The stressosome, a caspase‐8‐activating signalling complex assembled in response to cell stress in an ATG5‐mediated manner

Abstract Stress‐induced apoptosis is mediated primarily through the intrinsic pathway that involves caspase‐9. We previously reported that in caspase‐9‐deficient cells, a protein complex containing ATG5 and Fas‐associated death domain (FADD) facilitated caspase‐8 activation and cell death in response to endoplasmic reticulum (ER) stress. Here, we investigated whether this complex could be activated by other forms of cell stress. We show that diverse stress stimuli, including etoposide, brefeldin A and paclitaxel, as well as heat stress and gamma‐irradiation, caused formation of a complex containing ATG5‐ATG12, FADD and caspase‐8 leading to activation of downstream caspases in caspase‐9‐deficient cells. We termed this complex the ‘stressosome’. However, in these cells, only ER stress and heat shock led to stressosome‐dependent cell death. Using in silico molecular modelling, we propose the structure of the stressosome complex, with FADD acting as an adaptor protein, interacting with pro‐caspase‐8 through their respective death effector domains (DEDs) and interacting with ATG5‐ATG12 through its death domain (DD). This suggests that the complex could be regulated by cellular FADD‐like interleukin‐1β‐converting enzyme–inhibitory protein (cFLIPL), which was confirmed experimentally. This study provides strong evidence for an alternative mechanism of caspase‐8 activation involving the stressosome complex.


| INTRODUC TI ON
Caspase-8 is a cysteine protease that was first identified as the apical caspase in the induction of extrinsic apoptosis. 1 Since then, a role for caspase-8 has emerged in diverse cellular processes in cancer biology and anticancer therapies, such as necroptosis, 2 autophagy-dependent cell death, 3 inflammation and cytokine release, 4 angiogenesis, 5 cell adhesion and migration 6 and immune cell homeostasis. 7 Thus, caspase-8 is important for tumour development, progression, therapy resistance and the development of the tumour microenvironment.
Pro-caspase-8 comprises a pro-domain at the N-terminus containing tandem death effector domains (DED1 and DED2), connected to a long domain and a short domain at the C-terminus. 8 The canonical mechanism of caspase-8 activation during initiation of extrinsic (or death receptor-mediated) apoptosis is through ligand-induced oligomerization of death receptors (DR) such as Fas or tumour necrosis factor-related apoptosis-inducing ligand receptors. 8 This leads to recruitment of Fas-associated death domain (FADD), which forms homotypic interactions with the DRs through their respective death domains (DDs) and with pro-caspase-8 through their respective DEDs. Together these form the death-inducing signalling complex where caspase-8 undergoes proximity-induced auto-activation. 8 Several non-canonical or alternative platforms for activation of caspase-8 have been identified, many of which demonstrate a requirement for FADD as an adaptor protein. 3,[9][10][11] Several studies describe an involvement of components of the autophagy pathway such as p62 3 or ATG5, 3,9,10 while some report a role for autophagosomal membranes. 3,12,13 Furthermore, unmitigated endoplasmic reticulum (ER) stress has been shown to activate caspase-8 by inducing ligand-independent intracellular DR5 activation, which is regulated by the integrated stress response (ISR) mediator, CHOP. 14 We reported that caspase-9 deficiency and blockade of the intrinsic apoptosis pathway can unmask an alternative caspase-8-dependent cell death pathway in response to ER stress. This caspase-8 activation was independent of external DR ligation, but required interaction with FADD and ATG5 to form an intracellular caspase-8-activating complex. 10 Here, we investigated whether ER stress is a unique inducer of the ATG5-ATG12:FADD:pro-caspase-8 complex formation in caspase-9-deficient cells or whether it can assemble as a result of other cellular stresses. We demonstrate that in Casp9 −/− cells, a range of stress stimuli can activate caspase-8 through the formation of a ATG5-ATG12:FADD:pro-caspase-8 complex, which we termed the stressosome. Using molecular modelling, we generated a model supporting the interaction of caspase-8 with FADD and interaction of the DD region of FADD with ATG5-ATG12. Caspase-8 activation required ATG5 but was independent of autophagy. Although activation of the ISR was a common feature of the various stress stimuli, it did not mediate pro-caspase-8 activation. We further demonstrated that stressosome-induced activation of caspase-8 was inhibited by cellular FLICE-inhibitory protein (cFLIP).
To induce cellular stress, MEFs were seeded at the required density 24 h prior to treatment. The culture medium was then replaced with fresh medium containing 0.3 μg/ml brefeldin A (Sigma-Aldrich, B6542), 50 μM etoposide (Sigma-Aldrich, E1383) or 1 μM paclitaxel (Sigma-Aldrich, T7402). Chloroquine (Sigma-Aldrich, C6628) was used at 20 μM. ISRIB (Sigma-Aldrich, SML0843) was added every 48 h at 200 nM. Spautin-1 (Sigma-Aldrich, SML0440) and rapamycin (LC Laboratories, Woburn, MA, USA, R-500) were added every 24 h at 10 μM and 400 nM, respectively. For γ-irradiation, cells were trypsinized and kept in suspension during γ-irradiation (33 Gγ) with a caesium-137 source (Mainance, Hampshire, UK). Cells were collected by centrifugation at 300 × g for 5 min and then seeded at the required density. For heat shock, cells were trypsinized and resuspended in DMEM into 50 ml tubes, which were immersed completely for 45 min in a preheated water bath with a circulating pump at 43.5 °C. Afterwards, they were pelleted at 300 × g for 10 min, resuspended in fresh DMEM and seeded at required densities.

