Programmed necrosis or necroptosis is controlled by the action of two serine/threonine kinases, RIP1 (receptor-interacting serine/threonine protein kinase 1; also known as RIPK1) and RIP3. The phosphorylation of RIP1 and RIP3 is critical for assembly of the necrosome, an amyloid-like complex that initiates transmission of the pro-necrotic signal. In the present study, we used site-directed mutagenesis to systematically examine the effects of putative phosphoacceptor sites on RIP1 and RIP3 on TNF (tumour necrosis factor)-induced programmed necrosis. We found that mutation of individual serine residues in the kinase domain of RIP1 had little effect on RIP1 kinase activity and TNF-induced programmed necrosis. Surprisingly, an alanine residue substitution for Ser89 enhanced RIP1 kinase activity and TNF-induced programmed necrosis without affecting RIP1–RIP3 necrosome formation. This indicates that Ser89 is an inhibitory phosphoacceptor site that can dampen the pro-necrotic function of RIP1. In addition, we show that a phosphomimetic mutant of RIP3, S204D, led to programmed necrosis that was refractory to RIP1 siRNA and insensitive to necrostatin-1 inhibition. Our results show that programmed necrosis is regulated by positive and inhibitory phosphorylation events.
Programmed necrosis/necroptosis is an inflammatory form of cell death with important functions in certain viral infections and trauma-induced tissue injury. TNF (tumour necrosis factor) and related death ligands are strong inducers of programmed necrosis. Induction of programmed necrosis requires the assembly of the amyloid-like RIP1 (receptor-interacting serine/threonine protein 1; also known as RIPK1)–RIP3 necrosome [1,2]. The necrosome recruits and activates downstream RIP3 substrates such as the MLKL (mixed lineage kinase domain-like) and Pgam5 (phosphoglycerate mutase family member 5) to effect programmed necrosis [3–5]. Necrosome assembly and activation is controlled by three major types of post-translational modifications: caspase 8-mediated proteolytic cleavage, protein ubiquitination and phosphorylation . Following TNF signalling, RIP1 bound to the TNFR-1 (tumour necrosis factor receptor 1) complex undergoes polyubiquitination. Polyubiquitinated RIP1 binds NEMO (nuclear factor κB essential modulator) and other adaptors that are important for NF-κB (nuclear factor κB) signalling and is sterically restricted from engaging cytoplasmic death-inducing factors including RIP3 . The deubiquitinase CYLD (cylindromatosis) promotes programmed necrosis by facilitating removal of the ubiquitin chains on RIP1 [8,9].
Caspase-8 and its upstream adaptor FADD (Fas-associated death domain) are crucial for caspase-dependent apoptosis induced by TNF-like death cytokines. They inhibit programmed necrosis by cleavage and inactivation of essential necrosis mediators including RIP1, RIP3 and CYLD [10–12]. When caspase activity is inhibited, such as that in mice with genetic deletion of caspase-8 or its upstream adaptor FADD, extensive necrosis ensued, giving rise to extensive injury-induced tissue inflammation and embryonic lethality [13–19].
The kinase function of RIP1 and RIP3 is essential for assembly and activation of the necrosome [20–22]. RIP1 is generally believed to be the apical kinase that phosphorylates RIP3 to initiate necrosis signalling. Consistent with this notion, Nec-1 (necrostatin-1), a RIP1 inhibitor , blocks necrosis-specific RIP1 and RIP3 phosphorylation . However, neither RIP1 nor RIP3 was phosphorylated under necrosis inducing conditions in RIP3−/− cells . These results raise the possibility that RIP3 might in fact act upstream to promote RIP1 activation. Identifying the key phosphoacceptor sites on RIP1 and RIP3 that regulate their activities will help clarify the hierarchical relationship between RIP1 and RIP3 in the necrosis signalling pathway. To this end, several serine residues in the kinase domains of RIP1 and RIP3 have been identified [3,21,23]. However, their roles in programmed necrosis have not been carefully investigated. It is also not known whether phosphatases might counter necrosis by inhibiting the function of the RIPs.
