Introduction
Autophagy is a major pathway for delivery of proteins and organelles to lysosomes/the vacuole, where they are degraded and recycled. Autophagy plays an essential role in differentiation and development, as well as in cellular response to stress. It is activated during amino‐acid deprivation and has been associated with neurodegenerative diseases, cancer, pathogen infections and myopathies (reviewed in
Cuervo, 2004;
Shintani and Klionsky, 2004). Autophagy is initiated by the surrounding of cytoplasmic constituents by the crescent‐shaped isolation membrane/phagophore, which forms a closed double‐membrane structure, called autophagosome. Finally, the autophagosome fuses with a lysosome to become an autolysosome, and its content is degraded by lysosomal hydrolases.
The molecular mechanism underlying autophagy has been extensively researched in the past decade, and the genes participating in this process, denoted ATGs (
au
topha
gy‐related;
Klionsky et al, 2003), were found to be conserved from yeast to man (
Ohsumi, 2001;
Huang and Klionsky, 2002). Yet, the roles played by the different gene products and their modes of action are to be resolved. A hallmark event in the autophagic process is the reversible conjugation of the Atg8 family of proteins to the autophagosomal membrane (
Ichimura et al, 2000). This ubiquitin‐like (UBL) protein family, which has been implicated in a variety of cellular processes, includes Atg8p in yeast (
Lang et al, 1998;
Ichimura et al, 2000) and GATE‐16 (Golgi‐associated ATPase enhancer of 16 kDa;
Sagiv et al, 2000), LC3 (light‐chain 3;
Mann and Hammarback, 1994) and GABARAP (GABA receptor‐associated protein;
Wang et al, 1999) in mammals. All Atg8 homologues (hereafter termed Atg8s) serve as substrates for the Atg4 family of cysteine proteases (
Kirisako et al, 2000;
Hemelaar et al, 2003;
Marino et al, 2003;
Scherz‐Shouval et al, 2003). Atg4s cleave Atg8s near the C‐terminus, downstream of a conserved glycine. This cleavage allows the conjugation of Atg8 to phosphatidylethanolamine (PE) through the exposed glycine, a process mediated through a ubiquitination‐like mechanism. Atg8‐PE serves as another substrate for Atg4, which cleaves Atg8 and releases it from the membrane. Notably, this conjugation process must be preceded by another ubiquitination‐like process, conjugating Atg12 to Atg5 (
Tanida et al, 2004b). In mammalian cells, amino‐acid deprivation induces lipidation of LC3 (
Kabeya et al, 2000) as well as of GATE‐16 and GABARAP (
Kabeya et al, 2004). Lipidated Atg8s associate with phagophaores and autophagosomes and remain there until fusion with lysosomes, at which point intra‐autophagosomal Atg8 is probably degraded (
Kabeya et al, 2000,
2004).
At least four Atg4 mammalian homologues have been reported based on sequence homology to the yeast
Saccharomyces cerevisiae (Sc)Atg4. Two of the homologues, HsAtg4A and HsAtg4B, were shown to cleave the three mammalian Atg8s with different efficiencies: HsAtg4A cleaves mainly GATE‐16, whereas HsAtg4B cleaves all three homologues (GATE‐16, GABARAP and LC3), with the highest efficiency for LC3 (
Kabeya et al, 2004).
Atg4s act both as conjugating and deconjugating enzymes and therefore their activity is expected to be tightly regulated. Hence, in the process of autophagy, following the initial cleavage of Atg8‐like proteins, Atg4 must become inactive so as to ensure the conjugation of Atg8 to the autophagosomal membrane. Later on, as the autophagosome fuses with the lysosome, Atg4 might be locally re‐activated in order to delipidate and recycle Atg8. How is this regulation achieved? We have previously shown that recombinant HsAtg4A is active as a cysteine protease of GATE‐16 only in the presence of the reducing agent DTT (
Scherz‐Shouval et al, 2003). This could implicate that,
in vivo, the delipidating activity of Atg4 is regulated through changes in redox potentials that take place under different conditions and at specific subcellular microenvironments. Such regulation could be inflicted, among other factors, by reactive oxygen species (ROS). At high levels, ROS are deleterious to cells, leading to programmed cell death (PCD) (
Jabs, 1999;
Lee et al, 2003;
Macip et al, 2003). At low levels, however, ROS can serve as signaling molecules, by oxidizing factors in a variety of pathways that lead to growth and survival. Here, we report for the first time the involvement of ROS as signaling molecules in nutrient starvation‐induced autophagy, which is essentially a survival pathway. We show that under starvation conditions, cells form ROS, specifically H
2O
2, which is essential for autophagosome formation and autophagic degradation. The oxidative signal is partially PI3K dependent and leads to inhibition of Atg4. Using an
in vitro assay, we could demonstrate that H
2O
2 directly regulates HsAtg4A. Moreover, we identified Cys
81, situated four amino acids downstream of the active site, as an essential residue for the redox regulation of HsAtg4A. Expression of HsAtg4A mutated in Cys
81, or the corresponding mutant of HsAtg4B, inhibited the formation of GATE‐16‐ or LC3‐labeled autophagosomes in cells.