| Immunoblotting
Cells were lysed in whole cell lysis buffer (100 mM Tris-HCl pH 6.8,

| shRNA knockdown
Lentivirus for pLKO empty vector, shRNA vector against mouse

| cFLIP overexpression
The MEFs were transfected with a pME18S-Flag vector containing a CDS for mouse cFLIP long isoform (kindly provided by Prof Kazuhiro Sakamaki, Kyoto University, Japan) using Lipofectamine 3000 transfection reagent (Thermo Fisher, #L3000008) at DNA-to-lipid ratio of 1:2. The media was removed 6 h after transfection. Transfected cells were allowed to recover for 24 h prior to any additional treatment. For heat shock treatment and γ-irradiation exposure, cells were treated, seeded, let to recover for 4 h and followed with transfection with Lipofectamine 3000 which was performed as described above.

| Molecular dynamics simulation
Desmond molecular dynamic (MD) simulation engine 19 was used to investigate how the FADD Death Domain behaved over the time in the ATG5-ATG12 complex. The OPLS3e force field was used. 20 Complex solvation was performed using TIP3P water models 21 positioned in a cubic box extending 10 Å from nearest protein atom. Na + and Cl − ions were added to balance the system charge and to obtain a final physiological concentration of 150 mM of NaCl. The temperature and the pressure were kept constant at 300 K and 1 atmosphere, respectively.
The simulations were run in triplicate for 100 ns under periodic boundary conditions.

| Residue scanning
Residue hot spots were identified using Schrodinger-BioLuminate residue scanning calculation. 22 Mutated residues inducing the largest change in protein binding affinity (ΔAffinity, in kcal/mol) indicated that they have a significant contribution to the interaction between FADD and ATG5-ATG12.

| Statistical analysis
Images are representative of at least three independent experiments. Graphs show an average of at least three biological repeats.
Error bars represent standard error of mean (SEM). Significance was determined using two-way ANOVA, with p value <0.05 being considered significant and annotated by *.

