosis. Fadd^−/−MEFs reconstituted with either wild-type FADD or FADD-S191D were treated with IFN-γ (5 ng/mL) in the presence of low-dose z-VAD (5 μM), and viability was quantified 36 h posttreatment. As controls, fadd^+/+ MEFs treated with IFN-γ plus z-VAD (50 μM), or fadd^−/− MEFs treated with IFN-γ (5 ng/mL), each for 48 h, are shown. Error bars represent mean ± SD; n = 3. **P < 0.005. (G) Schematic of IFN-activated necrosis. Both type I (predominantly α/β) and type II (γ) IFNs initiate pronecrotic signaling by Jak/STAT-dependent transcriptional activation of the latent kinase PKR (ostensibly through up-regulation of one or more PKR-activating ISG mRNAs, such as PKR’s own mRNA). Once activated, PKR initiates the formation of the “PKR necrosome,” a kinase complex comprising PKR, RIP1, and RIP3, that executes necrosis. FADD and caspases inhibit formation of the PKR necrosome, and this inhibition is relieved when FADD is phosphorylated and/or when caspases are inactivated. Small-molecule inhibitors used in this study are shown in blue.

Discussion

In this study, we show that both type I and type II IFNs activate RIP1–RIP3-dependent necrosis in MEFs when the adaptor protein FADD is either absent or inactivated by phosphorylation, or when caspases are inactivated (see Fig. 7G for schematic). Although both IFN and TNF-α use the kinases RIP1 and RIP3 to activate necrotic death, IFN-triggered necrosis differs from that activated by TNF-α in three important ways: (i) IFN-activated RIP1-RIP3 complex formation requires Jak1/STAT1-mediated active transcription, whereas TNF-α assembles the RIP1–RIP3 kinase complex mainly by cytoplasmic mechanisms; (ii) IFN- but not TNF-α–induced RIP1–RIP3 complex formation required the RNA-responsive kinase PKR; and (iii) IFN-induced necrotic death was revealed when FADD is absent or disabled, whereas TNF-α requires FADD to activate necrosis in MEFs.

Of particular interest was the finding that the RNA-responsive kinase PKR initiates IFN-activated RIP1–RIP3 complex formation by a mechanism requiring ongoing transcription. In its best-described role, latent PKR is activated by viral RNAs during acute virus infections, following which it shuts down viral (and cellular) protein synthesis by phosphorylation of the translation initiation factor eIF2α (21). IFNs, however, activate PKR in the absence of any virus infection, and do so without increasing levels of PKR protein. Based on these observations, we suggest that IFNs induce expression of an mRNA(s), which then activates latent PKR that is already present in the cytoplasm. In support of this hypothesis, we find that exogenously supplied dsRNA can activate PKR and trigger necrosis without need for mRNA transcription. Activation of PKR by cellular RNAs is not without precedent: the 3′-untranslated regions (UTRs) of several cytoskeletal and cytokine mRNAs have been shown to activate this kinase in various physiological contexts (22–26). Indeed, such endogenous mRNAs can activate PKR even more potently than dsRNA (27). A strong such candidate initiator mRNA is PKR’s own transcript, which is induced by IFN, contains a long 3′-UTR, and is capable of stimulating PKR activity in cells (28).

Once activated, PKR complexes with RIP1 to promote formation of the nascent RIP1–RIP3 necrosome. Our data and those of others (14) indicate that PKR can physically associate with RIP1 in vitro, suggesting that the most likely mechanism by which PKR activates necrosome formation is by directly phosphorylating RIP1 after activation by IFNs. In agreement, we find that active PKR can phosphorylate RIP1 in vitro, and a specific inhibitor of PKR kinase activity not only prevents PKR–RIP1 interaction, but, unlike Nec-1, also abolishes IFN-γ–driven phospho-RIP1–RIP3 association.

Downstream of the PKR-activated necrosome, IFN-triggered necrosis most likely proceeds by mechanisms common to other RIP kinase-reliant necrotic pathways, albeit with notably slower kinetics, given the progressive transcription-dependent mechanism by which IFNs activate the necrosome. For example, we have previously shown that IFN-activated necrosis proceeds via RIP1-dependent ROS generation and consequent mitochondrial dysfunction, sequelae that are also key to RIP kinase-dependent necrosis triggered by stimuli such as TNF-α (2, 9). We speculate that the Nec-1–sensitive step in IFN-triggered necrosis occurs between RIP1 and ROS generation, as Nec-1 fails to inhibit IFN-driven necrosome formation (Fig. 2), but efficiently prevents ROS buildup after IFN exposure (Fig. S8 and ref. 9).