Discussion
Autophagy is a unique pathway of intracellular trafficking activated in response to extracellular signals. Although many of the proteins involved in this process have been identified, the signal transduction pathway leading to activation of autophagy is not fully solved. Here we demonstrate, for the first time, the involvement of ROS as signaling molecules in starvation‐induced autophagy. We show that starvation triggers accumulation of ROS, most probably H2O2, which is necessary for autophagosome formation and the resulting pathway of degradation. The oxidative signal is partially PI3K dependent, located upstream of the Atg5‐dependent conjugation machinery. Furthermore, we identify a direct target for oxidation by H2O2—the protease Atg4.
Based on our results, we propose the following model, illustrated in
Figure 10: nutrient starvation induces complex formation between class III PI3K and Beclin 1, which in turn leads, together with other signals, to a local rise in H
2O
2 in the vicinity of mitochondria. This oxidative signal leads to inactivation of Atg4 at the site of autophagosome formation, thereby promoting lipidation of Atg8, an essential step in the process of autophagy. As the autophagosome matures towards fusion with the lysosome, its localization apparently changes to an H
2O
2‐poor environment where Atg4 is active and can delipidate and recycle Atg8.
How does H
2O
2 regulate Atg4? Atg4s are cysteine proteases that share several conserved cysteine residues. One of these is the catalytic residue, found previously in other members of the Atg4 family and shown here to be Cys
77 in HsAtg4A. Another conserved motif is a cysteine residue located four amino acids from the catalytic residue, Cys
81 in HsAtg4A. Remarkably, we found this residue to be crucial for the redox regulation of HsAtg4A. Mutation of this residue to serine significantly impaired the sensitivity of the protein to H
2O
2 in vitro, and abolished the formation of GATE‐16‐labeled autophagosomes in cells. The essentiality of this residue for proper redox regulation was further confirmed by the finding that the corresponding HsAtg4B mutant (HsAtg4B
C78S) exhibited a similar autophagic defect. Our
in vitro study shows that H
2O
2 directly inactivates Atg4. This could be accomplished either by binding of H
2O
2 to Cys
77 or to Cys
81, forming reversible sulfenic acid that shields Cys
77, or by oxidation, which leads to other modifications such as formation of a disulfide bond between Cys
77 and Cys
81, again shielding Cys
77. Such a disulfide bond, however, was not detected in our experiments (unpublished data). Alternatively, H
2O
2 might be affecting a different site, which is altered when Cys
81 is mutated. The active site, Cys
77, is conserved from yeast to man, whereas Cys
81 is conserved only in tetrapod homologues of Atg4A and Atg4B. Lower species harbor a conserved Ser or Thr in this residue and hence the choice of a C81S mutant. We thus propose that the redox regulation may be mediated via Cys
77 in lower species, whereas tetrapods evolved a more complex mechanism of regulation that requires both Cys
77 and Cys
81, where the latter provides tetrapods with higher sensitivity to ROS than lower species. The recently solved three‐dimentional structure of HsAtg4B (
Sugawara et al, 2005;
Kumanomidou et al, 2006) supports our model, as it shows Cys
77 and Cys
81 to be situated at the beginning of an α‐helix, both facing the same plane of the helix. With such close proximity between the two cysteines, modifications on Cys
81 could indeed affect the active site. Notably, both crystals were prepared under reducing conditions, and consequently no disulfide bridges or other modifications were reported for any of the cysteine residues.
Cysteine‐harboring proteins serve as redox sensors. They can undergo rapid, variable and, most importantly, reversible post‐translational modifications in response to changes in the oxidative conditions of the environment (
Sitia and Molteni, 2004). Several cysteine proteases, including cathepsin D, cathepsin B and the cytosolic caspase‐3 (
Chandler et al, 1998) and calpains (
Glading et al, 2001), have been shown to undergo redox regulation through a direct modification of the active site (reviewed in
Giles et al, 2003). Redox regulation of Atg4 can provide rapid activation and inactivation of this protease, as part of the signaling pathway, leading to induction of the autophagic process.
A growing body of evidence suggests that autophagy plays a dual role, both in survival and in death‐related pathways (
Cuervo, 2003;
Codogno and Meijer, 2005;
Reef et al, 2006). The data available till now, linking ROS with autophagy, describe the death‐related pathway, known as type II PCD, which is induced by high levels of ROS (i.e. oxidative stress;
Yu et al, 2004;
Kiffin et al, 2006). Recently, it has been shown that caspase inhibition leading to cell death through autophagy involves accumulation of ROS, owing to catalase‐specific autophagic degradation. In that study, however, starvation did not lead to a similar effect (
Yu et al, 2006). In our system, the rise in ROS is both local and reversible; it is not deleterious to cells and serves to oxidize a specific target. Therefore, our study provides the first indication for the involvement of ROS in starvation‐induced autophagy as signaling molecules in a survival pathway.