| Diverse inducers of cell stress activate caspase-3 in Casp9 −/− MEFs in a manner that is dependent on caspase-8 and ATG5
We previously reported that prolonged ER stress induced by thapsigargin (Tg) or tunicamycin (Tm) leads to processing of pro-caspase-8 and pro-caspase-3 in Casp9 −/− MEFs. 10 Here, we investigated whether ER stress (BFA), DNA damage (etoposide and γ-irradiation), microtubule stabilization (paclitaxel) and heat shock could similarly activate caspase-8 in the absence of caspase-9 ( Figure 1A). We observed that all treatments increased the levels of the cleaved form of caspase-8 and of caspase-3 ( Figure 1B) and induction of cell death, albeit with slower kinetics than for Casp9 +/+ MEFs ( Figure S1).
We have also previously reported that ER stress-induced caspase-3 processing in Casp9 −/− MEFs is dependent on caspase-8 and ATG5. 10 Here, shRNA-mediated silencing of caspase-8 in shRNA-mediated silencing of ATG5 produced a large reduction in the ATG5-ATG12-conjugated protein, the most abundant form of cellular ATG5 23 ( Figure 1D). In control cells (empty vector), stress stimuli led to an increase in the levels of ATG5-ATG12 and also increased the processing of pro-caspase-8 ( Figure 1D). By contrast, in shAtg5 MEFs, there was almost complete inhibition of procaspase-8 (and pro-caspase-3) processing, indicating a requirement for ATG5 in caspase-8 and caspase-3 activation under diverse stresses ( Figure 1D).

| Computational modelling indicates how the stressosome components interact
Protein docking studies were then used to identify the mode of interactions between stressosome components. We undertook a blind protein-protein docking approach using PatchDock and   Figure 5A). To assess a potential role for the ISR in activation of caspase-8, we used pharmacological and genetic approaches to inhibit ISR signalling. We found that inhibition of the ISR using ISRIB 26 reduced expression of ATF4 but did not affect activation of caspase-8 and −3 by BFA, etoposide or paclitaxel ( Figure 5B).

| Lack of dependence of caspase-8 activation on autophagy or the integrated stress response
Knockdown of Atf4 ( Figure 5C) or Ddit3 ( Figure 5D) in Casp9 −/− MEFs also did not affect processing of caspase-8 and caspase-3 induced by indicated stressors. Together our data suggest that the ISR does not contribute to the downstream activation of caspases.

| Effect of caspase-8 and ATG5 knockdown on cell death induction by various stressors
Due to the role of caspases in apoptosis, we next examined whether stressosome-dependent caspase-8 activation mediates the cell death observed upon exposure of Casp9 −/− MEFs to various stresses (Fig. S1). We knocked down caspase-8 or ATG5 to analyse cell death via PI uptake and long-term survival by looking at colony formation.
While knockdown of Casp8 significantly reduced cell death due to BFA, γ-irradiation, heat shock and a small extent paclitaxel, it did not affect cell death induced by etoposide ( Figure 6A). Atg5 knockdown cells displayed reduced sensitivity to cell death induced by BFA, etoposide and heat shock but similar sensitivity to cell death induced by paclitaxel and γ-irradiation as control counterparts ( Figure 6B).
We also observed that, upon depletion of caspase-8 or ATG5, more to γ-irradiation, but they did not form colonies due to the primary inhibitory effect of these treatments on proliferation ( Figure 6C).
Following treatment with etoposide, there was no long-term survival of cells.

| Stressosome-mediated caspase-8 activation is regulated by cFLIP L
The molecular model ( Figure 3A) indicated that cFLIP could prevent caspase-8 activation within the stressosome complex by interacting with the α1/α4 helices of caspase-8 DED2, thereby blocking filament formation or, alternatively, by direct binding to FADD DED instead of caspase-8. To test this, we overexpressed FLAG-tagged the long form of cFLIP (cFLIP L ) in Casp9 −/− MEFs ( Figure 7A) and showed that cFLIP L -overexpressing cells displayed reduced processing of procaspase-8 and pro-caspase-3 following treatment with BFA, etoposide, paclitaxel or exposure to γ-irradiation or heat shock compared to empty vector counterparts ( Figure 7B). cFLIP L was also able to reduce cell death in response to BFA, paclitaxel and heat shock but not in response to etoposide or γ-irradiation ( Figure 7C).

