TNF-induced necroptosis initiates early autophagy events via RIPK3-dependent AMPK activation, but inhibits late autophagy

ABSTRACT Macroautophagy/autophagy and necroptosis represent two opposing cellular s tress responses. Whereas autophagy primarily fulfills a cyto-protective function, necroptosis is a form of regulated cell death induced via death receptors. Here, we aimed at investigating the molecular crosstalk between these two pathways. We observed that RIPK3 directly associates with AMPK and phosphorylates its catalytic subunit PRKAA1/2 at T183/T172. Activated AMPK then phosphorylates the autophagy-regulating proteins ULK1 and BECN1. However, the lysosomal degradation of autophagosomes is blocked by TNF-induced necroptosis. Specifically, we observed dysregulated SNARE complexes upon TNF treatment; e.g., reduced levels of full-length STX17. In summary, we identified RIPK3 as an AMPK-activating kinase and thus a direct link between autophagy- and necroptosis-regulating kinases. Abbreviations: ACACA/ACC: acetyl-CoA carboxylase alpha; AMPK: AMP-activated protein kinase; ATG: autophagy-related; BECN1: beclin 1; GFP: green fluorescent protein; EBSS: Earle’s balanced salt solution; Hs: Homo sapiens; KO: knockout; MAP1LC3/LC3: microtubule associated protein 1 light chain 3; MEF: mouse embryonic fibroblast; MLKL: mixed lineage kinase domain like pseudokinase; Mm: Mus musculus; MTOR: mechanistic target of rapamycin kinase; MVB: multivesicular body; PIK3C3/VPS34: phosphatidylinositol 3-kinase catalytic subunit type 3; PIK3R4/VPS15: phosphoinositide-3-kinase regulatory subunit 4; PLA: proximity ligation assay; PRKAA1: protein kinase AMP-activated catalytic subunit alpha 1; PRKAA2: protein kinase AMP-activated catalytic subunit alpha 2; PRKAB2: protein kinase AMP-activated non-catalytic subunit beta 2; PRKAG1: protein kinase AMP-activated non-catalytic subunit gamma 1; PtdIns3K: phosphatidylinositol 3-kinase; PtdIns3P: phosphatidylinositol-3-phosphate; RIPK1: receptor interacting serine/threonine kinase 1; RIPK3: receptor interacting serine/threonine kinase 3; SNAP29: synaptosome associated protein 29; SNARE: soluble N-ethylmaleimide-sensitive factor attachment protein receptor; SQSTM1/p62: sequestosome 1; STK11/LKB1: serine/threonine kinase 11; STX7: syntaxin 7; STX17: syntaxin 17; TAX1BP1: Tax1 binding protein 1; TNF: tumor necrosis factor; ULK1: unc-51 like autophagy activating kinase 1; VAMP8: vesicle associated membrane protein 8; WT: wild-type.

Both autophagy and necroptosis balance cell death and survival. Although there are some reports indicating a crosstalk between these two stress responses, the mechanistic details remain poorly understood so far. Here, we show that TNF-induced necroptosis blocks the lysosomal degradation of autophagosomes, presumably via the dysregulation of the autophagosome-lysosome fusion-mediating SNARE (soluble N-ethylmaleimide-sensitive factor attachment protein receptor) proteins. In detail, we observe reduced interactions of VAMP8 (vesicle associated membrane protein 8), STX17 (syntaxin 17) and STX7 (syntaxin 7) with SNAP29 (synaptosome associated protein 29), and a decrease of full-length STX17. In contrast, it appears that early autophagy signaling events such as ULK1 and BECN1 phosphorylation are induced under pro-necroptotic conditions and in a RIPK3dependent manner. We observe that RIPK3 interacts with AMPK and that RIPK3 phosphorylates the catalytic subunit PRKAA1 at T183 in order to activate AMPK. Subsequently, pro-autophagic AMPK substrates such as ULK1 and BECN1 become phosphorylated. Collectively, we have identified RIPK3 as AMPK-activating kinase and thus a molecular crosstalk between the autophagy-and necroptosis-regulating kinases.

TNF-induced necroptosis is accompanied by AMPK activation and phosphorylation of ULK1 and BECN1
Treatment of murine L929 fibroblasts with TNF in combination with the caspase inhibitor QVD (hereafter abbreviated as TQ) leads to the induction of necroptosis ( Figure S1A). In order to investigate the effect of necroptosis on autophagy, we initially monitored early autophagy signaling events under these pro-necroptotic conditions. We observed increased activity of AMPK, as detected by immunoblotting for phospho-PRKAA1 (T183), phospho-ACACA (S79), phospho-ULK1 (S555, murine sequence; corresponding to S556 in human ULK1), and phospho-BECN1 (S91, murine sequence; corresponding to S93 in human BECN1) ( Figure  1A). These increased phosphorylation events were clearly dependent on RIPK3 signaling, since they were abolished by siRNA-mediated RIPK3 knockdown. Similar results were obtained when we used a RIPK3 inhibitor (GSK'872 [40]) instead of Ripk3 siRNA ( Figure S1B). In order to confirm these observations in an alternative cellular model system, we made use of wild-type (WT) and ripk3 KO murine embryonic fibroblasts (MEFs). Here, we induced necroptosis by treatment with TNF in combination with a DIABLO/Smac mimetic (here: Birinapant) and the pancaspase inhibitor zVAD (hereafter abbreviated as TSZ) as previously reported [37], and neither of these components alone affected AMPK activity ( Figure 1B, left panels). TSZ induced AMPK activation and phosphorylation of the AMPK substrates ACACA, ULK1 and BECN1 in wild-type MEFs, whereas these processes were inhibited in ripk3 KO MEFs ( Figure 1B, right panels).
