Calpain-mediated cleavage of Atg5 switches autophagy to apoptosis Shida Yousefi1, Remo Perozzo2, Inès Schmid1, Andrew Ziemiecki3, Thomas Schaffner4, Leonardo Scapozza2, Thomas Brunner4 and Hans-Uwe Simon1,5 Autophagy-related gene (Atg) 5 is a gene product required The process of autophagosome formation is regulated by several for the formation of autophagosomes. Here, we report that autophagy genes (Atgs)1,2. Atg5- (ref. 6) or Atg6- (also known as Beclin- Atg5, in addition to the promotion of autophagy, enhances 1)7,8 deficient mouse embryonic stem cells contain decreased numbers susceptibility towards apoptotic stimuli. Enforced expression of of autophagic vesicles, suggesting that at least some Atgs are essential Atg5-sensitized tumour cells to anticancer drug treatment both for autophagy. We were interested in the effect of elevating Atg5 levels in vitro and in vivo. In contrast, silencing the Atg5 gene with in cells and used two techniques for this purpose: first, HeLa cells were short interfering RNA (siRNA) resulted in partial resistance to stably transfected with the GFP-tagged full-length human Atg5 (NCBI; chemotherapy. Apoptosis was associated with calpain-mediated accession number, 12803724), GFP-tagged cytochrome c9, GFP-tagged Atg5 cleavage, resulting in an amino-terminal cleavage product microtubule-associated protein light chain (LC3, Atg8) or no insert. with a relative molecular mass of 24,000 (Mr 24K). Atg5 Second, Jurkat cells received Atg5 using lentiviral gene transfer. As con- cleavage was observed independent of the cell type and the trol, the same vector with a GFP cDNA insert was used. In both systems, apoptotic stimulus, suggesting that calpain activation and Atg5 isolated exogenous Atg5 gene transfer resulted in cells exhibiting signs of cleavage are general phenomena in apoptotic cells. Truncated increased autophagy (see Supplementary Information, Fig. S1). Atg5 translocated from the cytosol to mitochondria, associated We next investigated whether elevated Atg5 levels and subsequent with the anti-apoptotic molecule Bcl-xL and triggered increased autophagic activity were associated with changes in the sus- cytochrome c release and caspase activation. Taken together, ceptibility of cells to undergo death receptor-mediated or stress-medi- calpain-mediated Atg5 cleavage provokes apoptotic cell death, ated death. Compared with GFP-transfected (or untransfected) cells, therefore, represents a molecular link between autophagy and an increased susceptibility to undergo death was observed in cells apoptosis — a finding with potential importance for clinical with enforced expression of Atg5. Atg5 sensitized cells to CD95-, cera- anticancer therapies. mide- and anticancer drug- (etoposide and doxorubicin) mediated cell death, independent of the cell type. As an additional control, LC3 was Autophagy is a highly conserved process that occurs in yeast, plants overexpressed in HeLa cells and no increased susceptibility to death and in mammalian cells. By means of autophagy, the cell keeps a con- was observed (Fig. 1a). Increased Atg5 levels were also associated with stant balance between biosynthetic and catabolic processes, which increased caspase-3 activity after CD95 stimulation of HeLa cells, as also include the degradation of entire organelles. A portion of the well as with increased apoptotic morphology and phosphatidylinositol cytoplasm is sequestered in an autophagosome and its contents are redistribution within the plasma membrane in anti-CD95 antibody and degraded on fusion with lysosomes1,2. Although autophagy promotes staurosporine treated Jurkat cells, suggesting that the type of death was a cell-survival response3, morphological features of autophagy have apoptosis (Fig. 1b, c). also been observed in dying cells, where caspases (key enzymes in To further understand the requirement for Atg5 in induction of apop- the process of apoptosis) were suppressed or not sufficiently acti- tosis, Atg5 levels were reduced using short interfering RNA (siRNA; vated. Therefore, autophagic cell death has been considered as a Fig. 1d). Atg5-deficient HeLa and breast cancer MDA-MA-231 cells “fail-safe” mechanism under conditions in which apoptosis does exhibited markedly reduced sensitivity towards the cytotoxic agents not work4,5. Thus, a switch from apoptosis to autophagic cell death staurosporine and doxorubicin, respectively. It should be noted, how- can occur. However, whether autophagy could promote apoptosis ever, that Atg5-deficient cells die in long-term cultures (5–7 days), pos- was not known. sibly due to insufficient removal of defective cell organelles that may a Jurkat cells Immature neutrophils MDA-MA-231 cells 100 100 50 ** * 80 80 40 60 60 30 40 40 20 20 20 10 0 0 0 Cell death (percentage) Cell death (percentage) Cell death (percentage) GFP Atg5 GFP Atg5 Atg5 Control + ceramide+ ceramide + anti-CD95+ anti-CD95 Doxorubicin + doxorubicin GFP Atg5 GFP Atg5 Atg HeLa cells