Going up in flames: necrotic cell injury and inflammatory diseases
- First Online:
- Cite this article as:
- Challa, S. & Chan, F.K. Cell. Mol. Life Sci. (2010) 67: 3241. doi:10.1007/s00018-010-0413-8
- 320 Views
Recent evidence indicates that cell death can be induced through multiple mechanisms. Strikingly, the same death signal can often induce apoptotic as well as non-apoptotic cell death. For instance, inhibition of caspases often converts an apoptotic stimulus to one that causes necrosis. Because a dedicated molecular circuitry distinct from that controlling apoptosis is required for necrotic cell injury, terms such as “programmed necrosis” or “necroptosis” have been used to distinguish stimulus-dependent necrosis from those induced by non-specific traumas (e.g., heat shock) or secondary necrosis induced as a consequence of apoptosis. In several experimental models, programmed necrosis/necroptosis has been shown to be a crucial control point for pathogen- or injury-induced inflammation. In this review, we will discuss the molecular mechanisms that regulate programmed necrosis/necroptosis and its biological significance in pathogen infections, drug-induced cell injury, and trauma-induced tissue damage.
Non-apoptotic cell death comes in different flavors
Metazoan homeostasis is achieved through balancing cell proliferation and cell death. The critical role of apoptosis is demonstrated by the many experiments where genetic ablation of components of the apoptosis pathway leads to blunted embryonic development. Work in recent years shows that cell death in multi-cellular organisms comes in many flavors. For instance, paraptosis is a form of non-apoptotic cell death marked by mitochondrial/ER swelling and vacuolation. It was first described as a form of neuronal cell death [1, 2], but has recently been observed in certain tumors treated with the chemotherapeutic drug taxol . Pyroptosis is characterized by potassium efflux and caspase-1-mediated activation of the inflammasome, and the rapid loss of plasma membrane integrity . It is often activated during bacterial infections and thus is an important host defense against pathogens . Autophagy, a cellular survival pathway in response to nutrient deprivation , has been shown to cause cell death in certain situations [7–9]. Although these non-apoptotic cell-death pathways often exhibit overlapping morphological features, it is unclear if they also utilize similar molecular machineries to execute the demise of the cell.
The term “necrosis” was once used to describe cell death from trauma such as that induced by heat shock or repeated cycles of freeze/thaw. In tissue culture, cells often undergo late-stage “secondary necrosis” in response to apoptotic stimuli. This observation further contributed to the notion that necrosis is a by-product of apoptosis. However, it is now clear that necrosis can be triggered as a direct result of “extrinsic” stimulation by death cytokines in the TNF superfamily (e.g., TNF, Fas ligand (FasL) and TRAIL (TNF-related apoptosis inducing factor) ) or by “intrinsic” signals such as DNA damage [11, 12]. Emerging evidence indicates that specific signaling pathways are required to mediate necrosis induced by different death signals. Because specific triggers (e.g., death cytokines and their corresponding death receptors) and an elaborate biochemical signaling cascade are involved, terms such as “programmed necrosis”  or “necroptosis”  have been used to highlight the fact that necrotic cell injury is not merely an “accidental death”, but rather a “programmed” form of cellular demise. In this review, we will review recent advances in the molecular regulation of programmed necrosis and discuss the physiological functions of this cell injury pathway in mediating the inflammatory responses against pathogens and other cellular trauma.
Morphological features of programmed necrosis
The pleiotropic adaptor RIP1 mediates programmed necrosis
TNF was recognized as a host factor that could induce tumor regression in the form of necrosis in the 1970s . Since that time, most research had focused on how TNF induces apoptosis. It took almost 30 years for TNF to regain its identity as a necrosis-inducing factor, when it was found that caspase inhibition enhanced cellular sensitivity to TNF-induced necrosis . In 2000, Holler and colleagues reported that in Jurkat T cells that are deficient for FADD or caspase-8 or wild-type T cells treated with the pan-caspase inhibitor zVAD-fmk, the death cytokines TNF, FasL, and TRAIL induce necrosis through the serine/threonine kinase RIP1 . These results were surprising because early experiments indicate that RIP1 is essential for TNF-induced NF-κB activation [19, 20], which confers pro-survival signals that counter the cytotoxic effects of TNF . Subsequent investigations show that RIP1 was also required for NF-κB activation by Toll-like receptor (TLR)-3 and TLR-4 [22, 23].