Generally, AMPK-dependent phosphorylation of ULK1 at S555 (human S556) and of BECN1 at S91 (human S93) have been associated with the induction of autophagy [10,16,41,42]. Accordingly, we speculated that initial autophagy signaling events are activated upon TQ-or TSZ-induced necroptosis in a RIPK3-dependent manner. To test this hypothesis, we tested downstream markers of autophagy by immunofluorescence in L929 cells, e.g. ATG14 ( Figure 2A) and ATG16L1 ( Figure 2B). We observed that TQ treatment induces ATG14 Figure 1. TNF treatment induces activation of AMPK. (A) L929 cells were transfected with non-targeting (siCtrl) or Ripk3 siRNAs (siRipk3). 48 h post transfection, cells were exposed to 10 ng/ml TNF and 30 µM QVD (TQ) for the indicated times. Then, whole cell lysates were subjected to SDS-PAGE and immunoblotting for indicated proteins. A compilation of representative immunoblots is shown; three ACTB immunblots are shown, but each protein was normalized to its corresponding loading control. The density of each protein band was divided by the average of the density of all bands from the same protein on the membrane. Fold changes were calculated by dividing each normalized ratio (protein to loading control) by the average of the ratios of the control lane (scr, 0 h TQ). Results are mean + SD from at least 3 independent experiments. Statistical analysis was done by repeated measures two-way ANOVA (corrected by Sidak's multiple comparisons test between siRNAs and corrected by Tukey's multiple comparisons test between time points). Statistically significant differences within non-targeting siRNA-transfected cells (compared to scr, 0 h TQ) are depicted as letters directly above the bars. * or a: P < 0.05, ** or b: P < 0.01, *** or c: P < 0.001, **** or d: P < 0.0001. (B) Ripk3 WT and KO MEFs were exposed to indicated treatments (medium [  puncta formation, and that this effect was significantly blocked by GSK'872-mediated inhibition of RIPK3 ( Figure  2A). Similarly, TQ-induced ATG16L1 puncta formation was inhibited in ripk3 KO L929 cells or in WT L929 cells treated with GSK'872 ( Figure 2B). Of note, this ATG16L1 puncta formation was clearly dependent on AMPK, because both Prkaa1/2 siRNA and treatment with the AMPK inhibitor dorsomorphin prevented TQ-induced ATG16L1 puncta formation ( Figure 2C and Figure 2D). Taken together, it appears that TQ or TSZ treatment induces early autophagy signaling events via a RIPK3-AMPK signaling axis. We also investigated whether RIPK3 is also involved in canonical starvationinduced autophagy, but we did not observe a significant difference of EBSS-induced AMPK activation or LC3 lipidation between WT and ripk3 KO L929 cells ( Figure S2A). The RIPK3-dependent induction of early autophagy signaling events upon TQ or TSZ treatment might pursue a cytoprotective function and thus slow down the execution of necroptosis. Along these lines, necroptosis-inhibitory functions have been attributed to AMPK and to ULK1 [43][44][45]. Accordingly, we observed increased TQ-induced cell death in L929 cells upon PRKAA1/2 knockdown ( Figure S2B) and increased TSZ-induced cell death in prkaa1 prkaa2 or ulk1 ulk2 double-knockout MEFs compared to wild-type control cells ( Figures S2C and S2D).

RIPK3 interacts with AMPK
Since we observed that the induction of necroptosis affected AMPK-dependent signaling and early autophagy events, we next investigated the crosstalk between AMPK and the pronecroptotic RIPK3. In a first approach, we performed size exclusion chromatography of S100 lysates derived from MEFs ( Figure 3A). We detected the ULK1 complex in high molecular mass fractions of approximately 3 MDa as previously described [46][47][48]. Of note, AMPK and both RIPK1 and RIPK3 were present in fractions corresponding to a lower molecular mass range of 14-158 kDa ( Figure 3A). Since the presence of proteins in the same fractions does not prove a direct interaction, we next performed immunopurification experiments. Since we observed a RIPK3-dependent induction of early autophagy signaling events, we focused on RIPK3. We transiently transfected cDNA encoding 3xFLAG-HsRIPK3 into HEK293 cells. Following anti-FLAG immunopurification, we detected co-purification of endogenous AMPK ( Figure  3B). In a similar approach, we co-purified 3xFLAG-HsRIPK3 with purified AMPK ( Figure 3C). Furthermore, we also immunopurified endogenous AMPK with endogenous RIPK3 ( Figure 3D). The direct interaction between RIPK3 and AMPK was also confirmed by affinity purification using recombinant proteins ( Figure 3E). Finally, we confirmed the interaction between these two kinases on the cellular level by a proximity ligation assay (PLA). In this assay, single proteinprotein interactions can be detected using antibody-recognition combined with exponential signal amplification by PCR. We transfected ripk3 KO MEFs with cDNA encoding 3xFLAG-MmRIPK3. 3xFLAG-MmRIPK3 was stained with rabbit anti-RIPK3 antibodies and AMPK with mouse anti-PRKAA1/2 antibodies. As control, ripk3 KO MEFs were transfected with empty vector. Cells reconstituted with 3xFLAG-MmRIPK3 displayed strong signals with significant difference to control cells ( Figure 3F). Collectively, these data indicate that AMPK can associate with RIPK3.

