Autophagosomal Membrane Serves as Platform for Intracellular Death-inducing Signaling Complex (iDISC)-mediated Caspase-8 Activation and Apoptosis

Background: It remains a matter of debate whether autophagy contributes to apoptosis.

Results: Atg5 and p62 are required for an intracellular death-inducing signaling complex (iDISC) formation on autophagosomal membranes for caspase-8 self-processing.

Conclusion: Autophagosome serves as a platform for the intracellular activation of caspase-8.

Significance: Induction of iDISC formation may shift cytoprotective autophagy to apoptosis for more effective cancer therapies.

Abstract

Autophagy and apoptosis are two evolutionarily conserved processes that regulate cell fate in response to cytotoxic stress. However, the functional relationship between these two processes remains far from clear. Here, we demonstrate an autophagy-dependent mechanism of caspase-8 activation and initiation of the apoptotic cascade in response to SKI-I, a pan-sphingosine kinase inhibitor, and bortezomib, a proteasome inhibitor. Autophagy is induced concomitantly with caspase-8 activation, which is responsible for initiation of the caspase cascade and the mitochondrial amplification loop that is required for full execution of apoptosis. Inhibition of autophagosome formation by depletion of Atg5 or Atg3 results in a marked suppression of caspase-8 activation and apoptosis. Although caspase-8 self-association depends on p62/SQSTM1, its self-processing requires the autophagosomal membrane. Caspase-8 forms a complex with Atg5 and colocalizes with LC3 and p62. Moreover, FADD, an adaptor protein for caspase-8 activation, associates with Atg5 on Atg16L- and LC3-positive autophagosomal membranes and loss of FADD suppresses cell death. Taken together, these results indicate that the autophagosomal membrane serves as a platform for an intracellular death-inducing signaling complex (iDISC) that recruits self-associated caspase-8 to initiate the caspase-8/-3 cascade.

Introduction

Programmed cell death (PCD)⁴ plays a key role in development and homeostasis (1). In addition, dysregulation of PCD is implicated in the pathogenesis of a number of diseases, including cancer, neurodegenerative disorders, and diabetes. Apoptosis is a well characterized mechanism of PCD that is defined by the activation of a family of cysteine proteases known as caspases. Caspase activation occurs through extrinsic and intrinsic signaling pathways (2, 3). Activation of the extrinsic pathway of apoptosis is initiated at the plasma membrane by the ligation of a death receptor belonging to the tumor necrosis factor receptor superfamily. Upon activation, the multimerization of death receptors stimulates the recruitment of Fas-associated death domain (FADD) and the initiator caspase-8 to form the death-inducing signaling complex (DISC). The formation of DISC results in the oligomerization of caspase-8 and thus facilitates its autoactivation through self-cleavage. In contrast, the intrinsic pathway of apoptosis is induced in response to intracellular stress signals, such as DNA damage or cytotoxic stress, and is characterized by mitochondrial outer-membrane permeabilization followed by the release of apoptogenic factors including cytochrome c. The initiator caspase-9 is activated upon association with cytochrome c and the apoptotic protease-activating factor 1 (Apaf-1) in a multiprotein complex known as the apoptosome. Both apoptotic signaling pathways converge upon the activation of effector caspases (caspase-3, -6, and -7), which cleave a number of cytosolic and nuclear substrates to execute the cell death pathway. Activated effector caspases also directly cleave caspase-8 to amplify the caspase cascade. Furthermore, the extrinsic pathway can initiate the mitochondrial pathway through the caspase-8-mediated cleavage of the BH3-only pro-apoptotic protein Bid.

Macroautophagy, hereafter referred as autophagy, is a catabolic process in which cytoplasmic components are sequestered within membrane-enclosed autophagosomes and delivered to lysosomes for degradation. The degradation through autophagy is generally considered to be a cytoprotective mechanism that maintains homeostasis under exposure to environmental stresses, such as nutrient deprivation or hypoxia (4, 5). Paradoxically, many studies have shown that the induction of autophagy can also contribute to caspase-dependent or -independent PCD (6–8). The formation of autophagosomes begins with the sequestration of cytoplasmic constituents into cup-shaped membrane structures, known as isolation membranes, which are expanded and eventually sealed to form double-membrane vesicles. Although the origin of autophagosomal membranes remains unclear, the elongation, expansion, and closure of autophagosomal membranes have been shown to require the Atg12-Atg5 and LC3-phosphatidylethanolamine (PE) ubiquitin-like conjugation systems (9). The ubiquitin-like conjugations are mediated by a common E1-like activating enzyme, Atg7, followed by E2-like conjugating enzymes, Atg10 for Atg12-Atg5 and Atg3 for LC3-PE. The Atg12-Atg5 conjugate further forms a complex with Atg16L, which acts as an E3-like enzyme to determine the site of LC3-PE conjugation (10). Moreover, loss of Atg3 has been shown to result in a marked reduction of the Atg12-Atg5 conjugates (11). Thus, the two conjugation systems work in concert to expand autophagosomal membranes.

The cross-talk between apoptosis and autophagy exists to regulate cell death (12, 13). Recent studies have shown that several molecules required for autophagy also play a key role in the regulation of apoptosis. For example, calpain-mediated cleavage of Atg5 generates a pro-apoptotic protein fragment that translocates to the mitochondria and interacts with the anti-apoptotic Bcl-2 family protein Bcl-xL to stimulate the mitochondrial pathway of apoptosis (14). In addition, Atg5 has been shown to directly interact with FADD to stimulate caspase-dependent cell death (15). Beclin 1, an essential autophagy-related protein that regulates the nucleation of autophagosomal membranes, is cleaved by caspases and translocates to the mitochondria to enhance apoptosis (16–18). However, the functional relationship between apoptosis and autophagy remains to be further explored. Here, we use the sphingosine kinase (SK) inhibitor SKI-I and the proteasome inhibitor bortezomib to demonstrate the cross-talk between apoptosis and autophagy. SKI-I is a non-lipid pan-SK inhibitor that inhibits both SK1 and SK2 to suppress the production of pro-mitogenic sphingosine 1-phosphate and promote cell death (19–21). We provide evidence that the autophagosomal membrane serves as a platform for the intracellular activation of caspase-8 to initiate caspase cascade and apoptotic cell death.

