RIP3, an Energy Metabolism Regulator That Switches TNF-Induced Cell Death from Apoptosis to Necrosis

See allHide authors and affiliations

Science  17 Jul 2009:
Vol. 325, Issue 5938, pp. 332-336
DOI: 10.1126/science.1172308


Necrosis can be induced by stimulating death receptors with tumor necrosis factor (TNF) or other agonists; however, the underlying mechanism differentiating necrosis from apoptosis is largely unknown. We identified the protein kinase receptor-interacting protein 3 (RIP3) as a molecular switch between TNF-induced apoptosis and necrosis in NIH 3T3 cells and found that RIP3 was required for necrosis in other cells. RIP3 did not affect RIP1-mediated apoptosis but was required for RIP1-mediated necrosis and the enhancement of necrosis by the caspase inhibitor zVAD. By activating key enzymes of metabolic pathways, RIP3 regulates TNF-induced reactive oxygen species production, which partially accounts for RIP3’s ability to promote necrosis. Our data suggest that modulation of energy metabolism in response to death stimuli has an important role in the choice between apoptosis and necrosis.

Cell death occurs through morphologically distinct processes of apoptosis and necrosis (1). Some necrosis is regulated, via pathways differing from those controlling classical apoptosis, although necrosis/apoptosis interconnectivity has been observed (14). Caspase inhibition, which distinguishes apoptotic and nonapoptotic cell death, sometimes shifts apoptosis to necrosis or enhances necrosis (58). Receptors containing death domains can induce a form of regulated necrosis through kinase activity of RIP1 (receptor-interacting protein 1) (3, 5, 6). Mitochondrial generation of reactive oxygen species (ROS) is essential for this type of necrosis (9). A genome-wide small interfering RNA (siRNA) screen has identified a number of genes involved in the signaling network of death domain receptor–mediated necrosis (10), but the precise mechanisms underlying programmed necrosis and the apoptosis/necrosis molecular switch remain unclear.

Although NIH 3T3 cells typically undergo apoptosis in response to tumor necrosis factor (TNF) stimulation (10), caspase-independent cell death has been reported in one NIH 3T3 line (termed N cells here) (11). We compared the caspase-dependence of TNF-induced cell death between NIH 3T3 cells obtained from American Type Culture Collection (termed A cells here) and N cells. Benzyloxycarbonyl-Val-Ala-Asp-fluoromethylketone (zVAD) inhibited TNF-induced cell death in A cells (Fig. 1A) but enhanced it in N cells (Fig. 1B). TNF- or TNF+zVAD–treated N cells had necrotic morphologies. Because both cell lines were derived from the same cell population, gene expression changes in the N cells could explain the switch from apoptosis to necrosis in the NIH 3T3 cells. Affymetrix microarray analysis revealed that a few genes that have the potential to regulate cell death or TNF signaling were expressed differently in A and N cells (table S1). The expression of RIP3, Calpain 5, Serpinb1a, and Serpinb9b increased in N cells, whereas expression of Calpain 6, Aifm1, and Serpine2 decreased. TNF or zVAD treatment had no effect on the expression of these genes in either A or N cells. We decreased expression of the increased genes and increased expression of the decreased genes in N cells, then analyzed TNF-induced cell death and zVAD enhancement of TNF-induced cell death. Only the siRNA-knockdown of RIP3 in N cells inhibited zVAD’s enhancement of cell death (fig. S1). We therefore focused on RIP3.