RIPK3 directly phosphorylates PRKAA1 at T183 and thus activates AMPK
Because we observed that RIPK3 interacts with AMPK, we next tested if RIPK3 can phosphorylate AMPK. In a first approach, we used recombinant RIPK3 as kinase and GST-HsPRKAA1 (1-278) or GST-HsPRKAA1 (279-559) as substrates in an in vitro kinase assay. We observed that RIPK3 can phosphorylate both truncated versions of PRKAA1 ( Figure 4A). Since we observed that RIPK3 can regulate AMPK activity upon TQ or TSZ treatment ( Figure 1A, Figures 1B and S1B), we wondered if RIPK3 could directly phosphorylate PRKAA1 at T183 to regulate AMPK activity. To test this hypothesis, we generated the mutant GST-HsPRKAA1 (1-278) T183A and repeated the in vitro kinase assay with cold ATP. We performed immunoblotting for phospho-PRKAA1 T183 and observed a specific band for GST-HsPRKAA1 (1-278), but not for the GST-HsPRKAA1 (1-278) T183A mutant ( Figure 4B). This RIPK3-dependent phosphorylation of PRKAA1 at T183 was sensitive to both alkaline phosphatase and RIPK3 inhibitor GSK'872 treatment ( Figure  4C and Figure 4D). To test if RIPK3 phosphorylates AMPK in intact cells, we treated HEK293 cells that express either 3xFLAG-HsRIPK3 WT or a kinase-dead (KD) variant with TQ for 24 h. We observed strong phospho-PRKAA1 T183 signal in HEK293 cells expressing 3xFLAG-HsRIPK3 compared to cells expressing kinase-dead 3xFLAG-HsRIPK3 ( Figure 4E). Using again a proximity ligation assay, we also detected stronger PRKAA phosphorylation in ripk3 KO MEFs reconstituted with 3xFLAG-MmRIPK3 compared to non-reconstituted ripk3 KO MEFs ( Figure 4F). Although these assays confirm the direct phosphorylation of AMPK by RIPK3, we wanted to investigate whether an additional mechanism contributes to AMPK activation during necroptosis. It has previously been described that the induction of necroptosis results in reduced cellular ATP levels [49,50]. In order to analyze the AMP-dependency of the observed AMPK activation, we expressed 3xFLAG-tagged wildtype PRKAG1 and an R299G variant in L929 cells. This mutation affects the binding of AMP at site 3 in human PRKAG1 [51]. Upon treatment with TQ and subsequent anti-FLAG immunopurification, we did not observe any differences in PRKAA1/2 T183/T172 phosphorylation between cells expressing wild-type or PRKAG1 R299G ( Figure 4G). These data indicate that an intact AMP-binding site 3 on PRKAG is not required for TQ-induced AMPK activation. Altogether, our antibodies. Puncta quantification was done using ImageJ software. Data represent mean + SD. A minimum of 122 cells was analyzed. (A-D) Statistical analysis was performed using ordinary one-way ANOVA (corrected by Tukey's multiple comparisons test) for A, B and D; or two-way ANOVA (corrected by Tukey's multiple comparisons test) for C. For B, statistical analysis was additionally performed using unpaired t test with Welch's correction (TQ treatment of WT vs. ripk3 KO cells). ****P < 0.0001. Scale bar: 20 µm.    and the density of each protein band was divided by the average of the density of all bands from the same protein on the membrane. (B) HEK293 cells were left untransfected or were transfected with a vector encoding 3xFLAG-HsRIPK3 for 24 h. Then, cells were lysed and cleared cellular lysates were subjected to immunopurification using anti-FLAG beads. Purified proteins were subjected to SDS-PAGE and analyzed by immunoblotting for FLAG and AMPK. (C) HEK293 cells were left untransfected or were transfected with a vector encoding 3xFLAG-HsRIPK3 for 24 h. Then, cells were lysed and cleared cellular lysates were subjected to immunopurification using anti-AMPK antibodies. Purified proteins were subjected to SDS-PAGE and analyzed by immunoblotting for FLAG and AMPK. (D) L929 cells were lysed and cleared cellular lysates were subjected to immunopurification using anti-IgG or anti-RIPK3 antibodies. Purified proteins were subjected to SDS-PAGE and analyzed by immunoblotting for AMPK and RIPK3. (E) GST or GST-MmRIPK3 immobilized on glutathione-Sepharose beads was incubated with His-AMPK [His-HsPRKAA1 (11-559) + HsPRKAB2 (1-272) + HsPRKAG1 (1-331)] overnight. After washing the beads, bound proteins were eluted by boiling for 10 min at 95°C. Proteins were subjected to SDS-PAGE and analyzed by results indicate that RIPK3 phosphorylates PRKAA1 at T183 to regulate AMPK activity upon TNF-induced necroptosis. This RIPK3-dependent activation of AMPK then induces early autophagy events.