How does RIP1 mediate pro-necrotic and pro-survival signals? Upon binding to TNF-R1, the E3 ligases cIAP-1 and cIAP-2 mediates K63 and non-K63-dependent RIP1 polyubiquitination at lysine 377 [24–26]. Polyubiquitinated RIP1 binding to NEMO, the regulatory subunit of the IKK complex, is a critical event for NF-κB activation [27, 28]. It is noteworthy that mouse embryonic fibroblasts (MEFs) isolated from RIP1−/− mice were recently shown to exhibit normal TNF-induced NF-κB activation . The basis for these discrepant results is unclear at the moment, but they do raise the possibility that NF-κB activation could occur in a RIP1-independent manner. In addition to RIP1, the linear ubiquitin chain assembly complex (LUBAC) comprised of the RING finger ligases HOIL-1 and HOIP promote linear ubiquitination of NEMO and NF-κB activation [30, 31]. HOIL-1−/− cells fail to activate NF-κB in response to TNF . Interestingly, recruitment of HOIL-1 to the TNF-R1 signaling complex is independent of RIP1 [30, 31]. It remains to be tested whether LUBAC might cooperate with RIP1 to promote TNF-induced NF-κB induction in certain cell types. Nonetheless, these results indicate that RIP1-driven NF-κB activation plays an important role in cell survival in most situations.
RIP3: a critical determinant for programmed necrosis
Protein ubiquitination and deubiquitination in programmed necrosis
RIP1 recruited to TNF-R1 is heavily modified by ubiquitination. By contrast, RIP1 found in the cytoplasmic RIP1-RIP3 complex was largely unmodified by ubiquitination [35, 36]. The differential ubiquitination pattern implies that RIP1 might undergo deubiquitination as it transitions from the membrane receptor-signaling complex to the cytoplasmic signaling complex. The deubiquitinating enzymes A20 and cylindromatosis (CYLD) have been shown to facilitate hydrolysis of the polyubiquitin chain on RIP1 [38, 39]. Interestingly, CYLD was recently identified in a genome-wide RNAi screen as a positive regulator for programmed necrosis . CYLD was also found to promote the assembly of an alternative caspase-8 activating complex containing RIP1 and FADD . Although the biochemical mechanism is still vague at present, it is tempting to speculate that CYLD or A20 might facilitate programmed necrosis by removing the polyubiquitin chain from RIP1 (Fig. 3). According to this model, polyubiquitination of RIP1 might interfere with programmed necrosis by sterically hindering the recruitment of RIP3 or other components of the pro-necrotic signaling complex. In fact, such a mechanism appears to be important for the anti-apoptotic function of RIP1. Ting and colleagues showed that RIP1 confers protection against apoptosis in cells that stably express a dominant negative mutant of IκBα, especially early during TNF stimulation . Strikingly, mutation at the RIP1 polyubiquitination site, K377, abolished the early protective effect of RIP1 . Polyubiquitinated RIP1 binds to NEMO, which prevents the interaction between RIP1 and caspase-8 and blocks the activation of caspase-8 by RIP1 [41, 42]. Although it is unclear if deubiquitinases like A20 or CYLD are also involved in controlling this anti-apoptotic function of RIP1, these results are consistent with the notion that ubiquitination and deubiquitination of RIP1 plays a critical role in controlling its pro-survival and pro-death cell death activities (Fig. 3).