In the present study, we show that the previously identified phosphorylation sites on RIP1 had minimal effects on RIP1-dependent programmed necrosis. Instead, we identified Ser89 as a novel regulatory serine residue on RIP1. An alanine residue substitution for Ser89 resulted in hyperactive RIP1 kinase activity and increased TNF-induced programmed necrosis. However, necrosome assembly was not altered. We further show that cells expressing the phosphomimetic gain-of-function RIP3 mutant S204D underwent programmed necrosis that was partially independent of RIP1. These results indicate that RIP1 activity is regulated by positive and inhibitory phosphorylation. Moreover, phosphorylation of RIP3 at Ser204 is likely to be a critical mechanism by which RIP1 promotes necrosis signalling.
MATERIALS AND METHODS
Antibodies against GFP (Roche), RIP1, TRAF2 (tumour-necrosis-factor-receptor-associated factor 2), β-actin (BD Pharmingen), p- (phospho-) IκBα (inhibitor of NF-κBα) and total IκBα (Cell Signaling Technology), TNFR-1 (R&D Systems and Abcam) and RIP3 (ProSci) were used. Jurkat cells and fibroblasts were grown in RPMI 1640 medium and DMEM respectively, supplemented with 10% FBS (Invitrogen). GFP-tagged RIP1 and RIP3 have been described previously . All mutants were generated by QuikChange® site-directed mutagenesis (Stratagene) on the GFP-tagged RIP1 and RIP3 backbone. The RIP1 clones were also tagged at their N-termini with HA (haemagglutinin). The mutations and sequence integrity were confirmed by DNA sequencing.
Immunoprecipitations and Western blotting
Ten million Jurkat cells were resuspended in 1 ml of medium, pre-treated with 50 mM Z-VAD-FMK (benzyloxycarbonyl-Val-Ala-DL-Asp-fluoromethylketone; Enzo Life Sciences), and stimulated with 100 ng/ml TNF for the times indicated. After washing with ice-cold PBS, cells were lysed with 150 mM NaCl, 0.2% Nonidet P40, 1 mM EDTA, 3 mM NaF, 1 mM 2-glycerophosphate, 1 mM sodium orthovanadate, and 10% glycerol supplemented with 1× protease and 1× phosphatase inhibitor cocktails (Sigma and Roche). After pre-clearing with Sepharose 6B beads (Sigma), lysates were incubated with anti-TNFR-1 antibody (R&D Systems) or anti-RIP3 antibody (ProSci) overnight. The resulting immune complexes were washed and resolved on 10% Tris/glycine gels (Invitrogen). For p-IκBα Western blot analysis, 10 ng/ml TNF was used.
In vitro kinase assays
HEK (human embryonic kidney)-293T cells were transfected with different GFP-tagged RIP1 or RIP3 mutants using FuGENE HD (Promega) according to the manufacturer's instructions. At 24 h post-transfection, cells were washed with PBS, lysed in buffer containing 150 mM NaCl, 20 mM Tris/HCl, pH 7.5, 1% Nonidet P40, 1 mM EDTA, 1 mM sodium orthovanadate, 3 mM NaF and 2 mM 2-glycerophosphate, supplemented with Complete™ protease inhibitor cocktail (Roche). The lysates were incubated with anti-GFP antibody and Dynabead Protein G magnetic beads for 4 h to overnight. The immune complex was washed and resuspended in kinase buffer (20 mM 2-glycerophosphate, 20 mM MgCl2, 20 mM MnCl2, 1 mM EDTA, 100 μM unlabelled ATP and 10 μCi [γ-32P]ATP). After incubation at 30°C for 30 min, the complex was boiled and resolved by SDS/PAGE. Kinase activity was determined by autophosphorylation and radiography.