TNF-induced necroptosis inhibits lysosomal autophagosome degradation
Generally, the above-described formation of ATG14 or ATG16L1 puncta upon TQ treatment might also be caused by a blockade of the autophagic flux. Accordingly, we also aimed at investigating later steps of the autophagic pathway. Following TQ treatment, we monitored levels of MAP1LC3/ LC3-II, which is the phosphatidylethanolamine-conjugated form of LC3 and represents a marker protein for autophagosomes. We observed that LC3-II levels increased with time upon TQ treatment ( Figure 5A). TQ-induced LC3-II levels did not further increase with simultaneous bafilomycin A 1 treatment, indicating that lysosomal LC3-II (and thus autophagosome) degradation was rather blocked than induced upon TNF-induced necroptosis ( Figure 5A).
Next to LC3 turnover, we also analyzed levels of the autophagy receptors SQSTM1/p62 and TAX1BP1, respectively. Of note, these autophagy receptors decreased upon TQ treatment, and this reduction was sensitive to bafilomycin A 1 treatment ( Figure 5A). It has previously been observed that starvation induces the rapid degradation of autophagy receptors by endosomal microautophagy, i.e., a lysosomedependent but macroautophagy-independent pathway [52]. To further evaluate this observation, we made use of the PIK3C3/VPS34 inhibitor SAR405, which blocks early steps of macroautophagy. Indeed, autophagy receptors still became degraded in the presence of SAR405 ( Figure S3A), indicating that this reduction is not caused by the execution of macroautophagy. Previously, the starvation-induced degradation of autophagy receptors by endosomal microautophagy has been associated with late endosomes including multivesicular bodies (MVBs) [52]. Interestingly, we also observed MVBlike structures in L929 cells treated with TQ ( Figure S3B). Furthermore, the degradation of autophagy receptors by endosomal microautophagy was reported to depend on the endosomal sorting complex required for transport III (ESCRT-III) component CHMP4B (charged multivesicular body protein 4B) [52]. Similarly, we observed that the TQinduced degradation of SQSTM1/p62 and TAX1BP1 was abolished upon siRNA-mediated depletion of CHMP4B ( Figure S3C).
The TQ-induced accumulation of LC3-II was clearly dependent on RIPK3-dependent necroptosis, since we did not detect LC3-II accumulation in RIPK3-or MLKLdeficient L929 cells ( Figure 5B), or in cells treated with Ripk3 siRNA or the RIPK3 inhibitor GSK'872 ( Figures S4A  and S4B). In order to confirm our observation of blocked LC3 turnover by an independent approach, we performed LC3 immunofluorescence in L929 cells. Treatment with TQ increased LC3-positive puncta ( Figure S4C). Similar to the LC3 turnover analysis by immunoblot, this increase in LC3 puncta was not further increased by parallel bafilomycin A 1 treatment but was reduced by RIPK3 knockdown. To further confirm our observation that the autophagic flux is indeed blocked on the lysosomal level, we made use of the mRFP-EGFP-rLC3 tandem construct [53,54]. The GFP signal is sensitive to the acidic and/or proteolytic environment of lysosomes, whereas the mRFP signal is more stable [53,54]. We detected a strong accumulation of the tandem fluorescencetagged LC3 upon TQ or medium/bafilomycin A 1 treatment ( Figure 5C), and again the TQ-induced mRFP-EGFP-rLC3 accumulation was reduced upon siRNA-mediated knockdown of RIPK3 ( Figure 5C). Finally, we assessed autophagic flux by monitoring mCitrine-LC3 degradation by flow cytometry. Whereas starvation induced a clear reduction of mCitrine-LC3 levels, mCitrine-dependent fluorescence remained unaltered or even increased upon TQ treatment ( Figure 5D). Taken together, we showed by different assays that TQinduced necroptosis blocks lysosomal autophagosome degradation.

TNF inhibits autophagic flux via dysregulating SNARE-mediated autophagosome-lysosome fusion
In order to characterize the lysosomal blockade described above, we first measured CTSB (cathepsin B) and CTSL (cathepsin L) activities. However, we did not detect any differences between TQ-treated or control cells, whereas this was clearly the case for bafilomycin A 1 treatment ( Figure S5). The fusion of autophagosomes with lysosomes is a critical step during the autophagic pathway, and this step is mediatedamong other factors-by autophagosomal and lysosomal SNARE proteins. First, we investigated the stability of STX17-SNAP29-VAMP8 and YKT6-SNAP29-STX7 complexes, respectively, since these SNARE complexes have been implicated in the autophagosome-lysosome fusion process [55,56]. For that, we overexpressed GFP-SNAP29 in L929 cells and investigated the interactions with the SNARE proteins upon TQ treatment ( Figure 6A). We observed reduced binding to VAMP8, STX17 and STX7 upon treatment with TQ. These data indicate that the SNARE complexes responsible for the fusion of autophagosomes with lysosomes become destabilized upon TQ treatment. Furthermore, we also observed that the destabilization of these interactions could be prevented by RIPK1 and RIPK3 inhibitors, respectively ( Figure  6B). We also analyzed general STX17 expression upon TQ treatment. We observed that this treatment induces the reduction of full-length STX17 ( Figure 6C). Furthermore, we simultaneously observed the appearance of an additional band at a lower molecular weight, suggesting a possible cleavage of immunoblotting for AMPK and GST. (F) ripk3 KO MEFs were retrovirally transfected with empty vector or cDNA encoding 3xFLAG-MmRIPK3. Cells were seeded onto glass coverslips. The next day, cells were fixed and analyzed using proximity ligation assay as described in the material and methods section (anti-RIPK3: Prosci, 2283; anti-PRKAA1/2: Cell Signaling Technology, 2793). Nuclei were stained with DAPI. Signals and nuclei per image were counted and the signal:nuclei ratio was calculated. Data represent mean + SD. A minimum of 216 cells was analyzed. Statistical analysis was performed using an unpaired t test with Welch's correction. ****P < 0.0001. Scale bar: 20 µm.  