Fig. 1

Effects of RIP3 on type of cell death. (A) Viabilities of A cells treated with medium (Ctrl), zVAD (20 μM), TNF (30 ng/ml), or TNF+zVAD. (B) Viabilities of N cells treated as described in (A). (C) Immunoblotting analysis with antibody to RIP3. (D) Viabilities of N cells treated with or without (–) a lentiviral vector encoding LacZ shRNA (Ctrl) or RIP3-shRNAs for 48 hours then stimulated with or without TNF or TNF+zVAD for 24 hours. RIP3 protein amounts in these cells were analyzed by means of immunoblotting 48 hours after infection. (E) Viabilities of A cells infected with a lentivirus encoding nothing (vector) or RIP3 for 36 hours then stimulated with or without TNF or TNF+zVAD for 24 hours. RIP3 protein amounts were analyzed 36 hours after infection. (F) Viabilities of wild-type and RIP3−/− mouse peritoneal macrophages treated with medium (Ctrl), zVAD, TNF, TNF+zVAD, LPS (100 ng/ml), or LPS+zVAD for 24 hours. Data are the mean ± SD of triplicates [(A), (B), and (D) to (F)] and are representative of two to five experiments [(A) to (F)].

RIP3 is a protein kinase that has an N-terminal kinase domain similar to that found in RIP1, RIP2, and RIP4, but its C-terminal domain has no sequence similarity to any known protein domains (12, 13). Transient expression of RIP3 in some cell lines causes cell death (12). Cleavage of RIP3, RIP1, and RIP4 by caspases was observed after activation of death receptors (8, 13, 14). Modulating nuclear factor κB (NF-κB) activation by means of RIP3 overexpression was observed (15, 16); however, deletion of RIP3 caused no signs of deregulated NF-κB signaling (16). The biological function of RIP3 is thus largely unknown.

Differing expression of RIP3 in A and N cells was confirmed by means of Western blotting (Fig. 1C) (17). To study the role of RIP3, we used lentiviral vectors to deliver short hairpin RNA (shRNA) or cDNA into A, N, or other cells because lentivirus can infect with 100% efficiency. RIP3 knockdown by means of siRNA in N cells appeared to switch the type of cell death to that of A cells because TNF-induced cell death in RIP3-knockdown N cells was not enhanced but rather inhibited by zVAD (Fig. 1D). A lacZ-targeting siRNA (Ctrl) showed no effect on the type of cell death in N cells (Fig. 1D). RIP3 expression in A cells switched TNF-induced apoptosis to caspase-independent cell death (Fig. 1E). zVAD did not inhibit but rather enhanced cell death (Fig. 1E). RIP3 kinase activity was required to change the type of cell death because a catalytically inactive RIP3 mutant (RIP3-K51A) did not cause A cell death to become caspase-independent (fig. S2). Deletion and point mutations revealed that the RIP1-interacting RHIM (RIP homotypic interaction motif) domain in RIP3 is required for necrosis (fig. S2).

Although zVAD may also participate in the induction of cell death in some cell lines (10, 18), we established that RIP3 has a role in necrosis in the absence of zVAD because it is required for TNF-induced necrosis in L929 cells (fig. S3A). We then analyzed whether RIP3 is required for zVAD-mediated necrosis. zVAD can trigger necrosis of lipopolysaccharides (LPS)– or TNF-stimulated primary peritoneal macrophages (19) but failed to promote cell death of RIP3−/− macrophages stimulated with TNF or LPS (Fig. 1F). It appears that RIP3 functions in necrosis induced by physiological stimuli and is required for zVAD to induce or enhance necrosis. Thus, RIP3 is essential for some forms of necrosis.

RIP1 is required for death receptor agonists to activate necrosis (5, 6, 8, 18). Although we detected no difference in RIP1 expression in A and N cells, depletion of RIP1 with shRNA in N cells inhibited TNF+zVAD–induced necrosis (Fig. 2A). Similarly, depletion of RIP1 from L929 cells also blocked TNF-induced necrosis (fig. S3B). Therefore, both RIP1 and RIP3 are required for TNF-induced and zVAD-enhanced necrosis.