EXPERIMENTAL PROCEDURES

Reagents

SKI-I (N′-[(2-hydroxy-1-naphthyl)methylene]-3-(2-naphthyl)-1H-pyrazole-5-carbohydrazide) was synthesized as described (21). Antibodies were obtained from the following sources: rabbit polyclonal anti-LC3 (Novus Biologicals, NB100-2220 for immunoblot analyses; MBL International, PM046 for immunostaining), rabbit polyclonal anti-cleaved caspase-3 (Cell Signaling, 9661), rabbit polyclonal anti-poly(ADP-ribose) polymerase (PARP) (Cell Signaling, 9542), rabbit polyclonal anti-caspase-8 (Cell Signaling, 4927S; R & D Systems, AF1650), mouse monoclonal anti-caspase-8 (Cell Signaling, 9746), rabbit polyclonal anti-Atg16L (MBL International, PM040), mouse monoclonal anti-Atg5 (MBL International, M153-3), mouse monoclonal anti-Bcl-xL (Sigma, B9429), guinea pig polyclonal anti-p62 (American Research Products, Inc., 03-GP62-C), mouse monoclonal anti-FADD (Enzo Life Sciences, AAM-212), rabbit polyclonal anti-DsRed (Clontech, 632496), and mouse monoclonal anti-β-actin (Sigma, A5441).

Cell Culture

SV40 large T antigen immortalized Atg5^+/+ and Atg5^−/− mouse embryonic fibroblasts (MEF) cell lines were provided by Dr. Noboru Mizushima (Tokyo Medical and Dental University, Tokyo, Japan). SV40 large T antigen immortalized Atg3^+/+ and Atg3^−/− MEF cell lines were provided by Dr. Shengkan (Victor) Jin (University of Medicine and Dentistry of New Jersey-Robert Wood Johnson Medical School, NJ). FADD^+/+ and FADD^−/− MEF cell lines were established using the 3T3 protocol. KG-1 cells were obtained from ATCC (Manassas, VA). MEFs and KG-1 cells were cultured in Dulbecco's modified Eagle's medium and RPMI 1640 medium, respectively, supplemented with 10% fetal bovine serum (FBS), 100 μg/ml of streptomycin, 100 units/ml of penicillin, and 250 ng/ml of amphotericin B.

Virus Production and Transduction

The pLKO.1-based lentiviral shRNAs targeting ATG5 (TRCN0000150940) and p62/SQSTM1 (MEF, TRCN0000098616; KG-1, TRCN0000098618) were purchased from Open Biosystems (Huntsville, AL). The scrambled non-targeting shRNA was purchased from Addgene (number 1864). The mStrawberry-Atg4B(C74A) cDNA was purchased from Addgene (number 21076) and subcloned into pCDH1-MCS1-EF1-puro lentiviral vector (NheI and SwaI). Recombinant lentiviruses were produced using the ViraPower lentiviral expression system (Invitrogen) and transduced into targeted cells as described previously (22). The pK1-Bcl-xL-IRES-Puro and control pK1-IRES-Puro retroviral vectors were transfected to Amphotropic 293T cells and retroviruses encoding each gene were generated and transduced into targeting cells as described previously (23).

Cell Death Assays

Cell viability was measured using a luciferase-coupled ATP quantification assay according to the manufacturer's protocol (CellTiter-Glo viability assay, Promega G7570). Luminescence intensity was measured using a Victor X5 plate reader (PerkinElmer Life Sciences). To determine apoptosis, cells were stained with annexin V-PE (BD Bioscience, number 559763) and analyzed using a Guava EasyCyte Plus Flow Cytometry System (Millipore).

Caspase Activity Assays

To determine caspase activity, total cellular lysates were prepared in CHAPS lysis buffer (150 mm NaCl, 10 mm HEPES, pH 7.4, 1% CHAPS) containing protease inhibitors. Caspase-3 activity was determined by measuring the DEVDase activity as described previously (24). Caspase-8 activity was determined by measuring the IETDase activity using a caspase-8 assay kit (Sigma, number CASP8F) according to the manufacturer's protocol.

Bimolecular Fluorescence Complementation (BiFC) Assay

Mouse Atg5 (SalI and KpnI) and mouse FADD (EcoRI and XhoI) cDNAs were subcloned into the N- and C-terminal Venus fragment expression vectors, pBiFC-VN155(I152L) and pBiFC-VC155, respectively. Human caspase-8 (C360A) catalytic mutant (Addgene, 11818) (SalI and KpnI) was subcloned into pBiFC-VN155 (I152L) and pBiFC-VC155 vectors. MEFs were grown on gelatinized coverslips or NUNC Lab-Tek II chamber slides and transfected with BiFC pairs using FuGENE HD (Promega) or Lipofectamine LTX (Invitrogen) according to the manufacturer's protocol. For Atg5-VN and FADD-VC co-transfection, 50 μm Z-VAD-fmk was added 1 h post-transfection to prevent caspase-dependent cell death. The BiFC signals were detected using an OLYMPUS IX81 deconvolution microscope and analyzed using SlideBook 5.0 software (Intelligent Imaging Innovations).

Coimmunoprecipitation Assay

Total cell lysates prepared in DISC IP lysis buffer (1% Triton X-100, 150 mm NaCl, 10% glycerol, 30 mm Tris-HCl, pH 7.5) containing protease inhibitors were precleaned by incubating with anti-mouse IgG-conjugated agarose beads (Sigma, A0919) for 2 h at 4 °C and subjected to immunoprecipitation with anti-GFP monoclonal antibodies (Roche Diagnostics, number 11814460001). The resulting immunocomplexes were washed three times with the lysis buffer and subjected to immunoblot analyses.