Generation of stable Jurkat and fibroblasts
RIP1-deficient Jurkat cells stably transfected with TNFR-2 were described previously . WT (wild-type) and mutant RIP1 (HA-tagged at the N-terminus and GFP-tagged at the C-terminus) were transfected to these TNFR2+ RIP1-deficient Jurkat cells along with pEF6/Myc–HisA (Invitrogen, for blasticidin selection) by electroporation as described previously . For fibroblasts, RIP3−/− fibroblasts were transfected with RIP3 tagged with GFP at the C-terminus with FuGENE HD. Clones were selected with 1 mg/ml geneticin (Gibco®, Invitrogen). Clones with comparable expression were selected for experiments.
Cell death assay
Jurkat cells (50000) were pre-treated with 20 μM Z-VAD-FMK for 1 h before treatment with 100 ng/ml recombinant human TNF (Biosource). Cell death was measured after 16 h by flow cytometry using uptake of PI (propidium iodide) as indicator of cell death. For fibroblasts, cells were pre-treated with 20 μM Z-VAD-FMK for 1 h before 100 ng/ml (or otherwise indicated) of mouse TNF was added. Cell death was measured 18 h later by flow cytometry and PI uptake or by MTS assay (Promega) according to the manufacturer's instructions. Percentage cell loss was calculated using the formula:% cell loss = [1−(number of live cells in treated sample/number of live cells in control sample)]×100%. Results are presented as means±S.E.M. for triplicates of at least two representative experiments. Student's t test or two-way ANOVA analysis was used for statistical analysis.
Ser161 controls sensitivity of RIP1 to Nec-1
During programmed necrosis, RIP1 and RIP3 form filamentous punctate structures that exhibit amyloid-like properties. Amyloid assembly requires the RHIM (RIP homotypic interaction motif) that is present in both RIP1 and RIP3 . Interestingly, expression of truncated RIP1 lacking the N-terminal kinase domain, but not full-length RIP1, resulted in spontaneous clustering of RIP1 into amyloid fibrils . These results suggest that the kinase domain may mask the RHIM to prevent it from forming amyloid filaments. Upon activation, the negative charge generated from phosphorylation of the kinase domain may relieve this inhibition to permit RHIM–RHIM interaction and amyloid assembly. Indeed, the kinase domain of RIP1 was reported to be heavily phosphorylated at multiple serine and threonine residues . In particular, structural modelling of RIP1 on B-Raf suggested that phosphorylation of Ser161 might regulate the pro-necrotic kinase activity of RIP1 . We tested the function of Ser161 and found that an alanine residue substitution only reduced RIP1 kinase activity by approximately 20% (Figure 1A, compare lanes 2 and 5). We previously showed that reconstitution of WT RIP1 into RIP1-deficient Jurkat cells stably expressing TNFR-2 restored TNF-induced programmed necrosis and enhanced TNF-induced apoptosis . Using this system, we found that S161A-RIP1 enhanced TNF-induced apoptosis and restored programmed necrosis to similar levels achieved by WT RIP1 (Figure 1B). As control, kinase inactive RIP1 (K45A) only enhanced TNF-induced apoptosis, but did not restore TNF-induced programmed necrosis (Figure 1B). Therefore despite the minor reduction in kinase activity, S161A-RIP1 is functionally competent to mediate TNF-induced programmed necrosis.
Effect of serine phosphorylation on RIP1 kinase activity
Because Ser161 resides in a potential regulatory loop that contacts the RIP1 inhibitor Nec-1 , we asked whether this mutation might instead affect the sensitivity of RIP1 to Nec-1. We generated RIP1-deficient Jurkat cells that stably express WT RIP1–GFP or S161A-RIP1–GFP. Dose titration showed that TNF-induced programmed necrosis was inhibited by 1 μM Nec-1 in cells expressing WT RIP1, but not S161A-RIP1 (Figure 1C). In contrast, higher doses of Nec-1 inhibited programmed necrosis in both cell types. Hence Ser161 was not critical for the pro-necrotic function of RIP1. Rather, it contributes to Nec-1-mediated inhibition of RIP1 kinase activity.