Figure 6C). This TQ-induced STX17 fragment was also observed for overexpressed GFP-STX17 ( Figure S6A). Furthermore, we confirmed that the detected fragment derives from STX17 by siRNA ( Figure S6B). The appearance of this truncated STX17 was clearly dependent on the execution of necroptosis, since we did not observe this fragment in ripk3 or mlkl KO L929 cells ( Figure 6D) or in cells treated with GSK'872 ( Figure 6E). In contrast, the STX17 fragment was detectable in cells with a knockdown of PRKAA1/2 ( Figure  6E), indicating that induction of early autophagy events via the RIPK3-AMPK and the presumable cleavage of STX17 are independent events. This was also confirmed by two alternative approaches. First, we overexpressed 3xFLAG-MmRIPK3 in wild-type L929 cells. In these cells, phosphorylation of AMPK and of its downstream substrates was induced, whereas the shorter STX17 fragment was not detected ( Figure 6F). Second, the combination of AMPK activating compounds with bafilomycin A 1 induced AMPK activation, phosphorylation of AMPK substrates, and LC3 accumulation ( Figure S6C). However, neither STX17 was cleaved nor cell death induced upon this treatment ( Figure S6C and S6D). Taken together, early autophagy events can be mediated by RIPK3, but additional pro-necroptotic, TNF-dependent signaling events are required for the dysregulation of SNAREmediated autophagosome-lysosome fusion.

Discussion
Autophagy, apoptosis and necroptosis are three main stress responses regulating life and death of a cell. Whereas the crosstalk between autophagy and apoptosis is well established, the relationship between autophagy and necroptosis awaits further clarification. Although recent reports suggest the existence of this crosstalk [57][58][59][60], mechanistic details are rudimentary. Here, we report that necroptosis induced by TNF blocks late events of the autophagic pathway (i.e. the degradation of autophagosomes by lysosomes) at least in part via the dysregulation of SNARE-mediated autophagosome-lysosome fusion. We also found that early signaling events of autophagy -for example AMPK-dependent phosphorylation of ULK1 and BECN1 or formation of ATG14 or ATG16L1 punctaare induced upon TQ or TSZ treatment. Remarkably, we found that the pro-necroptotic RIPK3 directly phosphorylates PRKAA1/2 at T183/T172 and thus activates it. This is especially noteworthy since only limited number of examples exist for both RIPK3 substrates and AMPK-activating kinases.
At first glance, the promotion of early autophagy events and the simultaneous inhibition of late events appear to be contradictory. We observed that autophagy-specifically the lysosomal degradation of autophagosomes-is blocked upon TQ treatment and necroptosis induction. Additionally, we observed that the lysosomal CTSB and CTSL remain functional upon TQ treatment, which was not the case for bafilomycin A 1 treatment. We found that the blockade of autophagosome degradation is-likely among additional mechanisms-caused by the dysregulation of proteins that participate in the fusion of autophagosomes with lysosomes. Specifically, we observed destabilized interactions between VAMP8, STX17, STX7 and SNAP29, a reduction of fulllength STX17, and the appearance of a shorter STX17derived fragment upon TQ treatment. An increased lipidation of LC3 due to the inhibition of the autophagic flux has also been postulated by Frank et al. recently [57]. They postulate that activated MLKL has to associate with intracellular membranes in order to inhibit autophagy [57]. Further studies will have to determine whether this function of MLKL is connected to the observed STX17 modification. However, we found that the presence of RIPK3 and MLKL are clearly required for both LC3 accumulation and STX17 fragmentation upon TQ treatment. Generally, proteolytic events control several forms of regulated necrosis [61]. Since our stimuli contain the caspase inhibitors QVD or zVAD, we can exclude the involvement of caspases in the proteolytic cleavage of STX17. It is well established that inhibitors of serine proteases such as tosyl phenylalanyl chloromethyl ketone (TPCK) can block necroptosis [61], but the specific substrates that become cleaved during necroptosis await further clarification. In our case, we could not block STX17 truncation upon TQ treatment using TPCK (data not shown). Nevertheless, the dysregulation of SNARE-mediated autophagosome-lysosome fusion and the resulting blockade of autophagy has been observed for other stimuli, e.g. enteroviral infections [62,63].