RESULTS

SKI-I Induces Caspase-dependent Cell Death Accompanied with Induction of Autophagy

Although the antitumor effects of SKI-I have been largely attributed to the induction of apoptosis, the precise mechanism by which SKI-I induces apoptosis is unknown. To investigate the detailed mechanism of SKI-I-induced cell death, SV40 large T antigen-immortalized MEFs were treated with SKI-I. MEFs were selected based on the availability of genetic models that permit the study of apoptosis and autophagy. As shown in Fig. 1A, significant cleavage of caspase-3 was detected 24 h after SKI-I treatment, supporting the previous finding that SKI-I triggers the induction of apoptosis (19). Interestingly, we found that exposure of MEFs to SKI-I resulted in the accumulation of the autophagy marker LC3-II in a time-dependent manner. Moreover, SKI-I treatment increased the number of GFP-LC3 puncta (Fig. 1B), a well characterized marker used to visualize autophagosomes (25). As a portion of LC3-II is degraded by autophagosome fusion with lysosomes (25), we next examined whether the accumulation of LC3-II observed during SKI-I treatment was due to an increase in autophagy induction or a decrease in autophagic degradation by treating cells in the presence or absence of lysosomal inhibitors including bafilomycin A1, chloroquine (CQ), and ammonium chloride (NH₄Cl). We found that inhibition of lysosomal degradation resulted in a further accumulation of LC3-II in response to SKI-I (Fig. 1C). This clearly indicates that SKI-I promotes the induction of autophagy. Furthermore, the addition of lysosomal inhibitors significantly enhanced SKI-I-induced caspase-3 processing (Fig. 1C). Notably, cleaved caspase-3 does not colocalize with the lysosomal marker Lamp1 during SKI-I treatment (Fig. 1, D and E), thereby suggesting that lysosomal degradation of autophagosomes rather than active caspase-3 serves as a cell survival mechanism during SKI-I treatment. In contrast, SKI-I-induced cell death was significantly suppressed by co-treatment with a pan-caspase inhibitor, Z-VAD-fmk (Fig. 1F). Taken together, these results indicate that SKI-I-induced cell death occurs primarily through the caspase-dependent apoptotic pathway.

FIGURE 1.

SKI-I simultaneously induces autophagy and caspase-dependent cell death. A, MEFs were treated with 2.5 μm SKI-I for the indicated periods of time and subjected to immunoblot analyses using the indicated antibodies. B, MEFs stably expressing GFP-LC3 were treated with 2.5 μm SKI-I or control DMSO for 12 h and the number of GFP-LC3 dots per cell area (1000 μm²) was determined using a fluorescence microscope (mean ± S.D.; n = 36). Statistical significance was determined by Student's t test. C, MEFs were treated with 2.5 μm SKI-I or control DMSO for 8 h followed by a 4-h co-treatment with 100 nm bafilomycin A1, 25 μm CQ, 20 mm NH₄Cl, or control PBS and subjected to immunoblot analyses using the indicated antibodies. D, MEFs were treated with 2.5 μm SKI-I for 12 h, stained with anti-Lamp1 monoclonal and anti-active caspase-3 (C-Casp-3) polyclonal antibodies, and analyzed by fluorescence deconvolution microscopy. Magnified images are shown as insets. E, the fluorescence intensities along the dotted line in D were quantified using SlideBook software. The values of the vertical axis represent fluorescence intensity units (ADU). The horizontal axis represents distance (S, start point; E, end point). F, MEFs were treated with control DMSO or 2.5 μm SKI-I in the presence of 20 μm Z-VAD-fmk or control DMSO for 24 h and cell viability was assessed by measuring cellular ATP levels (mean ± S.D.; n = 3). The scale bars represent 10 μm in A and D, and 1 μm in the insets in D.

The LC3 Conjugation System Is Required for SKI-I-induced Apoptosis

Although autophagic degradation functions as a pro-survival mechanism to maintain cellular homeostasis through nutrient recycling and the removal of aggregated proteins and malfunctioning organelles (4, 5), autophagy has also been shown to contribute to caspase-dependent and/or -independent cell death (6–8). Thus, we next examined whether the formation of autophagosomes is involved in cell death induced by SKI-I. To this end, Atg5^+/+ and Atg5^−/− MEFs were treated with SKI-I or control DMSO. Atg5 is a member of the autophagy-related (Atg) protein family that covalently binds to Atg12, an ubiquitin-like protein, and is required for the formation of autophagosomes (26, 27). Consistently, we found that SKI-I-induced LC3-II conversion was completely abrogated by loss of Atg5 (Fig. 2A). Notably, SKI-I-induced cleavage of caspase-3 and PARP, a substrate for caspases, was also suppressed by depletion of Atg5. Furthermore, a significant reduction of apoptotic cell death was observed in SKI-I-treated Atg5^−/− MEFs as compared with their control wild-type cells (Fig. 2, B, E, and F).

FIGURE 2.

Inhibition of autophagosome formation by depletion of Atg5 or Atg3 suppresses SKI-I-induced apoptosis. MEFs with the indicated genotypes were treated with 2.5 μm SKI-I or control DMSO for 24 h. A and C, total cell lysates were subjected to immunoblot analyses using the indicated antibodies. B and D, representative micrographs are shown. E, cell viability was determined by ATP assays (mean ± S.D.; n = 3). Data are shown as percentage of control DMSO treatment. F, the induction of apoptosis was determined by annexin V staining. Drug-specific apoptosis was calculated by subtracting the percentage of annexin V-positive cells in DMSO-treated samples (mean ± S.D.; n = 3).

