The Prevalence of TNFα-Induced Necrosis over Apoptosis Is Determined by TAK1-RIP1 Interplay

Death receptor-induced programmed necrosis is regarded as a secondary death mechanism dominating only in cells that cannot properly induce caspase-dependent apoptosis. Here, we show that in cells lacking TGFβ-activated Kinase-1 (TAK1) expression, catalytically active Receptor Interacting Protein 1 (RIP1)-dependent programmed necrosis overrides apoptotic processes following Tumor Necrosis Factor-α (TNFα) stimulation and results in rapid cell death. Importantly, the activation of the caspase cascade and caspase-8-mediated RIP1 cleavage in TNFα-stimulated TAK1 deficient cells is not sufficient to prevent RIP1-dependent necrosome formation and subsequent programmed necrosis. Our results demonstrate that TAK1 acts independently of its kinase activity to prevent the premature dissociation of ubiquitinated-RIP1 from TNFα-stimulated TNF-receptor I and also to inhibit the formation of TNFα-induced necrosome complex consisting of RIP1, RIP3, FADD, caspase-8 and cFLIPL. The surprising prevalence of catalytically active RIP1-dependent programmed necrosis over apoptosis despite ongoing caspase activity implicates a complex regulatory mechanism governing the decision between both cell death pathways following death receptor stimulation.

Ligation of TNF-RI by TNFa also leads to the assembly of a second, cytoplasmic multiprotein complex. Deubiquitination of RIP1 is prerequisite for the formation of complex-II comprising RIP1, RIP3, FADD and caspase-8, which is then activated by auto-cleavage and initiates apoptosis. Caspase-8 also cleaves RIP1 and inactivates it within complex-II [1,3]. TNFa-induced apoptosis is mainly studied in the absence of de novo protein synthesis or in cells lacking crucial activators of the NF-kB pathway, since cells that can trigger TNFa-dependent NF-kB activation express multiple anti-apoptotic genes and block complex-II-mediated apoptosis initiation [4,5,6,7].
TNFa-induced programmed necrosis, or necroptosis, is a recently defined alternative cell death pathway absolutely requiring RIP1 kinase activity and is described to dominate only when dying cells cannot activate caspase-8 [1,3]. Under these conditions, TNFa-induced complex-II acts as the pre-necrotic 'necrosome' complex, where catalytically active RIP1 and RIP3 trigger rapid reactive-oxygen-species (ROS) accumulation and subsequent cell death with morphological features reminiscent of necrosis [1,3,8,9,10]. The catalytic activity of RIP1 is required for the formation of the necrosome complex [8]. Since RIP1 is inactivated by caspase-8, prevention of RIP1 cleavage by caspase inhibition is thought to be required for efficient death receptormediated necrosis induction.
Recent studies indicated a role for cFLIP L , the catalytically inactive homologue of caspase-8, in programmed necrosis inhibition [11,12]. Hetero-dimerisation of caspase-8 and cFLIP L was shown to be necessary for caspase-8 mediated protection from necrosis [12]. However, depending on its expression level, cFLIP L may inhibit or augment caspase-8 activity [13], hence its mechanistic contribution in necrosis is not completely clear.
As TAK1 is a critical component of TNFa-induced NF-kB activation, cells lacking TAK1 expression undergo cell death following TNFa stimulation [14,15,16,17,18]. However, TAK1 knock-out (KO) cells are remarkably more sensitive to TNFainduced cytotoxicity than other types of cells that cannot activate NF-kB [18,19]. In addition, TAK1 KO mice die at an earlier embryonic stage than mice lacking IKKb or NEMO expression [4,5,19,20]. This shows that TAK1 has additional, NF-kBindependent functions during embryonic development. It has been reported that TNFa stimulation of TAK1 KO keratinocytes led to a ROS-mediated, rapid cell death [18]. Moreover, it has recently been shown that in L929 cells, a murine fibrosarcoma cell line that undergoes caspase-independent programmed necrosis upon TNFa stimulation and is commonly used as a model system for this type of cell death, down-regulation of TAK1 augmented the ongoing necrotic response [21]. Here, we dissect the molecular mechanism of TNFa-induced death of TAK1 deficient cells and show that ablation of TAK1 expression alone is sufficient for the induction of TNFa-mediated programmed necrosis in cells that are otherwise resistant to TNFa-induced cytotoxicity. We identified RIP1 as a critical mediator of TNFa-induced ROS accumulation and cell death in the absence of TAK1, indicating the induction of the necrotic pathway. TAK1 mediated the stabilization of polyubiquitinated RIP1 in complex-I and prohibited RIP1-dependent rapid necrosis, unravelling a novel functional connection between these two kinases in addition to their established cooperation in TNFa-induced NF-kB signalling. Moreover, our results demonstrated that RIP1 kinase activitydependent cell death proceeds even in the presence of ongoing caspase activity.