Caspase-8 inhibits programmed necrosis by cleavage of RIP1 and RIP3
As we have discussed above, one of the early clues indicating a dedicated molecular circuitry for programmed necrosis came from the observation that caspase inhibition often sensitize cells to programmed necrosis. However, although caspase inhibition exacerbates the necrotic response, it is not an obligate requirement for programmed necrosis. For example, in Jurkat T cells expressing TNFR-1 and TNFR-2 , TNF stimulation causes both apoptosis and programmed necrosis . In fact, induction of necrosis-specific RIP1-RIP3 kinase activity could be detected in these cells . Caspase inhibition blocks apoptosis and renders programmed necrosis the predominant form of cell death [32, 44]. The sensitizing effect of caspase inhibition on programmed necrosis is at least in part due to the fact that RIP1 and RIP3 are substrates for caspase-8 [45, 46]. In fact, RIP1 cleavage could be detected within the caspase-8-associated complex in apoptotic cells, but not in cells undergoing programmed necrosis . The caspase cleavage site for RIP1 has been mapped to the boundary between the kinase domain and the carboxyl end of the protein at D324 (Fig. 2) . Thus, caspase-8-mediated cleavage of RIP1 will result in the release of the kinase domain from the protein–protein interaction domains of RIP1. In this scenario, access and phosphorylation of downstream substrates of RIP1 will likely be blocked (Fig. 3).
Compared to the results on RIP1 cleavage, the evidence supporting RIP3 cleavage and inactivation by caspase-8 is weaker. Although RIP3 cleavage at D328 has been observed in transfection studies , it has not been detected in endogenous signaling complexes. Nonetheless, a non-cleavable mutant of RIP3 is sufficient to abrogate the necrosis-enhancing effect of the pan-caspase inhibitor zVAD-fmk . Taken together, these results strongly support the model that caspase inhibitors such as zVAD-fmk facilitate programmed necrosis in part by preserving the integrity of RIP1 and RIP3 (Fig. 3).
Reactive oxygen species: critical effectors for programmed necrosis
The molecular events that drive the morphological changes associated with programmed necrosis, namely intracellular vacuoles, organelle and cell swelling, and eventually rupture of the plasma membrane, are poorly defined. A widely accepted view is that reactive oxygen species (ROS) play crucial roles in the execution of programmed necrosis . For example, high levels of ROS could be detected in necrotic cell death. ROS scavengers such as butylated hydroxyanisole (BHA) are effective inhibitors of programmed necrosis in many cell types . At present, it is not clear how ROS cause the cellular damages observed in necrosis. High levels of ROS could induce lipid membrane peroxidation and plasma membrane leakage . However, since ROS also regulate other biological processes including apoptotic death [50, 51], additional factors are clearly required to synergize with ROS to promote necrosis.
ROS are produced as by-products of mitochondrial oxidative phosphorylation. Inhibitors that block mitochondrial Complex I, a major site of ROS production, was effective against TNF-induced programmed necrosis in L929 cells [48, 52]. Several reports indicate that the necrosis-specific kinases RIP1 and RIP3 act upstream to regulate ROS production [35, 53]. Although the results are somewhat controversial, RIP1 and RIP3 have both been shown to co-localize with the mitochondria [53, 54]. Interestingly, several mitochondrial metabolic enzymes were reported to be possible substrates for RIP3 . Thus, these results support a role for the cytoplasmic RIP1-RIP3 complex in ROS production, possibly by acting directly on the mitochondrial machinery that regulates oxidative phosphorylation. In contrast to these results, several recent reports show that plasma membrane-associated complex containing TNFR-1, riboflavin kinase (RFK) and NADPH oxidase 1 (Nox1) is essential for ROS production in response to TNF-induced programmed necrosis [55, 56]. These discrepant results might be reconciled by different mechanisms of ROS generation in different cell types. Alternatively, membrane-induced ROS could synergize with mitochondrial-produced ROS to drive programmed necrosis. Clearly, more work is needed to clarify the mechanism that generates ROS during necrotic cell injury.