In addition to its role in autophagosome formation, Atg5 has been shown to promote mitochondrial apoptosis by interacting with Bcl-xL (14). To determine whether the SKI-I-resistant phenotype we observed in Atg5-deficient cells is specifically due to the impairment of autophagosome formation, we next treated Atg3^+/+ and Atg3^−/− MEFs with SKI-I. Atg3 is an E2-like enzyme for the conjugation of LC3-I and PE to form LC3-II, which is indispensable for elongation and closure of autophagosomal membranes (11). Similar to the results obtained using Atg5-deficient cells, loss of Atg3 resulted in a marked inhibition of the cleavages of caspase-3 and PARP and apoptotic cell death in response to SKI-I (Fig. 2, C–F). Collectively, these results indicate that the expansion of autophagosomal membranes is critical for the activation of caspase-3 and the subsequent cell death pathway upon SKI-I treatment. Importantly, SKI-I-induced apoptosis was also inhibited by loss of the pro-apoptotic Bcl-2 family genes, Bax and Bak, suggesting that the mitochondrial pathway is important for mediating SKI-I-induced apoptosis (Fig. 2, E and F).

Activation of Caspase-8 Is Critical for Initiation of Atg5-dependent Caspase Cascade during SKI-I Treatment

Our results shown above clearly indicate that inhibition of the early stages (i.e. autophagosome formation) but not the late stages of autophagy (i.e. lysosomal degradation) attenuates SKI-I-induced apoptosis. It has recently been reported that autophagy induction plays a role in regulating caspase-8 activation and subsequent cell death, although the precise role of autophagy in this process is not well defined (6, 15, 28, 29). We therefore next investigated whether activation of caspase-8 is involved in SKI-I-induced cell death. We found that SKI-I treatment resulted in cleavages of caspase-8 and -3 in wild-type MEFs (Fig. 3A). SKI-I-induced procaspase-8 (pro-Casp-8) processing was inhibited by the addition of Z-VAD-fmk, indicating that this process is mediated by caspases. Moreover, SKI-I-induced caspase-3 activation was blocked not only by Z-VAD-fmk but also by Z-IETD-fmk, a caspase-8-specific inhibitor (Fig. 3B). Importantly, loss of Atg5 drastically suppressed SKI-I-induced caspase-8 cleavage (Fig. 3A). Taken together, these results suggest that caspase-8 activation is a critical step in Atg5-dependent apoptosis induced by SKI-I.

FIGURE 3.

Loss of Atg5 suppresses SKI-I-induced activation of mitochondrial amplification loop that is initiated by caspase-8. A, Atg5^+/+ and Atg5^−/− MEFs were treated with 1.5 μm SKI-I or control DMSO in the presence of 20 μm Z-VAD-fmk or control DMSO for 24 h and subjected to immunoblot analyses using the indicated antibodies. The cleaved caspase-8 was quantified as the percentage of total caspase-8. B, Atg5^+/+ MEFs were treated with 1.5 μm SKI-I or control DMSO in the presence of 20 μm Z-IETD-fmk, 20 μm Z-VAD-fmk, or control DMSO for 24 h and the caspase-3/7-like DEVDase activity was measured (mean ± S.D.; n = 3). C–E, Atg5^+/+ and Atg5^−/− MEFs were infected with retroviruses encoding Bcl-xL or empty control. After selection with 1 μg/ml of puromycin, the resultant stable transfectants were subjected to treatment with 1.5 μm SKI-I or control DMSO for 24 h. C, total cell lysates were subjected to immunoblot analyses using the indicated antibodies. Asterisks indicate nonspecific bands. D, the caspase-3/7-like DEVDase activity and E, the caspase-8-like IETDase activity were measured (mean ± S.D.; n = 3).

Because SKI-I-induced cell death was also suppressed by loss of Bax and Bak (Fig. 2E), we next investigated whether the mitochondrial pathway is required for the initiation or amplification of the caspase cascade triggered by SKI-I. To this end, Atg5^+/+ and Atg5^−/− MEFs were infected with recombinant retroviruses encoding anti-apoptotic Bcl-xL or empty vector. After selection with puromycin, the expression of Bcl-xL was confirmed by Western blot analyses (Fig. 3D). SKI-I-induced cleavage and activation of caspase-3 were greatly suppressed by overexpression of Bcl-xL in both Atg5^+/+ and Atg5^−/− cells (Fig. 3, C and D). Moreover, Bcl-xL overexpression also significantly diminished SKI-I-induced caspase-8 cleavage and activation in wild-type cells (Fig. 3, C and E), indicating that the mitochondrial amplification loop is required for the full activation of caspase-8 upon SKI-I treatment. Notably, little effect of Bcl-xL overexpression was observed on the activation of caspase-8 in Atg5-deficient cells, indicating that Atg5 acts upstream of the mitochondrial pathway to initiate the caspase cascade in response to SKI-I (Fig. 3E).

SKI-I Promotes Self-association of Caspase-8 in a Manner Independent of Atg5

Oligomerization is a key step for caspase-8 processing and activation (30). To determine whether Atg5-containing autophagosomal membranes are required for the recruitment and/or oligomerization of caspase-8, we performed BiFC assays. BiFC is based on the formation of a fluorescent complex by two non-fluorescent fragments of Venus, VN155(I152L) (VN) and VC155 (VC), brought together by association of proteins fused with each Venus fragment (31). To this end, pro-Casp-8 (C360A) was fused with each Venus fragment and transfected to Atg5-deficient and control wild-type cells. We utilized a proteolytically inactive C360A mutant of pro-Casp-8 to prevent apoptosis induction due to overexpression of caspase-8 (32). SKI-I induced numerous fluorescent puncta as well as large aggregates throughout the cytoplasm of wild-type cells, whereas little signal was detected before the treatment (Fig. 4A). This result indicates that SKI-I promotes self-association and oligomerization of pro-Casp-8. Importantly, such fluorescent signals were not induced upon SKI-I treatment in cells transfected with control empty VN and VC (data not shown). To our surprise, loss of Atg5 did not suppress SKI-I-induced caspase-8 self-association. Taken together, these results indicate that, whereas SKI-I-induced activation of caspase-8 requires Atg5, the autophagosomal membrane is dispensable for caspase-8 self-association.