Ser89 phosphorylation limits RIP1 kinase activity
The lack of effect of S161A-RIP1 on programmed necrosis prompted us to evaluate whether other reported phosphoserine residues in the kinase domain might regulate RIP1 function. We introduced alanine substitutions at other previously reported phosphoacceptor sites and found that the majority of the single-site mutants as well as many compound mutations had little effect on RIP1 kinase activity (Figure 1A, compare lanes 1–9, 12 and 14). Moreover, expression of these mutants in RIP1-deficient Jurkat cells restored TNF-induced programmed necrosis to similar levels achieved by WT RIP1 (Figure 1D). These results show that individually, the previously identified phosphorylation sites on RIP1 have a minimal role in programmed necrosis.
During our analysis, we noticed that S89A-RIP1 was hyperactive compared with WT RIP1 and other RIP1 mutants in the in vitro kinase assay (Figure 1A, lane 13). Ser89 in the kinase domain of RIP1 is conserved in divergent species including Xenopus and zebrafish (Figure 2A). We suspected that phosphorylation of Ser89 might inhibit RIP1 kinase function and programmed necrosis. To test our hypothesis, we introduced aspartic acid substitutions to mimic the negative charge generated from phosphorylation. In contrast with S89A-RIP1, S89D-RIP1 showed reduced kinase activity compared with WT RIP1 (Figure 2B, compare lanes 1–4). Transient expression into RIP1-deficient Jurkat cells revealed that S89A-RIP1 restored programmed necrosis to a higher level compared with that by S89D-RIP1 (Figure 2C). To further elucidate the mechanism by which Ser89 regulates RIP1 function, we generated RIP1-deficient Jurkat cells that stably express WT RIP1, S89A-RIP1 or S89D-RIP1 (Figure 2D). Importantly, Jurkat cells reconstituted with S89A-RIP1 restored TNF-induced programmed necrosis to a higher level than that achieved by S89D-RIP1 or WT RIP1 (Figure 2E). Despite slightly lower expression level, cells expressing S89D-RIP1 responded similarly to TNF-induced programmed necrosis compared with WT RIP1 expressing cells (Figure 2E). This suggests that the reduced kinase activity in S89D-RIP1 was sufficient to initiate programmed necrosis (Figure 2B, lane 4). In contrast, TNF-induced apoptosis was comparable in all cell lines (Figure 2F). These results strongly implicate that phosphorylation of Ser89 inhibits RIP1 kinase activity to limit RIP1-dependent programmed necrosis.
S89A-RIP1 promotes necrosome assembly and programmed necrosis
Ser89 does not affect assembly of the RIP1–RIP3 necrosome
NF-κB activation protects cells against the cytotoxic effects of death cytokines. RIP1 polyubiquitination at the TNFR-1 complex has been shown to inhibit cell survival through NF-κB-dependent and -independent mechanisms . We compared RIP1 ubiquitination in TNFR-1 complexes and found that RIP1 polyubiquitin chains were somewhat shorter in S89A-RIP1 cells compared with those in WT RIP1 cells (Figure 2G, compare lanes 2 and 5). This led us to further examine whether NF-κB activation was impaired. We confirmed that RIP1-deficient Jurkat cells were defective in early, but not late, IκBα phosphorylation (Figure 2H, lanes 1–5). Cells reconstituted with either WT or S89A-RIP1 fully restored this early IκBα phosphorylation deficiency (Figure 2H, compare lanes 2, 7 and 12). In fact, IκBα phosphorylation was more robust in S89A-RIP1 cells at all time points tested (Figure 2H). Moreover, we did not observe detectable level of p100 in Jurkat cells (results not shown), indicating that non-canonical NF-κB activation also did not contribute to the increased necrosis in S89A-RIP1-expressing cells. Assembly of the RIP1–RIP3 necrosome is critical for programmed necrosis . We found that formation of the RIP1–RIP3 necrosome moderately increased at 90 min post-TNF stimulation. However, the overall level of RIP1–RIP3 necrosome was not greatly altered by 3 h after TNF stimulation (Figure 2I, compare lanes 1–3 with 4–6). Hence enhanced programmed necrosis in S89A-RIP1 cells was not due to reduced NF-κB activation or necrosome assembly. Rather, these results suggest that phosphorylation of Ser89 inhibits the kinase activity of RIP1 at a step subsequent to necrosome assembly.