Collectively, these observations allow us to order IFN-triggered necrosis into the following sequence: (i) transcriptional activation of PKR, (ii) PKR-driven necrosome formation and activation, and (iii) necrosome-dependent ROS generation and ultimate mitochondrial dysfunction (Fig. 7G). The second of these steps is negatively regulated by FADD and caspases; when FADD is absent or disabled, or when caspases are inactivated, RIP kinases are hyperactivated, ROS is overproduced and cannot be effectively quenched in mitochondria, and the cell undergoes necrotic death. Given that (i) we isolated RIP1 as a FADD-interacting partner by yeast-two hybrid screening, (ii) FADD and RIP1 associate via their death domains in other contexts, and (iii) RIP1–PKR and RIP1–RIP3 complexes form spontaneously when FADD is absent, the simplest mechanism by which FADD prevents RIP kinase-driven necrosis is by physically associating with RIP1 to prevent it from interacting with either PKR or RIP3.

We have identified phosphorylation of FADD as a putative mechanism that permits IFN-activated necrosis, but how and when does phosphorylation of FADD license necrosis? Phosphorylation on S191 may disrupt the otherwise-strong association between FADD and RIP1, allowing a fraction of RIP1 to now associate with its pronecrotic partners, PKR and RIP3, following exposure to IFN. In agreement, Kim and colleagues (29) have reported that a phosphomimetic mutant of FADD interacts poorly with RIP1 and cannot inhibit its kinase activity. This fraction of RIP1 is likely quite small, as (i) the magnitude of necrotic death induced by IFNs in cells containing FADD-S191D is considerably less that that seen in cells completely lacking FADD expression; (ii) unlike fadd^−/− MEFs, FADD-S191D–expressing cells do not display spontaneous necrosome formation; and (iii) the formation of the TNF-α–induced necrosome—an event requiring association between FADD and RIP1 (10, 30)—is not significantly affected by the S191D mutation (Fig. S9).

Thus, phosphorylation on S191 is not solely sufficient for IFN-driven necrosome formation and requires additional collaborative events for full execution of the necrotic program. Caspase inactivation is an attractive such event, because we find that low doses of the pancaspase inhibitor z-VAD sensitize necrosome formation in G₂/M-arrested wild-type MEFs upon IFN treatment (Fig. 6G), and collaborate with the phosphomimetic FADD-S191D mutant to fully license IFN-induced necrosis (Fig. 7F). Indeed, higher doses of z-VAD sensitize wild-type MEFs to IFN-induced necrotic death with molecular features largely indistinguishable from that seen in fadd^−/− MEFs, and caspase 8^−/− MEFs are also susceptible to IFN-induced cell death (Fig. 7). Although how caspases (such as caspase 8) regulate IFN-induced necrosis remains to be determined, our observation that caspase inactivation, like FADD ablation, induces spontaneous assembly of the RIP1–RIP3 necrosome (Figs. 1G, lane 5, and 7B, lane 2) suggests that caspases function at the level of necrosome assembly, perhaps at the same step controlled by FADD.

The best-studied physiological setting in which phosphorylation of FADD on S191 occurs is during progression through the G₂/M phase of the cell cycle, raising the intriguing possibility that phosphorylation of FADD may represent a host-defense checkpoint that licenses IFN-mediated elimination of cells before they divide. In the context of a virus infection, activating PKR (either directly by virus RNA or indirectly via production of type I IFNs) can initiate necrosis in G₂/M when FADD is phosphorylated and consequent PKR necrosome formation is licensed. In a similar manner, cells in which FADD/caspase expression or function have been compromised by virus infection [for example, by virally encoded inhibitors of apoptosis (31)] may also permit IFNs and viral RNA to activate necrotic death. Thus, when FADD is present and functional, apoptotic and innate immune mechanisms help clear a viral infection [e.g., PKR-induced apoptosis requires FADD, and FADD itself participates in antiviral innate-immune signaling (5, 12)]. However, if FADD is inactivated by phosphorylation (or the FADD–caspase 8 apoptotic axis is otherwise disabled by virus-encoded proteins), then IFNs trigger a necrotic cell death program to eliminate infected cells.

Recent studies have shown that FADD—in addition to regulating IFN-activated necrosis as reported in this study—also controls the activity of RIP1/3 kinases to (i) restrict fulminant inflammatory tissue damage in the gut and skin, (ii) prevent inopportune T-cell necrosis during development and immune responses, and (iii) allow normal progression through embryogenesis (16, 32–36). In each of these scenarios, the upstream signals that activate RIP1/3 when FADD is absent remain vague (4). Our data show that IFNs are likely not the trigger of these kinases during embryogenesis, but whether IFNs contribute to necrosis during T-cell ontogeny and function, or during RIP kinase-driven tissue inflammation, is still unknown.

