|Home | About | Journals | Submit | Contact Us | Français|
Autophagy is a potent intracellular degradation process with pivotal roles in health and disease. Atg8, a lipid-conjugated ubiquitin-like protein, is required for the formation of autophagosomes, double-membrane vesicles responsible for the delivery of cytoplasmic material to lysosomes. How and when Atg8 functions in this process, however, is not clear. Here we show that Atg8 controls the expansion of the autophagosome precursor, the phagophore, and give the first real-time, observation-based temporal dissection of the autophagosome formation process. We demonstrate that the amount of Atg8 determines the size of autophagosomes. During autophagosome biogenesis, Atg8 forms an expanding structure and later dissociates from the site of vesicle formation. On the basis of the dynamics of Atg8, we present a multistage model of autophagosome formation. This model provides a foundation for future analyses of the functions and dynamics of known autophagy-related proteins and for screening new genes.
Eukaryotic cells employ macroautophagy, hereafter referred to as autophagy, to eliminate objects ranging from soluble proteins to entire organelles (Yorimitsu and Klionsky, 2005 ; Xie and Klionsky, 2007 ). In multicellular organisms, autophagy has important roles in development, immune defense, programmed cell death, tumor suppression, and the prevention of neuronal degeneration (Levine and Klionsky, 2004 ; Shintani and Klionsky, 2004a ; Boland and Nixon, 2006 ; Schmid and Münz, 2007 ). In this pathway, cytoplasmic materials are sequestered into an expanding membrane sac, the phagophore, which subsequently matures into a double-membrane vesicle, the autophagosome. The site of autophagosome formation is termed the phagophore assembly site (PAS; Kim et al., 2002 ; Suzuki et al., 2001 ). Each autophagosome eventually fuses with a lysosome, resulting in the degradation of the inner membrane and the cargos. The formation of autophagosomes depends on the concerted actions of core autophagy machinery proteins (Yorimitsu and Klionsky, 2005 ; Xie and Klionsky, 2007 ).
Among the core machinery proteins is Atg8, a ubiquitin-like protein (Ichimura et al., 2000 ; Paz et al., 2000 ). Newly synthesized Atg8 is processed by a cysteine protease, Atg4, to expose its carboxyl terminal glycine residue (Kirisako et al., 2000 ; Kim et al., 2001 ; Hemelaar et al., 2003 ). It is then conjugated to phosphatidylethanolamine (PE) by the E1-like activating enzyme Atg7 and the E2-like conjugating enzyme Atg3 (Tanida et al., 1999 ; Ichimura et al., 2000 ; Tanida et al., 2001 , 2002 ). Conjugated Atg8 can also be deconjugated by Atg4 (Kirisako et al., 2000 ). A multimeric protein complex formed by Atg12, Atg5, and Atg16 facilitates the conjugation of Atg8, possibly by serving as an E3-like enzyme (Suzuki et al., 2001 ; Hanada et al., 2007 ). The formation of this complex itself also involves a conjugation reaction, in which the ubiquitin-like protein Atg12 is attached to Atg5 by Atg7 and the E2-like enzyme Atg10 (Mizushima et al., 1998a ,b , 2003 ; Kuma et al., 2002 ).
During autophagy, Atg8 localizes to the PAS. In addition to PE conjugation and the Atg12–Atg5-Atg16 complex, its proper localization requires Atg9, a transmembrane protein suggested as a membrane carrier, and the autophagy-specific phosphatidylinositol 3-kinase (PI3K) complex (Kametaka et al., 1998 ; Kirisako et al., 2000 ; Kim et al., 2001 ; Suzuki et al., 2001 , 2007 ; Reggiori et al., 2005 ; Young et al., 2006 ). The PAS population of Atg8 is presumably associated with the phagophore (Kirisako et al., 1999 ; Kabeya et al., 2000 , 2004 ). When the phagophore matures into an autophagosome, some Atg8 is trapped inside and eventually degraded (Kirisako et al., 1999 ; Huang et al., 2000 ; Kabeya et al., 2000 ; Tanida et al., 2005 ). The absence of Atg8 does not apparently affect the function of any other core machinery proteins (Suzuki et al., 2001 , 2007 ). Although Atg8 has been widely used as a marker for recognition of autophagosomes (Klionsky et al., 2007 ), its exact in vivo function is still not clear.