Fig. 2

Role of RIP3 and RIP1 in apoptosis and necrosis. (A) Viabilities of N cells treated with or without (–) a lentiviral vector encoding control or RIP1-shRNAs for 48 hours then stimulated with or without TNF or TNF+zVAD for 24 hours. RIP1 protein amounts were analyzed by means of immunoblotting 48 hours after infection. (B) Wild-type, RIP3−/−, RIP1−/−*, and RIP1−/−RIP3d MEF cells infected with a lentiviral vector encoding RIP1 or RIP3, treated with or without zVAD. Viabilities and RIP1 and RIP3 expression were analyzed 36 hours after infection. (C) N and A cells were treated and analyzed as described in (B). (D) RIP1 and RIP3 expression levels and viabilities of N cells infected with lentiviral vectors expressing control shRNA and RIP3; RIP1-shRNA and RIP3; control shRNA and RIP1; or RIP3-shRNA and RIP1 and treated with or without zVAD. Data are the mean ± SD of triplicates and are representative of two to three experiments [(A) to (D)].

Transient RIP1 expression caused cell death in both wild-type and RIP3−/− mouse embryo fibroblast (MEF) cells (Fig. 2B). zVAD inhibited RIP1-induced cell death in RIP3−/− cells but not in wild-type cells, so RIP3 may have a role in determining the type of RIP1-induced cell death. Consistently, RIP3-induced cell death was not inhibited by zVAD. Analyzing RIP1 and RIP3 expression in the knockout cell lines revealed that the previously described RIP1−/− MEF line (6) is defective in both RIP1 and RIP3 expression, and we therefore named it RIP1−/−RIP3d (Fig. 2B, top). We reconstituted RIP3 expression in this cell line and named it RIP1−/−*. RIP1- and RIP3-overexpression–induced death in RIP1−/−* cells was similar to that observed in wild-type MEF cells, and cell death in RIP1−/−RIP3d cells was similar to that in RIP3−/− cells (Fig. 2B), supporting the idea that RIP1 induces apoptosis in the absence of RIP3 and induces necrosis when RIP3 is present.

RIP1 overexpression induced apoptosis in A cells and necrosis in RIP3-rich N cells, whereas RIP3 overexpression induced necrosis in both A and N cells (Fig. 2C). Unlike MEF cells, necrosis in these cells was enhanced by zVAD, but only when RIP3 was present (Fig. 2C). We then found that zVAD did not enhance RIP1-induced N cell death when RIP3 was depleted (Fig. 2D). Thus, RIP3 is required for zVAD to enhance necrosis.

To find RIP3 targets, we identified interacting proteins during necrosis. Using a lentiviral vector, we expressed flag-tagged RIP3 in A cells and confirmed that flag-RIP3 changed the death type of A cells to that of N cells. Flag-RIP3 was then purified in a complex with its associated proteins from A cells treated with nothing or TNF+zVAD for 30 or 120 min. Protein constituents were identified by means of liquid chromatography–tandem mass spectrometry (LC-MS/MS). The raw data set containing nonredundant proteins (table S2) indicated dynamic changes in the RIP3 complex because proteins identified in TNF+zVAD–treated samples were different from the untreated sample. Seven metabolic enzymes—glycogen phosphorylase (PYGL), glutamate-ammonia ligase (GLUL), glutamate dehydrogenase 1 (GLUD1), fructose-1,6-bisphosphatase 2 (FBP2), fumarate hydratase (FH), glycosyltransferase 25 domain containing 1 (GLT25D1), and isocitrate dehydrogenase 1 (IDH1)—were identified in the RIP3 complex in TNF+zVAD–treated cells. PYGL, GLUL, and GLUD1 were confirmed to interact with RIP3 by means of coimmunoprecipitation of coexpressed proteins in 293T cells (fig. S4).