FIGURE 4.

SKI-I promotes caspase-8 self-association in a p62-dependent but Atg5-independent manner. A, Atg5^+/+ and Atg5^−/− MEFs were transfected with pro-Casp-8 (C360A)-VN and pro-Casp-8 (C360A)-VC for 24 h, treated with 2.5 μm SKI-I for 0 and 6 h, and analyzed by fluorescence microscopy (top) in combination with differential interference contrast microscopy (bottom). B, Atg5^+/+ and Atg5^−/− MEFs were infected with lentiviruses encoding shScr or shp62 and selected with 1 μg/ml of puromycin. Cells were transfected with pro-Casp-8 (C360A)-VN and pro-Casp-8 (C360A)-VC for 24 h, treated with 2.5 μm SKI-I for 6 h, stained with guinea pig anti-p62 polyclonal antibodies, and analyzed by fluorescence deconvolution microscopy. Arrows indicate colocalization of caspase-8 complexes with p62. The scale bars represent 20 μm in A and 10 μm in B.

p62-mediated Self-association of Caspase-8 Plays a Regulatory Role in Atg5-dependent Caspase-8 Activation and Apoptosis Induced by SKI-I

p62 is an adapter protein that recruits polyubiquitinated and aggregated proteins to autophagosomes through its direct interaction with LC3-II (33, 34). Recent studies have shown that p62 interacts with polyubiquitinated caspase-8 and promotes oligomerization and activation of caspase-8 (29, 35). To determine whether p62 is involved in SKI-I-induced caspase-8 oligomerization and subsequent activation of the caspase cascade, Atg5^+/+ and Atg5^−/− MEFs stably expressing p62 shRNA (shp62) or control scrambled shRNA (shScr) were transfected with pro-Casp-8 (C360A)-VN and pro-Casp-8 (C360A)-VC, treated with SKI-I, and stained with anti-p62 antibodies. As shown in Fig. 4B, a portion of SKI-I-induced caspase-8 homocomplexes were co-localized with p62 in both wild-type and Atg5-deficient cells. These SKI-I-induced BiFC signals were greatly diminished by knockdown of p62, indicating that p62 plays a key role in mediating caspase-8 self-association during SKI-I treatment. Moreover, by co-transfecting the pro-Casp-8 (C360A) BiFC constructs with mRFP-LC3, we found that SKI-I-induced caspase-8 homocomplexes were also positive for the autophagosomal membrane marker LC3 (Fig. 5A). Furthermore, a portion of endogenous caspase-8 colocalized with p62 and LC3 upon SKI-I treatment (Fig. 5, B and C). Taken together, these results suggest that SKI-I promotes p62-dependent self-association of pro-Casp-8 and its recruitment to autophagosomal membranes.

FIGURE 5.

Caspase-8 homocomplex is recruited to the autophagosomal membrane during SKI-I treatment. A, Atg5^+/+ MEFs stably expressing shScr or shp62 were transfected with pro-Casp-8 (C360A)-VN, pro-Casp-8 (C360A)-VC, and mRFP-LC3. Twenty hours after transfection, the cells were treated with 2.5 μm SKI-I for 6 h and subjected to fluorescence deconvolution microscopic analyses. Nuclei were stained with DAPI. B, Atg5^+/+ MEFs were infected with retroviruses encoding GFP-LC3 and selected with 1 μg/ml of puromycin for 5 days. The resultant stable transfectants were treated with 2.5 μm SKI-I or control DMSO for 12 h, immunostained with anti-p62 and anti-caspase-8 antibodies, and analyzed by fluorescence deconvolution microcopy. C, the intensity profiles for each fluorescence along the dotted line in B are shown. Arrows indicate colocalization of caspase-8 complexes with mRFP-LC3 (A) or caspase-8 with p62 and GFP-LC3 (B). The scale bars represent 10 μm in A and B, and 1 μm in the inset in B.

We next examined the effect of p62 knockdown on activation of the caspase cascade during exposure to SKI-I. As expected, knockdown of p62 partially suppressed apoptotic cell death in wild-type cells treated with SKI-I (Figs. 6, A and E). Consistently, SKI-I-induced cleavage and activation of caspase-8 and -3 were significantly reduced by knockdown of p62 in wild-type cells (Fig. 6, B–D). Notably, despite nearly complete suppression of p62 expression (Fig. 6B), the effect of p62 knockdown on SKI-I-induced caspase-8 and caspase-3 activation was not as dramatic as that seen in Atg5 knock-out cells (Fig. 6, C and D). These results suggest that a portion of caspase-8 is activated through an additional Atg5-dependent mechanism. Interestingly, knockdown of p62 in Atg5-deficient cells failed to further reduce, but rather slightly enhanced, SKI-I-induced cell death and caspase-3 activation (Fig. 6, A–C). In addition to mediating caspase-8 self-association, p62 promotes the oligomerization of TRAF6 to enhance the activation of NF-κB (36). Furthermore, p62 positively regulates the transcription factor Nrf2, which is responsible for transcription of cytoprotective genes (37). Therefore, knockdown of p62 may also suppress cell survival pathways to enhance apoptosis in response to SKI-I.

FIGURE 6.