In the yeast Saccharomyces cerevisiae, the core machinery proteins are shared between starvation-induced autophagy and a second autophagy-like process, the cytoplasm to vacuole targeting (Cvt) pathway, which transports hydrolase precursors from the cytosol to the vacuole (a lysosome analogue) under nutrient-rich conditions (Baba et al., 1997 ; Xie and Klionsky, 2007 ). Among these proteins, only Atg8 has a significantly elevated protein level when autophagy is induced by starvation, making it a natural candidate for an autophagy regulator (Kirisako et al., 1999 ; Huang et al., 2000 ). Here we investigated the role of Atg8 through morphological and functional analyses.
The media used were as follows: rich medium (YPD): 1% yeast extract, 2% peptone, and 2% glucose; and nitrogen starvation medium (SD-N): 2% glucose, and 0.17% yeast nitrogen base without amino acids and ammonium sulfate.
Gene knockouts were performed as previously described (Yen et al., 2007 ). To construct Atg8 expression plasmids, promoters from genes with protein levels expected to be similar to that of Atg8 in rich medium were placed in front of the ATG8 open reading frame; the resulting Atg8 protein levels were tested by Western blotting. Three plasmids were chosen for this study based on the following two criteria: 1) Atg8 amounts in starvation were lower than that of the wild-type strain and 2) Atg8 amounts in rich media were similar to that of the wild-type strain. Plasmid pPATG27ATG8(406) contains 740 base pairs of ATG27 5′ sequence; plasmid pPVPS30ATG8-PATG18ATG8(406) contains 420 base pairs of VPS30 5′ sequence and 530 base pairs of ATG18 5′ sequence in front of two copies of ATG8 open reading frames, respectively; plasmid pPATG3ATG8(406) contains 380 base pairs of ATG3 5′ sequence. Plasmid pP1KGreen fluorescent protein (GFP)-ATG8(406) contains 990 base pairs of ATG8 5′ sequence in front of the GFP-Atg8 open reading frame. These Atg8-expressing plasmids or the corresponding empty vectors were linearized and integrated into the atg8Δ strain. The strains used in this study are listed in Table 1.
Sample preparation and image acquisition were performed as described previously (Yen et al., 2007 ). Autophagic body cross-sections with a clear limiting membrane were outlined by hand using Adobe Photoshop (San Jose, CA). The area values of outlined autophagic body cross-sections were obtained using the particle analysis function of ImageJ (http://rsb.info.nih.gov/ij/). The area values were then converted to radii for data presentation, using the formula: radius = square root of (area divided by pi).
Live cell fluorescence microscopy was performed as previously described (Yen et al., 2007 ) with the following modification: one side of the cover glass was coated with 1 mg/ml concanavalin A for 5 min and rinsed with water; 100 μl of yeast cell culture was placed on the treated side for 3 min to immobilize yeast cells on the cover glass; the cover glass was rinsed with water, placed on a concavity slide containing liquid medium, and observed under the microscope.
At each time point, a stack of images was collected along the z-axis to cover the entire cell. A projection of the image stack was created by calculating the sum of signal intensities. The intensity of GFP-Atg8 puncta was calculated as the difference between the absolute value of the GFP-Atg8 punctum and that of the local background, using the intensity of its adjacent area, as determined with softWoRx software (Applied Precision, Issaquah, WA).