PYGL catalyzes the rate-limiting step in the degradation of glycogen by releasing glucose-1-phosphate, thereby having a key role in using reserved glycogen as an energy source. We sought to determine whether RIP3 and PYGL interact during necrosis, but were unable to obtain an antibody to detect endogenous PYGL in N cells. We therefore established an N cell line stably expressing hemagglutinin (HA)–PYGL. Immunoprecipitation using antibodies to RIP3 revealed that TNF+zVAD treatment for 120 min induced interaction of RIP3 and HA-PYGL (Fig. 3A). RIP3 and PYGL interaction was induced by TNF but not zVAD (Fig. 3B). We incubated flag-PYGL (purified from 293T cells) with or without glutathione S-transferase (GST)–RIP3 or GST-RIP3-K51A (purified from Escherichia coli) in a kinase buffer and found that RIP3, but not RIP3-K51A, enhanced PYGL activity in vitro (Fig. 3C). Because RIP3 may directly target one or more proteins copurified with flag-PYGL, we currently can only conclude that RIP3 directly or indirectly activates PYGL in vitro. By measuring PYGL activity in N cells and RIP3- or RIP1-knockdown cells before and after TNF or TNF+zVAD treatment, we found that RIP3 and RIP1 are required for TNF- and TNF+zVAD–increased PYGL activity (Fig. 3D and fig. S5). The role of RIP3 in regulating PYGL was confirmed by use of RIP3−/− peritoneal macrophages (Fig. 3E). Additionally, expression of RIP3 with PYGL in 293T cells enhanced PYGL activity (fig. S6). To determine whether PYGL is involved in necrosis, we used siRNAs to deplete PYGL in N cells and found that reduction of PYGL partially inhibited TNF+zVAD–induced cell death (Fig. 3F). Thus, PYGL activation by RIP3 appears to contribute to necrosis.

Fig. 3

Activation of PYGL by RIP3 and the involvement of PYGL in necrosis. (A) A stable HA-PYGL–expressing N cell line was treated with TNF+zVAD for various time intervals. The cell lysates and antibody-to-RIP3 immunoprecipitates were immunoblotted with antibodies to HA and RIP3. (B) Same as in (A), except the treatments were nothing, zVAD, TNF, or TNF+zVAD. (C) PYGL activity after flag-PYGL was incubated with 0.25 mM adenosine monophosphate (AMP) for 10 min then with or without recombinant GST-RIP3 or GST-RIP3-K51A (inactive mutant) in a kinase buffer for 30 min at 30°C. (D) PYGL activity in N cells infected with a lentiviral vector expressing control shRNA or RIP3 shRNA for 48 hours, treated with nothing, TNF, or TNF+zVAD for 2 hours. (E) PYGL activity in wild-type and RIP3−/− mouse peritoneal macrophages, treated with nothing, TNF+zVAD, or LPS+zVAD for 2 hours. (F) Viabilities of N cells infected with a lentivirus expressing control shRNA or PYGL shRNAs for 48 hours then stimulated with or without TNF+zVAD for 24 hours. PYGL mRNA levels were determined by means of real-time polymerase chain reaction 48 hours after infection. Data are the mean ± SD of triplicates [(C) to (F)] and are representative of two experiments [(A) to (F)].

GLUL is a cytosolic enzyme that catalyzes the condensation of glutamate (Glu) and ammonia to form glutamine (Gln). Gln can transfer into the mitochondria to function as an energy substrate. GLUD1 is a mitochondrial matrix enzyme that converts Glu to α-ketoglutarate. GLUL and GLUD1 are essential for the use of amino acid Glu or Gln as substrates for adenosine triphosphate production by means of oxidative phosphorylation. Enzymatic pathways specifically used in the mitochondrial oxidation of Gln have been suggested to sensitize the mitochondria to TNF-induced perturbation (20). We detected increased interaction of endogenous RIP3 with GLUL and GLUD1 after N cells were treated with TNF or TNF+zVAD (figs. S7, A and B, and S8, A and B). Recombinant RIP3, but not RIP3-K51A, can directly increase GLUL activity (fig. S7C). We were unable to prepare a functional recombinant protein of GLUD1, but coexpression of RIP3 with GLUD1 in 293T cells enhanced activity of GLUD1 (fig. S8C). We were unable to detect the activities of endogenous GLUL and GLUD1 in N cells, and therefore were unable to address their activation in cells. Deletion of either GLUL or GLUD1 by means of siRNA reduced TNF+zVAD–induced cell death in N cells (figs. S7D and S8D), suggesting that GLUL- and GLUD1-mediated use of Glu or Gln as energy substrates contributes to necrotic cell death.