Ser204 regulates the pro-necrotic function of RIP3
Similar to RIP1, RIP3 is also phosphorylated during programmed necrosis. Sequence alignment revealed that Ser204 in the kinase domain of mouse RIP3 (Ser199 for human RIP3) is highly conserved among different species (Figure 3A). Previous work shows that the S199A-RIP3 mutant was unable to undergo necrosis-dependent phosphorylation . However, it was not clear whether that was due to the mutation causing protein misfolding or abolition of an important phosphoacceptor site. We confirmed that an alanine residue substitution for Ser204 (S199A for human RIP3 and S204A for mouse RIP3) abolished in vitro kinase activity of RIP3 (Figure 3B, lanes 2 and 5). By contrast, aspartic acid substitution mutants (S204D for mouse RIP3 and S199D for human RIP3) retained in vitro kinase activity (Figure 3B, lanes 3 and 6). When introduced into RIP3−/− fibroblasts, S204D-RIP3, but not S204A-RIP3, restored TNF-induced programmed necrosis to level similar to that achieved by WT RIP3 (Figure 3C). Similar results were obtained with human RIP3 mutants (Figure 3D). Consistent with the fact that RIP3 does not participate in death receptor-induced apoptosis, TNF- and cycloheximide-induced apoptosis was not affected by mutation of Ser204 (Figure 3E).
Ser204 critically regulates RIP3 function
Although WT and S204D-RIP3 restored TNF-induced programmed necrosis to similar levels, we found that cells expressing S204D-RIP3 were not protected from TNF-induced programmed necrosis by Nec-1. In contrast, cells transfected with WT RIP3 were completely protected by Nec-1 (Figure 3C). This surprising result indicates that RIP1 kinase function is not essential for S204D-RIP3-mediated programmed necrosis. To elucidate further the mechanism of programmed necrosis in S204D-RIP3 cells, we transfected siRNAs against RIP1 into WT RIP3- and S204D-RIP3-expressing fibroblasts. We found that WT RIP3-expressing cells were completely protected from TNF-induced programmed necrosis by two different RIP1 siRNAs despite incomplete silencing of RIP1 expression (Figure 3F). By contrast, RIP1 siRNAs only partially protected cells expressing S204D-RIP3 from TNF-induced programmed necrosis, despite more efficient silencing of RIP1 expression (Figure 3G). RIP3 siRNA equally protected both cell types from programmed necrosis (Figures 3F and 3G). These results strongly suggest that a primary function of RIP1 in programmed necrosis is to directly or indirectly facilitate RIP3 phosphorylation at Ser204. However, RIP1 phosphorylation of RIP3 at other sites or other substrates is required for full activation of programmed necrosis.