To explore the possible role of Atg8 in autophagy regulation, we generated strains (A8-1, A8-2, and A8-3) that express different levels of Atg8 (Figure 1A; Materials and Methods). Under nitrogen-starvation conditions, the amounts of both conjugated and unconjugated Atg8 in these strains were lower than those of the wild type (Figure 1A). We then quantified the levels of autophagy in these strains using the Pho8Δ60 assay (Noda et al., 1995 ). This assay is based on the autophagy-dependent delivery of a nonspecific cytosolic marker, the modified phosphatase precursor Pho8Δ60, from the cytosol to the vacuole, where it gets activated. The resulting alkaline phosphatase (ALP) activity thus indicates the total internal volume of the autophagosome population. In wild-type cells, nitrogen starvation led to a sharp increase of Pho8Δ60-dependent ALP activity (Figure 1B). In contrast, the atg8Δ strain showed a negligible increase. Strains A8-1, A8-2, and A8-3 showed intermediate increases, which correlated with their Atg8 protein levels. Importantly, these results were not caused by long-term low-level expression of Atg8 (i.e., a chronic defect), because in nutrient-rich media the levels of Atg8 in strains A8-1, A8-2, and A8-3 were similar to that of the wild-type strain (Figure 1A). This level of Atg8 was sufficient for the maturation of the Cvt pathway cargo, precursor aminopeptidase I (prApe1; Figure 1C). Thus, the levels of Atg8 directly determine levels of autophagy in the corresponding strains.
Normal recruitment of Atg8 to the PAS requires Atg9 (Suzuki et al., 2001 , 2007 ). It has been shown that slowing the anterograde trafficking of Atg9 leads to retarded autophagosome formation without affecting autophagosome size, possibly by reducing the membrane supply (Yen et al., 2007 ). To explore whether Atg8 functions in a similar manner, we analyzed the effect of lower Atg8 expression by transmission electron microscopy (EM).
Strains expressing variable levels of Atg8 were deleted for the PEP4 locus. Pep4-dependent vacuolar hydrolase activity is required for degradation of the inner vesicles of autophagosomes (termed autophagic bodies) in the vacuole. Hence, the absence of PEP4 allows the preservation of autophagic bodies. After 4 h of starvation, no autophagic bodies were observed in atg8Δ pep4Δ strains (Figure 2A), showing that atg8Δ cells are unable to produce autophagosomes; in A8-1, A8-2, A8-3, and wild-type strains deleted for PEP4, autophagic bodies abounded in the vacuoles (Figure 2A). Lower levels of Atg8 led to significant reductions in sizes of autophagosomes. The average cross-sectional radii of autophagic bodies in the A8-1, A8-2, and A8-3 pep4Δ strains were 115 ± 2, 125 ± 2, and 148 ± 2 nm, respectively, whereas that of the wild-type pep4Δ strain was 162 ± 2 nm (mean ± SEM, n > 200; Figure 2B). In contrast, the numbers of autophagic bodies were not affected by the reduced level of Atg8 at either 2 or 4 h of starvation (Figure 2C). These data suggest that even though Atg8 depends on Atg9 for its PAS localization, the role of Atg8 in autophagosome formation is distinct from that of Atg9, in that it specifically controls the size of autophagosomes.
To gain further insight into how Atg8 participates in autophagosome formation, we used fluorescence microscopy to observe the dynamics of Atg8 in live cells. GFP-Atg8 was expressed under the control of the endogenous ATG8 promoter in the atg8Δ background (strain GA8). When cells were examined under starvation conditions, we saw the constant emergence and disappearance of GFP-Atg8 punctate structures, each with a duration of ~10 min (Supplementary Movie S1, Supplementary Figure S1). The sizes of the GFP-Atg8 puncta expanded initially but then remained nearly the same when the fluorescence decreased (Supplementary Figure S2).
To determine whether this dynamic pattern correlates with autophagosome formation, we repeated the analysis in cells expressing mCherry-prApe1 (Figure 3A; Shaner et al., 2004 ). Precursor Ape1 forms a large oligomer in the cytosol, detectable as a single punctum, which becomes a cargo of an autophagosome in starvation conditions (Baba et al., 1997 ; Shintani et al., 2002 ). Upon vacuolar delivery, the large oligomer disassembles into dodecamers, which then display a diffuse fluorescence pattern within the vacuole lumen. In wild-type cells, the disappearance of each mCherry-prApe1 punctum was preceded by the emergence and disappearance of a colocalizing GFP-Atg8 punctum. In contrast, in atg1Δ cells, which are defective in autophagosome formation (Tsukada and Ohsumi, 1993 ), both GFP-Atg8 and mCherry-prApe1 persisted without a significant decrease in fluorescence in punctate structures (Figure 3B). These data suggest that the dynamic pattern of Atg8 puncta formation and disassembly is an integral part of normal autophagy and that it represents events that occur before the disassembly of the cargo complex in the vacuole.