Because ROS production is required for TNF- or TNF+zVAD–induced necrosis in L929 cells, MEFs, and macrophages (6, 9, 19), we hypothesized that RIP3 might increase energy metabolism–associated ROS production. We confirmed that ROS is required for necrosis of N cells but not apoptosis of A cells (fig. S9). We found that depletion of RIP3 in N cells reduced ROS concentration in TNF+zVAD–treated cells (Fig. 4A) and that increasing RIP3 levels in A cells increased TNF+zVAD–induced ROS production (Fig. 4B). RIP3 dependence of ROS production was also confirmed by use of RIP3−/− macrophages (fig. S10). In all cases, the amount of ROS correlated with caspase-independent cell death (Fig. 1, D to F). Depletion of PYGL, GLUL, and GLUD1 by means of siRNA reduced TNF+zVAD–induced accumulation of ROS (Fig. 4C and fig. S11), which correlated with cell death (Fig. 3F and figs. S7D and S8D). Experiments performed by use of energy metabolism inhibitors suggested that mitochondrial ROS generated at the ubisemiquinone site has a key role in TNF cytotoxicity in N cells (fig. S12), which is consistent with a previous report on L929 cells (21). Because increasing glucose by breaking down glycogen and promoting the use of Glu and Gln as energy substrates all function upstream of ubisemiquinone so as to increase energy metabolism, the role of RIP3 in apoptosis/necrosis switching should at least partly occur through increasing energy metabolism–associated ROS production.

Fig. 4

Activation of PYGL, GLUL, and GLUD1 by RIP3 contributes to TNF+zVAD–induced ROS production. (A) ROS levels in N cells treated with a lentiviral vector expressing control shRNA or RIP3-shRNAs for 48 hours then stimulated with TNF+zVAD for 12 hours. (B) ROS levels in A cells infected with or without a lentivirus encoding nothing or RIP3 for 36 hours then stimulated with TNF+zVAD for 6 hours. (C) ROS levels in N cells infected with a lentivirus expressing control shRNA, PYGL-shRNA, GLUL-shRNA, or GLUD1-shRNA for 48 hours then stimulated with TNF+zVAD for 12 hours. (D) Proposed mechanism for RIP3 in necrosis. When RIP3 is absent, RIP1-mediated cell death is apoptotic. High levels of RIP3 switch apoptosis to necrosis in some cell systems. RIP3 activates PYGL, increasing the availability of energy substrate glucose; RIP3 also activates GLUL and GLUD1, increasing Glu and Gln consumption as energy substrates. These lead to an increase in energy metabolism and subsequent overproduction of the oxidative metabolism product, ROS. ROS is at least in part responsible for RIP3-mediated necrosis.

A metabolic increase may be needed when cells are treated with a pleiotropic cytokine such as TNF. In cells that have sufficient RIP3 expression, the gateways to using glycogen and Gln or Glu are readily opened after TNF stimulation. zVAD may enhance RIP3 function because RIP3 and RIP1 are reported to be cleaved by caspases (fig. S13) (8, 14). The enhanced metabolism should be accompanied by increased ROS production, which is probably responsible in part for the function of RIP3 in mediating necrosis (Fig. 4D). The role of RIP3 in determining the type of cell death supports the idea that energy metabolism affects cell-death mechanisms (22, 23).

Necrosis occurs under various physiological and pathological conditions, and some might occur in vivo when caspases are inactivated by S-nitrosylation or other means. Because inhibition of necrosis by a RIP1 inhibitor has a beneficial effect on ischemic brain injury and other animal disease models (3, 24), inhibition of RIP3 could be beneficial for diseases that are associated with necrosis, such as diabetes and cerebral ischemia. Indeed, deletion of RIP3 improved the condition of acute pancreatitis in a mouse model (table S3). RIP3 is a potential drug target for necrosis-related diseases.

Supporting Online Material

Materials and Methods

Figs. S1 to S13

Tables S1 to S3


References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We thank K. Newton and V. M. Dixit for RIP3−/− mouse. This work was supported by grants NSFC-CIHR 30611120526 and 973 program 2009CB522200.
View Abstract

Stay Connected to Science

Navigate This Article