Type I IFNs have recently been shown to trigger premature RIP kinase-dependent necrotic death in macrophages during antibacterial responses (37). Taken together with the earlier finding that PKR was necessary for bacterial infection-induced macrophage death (38), these observations suggest that an IFN–PKR–RIP kinase axis may control phagocyte cell fate during antibacterial immune responses. It will be interesting to see whether FADD (and/or caspases) also regulate IFN-activated macrophage necrosis during bacterial clearance, given that a hypomorphic FADD mutation in humans has been reported to increase susceptibility to bacterial infections (39).

Materials and Methods

Cells, Viruses, and Reagents.

Early-passage fadd^+/+ and fadd^−/− MEFs were generated from timed crosses of fadd^+/− mice (gift from Tak Mak, University of Toronto, Toronto, ON, Canada). Caspase 8^+/+ and caspase 8^−/− MEFs were a kind gift from Andrew Oberst and Douglas Green (St. Jude Children’s Hospital, Memphis, TN). Cytokines and chemicals were from the following sources: IFN-α, IFN-β, IFN-γ (Pestka Biomedical Laboratories); TNF-α, TRAIL, IL-1β, IL-6 (R&D Systems); poly(I:C) (InvivoGen); Necrostatin-1 (Enzo Life Sciences); C16, C22, Jak inhibitor I, z-VAD.fmk, nocodazole (Calbiochem); and Act D (MP Biochemicals). Antibodies were purchased from Millipore (Jak1, FADD, STAT1, phospho-STAT1), BD Biosciences (RIP1), Santa Cruz (PKR, phospho-PKR), Cell Signaling (phospho-FADD), and ProSci (RIP3). All other reagents were from Sigma-Aldrich, unless otherwise noted. All cells were cultured in high-glucose DMEM containing 10% (vol/vol) FBS and antibiotics.

RNAi.

For acute RNAi, HeLa cells (6 × 10⁴/well) seeded into six-well plates were transfected with pools of four distinct proprietary siRNAs (SMARTpool; Dharmacon) to each target mRNA at 20 nM using Oligofectamine (Invitrogen) as a transfection reagent. As controls, nontargeting siRNA duplexes were used. Cells were used in experiments 72 h posttransfection. For stable RNAi, prepackaged shRNA-expressing lentiviral particles (Sigma) were used. At least four distinct shRNAs per target gene were independently tested for knockdown efficiency by immunoblot analysis, and the most efficient shRNAs were used to generate at least two individual stable populations of MEFs for subsequent experiments.

Coimmunoprecipitation.

MEFs (1 × 10⁶ per condition) were harvested in lysis buffer [1% (vol/vol) Triton X-100, 150 mM NaCl, 20 mM Hepes (pH 7.3), 5 mM EDTA, 5 mM NaF, 0.2 mM NaVO₃ (ortho), and Complete protease inhibitor mixture (Roche)]. After centrifugal clarification of lysates, 2 μg of antibody was added to each sample followed by overnight incubation with rotation at 4 °C. Samples were then supplemented with protein A/G-agarose slurry (40 μL) and incubated for an additional 2 h. IPs were washed four times with cold lysis buffer and eluted by boiling in SDS sample buffer, resolved by SDS/PAGE, and analyzed by immunoblotting.

Molecular Modeling.

A model of the RIP1-FADD death domain complex was constructed on the multimeric structure of the FADD death domain–FAS complex [Protein Data Bank (PDB) ID code 3EZQ]. The sequence of the RIP1 death domain was aligned to the Fas structure and a model built with side chains using MolIDE and SCWRL4 (40, 41). Additional modeling of FADD C-terminal tail residues were built with loop modeling routines in the program Yasara12 (42), and the resulting complex was optimized by steepest descent energy minimization methods.

In Vitro PKR Kinase Assay.

Recombinant, active human PKR (amino acids 252–551; SignalChem) was incubated with human wild-type or kinase-dead RIP1 (SignalChem) in kinase reaction buffer containing 10 mM Hepes, 10 mM MnCl₂, and 10 mM MgCl₂. To neutralize RIP1 autophosphorylation, recombinant RIP1 was preincubated with Nec-1 (50 μM) for 15 min before addition of PKR. To inhibit PKR activity, PKR was similarly preincubated with 2AP for 15 min before addition of kinase-dead RIP1 substrate. The kinase reaction was initiated by addition of 100 μM cold ATP and 5 μCi of [γ-³²P]ATP, and incubated at 30 °C for 1 h. The reaction was stopped by boiling in 30 μL of SDS/PAGE loading buffer, following which samples were resolved by SDS/PAGE and examined by autoradiography.

Statistical Analysis.

Student t test was used for comparison between two groups, and values of P < 0.05 were considered significant.