Knockdown of p62 suppresses SKI-I-induced caspase-3 and caspase-8 activation. Atg5^+/+ and Atg5^−/− MEFs stably expressing shScr or shp62 were treated with 2.5 μm SKI-I or control DMSO for 24 h. A, representative micrographs of SKI-I-treated cells are shown. B, total cell lysates were subjected to immunoblot analyses using the indicated antibodies. C, the caspase-3/7-like DEVDase activity and D, the caspase-8-like IETDase activity were measured (mean ± S.D.; n = 3). E, the cells were stained with annexin V, and drug-specific apoptosis was calculated by subtracting the percentage of annexin V-positive cells in DMSO-treated samples (mean ± S.D.; n = 3).

SKI-I Induces Translocation of Caspase-8 and FADD to Atg5-positive Autophagosomal Membranes

It has recently been suggested that the Atg12-Atg5 complex interacts with the death adaptor protein FADD to recruit caspase-8 (15). To determine whether association of FADD and the Atg12-Atg5 complex on the autophagosomal membrane serves as an additional mechanism for caspase-8 recruitment and activation in response to SKI-I, we first analyzed the spatial association of Atg5 and FADD by BiFC. We found that strong punctate signals of Venus were accumulated in the cytoplasm of Atg5^−/− MEFs transfected with Atg5-VN and FADD-VC, whereas diffuse cytoplasmic and nuclear signals were detected in cells expressing control empty VN and VC (Fig. 7A). The BiFC signals from Atg5-VN and FADD-VC were positive not only for LC3 but also for an isolation membrane marker, Atg16L. Interestingly, the majority of cells transfected with Atg5-VN and FADD-VC but not control vectors eventually underwent apoptotic cell shrinkage and death (data not shown). As the fluorescent complex formation in BiFC is essentially irreversible (31), formation of the stabilized Atg5·FADD complex may result in sustained recruitment and activation of caspase-8 leading to cell death. The interaction of Atg5, FADD, and procaspase-8 was confirmed by coimmunoprecipitation analyses (Fig. 7B). Notably, whereas the association of Atg5 with procaspase-8 was enhanced by SKI-I treatment, it decreased upon longer incubation with SKI-I. This suggests that activated/cleaved caspase-8 was released into the cytosol in a manner similar to observed upon TNF-α treatment (38). Indeed, whereas a large portion of FADD signals were found to be colocalized with Atg5 during SKI-I treatment, only a small fraction of caspase-8 signals were detected on Atg5 and FADD-positive structures (Fig. 7, C and D). Moreover, loss of FADD was found to suppress SKI-I-induced cell death (Fig. 7E). To further examine the role of Atg5-containing autophagosomal membranes in SKI-I-induced apoptosis, we generated MEFs stably overexpressing a dominant-negative mutant of Atg4B (mStrawberry-Atg4B (C74A)). The protease Atg4B is one of four mammalian Atg4 homologues required for processing pro-LC3 paralogues prior to lipidation, a key step in autophagosome biogenesis. In addition, Atg4 catalyzes the delipidation and release of LC3 from autophagosomal membranes during, or before, autophagosome-lysosome fusion. Previous studies have shown that overexpression of the inactive mutant Atg4B (C74A) blocks LC3 modification and autophagosome closure to result in the accumulation of unsealed, Atg5-positive membrane structures (39). Consistently, SKI-I-induced LC3 modification was completely suppressed in cells overexpressing dominant-negative Atg4B (Fig. 7F). Interestingly, we found that treatment with SKI-I significantly stabilized dominant-negative Atg4B protein expression through an unknown mechanism. Importantly, despite the absence of LC3-II, overexpression of dominant-negative Atg4B substantially enhanced SKI-I-induced caspase-8 and caspase-3 cleavage compared with wild-type cells (Fig. 7F), supporting the notion that the autophagosomal membrane serves as a platform for caspase-8 activation and induction of apoptosis. Taken together, these results suggest that the recruitment and activation of caspase-8 in response to SKI-I occurs not only in a p62·LC3-II-dependent manner but also through the association of Atg5·FADD at the autophagosomal membrane.

FIGURE 7.

SKI-I induces translocation of caspase-8 and FADD to Atg5-positive autophagosomal membranes. A, Atg5^−/− MEFs were transfected in combination with Atg5-VN and FADD-VC or control empty VN and VC for 18 h. Cells were fixed, immunostained with anti-LC3 and anti-Atg16L antibodies, and analyzed by fluorescence microcopy. B, Atg5^−/− MEFs were infected with retroviruses encoding GFP-Atg5 and selected with 1 μg/ml of puromycin for 5 days. The resultant stable transfectants were treated with 2.5 μm SKI-I for the indicated times and subjected to immunoprecipitation with anti-GFP monoclonal antibodies or control mouse IgG followed by immunoblot analyses using the indicated antibodies. C, Atg5^−/− MEFs stably expressing GFP-Atg5 were treated with 2.5 μm SKI-I or control DMSO for 12 h, immunostained with anti-FADD and anti-caspase-8 antibodies, and analyzed by fluorescence deconvolution microcopy. Arrowheads and arrows indicate colocalization of Atg5 with FADD, and Atg5 and FADD-positive signals with caspase-8, respectively. D, the intensity profiles for each fluorescence along the dotted line in C are shown. The scale bars represent 10 μm in A and C, and 5 μm in the inset in C. E, FADD^+/+ and FADD^−/− MEFs were treated with 5 μm SKI-I or control DMSO for 24 h. Representative micrographs are shown. F, Atg5^+/+ MEFs were infected with lentiviruses encoding mStr-Atg4B(C74A) or empty control. After selection with 1 μg/ml of puromycin, the resultant stable transfectants were subjected to treatment with 2.5 μm SKI-I or control DMSO for 16 h. Total cell lysates were subjected to immunoblot analyses using the indicated antibodies. Asterisks indicate nonspecific bands. Cleaved caspase-8 was quantified as the percentage of total caspase-8. Cleaved caspase-3 was quantified as the relative expression after normalization to β-actin.