To test if the decrease and disappearance of the GFP-Atg8 signal represents the degradation of GFP-Atg8–containing autophagic bodies in the vacuole, we examined the dynamics of GFP-Atg8 in pep4Δ atg11Δ cells (ATG11 was knocked out to reduce the accumulation of Cvt bodies, (Shintani and Klionsky, 2004b ), which otherwise interfere in tracing the movement GFP-Atg8 puncta that are associated with autophagic bodies). A decrease of the GFP-Atg8 signal was again clearly observed (Figure 4A). After the decrease in GFP-Atg8 fluorescence, some GFP-Atg8 puncta persisted within the vacuole lumen, displaying a low amount of fluorescence, whereas the majority could no longer be detected. These data suggest that the majority of GFP-Atg8 molecules were released into the cytoplasm before autophagosome–vacuole fusion.
It should be noted that this large decrease of fluorescence was not an artifact due to photobleaching. We quantified the fluorescence of GFP-Atg8 puncta in live cells and compared them against the values obtained from fixed cells. In fixed cells, photobleaching caused less than a 10% reduction in fluorescence after 5 min (Figure 4, B and C). In contrast, in live pep4Δ atg11Δ cells and in GA8 cells, the fluorescence of GFP-Atg8 structures dropped well below 50% of the maximum intensities 5 min after the peak (Figure 4, B and C).
Next, we decided to test if the release of GFP-Atg8 fluorescence from the PAS required deconjugation. The deconjugation of Atg8–PE by Atg4 is necessary for normal autophagy (Kirisako et al., 2000 ), although its exact role is not clear. We expressed GFP-Atg8ΔR, which lacks the carboxyl terminal arginine residue (thus bypassing the first step of Atg4 processing), together with mCherry-prApe1 in atg4Δ atg8Δ cells. In the absence of Atg4, Atg8ΔR can be conjugated normally but cannot be deconjugated. We then examined these cells in starvation conditions. As in wild-type cells, we found GFP-Atg8 puncta colocalizing with mCherry-prApe1, indicating that deconjugation is not required for the PAS localization of Atg8 (Figure 4D). However, these puncta did not show a significant decrease in fluorescence, and the colocalizing mCherry-prApe1 remained visible (Figure 4D), suggesting that the release of GFP-Atg8 from the PAS that we are monitoring is mediated by Atg4 and that this reaction is a crucial step in autophagosome formation and/or completion.
Taken together, these results suggest that each cycle of appearance and disappearance of the GFP-Atg8 signal represents the recruitment and release of Atg8 involved in the formation and completion of an autophagosome.
The PAS is generally considered to be the site where the core machinery proteins, including Atg8, act to form autophagosomes. We next tested whether the regulation of the level of autophagy that occurs through modulating the size of the autophagosome is accomplished by controlling the amount of Atg8 at the PAS. GFP-Atg8 was expressed in atg8Δ cells under the control of either its own promoter (strain GA8) or the promoter used in strain A8-1 (strain A81-GA8; Figure 5A). After cells were incubated in starvation conditions for 1 h, we quantified the peak fluorescence of the GFP-Atg8 PAS puncta during the dynamic cycles. The average intensity value in strain A81-GA8 was ~50% of that in strain GA8 (Figure 5B). Consistent with this observation, the Pho8Δ60 activity of strain A81-GA8 was lower than that of strain GA8 (Figure 5C), indicating that the amount of Atg8 recruited to the PAS regulates the level of autophagy.