Programmed necrosis: a driver of inflammation
The identification of a molecular pathway that regulates programmed necrosis raises questions about its physiological functions. From an immunological perspective, apoptosis is a non-inflammatory form of cell death because apoptotic cells expose “eat-me” signals such as PS on the cell surface early during the cell death process. This ensures the rapid clearance of apoptotic cells by phagocytes before they progress to secondary necrosis and prevents the release of cellular adjuvants that can stimulate inflammation . The importance of timely removal of apoptotic cells is demonstrated by the induction of autoimmune inflammatory diseases when phagocytosis of apoptotic cells is inhibited . By contrast, cells undergoing necrosis rapidly lose plasma membrane integrity, which leads to the release of immuno-stimulatory adjuvants such as monosodium urate crystals and the DNA-binding factor HMGB1 [59–62]. Interestingly, specific receptors in the C-type lectin family have recently been implicated to facilitate necrotic cell detection and inflammation signaling [63, 64]. All in all, these results suggest a unique role for programmed necrosis in inflammatory processes. In the next few sections, we will provide several examples in which programmed necrosis plays crucial roles in various physiological and pathological responses.
Programmed necrosis in T-cell expansion
Immune homeostasis is maintained through the delicate balance between cell proliferation and cell death. Surprisingly, many components of the machinery that control lymphocyte cell death also possess the dual function of regulating lymphocyte proliferation [65, 66]. In particular, the essential adaptors for death cytokine-induced apoptosis, FADD and caspase-8, are required for CD4+ T-cell proliferation in response to antigen receptor activation [67, 68]. Lymphocytes from mice that lack FADD or caspase-8 or those that express a dominant negative mutant FADD (FADD-dn) exhibit defective proliferation and undergo death in response to antigen receptor stimulation [67, 69–71]. These results were corroborated by the immuno-deficiencies observed in human patients with caspase-8 mutations [68, 72]. Initial reports suggest that the defective activation in response to antigen receptor stimulation in caspase-8-deficient T-cells was due to the failure to activate NF-κB via the Bcl10-CARMA1-MALT1 complex [71, 72]. However, recent results indicate that NF-κB activation in caspase-8−/− CD4+ T cells is normal . Instead, antigen receptor-induced cell death and expansion of caspase-8−/− and FADD-dn T-cells were rescued by the RIP1-specific inhibitor necrostatin-1 [73, 74]. Necrostatin-1 specifically inhibits the pro-necrotic RIP1 kinase activity, but not the kinase activities of other kinases including RIP2 and RIP3 [75, 76]. These results imply that FADD and caspase-8 actively suppress RIP1-dependent programmed necrosis to ensure antigen-specific lymphocyte expansion. FADD and caspase-8 are also required for TLRs-induced B-cell proliferation [77, 78]. However, it is unclear if the defective proliferation in FADD−/− and caspase-8−/− B-cells was due to RIP1-dependent programmed necrosis. Interestingly, RIP3−/− T-cells are resistant to activation-induced cell death in the presence of caspase inhibitors . It will be interesting to determine whether RIP3 might similarly control antigen-induced lymphocyte expansion.
Programmed necrosis in poxvirus infection
One of the hallmarks of programmed necrosis is the rapid loss of plasma membrane integrity. For cytolytic viruses that require host-cell lysis to release new viral progenies, necrotic cell death might facilitate viral dissemination within the host. Many viral inhibitors potently inhibit the host apoptosis machinery . When apoptosis is inhibited, the host might employ programmed necrosis as a backup or alternative mechanism to eliminate the infected cells. Moreover, the release of endogenous adjuvants from the necrotic cells might facilitate innate inflammatory responses against the virus. This appears to be the case for certain poxviruses. Poxviruses encoded serpins such as CrmA that inhibit caspase-1, -6, and -8 by complexing with the p20 subunit of the caspases via thioester bonds . Despite inhibition of caspases and apoptosis, cells infected with the poxvirus vaccinia virus remain sensitive to the cytotoxic effects of TNF with classical necrotic morphologies . Deletion of RIP1 or RIP3 rescues vaccinia virus infected cells from TNF-induced programmed necrosis [32, 35]. Thus, programmed necrosis circumvents viral inhibition of host apoptosis to eliminate vaccinia virus-infected cells.