Interestingly, we observed the degradation of the selective autophagy receptors SQSTM1/p62 and TAX1BP1 upon TQ treatment, although autophagosome degradation is blocked. We think that this turnover is caused by a macroautophagyindependent mechanism, and this is clearly supported by our experiments using the PIK3C3/VPS34-specific inhibitor SAR405. It has previously been described that starvation induces the rapid degradation of selective autophagy receptors by endosomal microautophagy [52]. Although we do not necessarily think that these are identical phenomena, it is noteworthy that also TQ treatment induces the accumulation of MVBs and that the degradation of autophagy receptors was abolished upon knockdown of CHMP4B (Figures S3B and  S3C). In contrast to the LC3 levels reported for the immediate response to starvation [52], we observed a TQ-induced accumulation of LC3. Notably, we did not observe LC3-II accumulation in cells treated with TQ plus SAR405 and bafilomycin A 1 , whereas this was clearly the case for the combination of TQ with bafilomycin A 1 alone ( Figure S3A). We speculate that TQ plus SAR405 represents a strong cell death stimulus, which might also induce non-canonical and specifically PIK3C3/VPS34-independent LC3 lipidation [64].
coverslips. The next day, the cells were left untreated (medium, M) or treated with 30 ng/ml TNF + 100 nM SMAC-mimetic + 20 µM z-VAD (TSZ) for 3 h. Then cells were fixed and analyzed using proximity ligation assay as described in the material and methods section (anti-phospho-PRKAA1/2 T183/T172: Cell Signaling Technology, 2535; anti-PRKAA1/2: Cell Signaling Technology, 2793). Nuclei were stained with DAPI. Signals and nuclei per image were counted and the signal:nuclei ratio was calculated. Data represent mean + SD. A minimum of 107 cells was analyzed. Statistical analysis was performed using ordinary two-way ANOVA (corrected by Tukey's multiple comparisons test). ****P < 0.0001. Scale bar: 20 µm. (G) L929 cells were transiently transfected with cDNA encoding either 3xFLAG-HsPRKAG1 WT or R299G for 24 h. After that, cells were treated with or without 10 ng/ml TNF + 30 µM QVD (TQ) for 2 h. Then, cells were lysed and cleared cellular lysates were subjected to immunopurification using anti-FLAG beads. Purified proteins were subjected to SDS-PAGE and analyzed by immunoblotting for indicated proteins. In these pathways, LC3 conjugation occurs at single membranes [65]. For example, it has been reported that drugs possessing lysosomotropic or ionophore drugs are able to induce LC3 lipidation at single-membrane compartments of the endolysosomal pathway [66][67][68]. It has also been suggested that these non-canonical LC3 lipidations require vacuolar-type H + -ATPase (V-ATPase) activity [67,69], and accordingly treatment with bafilomycin A 1 in combination with TQ and SAR405 causes a decrease of LC3-II [69]. The strong cell death induction by the combination of TQ and SAR405 can also be deduced from the bafilomycin A 1insensitive degradation of the selective autophagy receptors SQSTM1/p62 and TAX1BP1, suggesting a caspase-and lysosome-independent demise of these proteins during cell death. Howsoever, in summary we observed a TNF-induced blockade of autophagosome degradation, and the dysregulation of SNARE-mediated autophagosome-lysosome fusion likely contributes to this blockade.
Generally, the blockade of cyto-protective pathways upon necroptosis induction is comprehensible and is similar to the crosstalk between apoptosis and autophagy, in which e.g. activated caspases cleave several autophagy-relevant proteins and thus inactivate their autophagic function [70]. In contrast, the induction of early autophagy signaling events upon TQ or TSZ treatment is less intuitive. We speculate that this phenomenon might be explained by two hypotheses: 1) an antinecroptotic and thus cyto-protective function of RIPK3, or 2) a pro-necroptotic function of early autophagy events. In the first scenario, it might be that RIPK3 regulates cyto-protective autophagy independently of its pro-death function during necroptosis. So far, RIPK3 has been attributed a central role in necroptosis signaling, and the best-characterized RIPK3 substrate MLKL is a main executor of necroptosis [31,32,39]. A possible role of RIPK3 in the regulation of autophagy has also been suggested by two other groups. Harris et al. suggest RIPK3 positively regulates autophagy since depletion of RIPK3 inhibits autophagic flux and leads to the accumulation of autophagosomes and amphisomes [59]. In contrast, Matsuzawa et al. reported that RIPK3 serves as a negative regulator of selective autophagy by binding to SQSTM1/p62 and thus regulating SQSTM1/p62-LC3 complex formation [71]. We observed that RIPK3 positively regulates AMPK activity. Activated AMPK phosphorylates both ULK1 and BECN1, and these two phosphorylation events have been associated with a positive regulation of autophagy [10,16,41,42]. We also observed increased ATG14 and ATG16L1 puncta formation in a RIPK3-dependent manner, although we cannot exclude that these puncta formations are caused by the TNF-mediated blockade of autophagosome degradation. The RIPK3-dependent induction of early autophagy signaling events might pursue a cyto-protective function and slow down the execution of necroptosis. Along these lines, necroptosis-inhibitory functions have been attributed to AMPK and to ULK1 [43][44][45], and this is confirmed by our data (Figure S2B-D).
In a second and opposing scenario, one might speculate that the induction of early autophagy signaling events supports necroptosis. Goodall et al. reported that inhibition of late stage autophagy enhanced TNFSF10/TRAIL-induced cell death. In contrast, genetic or pharmacological inhibition of early/mid-stage autophagy prevented cell death [58,72]. The authors hypothesized that components of the autophagy machinery mediate cell death by functioning as a scaffold for necrosome formation and activation. Specifically, they propose that the interaction between SQSTM1/p62 and RIPK1 localizes the necrosome on the autophagosome [58,72]. It is tempting to speculate that RIPK3 itself promotes early autophagy events to support its pro-necroptotic function. Furthermore, Goodall et al. also stated that apparently the turnover of cellular components does not mediate this pro-death function and that they observed necrosome structures on quite mature autophagosomes [58,72]. Our observed blockade of autophagosome degradation might point toward the same direction. We observed increased ATG14-, ATG16L1-or LC3-positive structures, which potentially reflect increased necrosome-activating platforms. Collectively, our observations might indicate that RIPK3 "feeds" the autophagic flux via AMPK-dependent phosphorylation of ULK1 and BECN1 in order to ensure sufficient autophagosome maturation and accumulation. AMPK becomes activated under stress and starvation conditions, but it has also been shown that AMPK can support autophagy under nutrient sufficiency [29]. Accordingly, it is conceivable that AMPK also supports autophagy induction upon TQ/TSZ treatment. Admittedly, our second model is rather speculative and especially kinetic and spatial aspects need further validation. As mentioned above, necroptosis is rather increased in ulk1 ulk2 or prkaa1 prkaa2 DKO MEFs compared to wild-type control cells, supporting a generally anti-necroptotic function of these two kinases.