Ser232 regulates RIP3 function in programmed necrosis, but not its kinase activity
Besides Ser204, phosphorylation of Ser232 (Ser227 in human RIP3) was recently implicated in promoting RIP3 function . Specifically, S227A-hRIP3 and S232A-mRIP3 were unable to bind MLKL [3,27]. As such, phosphorylation of Ser232 was thought to be critical for recruitment of MLKL. However, sequence alignment revealed that this serine residue is not conserved in sheep RIP3 (Figure 4A). This observation prompted us to re-evaluate Ser232 phosphorylation in programmed necrosis. Consistent with previous reports, S232A-RIP3 was unable to restore TNF- and Z-VAD-FMK-induced programmed necrosis in RIP3−/− fibroblasts (Figure 4B). Surprisingly, the phosphomimetic mutant S232E-RIP3 also failed to restore programmed necrosis in RIP3−/− fibroblasts (Figure 4B). In vitro kinase assay showed that S232A-RIP3 and S232E-RIP3 retained kinase activity (Figure 4C). These results show that Ser232 phosphorylation does not control RIP3 kinase activity and that the negative charge from Ser232 phosphorylation is not sufficient for MLKL binding and programmed necrosis. We reasoned that mutations at Ser232 might simply disrupt the binding surface for MLKL. To test this hypothesis, we introduced S232A or S232E into the gain-of-function S204D-RIP3 mutant and generated RIP3−/− fibroblasts that stably expressed these double mutants. Consistent with our hypothesis, both S204D/S232A-RIP3 and S204D/S232E-RIP3 failed to restore TNF- and Z-VAD-FMK-induced programmed necrosis in RIP3−/− fibroblasts (Figure 4D). Hence we conclude that the binding surface of RIP3 for MLKL is highly sensitive to perturbations and that the mere presence of negative charge is not sufficient to mediate binding to MLKL and programmed necrosis.
Ser232 does not regulate RIP3 function in programmed necrosis
In the present study, we show that Ser89 in RIP1 and Ser204 in RIP3 play critical roles in programmed necrosis signalling. Neither of these residues was identified as phosphoacceptor sites in MS analyses [3,23,27]. This might be due to the fact that only a small fraction of the total cellular pool of RIP1 and RIP3 is involved in TNF signalling and that the active RIP1–RIP3 complex is sequestered in detergent insoluble compartment . Our mutagenesis studies revealed several important mechanistic insights into the control of programmed necrosis. First, the S89A-RIP1 mutant revealed that programmed necrosis could be regulated by positive as well as inhibitory phosphorylation. At present, we do not know the identity of the kinase that mediates this phosphorylation. However, we can rule out PKA (protein kinase A), PKC (protein kinase C) and JNK (c-Jun N-terminal kinase) as the kinase that phosphorylates RIP1 at this site (results not shown). Conversely, dephosphorylation of Ser89 may be an important regulatory event for TNF-induced programmed necrosis. To this end, it is noteworthy that the Pgam5 identified recently promotes TNF-induced necrosis by acting as a phosphatase . In addition, the phosphatase WIP1 (wild-type p53-induced phosphatase 1) has been shown to negatively regulate NF-κB signalling [28,29]. It will be interesting to determine whether Pgam5 or other phosphatases have a similar regulatory role in promoting RIP1 dephosphorylation and necrosis. Secondly, our results suggest that a key function of RIP1 is to facilitate RIP3 phosphorylation at Ser204. Our results are consistent with recent observations that RIP3-dependent, but RIP1-independent, necrosis could occur under certain conditions [30,31]. Interestingly, RIP3 was not phosphorylated by RIP1 in in vitro kinase assays . These discrepant results raise the possibility that a third kinase may be involved in RIP3 activation. Finally, our results show that the mere presence of negative charge at Ser232 is not sufficient to mediate necrosis signalling. Rather, phosphorylation of Thr231/Ser232 likely creates a conformation-sensitive binding surface that is conducive for recruitment of downstream factors such as MLKL .
Fas-associated death domain
human embryonic kidney
inhibitor of NF-κBα
mixed lineage kinase domain-like
nuclear factor κB essential modulator
nuclear factor κB
phosphoglycerate mutase family member 5
RIP homotypic interaction motif
receptor-interacting serine/threonine protein kinase
tumour necrosis factor
Thomas McQuade and YoungSik Cho performed the experiments. Francis Ka-Ming Chan designed and supervised the research and wrote the paper.
We thank Kenta Moriwaki for critically reading the paper.
This work was supported by National Institutes of Health [grant numbers AI083497 and AI088502]. F.K.-M.C. is a member of the University of Massachusetts Diabetes Endocrine Research Center (DERC) [grant DK32520].
These authors contributed equally to this work.
Present address: College of Pharmacy, Keimyung University, Dalseo-Gu, 704-701, Korea.