SKI-I Induces Atg5-dependent Apoptosis in KG-1 Cells Enhanced by Stabilization of p62 through Proteasome Inhibition

The apoptotic and antitumor effects of SKI-I have previously been demonstrated in several human cancer cell lines as well as a murine model of mammary adenocarcinoma (19, 40). We sought to investigate whether SKI-I induces an autophagy-dependent activation of the apoptotic cascade in human acute myeloid leukemia (AML), a model in which sustained SK activity has been associated with chemoresistance (41). To this end, we treated several human AML cell lines including KG-1, HL-60, and multidrug-resistant HL-60/VCR with SKI-I. We found that SKI-I concomitantly induces autophagy and apoptosis in all cell lines tested (data not shown). To determine whether Atg5 and p62 are responsible for the induction of apoptosis triggered by SKI-I, KG-1 cells stably expressing ATG5 shRNA (shATG5), shp62, or control shScr were generated using a lentiviral transduction system and treated with SKI-I or control DMSO. We chose KG-1 cells for further studies as, in addition to the capability to induce the most prominent autophagy, the most effective knockdown of the targeted genes was achieved in this cell line among all AML cell lines we tested (data not shown). Consistent with the results obtained using MEFs, knockdown of either ATG5 or p62 suppressed SKI-I-induced caspase-8, caspase-3, and PARP cleavages and the induction of apoptosis in KG-1 cells (Figs. 8, A and D). Moreover, co-treatment with CQ enhanced SKI-I-induced apoptosis in KG-1 cells (Fig. 8B), supporting the notion that lysosomal degradation of autophagosomes serves as a cell survival mechanism upon SKI-I treatment. Notably, extensive degradation of p62 was observed in KG-1 cells upon SKI-I treatment (Fig. 8A). Degradation of p62 occurs by autophagy and the proteasome system (42). As knockdown of ATG5 did not prevent, but rather slightly enhanced, p62 degradation in response to SKI-I, SKI-I-induced p62 degradation was suggested to occur in a proteasome-dependent manner (Fig. 8A). The enhanced degradation of p62 observed in KG-1 cells upon SKI-I treatment compared with that of MEFs correlates with a significant reduction of SKI-I-induced apoptosis in this cell line (14 versus 29%; Figs. 2F and 8B). Therefore, the rapid turnover of p62 may affect the efficacy of SKI-I in KG-1 cells.

FIGURE 8.

Inhibition of proteasome-mediated p62 degradation enhances the efficacy of SKI-I to induce Atg5-dependent apoptosis in AML cells. KG-1 cells were infected with lentiviruses encoding shScr, shATG5, or shp62 and selected with 1 μg/ml of puromycin. KG-1 cells were infected with lentiviruses encoding shScr, shATG5, or shp62 and selected with 1 μg/ml of puromycin. A and C, the resultant stable transfectants were treated with or without 2.5 μm SKI-I in the presence or absence of 3 nm bortezomib (Bort) for 24 h and subjected to immunoblot analyses using the indicated antibodies. B and D, parental KG-1 cells (B) or the stable transfectants (D) were treated with 2.5 μm SKI-I or control DMSO in the presence or absence of 25 μm CQ for 24 h (B) or 3 nm bortezomib for 48 h (D), stained with annexin V and 7-amino-actinomycin D (7-AAD), and analyzed flow cytometry. The percentage of annexin V-positive cells is shown. Data shown are representative of two independent experiments. E, the stable transfectants were treated with 5 nm bortezomib for 24 h and subjected to immunoblot analyses.

To determine whether stabilization of p62 could enhance SKI-I-induced apoptosis in KG-1 cells, we next treated cells in combination with SKI-I and the proteasome inhibitor bortezomib. As expected, treatment with bortezomib resulted in the accumulation of p62, indicating that p62 degradation during SKI-I treatment occurs mainly through the proteasome system rather than autophagy (Fig. 8C). Accordingly, a combination treatment with SKI-I and bortezomib increased the cleavages of caspase-8, caspase-3, and PARP and thus greatly enhanced apoptosis in control shScr-expressing KG-1 cells (Fig. 8, C and D). Despite the accumulation of p62 upon combination treatment, this effect was blocked by knockdown of ATG5 (Fig. 8, C and D). Taken together, these results indicate that SKI-I induces Atg5-dependent apoptosis in AML cells and that this process is further enhanced by proteasome inhibition through the stabilization of p62. Notably, similarly to SKI-I, bortezomib-induced caspase-8 and -3 activation was also suppressed by knockdown of either p62 or ATG5 (Fig. 8, C and E), a result consistent with previous reports (6, 29).