Results and Discussion
TAK1 hinders TNFa-induced, RIP1-mediated rapid cell death To evaluate whether RIP1 plays a role in the death of TAK1 KO mouse embryonic fibroblasts (MEFs), cells were transfected with two different RIP1-targeting siRNA constructs along with a control siRNA. Loss of RIP1 completely prevented TNFa-induced cell death after 2.5 hours of stimulation and provided major protection after 6 hours (Fig. 1A). In addition, TNFa-induced ROS accumulation in TAK1 KO MEFs could be fully prevented by RIP1 down-regulation (Fig. 1B). A specific RIP1 kinase-and programmed necrosis-inhibitor, Necrostatin-1 (Nec-1) [22,23], and the ROS scavenger butylated hydroxyanisole (BHA) also completely blocked TNFa-mediated cytotoxicity (Fig. 1C) [18]. Moreover, Nec-1 was as effective as BHA in blocking TNFainduced ROS accumulation in the absence of TAK1 (Fig. 1D). Taken together, these experiments show that catalytically active RIP1 is essential for TNFa-dependent, ROS-mediated rapid cell death in the absence of TAK1, indicating the triggering of the necrotic pathway.

Necrosis overrules apoptosis in TNFa-treated TAK1 deficient cells
Death receptor-mediated programmed necrosis has been defined to predominate in the absence of caspase activity [1,3]. However, TNFa stimulation induced the cleavage of caspase-8, caspase-3 and the caspase-8 target, RIP1 in TAK1 KO MEFs ( Fig. 2A), similar to previous observations made in keratinocytes [18]. Treatment of TAK1 KO MEFs with the pancaspase inhibitor zVAD-FMK completely abrogated caspase-8 and caspase-3 cleavage (Fig. 2B) as well as the cleavage of RIP1 and the caspase-3 target, PARP1 (Fig. S1A), nevertheless provided only minor protection from TNFa-dependent death (Fig. 2C). Moreover, zVAD failed to block TNFa-induced ROS accumulation in TAK1 KO cells (Fig. 2D). Similar observations were obtained by using another broad-spectrum caspase inhibitor, Q-VD-OPh (Fig. S1). These results indicate that TNFa-dependent caspase activation plays only a marginal role in the death of TAK1 KO cells. Of note, Nec-1 and BHA treatment attenuated caspase-8 and caspase-3 cleavage following stimulation, indicating that TNFa-induced ROS accumulation augmented caspase activation (Fig. 2E) [18]. Nec-1 and BHA provided full protection from TNFa-mediated rapid cell death without a complete block on caspase activation ( Figure 1C and 2E), further supporting the conclusion that ROS-mediated necrosis is the primary cause of the rapid demise of TNFa-stimulated TAK1 KO cells, giving no time to caspasedependent apoptotic processes to mediate cell death.