HeLa cells HeLa cells 50 50 50 * * 40 40 40 30 30 30 n.s. 20 20 20 10 10 10 0 0 0 Cell death (percentage) Cell death (percentage) Cell death (percentage) LC3 GFP Atg5 GFP Atg5 Control Etoposide + etoposide Doxorubicin Doxorubicin + doxorubicin + doxorubicin Atg5 Atg LC3 bcJurkat cells Jurkat cells Medium Staurosporine VAD + Staurosporine 0.2 0.1 0.5 1.5 0.3 0.7 ) 405 HeLa cells A 0.2 Control 0.15 97.7 2.0 51.0 47.0 96.0 3.0 0.1 58 Control Atg5 0.05 0.3 0.7 0.8 0.2 0.5 0.5 0 Caspase-3 activity ( Atg5 GFP Atg5 Propidium iodide uptake Propidium 98.5 0.5 29.5 68.5 85.5 13.5 + anti-CD95+ anti-CD95 GFP Atg5 50 10 Phosphatidylserine redistribution Atg5 + anti-CD95 Atg5 + VAD + anti-CD95 d HeLa cells MDA-MA-231 cells 50 50 HeLa cells Atg5 40 40 GAPDH 30 30 MDA-MA-231 cells 20 20 Atg5 10 10 GAPDH Cell death (percentage) Cell death (percentage) 0 0 + + Control Control Control Atg5 siRNA Atg5 siRNA Atg5 siRNA Doxorubicin Atg5 siRNA Atg5 siRNA Staurosporine doxorubicin Staurosporine Figure 1 Atg5 regulates cytotoxicity of death triggers. (a) Cell death using flow cytometry. (c) Staurosporine induced redistribution of was analysed after the activation of death pathways in the presence phosphatidylserine was enhanced in Atg5 overexpressing Jurkat cells and and absence of enforced Atg5 expression (18 h cultures). Values are blocked by z-VAD-fmk. Numbers indicate the results of the quantitative means ± s.e.m. of three independent experiments. (b) The cell death analysis (percentage) of each dot blot. (d) Reduction of drug-induced cell regulated by Atg5 is apoptosis. All data presented are from 5 h cultures. death by silencing of Atg5. Values are means ± s.e.m. of three independent Caspase-3-like activity analysed in HeLa cells (graph). Morphology of experiments (18 h cultures). The effects of the Atg5 siRNA on Atg5 and Jurkat cells is also shown. The numbers in the lower right corner indicate GAPDH levels were analysed by immunoblotting. The asterisks indicate the percentage of cell death, which was determined in parallel experiments P <0.05. n.s., no significant difference. 2 accumulate in cells with reduced autophagic activity3. In addition to a these in vitro findings, we also obtained evidence for an apoptosis-sen- Neutrophils sitizing role forAtg5 in vivo (see Supplementary Information, Fig. S2). Time (h) 24 7 4 1 To test the hypothesis that Atg5 has a role in apoptosis pathways, the + −+−+−+− GM-CSF Fresh Control 10 model of spontaneous neutrophil apoptosis was used in initial experi- Mr(K) 33 ments. Interestingly, an additional anti-Atg5-reactive molecule with an Atg5 24 Mr of 24K was detected in blood neutrophils from healthy donors that had been cultured in the absence of survival factors for 7 h. In contrast, in the presence of the survival factor GM-CSF, the 24K protein was GAPDH not detected within this time period. Moreover, when the 24K form appeared, full-length 33K Atg5 was markedly reduced. In 24 h neu- b trophil cultures, the anti-Atg5 antibody recognized the 24K protein only Jurkat cells and the 33K protein was no longer detected (Fig. 2a). Next, Jurkat cells Anti-CD95 were stimulated with an anti-CD95 antibody known to induce apop- h h tosis in these cells. In this model, the 24K anti-Atg5-reactive molecule 7 5 3 h Untreated Act. D CHX Anti-CD95 Medium Mr(K) was rapidly detected (after 3 h stimulation) and its presence seemed to 33 increase over time (Fig. 2b). Atg5 24 Other apoptosis triggers and cell types were also used. For instance, apoptosis was induced by staurosporine in Jurkat, HL-60 and EOL GAPDH cells. Within 5 h of apoptosis induction, the 24K protein, which was associated with caspase-3 cleavage, was detected by immunoblot- Figure 2 Apoptosis is associated with Atg5 cleavage as assessed by ting (see Supplementary Information, Fig. S3a). In addition, anti- immunoblotting. (a) Spontaneous neutrophil apoptosis. Apoptosis was CD95 treatment for 5 h was associated with the appearance of the associated with the appearance of a 24K anti-Atg5-reactive protein. 33K 24K protein in eosinophils and this was accompanied by cleavage of Atg5 disappeared during apoptosis in a time-dependent manner. HeLa poly(ADP–ribose)polymerase (PARP), suggesting increased caspase- cells overexpressing Atg5 served as positive control. (b) Anti-CD95-induced apoptosis of Jurkat cells. Apoptosis was associated with the appearance of 3 activity in these cells (see Supplementary Information, Fig. S3b). a 24K anti-Atg5-reactive protein and a reduction of 33K Atg5 in a time- Cycloheximide (CHX) or actinomycin D treatment caused, similarly dependent manner. The 24K anti-Atg5-reactive protein was also visible in to CD95 activation, the appearance of 24K Atg5 in Jurkat cells, making Jurkat cells after treatment with actinomycin D (Act. D) or cycloheximide it unlikely that the molecule is newly synthesized during the process (CHX; 5 h treatment in both cases). GAPDH expression indicates equal loading in each panel.