Regulation of an ATG7-beclin 1 Program of Autophagic Cell Death by Caspase-8

See allHide authors and affiliations

Science  04 Jun 2004:
Vol. 304, Issue 5676, pp. 1500-1502
DOI: 10.1126/science.1096645


Caspases play a central role in apoptosis, a well-studied pathway of programmed cell death. Other programs of death potentially involving necrosis and autophagy may exist, but their relation to apoptosis and mechanisms of regulation remains unclear. We define a new molecular pathway in which activation of the receptor-interacting protein (a serine-threonine kinase) and Jun amino-terminal kinase induced cell death with the morphology of autophagy. Autophagic death required the genes ATG7 and beclin 1 and was induced by caspase-8 inhibition. Clinical therapies involving caspase inhibitors may arrest apoptosis but also have the unanticipated effect of promoting autophagic cell death.

Apoptosis is a well-studied pathway of programmed cell death conserved from Caenorhabditis elegans to humans (1). Caspases, a family of cystinyl aspartaterequiring proteases, produce the morphological changes associated with apoptotic death (2, 3). Nonapoptotic forms of cell elimination include those with features of necrosis and autophagy (47). Necrosis can result when cell metabolism and integrity are compromised by a nonphysiological insult. Recently, evidence has emerged that death receptors and receptor-interacting protein (RIP) can induce caspase-independent cell death that appears necrotic (6, 7). Autophagy promotes a cell survival response to nutritional starvation involving membrane-bound vacuoles that target organelles and proteins to the lysosome for degradation (8, 9). Two pathways functioning in autophagy contain ubiquitin-like genes that are highly conserved from yeast to humans (ATG genes). Certain examples of cell death have autophagic features, but a role for ATG genes in cell death has not been established (10).

In mouse L929 fibroblastic cells, tumor necrosis factor, oxidants, ceramide, and radiation can induce caspase-independent death (11). However, benzyloxycarbonylvalyl-alanyl–aspartic acid (O-methyl)–fluoro-methylketone (zVAD), a caspase inhibitor with broad specificity, also directly induced the death of L929 cells. Death began at 12 hours after zVAD treatment and was complete after 40 hours (Fig. 1, A and B). The dead cells appeared to be round and detached, and they had a convoluted plasma membrane permeable to vital dyes; this differed from apoptosis, in which nuclei are condensed and membrane integrity is preserved. Transmission electron microscopy (TEM) revealed intact mitochondria and endoplasmic reticulum, condensed osmophilic cytoplasm, and numerous large cytoplasmic inclusions that were membrane-bound vacuoles characteristic of autophagy (Fig. 1C). A time course revealed that vacuolated cells accumulated before cell death (Fig. 1D). Similar results were obtained in human U937 monocytoid cells (Fig. 1E and fig. S2). The zVAD treatment also induced cell death in mouse RAW 264.7 macrophage cells and primary mouse peritoneal macrophages (figs. S3 and S4).

Fig. 1.

Autophagy and ATG genes are required for zVAD-induced cell death. (A) L929 cells were treated with 1 μl of dimethyl sulfoxide (DMSO) (a and c) or 20 μM zVAD (b and d) for 24 hours and examined by phase contrast microscopy (a and b) or 4′,6′-diamidino-2-phenylindole–staining and fluorescent microscopy (c and d). Magnification, 200×. (B) Time course for zVAD-induced cell death of L929 cells. (C) TEM of L929 cells treated for 12 hours with DMSO (a) or zVAD (b to d). Arrows show membrane-bound vacuoles characteristic of autophagosomes. Scale bars, 1 μm (a and b); 0.1 μm (c and d). (D) Time course for zVAD-induced autophagy. The percentage of vacuolated cells is the fraction of cells that have 10 or more autophagic vacuoles by TEM. For statistical and morphometric analyses, see fig. S1 and table S1. (E) Reduction in cell number (mean value ± SD) for L929 or U937 cells treated with the PI-3 kinase inhibitors Wortmannin (WM) (0.1 μg/ml) or 3-MA (10 mM) for 1 hour and then with 20 μM zVAD or DMSO for 36 hours. (F) L929 cells were treated for 36 hours with zVAD or DMSO after transfection with beclin 1, ATG7(mAPG7) RNAi, or nonspecific (NS) oligoribonucleotides, and reduction in cell number (solid bars) and vacuolated cells (open bars) were quantified. Reductions in the corresponding proteins are shown by Western blot (inset). (G) U937 cells were activated by 10 ng/ml phorbol myristate acetate for 24 hours after transfection with beclin 1, ATG7(hGSA7) RNAi, or nonspecific (NS) oligoribonucleotides, and then reduction in cell number was measured after zVAD treatment for 36 hours. (H) Representative TEM photomicrographs of the L929 cells treated with zVAD (24 hours) and with the indicated RNAi preparations. Scale bars, 1 μm.