TAK1 blocks TNFa-dependent necrosome formation
TNFa-mediated necrotic cell death requires the formation of the necrosome complex, including RIP1, RIP3, FADD and caspase-8, whose catalytic activity has to be blocked by either small-molecule inhibitors or virally expressed anti-apoptotic proteins [1,3]. Using FADD immunoprecipitation, we investigated whether TNFa could induce necrosome formation in the absence of TAK1. As expected, WT MEFs required coadministration of zVAD together with TNFa and CHX for efficient necrosome formation (Fig. S2). On the other hand, in TAK1 KO cells, all known components of the necrosome complex immunoprecipitated with FADD within 30-45 minutes of TNFa stimulation (Fig. 3A). Necrosome formation was fully inhibited by Nec-1 treatment, implicating an absolute dependence on catalytically active RIP1 (Fig. 3B). Thus, our study presents TAK1 as a key mediator of TNFa signalling that specifically blocks RIP1 kinase activity-dependent necrosome formation.
Surprisingly, in both WT and TAK1 deficient MEFs, the necrosome complex contained cFLIP L (Fig. 3A, 3B and Fig. S2). Given its dual nature on the modulation of caspase-8 activation [13], further analysis of cFLIP L within the necrosome complex may shed light on the dynamics of necrosome formation and function.
It is of note that at later time points, predominantly the cleaved forms of caspase-8 and cFLIP L (43 kDa) were found in the necrosome complex, although the p43 fragment of caspase-8 was poorly identifiable in input samples ( Fig. 3A and 3B). This indicated that caspase-8 in the necrosome was catalytically active and processed itself as well as cFLIP L . Accordingly, treatment of TAK1 KO MEFs with TNFa and zVAD enhanced the incorporation of full length caspase-8, cFLIP L and RIP1 (Fig. 3B). Whether this translates into an enhanced activation of the necrosome complex is an open question, nevertheless, these experiments show that RIP1 kinase activity-dependent necrosome formation can occur in the presence of ongoing caspase activity. Also, caspase-8-mediated cleavage and the assumed inactivation of RIP1 ( Fig. 2A) [1,3] does not block programmed necrosis initiation following TNFa stimulation.

TAK1 stabilizes Ubi-RIP1 association with complex-I
In order to analyze whether TAK1 curbs the cytotoxic potential of RIP1 by regulating its incorporation into TNFa-induced complex-I, WT and TAK1 KO MEFs were stimulated with Flag-TNFa and complex-I was immunoprecipitated with an a-Flag antibody. Stimulus-dependent incorporation of Ubi-RIP1 [7,24] into complex-I was evident in both cell types, yet in TAK1 KO cells, the amount of Ubi-RIP1 decreased significantly at later time points compared to WT cells (Fig. 3C), indicating that TAK1 stabilizes Ubi-RIP1 in complex-I. In TNFa-treated TAK1 KO cells, the accelerated dissociation of RIP1 from complex-I was accompanied by RIP1-dependent necrosome formation within 30 minutes of stimulation. Hence, TAK1-mediated protection of cells from TNFa-induced rapid cytotoxicity might proceed via regulation of ubiquitinated-RIP1 incorporation into complex-I. In a similar manner, it has been shown that treatment of cells with Smac mimetic to diminish cIAP1/2 expression and Ubi-RIP1 association with complex-I also resulted in RIP1 kinase-activitydependent caspase activation and subsequent cell death following TNFa stimulation [7].