| DISCUSS ION
Here, we demonstrate that diverse cellular stresses can lead to the formation of an alternative caspase-8-activating platform, which we have termed the stressosome. Formation of this complex is unmasked when the intrinsic apoptosis pathway is blocked. driven by CHOP through the ISR signalling. 14 In our experiments, we did not observe a role for the ISR in pro-caspase-8 processing.
One intriguing observation from our data is that while different stress stimuli lead to stressosome formation and caspase activation, cell death induction was not universally a consequence of stressosome formation. One explanation may be that stressosomedependent caspase activation in certain scenarios could initiate other downstream signalling pathways. It has been reported previously F I G U R E 5 Inhibition of the ISR does not prevent caspase-8 activation. (A) Casp9 −/− MEFs were treated with brefeldin A (BFA), etoposide or paclitaxel for the indicated times. Whole cell lysates were immunoblotted for p-eIF2α, eIF2α, ATF4, CHOP and ACTIN. (b-d) Casp9 −/− MEFs were treated with 200 nM integrated stress response inhibitor (ISRIB) (B) or transfected with non-coding (NC) and Atf4 siRNA (C) or Ddit3 siRNA (D) followed by treatment with BFA, etoposide or paclitaxel for 24-72 h. Immunoblots show (B and C) ATF4, cleaved caspase-8, cleaved caspase-3 or (D) CHOP, cleaved caspase-8 and caspase-3. ACTIN was used as loading control F I G U R E 6 The role of the stressosome in cell death of Casp9 −/− MEFs is dependent on the type of stress stimuli. Casp9 −/− MEFs stably transduced with scrambled shRNA and (A, C) Casp8 shRNA or (B, C) Atg5 shRNA were treated with BFA, etoposide, paclitaxel for the indicated times or exposed to γ-irradiation or heat shock and allowed to recover for up to 72 h. (A) PI or (B) ToPro3 uptake was analysed at the indicated time points after treatment. (C) Following treatment for 72 h the culture medium was changed and cells were left to form colonies. Clonogenic survival assay was performed 10 days later that caspase-8 has additional, non-apoptotic functions such as regulation of tumour cell motility, regulation of the inflammasome and cleavage of inflammatory interleukin-1β. 32,33 Therefore, it is possible that formation of the stressosome and consequent activation of caspase-8 could lead to inflammatory signalling rather than cell death, depending on the nature of the stimulus.
Resistance to apoptosis is a hallmark of cancer. 34 Certain cancers express high levels of anti-apoptotic proteins such as anti-apoptotic Bcl-2 family members or cFLIP or conversely exhibit loss of caspase-9 and Apaf-1. [35][36][37][38] Such cells frequently exhibit reduced or delayed cell death when exposed to chemotherapeutic drugs. Based on our findings, we suggest that stimulation of stressosome formation could circumvent the chemoresistance and make those cells more sensitive to treatment. For example, our data suggest that targeting cFLIP L may be a strategy to sensitize cells to stressosome-mediated cell death. cFLIP L is frequently overexpressed in solid tumours and haematological cancers, and its high expression correlates with poor prognosis, 38 and several approaches are under development to prevent caspase-8 interaction with cFLIP L . 38,39 In conclusion, these findings highlight that cells resistant to intrinsic apoptosis can use alternative pathways to activate caspases. The stressosome complex, which consists of ATG5-ATG12:FADD:procaspase-8, is formed in response to diverse stresses. Formation of the complex is independent of the ISR, while its activity can be blocked by cFLIP L . The full implications of this to cellular outcomes in cancer remain to be elucidated as caspase-8 activation is involved in several cellular processes including cell death and inflammatory responses. However, having delineated the stressosome components and pathways regulating stressosome assembly may offer new targets to bypass chemoresistance in cancer.