The association of autophagic vacuoles with cell death has been observed in developing animals, but it has not been clear whether the process serves to rescue or condemn the cell (12). Drosophila cells manifesting autophagy and death have increased ATG gene transcripts (13, 14), but there is no known requirement for ATG genes in cell death. We sought evidence that autophagy was required for cell death by treating cells with two inhibitors of autophagy, 3-methyl-adenine (3-MA) and Wortmannin (9, 15). Both inhibitors arrested zVAD-induced cell death in all cell lines and in primary macrophages (Fig. 1E and figs. S3 and S4). However, these inhibitors are general phosphatidylinositol-3 (PI-3) kinase inhibitors and could independently affect autophagy and nonapoptotic cell death. We therefore tested whether ATG genes were required for cell death.

ATG7 (HsGSA7/mAPG7) is a key autophagy gene encoding a protein resembling E ubiquitin–activating enzyme that is used in both of the ubiquitin-like pathways required to form autophagic vacuoles in yeast (16, 17). We reduced expression of ATG7 by RNA interference (RNAi) and found that zVAD-induced cell death was almost completely inhibited (Fig. 1F). Another ATG gene, beclin 1, the mouse homolog of yeast ATG6, encodes a Bcl-2–interacting candidate tumor suppressor and antiviral protein (18, 19). Molecular alterations in beclin 1 are common in human cancers, and beclin 1 gene knockouts in mice cause a marked increase in epithelial and hematopoietic malignancies (20, 21). Reduction of the Beclin-1 protein by RNAi also decreased zVAD-induced death (Fig. 1F). Reduction of ATG7 and beclin 1 also inhibited zVAD-induced death in human U937 cells (Fig. 1G). TEM analyses of cells with reduced Atg7 or Beclin-1 protein levels showed a parallel inhibition of autophagic vacuole formation associated with reduced cell death (Fig. 1, F and H, and table S2). Thus, Atg7 and Beclin-1 are required for nonapoptotic cell death triggered by zVAD.

Death receptors can elicit nonapoptotic death through the RIP, a death domain–containing serine-threonine kinase (6, 7). We therefore reduced RIP expression by RNAi and observed decreased autophagy and decreased death (Fig. 2A). zVAD activated c-Jun N-terminal kinase (JNK), which is also activated by RIP in response to cytokines (Fig. 2B) (22). The mitogen-activated protein (MAP) kinases p38 and extracellular signal–regulated kinase (ERK) were not activated, indicating a specific role for JNK (23). Moreover, a JNK inhibitor, but not inhibitors against p38 or ERK, blocked zVAD-induced cell death, further indicating a specific role for JNK (Fig. 2C and fig. S5). The protein synthesis inhibitor cycloheximide (CHX) blocked cell death, indicating that protein synthesis was required. RNAi silencing of the JNK-activating kinase MAP kinase kinase 7 (MKK7) also completely prevented cell death and formation of autophagic vacuoles (Fig. 2D). RNAi suppression of the transcription factor c-Jun reduced but did not eliminate the c-Jun protein and inhibited autophagy and cell death by 45 to 50% (Fig. 2D). Thus, a signal pathway involving RIP, MKK7, JNK, and c-Jun appears to activate autophagy and cell death. The involvement of c-Jun and new protein synthesis implies that the transcription of target genes may also be required.