Here, we showed that RIPK3 is another AMPK-activating kinase. However, we do not believe that RIPK3-dependent phosphorylation is the only mechanism to activate AMPK during necroptosis. We have assessed the relevance of the AMP-binding site 3 in PRKAG1, since it has previously been described that the induction of necroptosis results in reduced cellular ATP levels [49,50]. We did not observe any difference in PRKAA1/2 coefficient using ImageJ software. Scale bar: 20 µm. (D) L929 cells were retrovirally transfected with cDNA encoding mCitrine-LC3B. Cells were left untreated (medium, M) or treated using 10 ng/ml TNF + 30 µM QVD (TQ) or EBSS with or without 20 nM bafilomycin A 1 (B) for indicated times. Cells were collected and mCitrine fluorescence intensity was measured by flow cytometry. The mean of fluorescence intensity for each sample was normalized to cells incubated in growth medium (M). Data represent mean + SD from two independent experiments. (A-D) Statistical analysis was done by repeated measures two-way ANOVA (corrected by Tukey's multiple comparisons test) for A, B and D, and by ordinary one-way ANOVA (corrected by Tukey's multiple comparisons test) for C. For C, statistical analysis was additionally performed using unpaired t test with Welch's correction (TQ treatment of non-targeting vs. Ripk3 siRNA). For A and B, statistically significant differences are only indicated for 6 h; for D, statistically significant differences are only indicated for 6 h vs. 6 h + bafilomycin A 1 . Statistically significant differences to control (medium, M) are depicted as letters directly above the bars. * or a: P < 0.05, ** or b: P < 0.01, *** or c: P < 0.001, **** or d: P < 0.0001.  Figure 6. Necroptosis induced by TNF destabilizes SNARE complexes and cleaves STX17 to block LC3 degradation. (A) L929 cells stably expressing GFP-SNAP29 were exposed to 10 ng/ml TNF and 30 µM QVD (TQ) for indicated times. Then, cells were lysed and cleared cellular lysates were subjected to immunopurification using anti-GFP beads. Purified proteins were subjected to SDS-PAGE and analyzed by immunoblotting for indicated proteins. (B) L929 cells stably expressing GFP-SNAP29 were left untreated (medium, M) or exposed to 10 ng/ml TNF + 30 µM QVD (TQ), TQ plus 5 µM GSK'872 (TQG), or TQ plus 5 µM necrostatin-1 (TQN) for 4 h. Then, cells were lysed and cleared cellular lysates were subjected to immunopurification using anti-GFP beads. Purified proteins were subjected to SDS-PAGE and analyzed by immunoblotting for indicated proteins. (C) L929 cells were left untreated (medium, M) or exposed to 30 µM QVD (Q), 20 nM bafilomycin A 1 (B), 10 ng/ml TNF (T), 10 ng/ml TNF + 30 µM QVD with or without 20 nM bafilomycin A1 (TQ or TQB) for indicated times. Then, cells were lysed and cleared cellular lysates were subjected to SDS-PAGE and immunoblotting for STX17 and ACTB. (D) L929 WT, ripk3 KO or mlkl KO cells were left untreated or exposed to 10 ng/ml TNF + 30 µM QVD (TQ) for 4 h. Then, cells were lysed and cleared cellular lysates were subjected to SDS-PAGE and immunoblotting for STX17, RIPK3, MLKL, and GAPDH. (E) L929 cells were transfected with non-targeting (siCtrl) or Prkaa1/Prkaa2 siRNAs (siPrkaa1/siPrkaa2). 48 h post transfection, cells were left untreated (medium, M) or exposed to 10 ng/ phosphorylation between cells expressing either wild-type PRKAG1 or the R299G variant. However, additional experiments are required to fully elucidate the mechanisms leading to AMPK activation during necroptosis. Nevertheless, we think that the RIPK3-dependent activation of AMPK represents another level of crosstalk between necroptosis and autophagy. The ultimate fate of a cell under stress conditions is determined by the integration of different cellular stress responses. Accordingly, a deeper understanding of the interplay between these stress responses is necessary in order to exploit these pathways for potential therapeutic approaches.