TAK1 has a kinase-independent cytoprotective function
In order to ensure that the sensitivity of TAK1 KO MEFs to TNFa-induced cytotoxicity was only due to the lack of TAK1 expression, cells were stably transfected with wild type TAK1. TAK1 KO MEFs were also reconstituted with a catalytically inactive mutant of TAK1 (TAK1-K63W [25]) to investigate PLoS ONE | www.plosone.org whether TAK1 kinase activity was required for the protection from necrosis. By using a TAK1 kinase inhibitor (NP-009245) at 1 mM, it has been suggested that inhibition of TAK1 catalytic activity further increased caspase-independent necrosis in L929 cells [21]. However, at this concentration, NP-009245 inhibited EGF and PMA-mediated JNK phosphorylation in WT and TAK1 KO MEFs equally well, indicating that NP-009245 did not enhance TNFa-mediated necrosis specifically via TAK1 inhibition (Fig. S3).
As expected, TAK1 KO cells reconstituted with TAK1-WT were capable of TNFa-induced IkBa degradation, which was impaired in cells expressing TAK1-K63W (Fig. 4A). Nevertheless, reconstitution of TAK1 KO cells with either construct resulted in a prolonged association of Ubi-RIP1 with TNFa-induced complex-I compared to empty vector-transfected cells (Fig. 4B). This suggested a scaffold function for TAK1 to restrict the premature dissociation of RIP1 from complex-I. In parallel with this observation, TNFa-dependent necrosome formation was inhibited upon reconstitution with either TAK1-WT or TAK1-K63W (Fig. 4C). Moreover, expression of TAK1-WT or TAK1-K63W in TAK1 KO cells provided complete protection from TNFa-dependent cytotoxicity (Fig. 4D) and also blocked TNFainduced caspase activation (Fig. 4E) as well as ROS accumulation (Fig. 4F). These data demonstrate that TAK1 acts as an adaptor molecule in complex-I to prevent premature dissociation of Ubi-RIP1 and also functions independently of its kinase activity to block TNFa-induced, RIP1-mediated necrosis. Therefore, our data suggest that TAK1 may suppress the pro-necrotic activity of RIP1 by sequestering it within complex-I, highlighting a novel non-catalytic function of TAK1 in complex-I in addition to its role as a signal transducer for TNFa-mediated NF-kB activation.
RIP1-dependent necrotic cell death is not only observed following death receptor stimulation, but also in acute pancreatitis [9,10], ischemia-induced brain injury [23] as well as myocardial infarction [26]. Hence, future studies may reveal whether TAK1 also modulates RIP1-dependent necrosis under these pathophysiological conditions and help further characterization of TAK1-RIP1 interplay in necrotic cell death.

Cell culture and transfections
WT and TAK1 KO MEF cells [19] were a gift of Dr. Shankar Ghosh (Columbia University, New York, USA) and maintained in DMEM-GlutaMAX (Gibco) supplemented with 10% FCS (Gibco) and penicillin/streptomycin (100 U/ml and 100 mg/ml). To reconstitute TAK1 deficient MEFs with pTRACER-TAK1-WT and pTRACER-TAK1-K63W along with empty pTRACER for empty vector transfection, cells were transfected by using Transfectin (Biorad) reagent according to manufacturer's instructions. Stably transfected monoclonal cells were selected with 80-100 mg/ml Zeocin (Invitrogen) and expanded for experimentation. To downregulate protein expression, cells were transfected with different siRNA (10-20nM) constructs with the help of lipids from Silence Therepeutics according to manufacturer's instructions and harvested 3 days after transfection. All cells were starved overnight before analysis.

Intracellular ROS measurement with carboxy-H 2 DCFDA and Propidium Iodide staining
MEFs were stimulated for indicated time points and were supplemented with 10 mM carboxy-H 2 DCFDA (Invitrogen) in the last 30 minutes of incubation at 37uC. Cells were then trypsinized, washed and re-suspended in 1 ml PBS. Intracellular ROS accumulation was analyzed by using the FL1 channel of a BD FACSCalibur flow cytometer. Propidium iodide (PI, Sigma) was added (2 mg/ml) to each sample just before measurement. PI staining was visualized by using the FL3 channel. Data was analyzed by using FlowJo software. In all experiments, cells were FSC/SSC gated before analysis. ROS accumulation was visualized in the living population (PI negative cells).

Crystal violet staining (CV)
Following medium removal, MEFs (in 12-well plates) were incubated for 5 minutes with 1 ml PFA/PBS (4%, RT). PFA/ PBS was replaced by 1 ml of 0.01% CV solution (Sigma, in ddH 2 O) and incubated for 20-30 minutes. Plates were washed 3-4x with tap water and let dry briefly. CV was solubilised with methanol, and its absorbance was measured at 540 nm. All treatments were done in at least triplicates. The graphs depict mean values 6 SEM %. were pre-treated as in (B) and stimulated with TNFa for 1 hour. Cells were lysed and immunoprecipitated with an a-FADD antibody to subsequently analyze RIP1 co-immunoprecipitation. Weak RIP1 co-immunoprecipitation with FADD was only observed at 1 mM of NP-009245. (TIF)