Fig. 2.

Requirement of RIP and JNK signaling pathways for autophagic death. (A) Cells were treated with zVAD or DMSO after transfection with RIP RNAi or nonspecific (NS) oligonucleotides. Reduction in cell number (solid bars) and the fractions of cells with autophagic features based on TEM (open bars) were quantitated as in Fig. 1. The amount of RIP is shown by Western blot (inset). (B) Western blot for phospho-JNK (left lanes) or total JNK protein (right lanes) after zVAD or DMSO treatment. (C) zVAD-induced death in cells treated with control, JNK inhibitor II (1 μg/ml), or CHX (2 μg/ml). (D) Reduction in cell number following zVAD or DMSO for 36 hours after transfection with MKK7 RNAi, c-Jun RNAi, or nonspecific oligoribonucleotides (solid bars) and the fractions of cells with autophagic features by TEM (open bars) were quantitated. The steady-state amounts of the corresponding proteins are shown by Western blot (inset).

Finally, we addressed how zVAD induced autophagic cell death. Active caspase-8 functions in lymphocyte receptor signaling pathways that do not cause cell death (24). We therefore used RNAi to progressively reduce caspase-8 expression over time and found that cell death was correspondingly increased (Fig. 3A). Cells in which caspase-8 was reduced showed features of autophagy (Fig. 3B). Other peptide caspase inhibitors and RNAi suppression of caspases 1, 2, 3, 9, and 12 had no ability to induce death (figs. S6 and S7). Because zVAD is a potent inhibitor of caspase-8, it likely exerted its death effect through the inhibition of caspase-8. Also, RIP was partially cleaved in a fragment characteristic of caspase-8 in unstimulated cells, and this was eliminated by zVAD treatment, implying low constitutive caspase-8 activity (fig. S8) (25).

Fig. 3.

Inhibition of the autophagic death pathway by caspase-8. (A) Time course of viability of L929 cells transfected with either nonspecific (NS) (open bars) or caspase-8–specific (solid bars) RNAi at 24, 96, and 110 hours after transfection. Panels below show the abundance of caspase-8 protein by Western blot. (B) Representative TEM pictures and quantification of the cells treated with either nonspecific or caspase-8–specific RNAi. Cells were harvested at 96 hours after RNAi transfection. (a) NS control cell. (b to d) Caspase-8 RNAi at different magnifications. Scale bars, 1 μM (a and b), 0.5 μM (c and d). Arrows indicate double-membrane autophagic vacuoles. (e) The fraction of cells with autophagic features based on TEM was quantified for NS control cells (open bar) and caspase-8 RNAi (solid bar) (for NS versus caspase-8, P < 0.0001, Mann-Whitney U test).

We have shown that two key autophagy genes, ATG7 and beclin 1, are necessary for a nonapoptotic death pathway in mammalian cells. This may explain other forms of nonapoptotic death (26). The conservation of autophagy genes throughout phylogeny suggests that this form of death might have a role in many eukaryotes. beclin 1 gene knockouts cause an unexplained increase in spontaneous tumors (21, 22), and it is possible that beclin 1 may act as a tumor suppressor by causing autophagic cell death. An interesting unresolved question is the molecular connection between RIP and the JNK signaling pathway (23). The suppression of autophagic death by caspase-8 in mammalian cells indicates that caspases can regulate both apoptotic and nonapoptotic cell death. We favor the idea that there is a low constitutive level of caspase-8 activation that carries out cellular regulatory processes (24). Because viral pathogens have caspase inhibitors, the autophagic pathway could be poised to counter infection as a “fail-safe” mechanism of nonapoptotic cell death. Caspase inhibitors are currently being developed as therapeutic agents (27, 28). Our findings indicate that caspase inhibition could have the untoward effect of exacerbating cell death and disease severity by activating the autophagic death pathway.

Supporting Online Material

Materials and Methods

Figs. S1 to S8

Tables S1 and S2


References and Notes

View Abstract

Navigate This Article