Immunofluorescence and proximity ligation assay (PLA)
Cells were seeded on glass coverslips overnight and exposed to indicated treatments at the next day. Then, cells were fixed using 4% paraformaldehyde for 10 min at room temperature and after washing three times with DPBS, cells were permeabilized with 50 µg/ml digitonin (Sigma-Aldrich, D141) for 5 min at room temperature. Cells were again washed with DPBS three times and incubated with 3% BSA (Roth, 8076) for 30 min at room temperature. The coverslips were transferred to a humidified chamber and incubated with indicated primary antibodies in 3% BSA for 1 h at room temperature. The cells were washed 3 times (3 min each time) with DPBS and incubated with indicated secondary antibodies in 3% BSA for 1 h at room temperature in the dark. The cells were washed 3 times (3 min each time) with DPBS and mounted on slide glass with ProLong® Gold Antifade Reagent with DAPI after rinsing the coverslips briefly in distilled water. After drying the coverslips, they were stored at 4°C. For proximity ligation assay, fluorescence signals were measured according to Duolink® PLA Fluorescence Protocol (Sigma-Aldrich, DUO82005 [antirabbit] and DUO82001 [anti-mouse]). Images were captured by Zeiss Axio Observer 7 (ApoTome extension, Objective Plan-Apochromat 40x/1,4 Oil DIC M27). For all immunofluorescence and PLA analyses, puncta quantification was done using ImageJ software.

In vitro kinase assay
For in vitro kinase assays, 1-2 μg substrate were incubated with 0.25-0.5 μg activated kinase in kinase buffer (

Protein dephosphorylation assay
Two duplicate samples were separated by the same SDS-PAGE and then transferred to the same PVDF membrane. Following blocking with 5% BSA in TBS-T (10 mM Tris-HCl, pH 7.6, 150 mM NaCl, 0.05% Tween-20 [Sigma-Aldrich, P1379]) for 1 h at room temperature, the PVDF membrane was washed with TBS-T twice (5 min each). The membrane was cut into two halves with duplicate samples on each membrane. The two membranes were placed into two tubes with or without 10 U alkaline phosphatase (2 ml reaction volume) and incubated for 1 h at 37°C with shaking (350 rpm). After washing twice with TBS-T (5 min each), the membranes were incubated with the corresponding primary antibodies at 4°C overnight.

Cell death assay
Total cells treated with indicated stimuli were trypsinized and collected. Cells were then incubated in propidium iodide (PI) solution (5 µg/ml in DPBS) at 4°C for 1 h. PI-positive cells were measured by flow cytometry (LSRFortessa TM , BD Biosciences).

Transmission electron microscopy (TEM)
L929 cells were seeded into 10 cm dishes and incubated in DMEM medium including 10% FCS overnight. After indicated treatments, cells were washed once using 1x DPBS. Then, cells were fixed using 3% glutaraldehyde in 0.1 M sodium cacodylate buffer, pH 7.2 for 2 h at room temperature (RT) in the hood. Then, cells were harvested with a cell scraper and centrifuged at approx. 4000 x g for 5 min. Pellets were washed twice with 0.1 M sodium cacodylate buffer, pH 7.2 (centrifugation at 4000 x g for 5 min). The pellets were heated to 40°C and embedded into 3% low melting agarose. Agarose was dissolved at 40°C in a water bath. The supernatant was aspirated and a volume of approx. 10 µl was left. The pellet was resuspended using the same volume of agarose. The mixture (agarose + pellet) was centrifuged at approx. 15,000-20,000 x g, 2 min. The samples were covered with 1% OsO 4 in sodium cacodylate buffer for 50 min, at RT. Then the samples were washed two times with sodium cacodylate buffer for 10 min at RT and once with 70% EtOH for 15 min at RT. Block contrast was applied using 1% uranyl acetate/1% phosphorotungstic acid in 70% EtOH (freshly made and filtered) for 1 h, at RT. The samples were dehydrated using graded ethanol series (90% EtOH, 96% EtOH, 100% EtOH) and embedded in SPURR epoxy resin (Serva, 21050). Polymerization was done at 70°C for 24 h. The 70-nm ultrathin sections were cut using an Ultracut EM UC7 (Leica, Wetzlar, Germany). TEM images were captured using an H7100 TEM (Hitachi, Tokyo, Japan) at 100 V equipped with Morada camera (EMSIS GmbH, Münster, Germany).

Statistical analysis
For PI uptake, shown data represent at least three independent experiments + SD. Absolute values are shown. For the quantification of immunoblots, the density of each protein band was divided by the average of the density of all bands from the same protein on the membrane. Subsequently, fold changes were calculated by dividing each normalized ratio (protein to loading control) by the average of the ratios of the control lane (as indicated in the corresponding figure legend). For Figure 1A, Figure 5A, Figure 5B, Figure 5D, S1A, S4A and S4B statistical analysis was performed using repeated measures two-way ANOVA (corrected by Tukey's multiple comparisons test). For Figure 1A, S1A and S4B, statistical analyses were additionally performed using repeated measures two-way ANOVA (corrected by Sidak's multiple comparisons test). For Figure 2C, Figure 4F, Figure 6C, S2B, S2C, S2D and S3C, statistical analyses were performed using ordinary two-way ANOVA (corrected by Tukey's multiple comparisons test). For Figure 2A, Figure 2B, Figure 2D, Figure 5C, S4C and S6D, statistical analysis was performed using ordinary one-way ANOVA (corrected by Tukey's multiple comparisons test). Additionally, for Figure  2B, Figure 3F, Figure 5C and S4C, statistical analysis was performed using an unpaired t test with Welch's correction. Compared treatments or cell lines are indicated in the corresponding bar diagrams and/or figure legends. P values < 0.05 were considered statistically significant. All statistical data were calculated with GraphPad Prism (version 7.01).