Our microscopy observation provided a novel method to test whether a lower amount of Atg8 limits the number of autophagosomes produced. Because the recruitment and release of the majority of Atg8 happens before the delivery of autophagosomes to the vacuole, the number of GFP-Atg8 signal peaks would reflect the number of autophagosomes formed. We compared the peak frequency in a strain (GA8) expressing GFP-Atg8 alone with that in a strain (GA8+A8) expressing additional Atg8 (Figure 6A). In strain GA8, the average number of peaks per cell in 15 min was 2.2 ± 0.1 (mean ± SEM). Strain GA8+A8 did not show a statistically significant difference in the frequency from strain GA8 (Figure 6B), even though the Pho8Δ60-dependent ALP activity was clearly higher in the former (Figure 6C), reflecting an increase in the total volume of the autophagosomes. This is consistent with our EM data, indicating that a lower amount of Atg8 did not limit the rate of autophagosome formation but did affect autophagosome volume. In contrast, the frequency was significantly lower in atg27Δ cells (Figure 6B), which are known to produce fewer autophagosomes and served as a control (Yen et al., 2007 ).
In this study, we showed that 1) the amount of Atg8 regulates the level of autophagy by specifically modulating the size of the autophagosomes, whereas the number of autophagosomes is unaffected, 2) each round of autophagosome formation involves a cycle of Atg8 trafficking in which Atg8 is first recruited to an expanding structure and later released from it, and 3) the release of Atg8 is essential for the completion of autophagosome formation and is mediated by deconjugation. In mammalian cells, the autophagosomes produced in response to group A Streptococcus invasion are larger than those seen in starvation conditions (Nakagawa et al., 2004 ). According to those data, where the signal is not saturated, the level of Atg8 induced by group A Streptococcus invasion is higher than that induced by starvation, suggesting that the regulation of autophagosome size achieved by controlling the amount of Atg8 may be a conserved mechanism.
For the first time, our data allowed temporal dissection of the autophagosome formation process based on real-time observations (Figures 3 and and4).4). Here we propose a five-stage model. In the first stage, cargos and factors required for the PAS recruitment of Atg8 arrive at the PAS; in stage 2, Atg8 arrives at the PAS and the Atg8-containing structure expands; stage 3 is a transitional step, allowing the completion of expansion and initiating release of Atg8 through deconjugation; in stage 4, additional Atg8 molecules are gradually released from the PAS; and in stage 5, the phagophore matures into an autophagosome, and some Atg8 molecules are trapped inside. Previously, the lack of a data-derived multistage model restricted most studies on autophagosome formation to the PAS recruitment-dependency of autophagy-related proteins. Our model provides the foundation for reanalysis of the functions of known autophagy-related proteins and for screening new genes whose products act at each stage. In addition, this model serves as a reference point to coordinate the dynamics of other autophagy-related genes. For instance, it would be interesting to find at which stage Atg9, a protein known to cycle between the PAS and peripheral sites, departs from the PAS (Reggiori et al., 2004 ; Young et al., 2006 ).
Our results suggest that the expansion and deformation of the phagophore happens concurrently or slightly after the recruitment of Atg8 to the PAS, given that Atg8 is a causal factor in determination of autophagosome size. When Atg8 is released, the Atg8-containing structures retained their sizes (Figure S2), indicating that at this moment the expansion of the phagophore should be near completion, but not fully closed so that Atg8 molecules attached to the concave side of the phagophore (that will become the inner membrane of the autophagosome) can leave the membrane.
At present, how Atg8 modifies the phagophore to produce different-sized autophagosomes is not clear; however, our data indicate that Atg8 is essential in autophagosome formation. In rare cases (in ~1% of the cells), small vesicular structures can be observed in atg8Δ pep4Δ cells (Supplementary Figure S3). Previously, small autophagosome-like structures have also been detected in atg8Δ cells (Abeliovich et al., 2000 ). Currently, the identity of these structures cannot be determined in the absence of an autophagic membrane marker other than Atg8. In addition, we note that similar rare structures can also be detected in atg1Δ pep4Δ vps4Δ cells, which are defective in autophagosome formation (Supplementary Figure S3). Further experiments are therefore needed to elucidate the nature of these vesicles.
The authors thank Dr. Roger Tsien (University of California, San Diego) for providing the mCherry-URA3 plasmid. This work was supported by National Institutes of Health Public Health Service Grant GM53396 to D.J.K.
This article was published online ahead of print in MBC in Press (http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E07-12-1292) on May 28, 2008.