Could programmed necrosis be more than a backup cell death response to apoptosis? Poxvirus infection elicits strong inflammatory responses. Recent evidence indicates that TNF-induced programmed necrosis might play a crucial role in eliciting the early inflammatory response in vaccinia virus infections. In mice, vaccinia virus infection induces TNF expression as early as 24 h . Mice deficient in TNFR-1 or TNFR-2 exhibited defective viral clearance [32, 82]. Vaccinia virus causes lethal infection in nude mice. However, nude mice were protected from death when recombinant vaccinia virus encoding mouse TNF was used . These results highlight a crucial role for TNF in the early innate immune control of poxvirus infections.
Programmed necrosis in MCMV infection
Cytomegaloviruses (CMVs) are betaherpesviruses that employ multiple strategies to subvert the host apoptosis machinery . Recent evidence indicates that productive infection with murine cytomegalovirus (MCMV) requires inhibition of RIP3-dependent programmed necrosis by the MCMV inhibitor M45 . M45 was originally identified as an apoptosis inhibitor in MCMV-infected endothelial cells . More recently, M45 was shown to also inhibit programmed necrosis through binding to RIP1 and RIP3 via the RHIM-like motif at its amino terminus [87, 88]. M45 also shares homology with mammalian ribonucleotide reductase (RNR) at the carboxyl terminus but exhibits no enzymatic activity . Thus, the carboxyl terminal RNR domain is unlikely to contribute to its anti-death activity.
The significance of inhibition of programmed necrosis in MCMV infection was beautifully demonstrated recently using mutant MCMV infection in RIP3−/− mice . An MCMV virus expressing a M45 mutant with tetra-alanine substitutions in the RHIM failed to establish productive infection in tissue culture as well as in mice due to extensive programmed necrosis (Fig. 4b). Strikingly, infection of the M45 mutant virus into RIP3−/− cells or RIP3−/− mice restored productive infection . Thus, productive infection by MCMV requires M45-mediated inhibition of RIP3-dependent programmed necrosis. Surprisingly, siRNA-mediated silencing of RIP1 or the RIP1-specific inhibitor necrostatin-1 did not protect mutant MCMV-infected cells from programmed necrosis . This last result is significant because it implies that RIP1 is dispensable for some types of programmed necrosis.
Although the evidence strongly favors M45 inhibition of RIP3-dependent programmed necrosis as an essential step in establishing productive MCMV infection, it is noteworthy that M45 could modulate other cellular processes to ensure productive infection. As we have discussed above, M45 was first identified as an apoptosis inhibitor in MCMV-infected endothelial cells . Inhibition of apoptosis could serve similar function to that of programmed necrosis inhibition to prevent premature death of infected cells. In addition, M45 was shown to inhibit NF-κB signaling by the DNA-dependent activator of interferon regulatory factor or Z-DNA binding protein 1 (DAI/ZBP1) . DAI mediates NF-κB activation through interaction with RIP1 and RIP3 via the RHIM . Inhibition of NF-κB by M45 might enhance viral replication by suppressing the innate inflammatory responses against the virus. However, it does not explain why the M45 mutant virus also failed to establish productive infection in tissue culture .
Viral inhibitors of programmed necrosis
The presence of a programmed necrosis inhibitor in MCMV strongly implicates that inhibition of programmed necrosis is an important viral immune evasion strategy. Indeed, this notion is further bolstered by the discovery of certain viral FLIPs (FLICE/caspase-8 inhibitory protein) that inhibit programmed necrosis . Viral FLIPs were first identified as inhibitors against death cytokine-induced apoptosis [92, 93]. They contain two tandem death effector domains (DEDs) that interact with the DEDs of caspase-8 and FADD . Interestingly, a subset of vFLIPs, including MC159 encoded by the poxvirus Molluscum contagiosum and E8 encoded by the Equine herpesvirus-2 (EHV-2), also potently inhibit TNF-induced programmed necrosis . Thus, the RHIM-containing M45 and the vFLIPs represent two distinct classes of viral inhibitors of programmed necrosis.