DISCUSSION

The induction of autophagy is generally considered an adaptive and a cytoprotective mechanism for the recycling of nutrients and the removal of cytotoxic materials (4, 5). However, mounting evidence has suggested that autophagy is also implicated in the induction of caspase-dependent and -independent cell death through pathways that need to be explored further (5, 13). In the present study, we have demonstrated a mechanism of apoptosis that is dependent on the autophagosomal membrane. We found that activation of caspase-8 is a critical step in the autophagy-dependent induction of apoptosis in response to SKI-I, a pan-SK inhibitor, and bortezomib, a proteasome inhibitor. Furthermore, we have shown that the expansion of autophagosomal membranes is essential for the intracellular activation of caspase-8. Our data demonstrates that SKI-I-induced self-association of caspase-8 is dependent on p62. Consistently, p62 has recently been shown to mediate caspase-8 oligomerization and activation in response to TRAIL or proteasome inhibition (29, 35). Importantly, we found that SKI-I-induced self-association of caspase-8 occurs independently of Atg5; however, Atg5 is required for the activation of caspase-8. These observations therefore suggest that self-associated caspase-8 is recruited to the autophagosomal membrane to form the proper higher order oligomer structure for activation. To support this concept, forced membrane localization and self-association of caspase-8 has been shown to dramatically promote apoptosis (43). Furthermore, it has recently been suggested that caspase-8 is recruited to the Atg12-Atg5 complex through FADD and that autophagic machinery is required for activation of caspase-8 (6, 15, 28). In this study, we show that the caspase-8·FADD complex associates with Atg5 on Atg16- and LC3-positive structures, indicating that the autophagosomal membrane serves as a platform for this interaction. Collectively, these results prompt us to propose a model in which the autophagosomal membrane serves as a platform for the formation of a dual-armed iDISC that facilitates the activation of caspase-8 and initiation of apoptosis. Elongation of autophagosomal membranes occurs in a manner that is dependent on the Atg12-Atg5 and LC3-PE ubiquitin-like conjugation systems. In our model, FADD is recruited to expanding autophagosomal membranes through interactions with Atg5. In a manner analogous to DISC formation upon death receptor ligation (45), FADD recruits caspase-8 to the autophagosomal membrane to promote self-activation of caspase-8. Additionally, caspase-8 self-association and recruitment to the autophagosomal membrane occurs in a p62-dependent manner. Here, the association of LC3-II and p62 mediates the recruitment of self-associated caspase-8 to the autophagosomal membranes for the formation of proper oligomer structures to facilitate caspase-8 self-activation. Furthermore, the mitochondrial amplification loop is indispensable for the full activation of the iDISC-mediated cell death pathway.

Our data clearly indicates that autophagy induction plays a key role in SKI-I-induced apoptosis. However, despite impaired autophagic activity, a slight induction of apoptosis is observed in Atg5- and Atg3-deficient MEFs. Inhibition of SK blocks the formation of pro-mitogenic sphingosine 1-phosphate (S1P) to accumulate pro-apoptotic precursors, sphingosine and ceramide. Ceramide accumulation has been shown to induce the intrinsic pathway of apoptosis through several mechanisms including direct permeabilization of the mitochondrial membrane, enhanced activation of Bax, and direct activation of the lysosomal protease cathepsin D (46). As a result, the accumulation of pro-apoptotic lipids, sphingosine and ceramide, is likely responsible for activation of the intrinsic pathway in autophagy deficient cells. Importantly, we also found that SKI-I promotes the degradation of p62 through proteasome in AML cells. Consistently, activation of the proteasome machinery in response to SK inhibition has been reported (47). As the addition of bortezomib stabilized p62 and enhanced the autophagy-dependent activation of caspase-8 and apoptosis upon SKI-I treatment, inhibition of proteasome may be a key factor to increase the efficacy of SKI-I-induced cell death. In addition, we detected a rapid autophagic flux in AML cell lines (data not shown). As the Atg12-Atg5 conjugate dissociates from the autophagosomal membrane upon completion (27), a rapid turnover of autophagosomes in cancer cells may limit the formation of iDISC and subsequent caspase-8 activation during SKI-I treatment. Furthermore, rapid autophagic flux may also lead to lysosomal degradation of the iDISC as well as damaged mitochondria associated with the autophagosomal membrane. Therefore, inhibiting autophagic flux is one approach to stabilize pro-apoptotic components on the autophagosomal membrane and shift autophagy to caspase-dependent cell death. To support this concept, inhibition of lysosomal degradation was found to enhance SKI-I-induced apoptosis in AML cells and caspase-3 cleavage in MEFs. Interestingly, chloroquine is currently being evaluated as an enhancing agent for cancer therapy due to its ability to selectively sensitize cancer cells to ionizing radiation and several anti-neoplastic drugs (48–52). Moreover, inhibition of the closure of autophagosomal membranes should stabilize the formation of iDISC and enhance apoptosis. Indeed, we have shown that inhibition of autophagosomal closure through the overexpression of a dominant-negative mutant of Atg4B substantially enhanced SKI-I-induced caspase-8 and caspase-3 cleavage. As there are currently no pharmacological inhibitors that target autophagosome completion, this remains an area to be addressed in future drug discovery.

The mechanism behind the induction of autophagy by SKI-I remains to be determined. Consistent with our results, inhibition of SK by other inhibitors, such as dimethylsphingosine, the pan-SK inhibitor SKI-2, and the sphingosine kinase-2-specific inhibitor ABC294640, have been shown to induce autophagy (53). The induction of autophagy in response to SK inhibition may arise from a decrease in S1P production or the accumulation of sphingosine and/or ceramide. S1P has been shown to mediate cell survival by activating the phosphoinositide 3-kinase/protein kinase B (PI3K/Akt) signaling pathway and nuclear factor κB (NF-κB), both of which exert inhibitory effects on autophagy (54, 55). As a result, suppressed S1P production may relieve the inhibitory effects of these pathways on autophagy. Additionally, S1P protects against apoptosis by blocking the activation of the stress-activated protein kinase Jun amino-terminal kinase (JNK), whereas ceramide activates JNK (56–58). JNK phosphorylates Bcl-2 to liberate Beclin 1 for the induction of autophagy (59). Moreover, JNK activates transcription factor c-Jun, which has been shown to mediate the up-regulation of Beclin 1 (44, 60). In addition, the induction of autophagy by ABC294640 has been reported to be accompanied not only by a decreased activation of Akt but also by an increased expression of Beclin 1 (53). However, no obvious Beclin 1 up-regulation was detected during SKI-I treatment in our system (data not shown). Consequently, the mechanism of autophagy induction in response to SK inhibition is currently under investigation.

In conclusion, the present findings, together with those of recent studies in T cell clonal expansion (15), proteasome inhibition (29), and oncolytic adenovirus (28), indicate that the autophagosomal membrane may serve as a platform for iDISC formation, which activates caspase-8 and the caspase cascade, leading to autophagy-dependent apoptosis. As many anticancer therapies induce both apoptosis and autophagy, establishment of an autophagy-dependent mechanism of caspase activation reveals a novel strategy to enhance therapeutic efficacy in tumor cells.