Structural studies indicate that the vFLIP MC159 binds to FADD to prevent FADD oligomerization . However, binding to FADD and caspase-8 is not sufficient for apoptosis inhibition, since this interaction was preserved in some MC159 mutants that fail to prevent apoptosis . This surprising finding might be reconciled by the apoptosis inhibitory function of a carboxyl terminal TRAF3 binding site in MC159 . By contrast, the mechanism by which vFLIPs inhibit programmed necrosis is less defined. MC159 has been shown to interact with many signal adaptors that regulate NF-κB activation including TRAF2 and RIP1 . As such, it is possible that MC159 and E8 inhibit programmed necrosis by targeting RIP1. Interestingly, we found that MC159 also enhances NF-κB induction (unpublished observation). Consequently, mice expressing a MC159 transgene under the ubiquitously expressed H2-Kb promoter  exhibited enhanced clearance of vaccinia virus infection (unpublished observation). These results are in contrast to an early report that shows an inhibitory effect of MC159 on NF-κB activation . These disparate results could be reconciled by the observation that MC159 modulates NF-κB activation in a dose-dependent manner. At low concentrations, MC159 enhances NF-κB activation, whereas at high expression level, it suppresses NF-κB activation (unpublished observation). Thus, vFLIPs such as MC159 might inhibit programmed necrosis through direct interference of the necrotic signaling pathway or by enhancing the NF-κB driven pro-survival signals. Regardless of the mechanisms, the discovery of viral inhibitors against programmed necrosis highlights the importance of this response in anti-viral immunity. Necrotic cell injury has also been reported in certain bacterial infections including Clostridium  Yersinia , and Pseudomonas . However, the role of RIP1 and RIP3 in bacteria-induced necrosis has yet to be tested.
Necrosis in chemical- and trauma-induced cell injury
Besides regulating pathogen-induced inflammation, the RIP1-RIP3 axis has also been implicated to regulate sterile inflammation. Cerulein causes injury to the pancreatic acinar cells and inflammation . Cerulein-induced pancreatitis might involve TNF signaling, since neutralization of TNF with the TNF blocker Etanercept attenuates cerulein-induced pancreas necrosis . Moreover, the induction of inflammatory effectors such as IL-6 and iNOS is reduced in cerulein-treated TNF−/− mice . Importantly, RIP3−/− mice were also resistant to cerulein-induced pancreatitis [36, 37]. However, at 10 h post-treatment, RIP3−/− mice exhibit similar tissue injury in the pancreas compared with their wild-type counterparts. The differential protection against pancreatic acinar cell necrosis was only apparent after 24 h of cerulein treatment in the RIP3−/− mice . Hence, RIP3-dependent necrosis appears to exacerbate rather than initiate cerulein-induced cell injury in the pancreas.
From a molecular perspective, we are only beginning to understand the biochemical pathway that controls programmed necrosis. Future work will no doubt focus on the identification of additional components that regulate this pathway. However, from the limited information we have at present, it is clear that programmed necrosis is not merely a backup form of cell suicide. Rather, programmed necrosis is linked with many disease pathologies, especially those that are associated with inflammation. We have described several examples in which inflammation induced by pathogens, cytotoxic drugs, and trauma requires an intact programmed necrotic response pathway. These studies raise the tantalizing possibility that programmed necrosis might have broader roles in controlling other inflammatory conditions such as autoimmune diseases or even cancers. Inhibitors that target components of the necrosis pathway, such as necrostatin-1, might have therapeutic value in treating these diseases.
This work is supported by NIH grants AI088502 and AI083497 (to F.K.C.). F.K.C. is a member of the UMass DERC (DK32520). We thank J. Upton, W. Kaiser and E. Mocarski for sharing their results ahead of print publication.