In vivo reconstitution of autophagy in Saccharomyces cerevisiae

Macroautophagy (hereafter autophagy) is a conserved degradative pathway in eukaryotic cells that also plays an essential role in development and differentiation and has been implicated in many human diseases (Levine and Klionsky, 2004; Mizushima, 2007). Autophagy involves de novo formation of a double-membrane vesicle (the autophagosome) in the cytosol, fusion of this vesicle with the lysosome/vacuole, and the subsequent breakdown and recycling of the cargo. With the identification of autophagy-related (ATG) genes in yeast, autophagy research has shifted from morphological analyses to the understanding of molecular mechanisms (Klionsky, 2007). 31 ATG genes have been identified in baker's and methylotrophic yeasts, and many of them have homologues in higher eukaryotes (Rubinsztein et al., 2007).

Autophagy is generally considered to be nonselective, but there are also selective types of autophagy. The cytoplasm-to-vacuole targeting (Cvt) pathway is an example of selective autophagy wherein the precursor form of the resident vacuolar hydrolase aminopeptidase I (prApe1) is selectively transported from the cytoplasm into the vacuole under vegetative conditions (Klionsky, 2005). The Cvt pathway shares most of the components needed for autophagy but utilizes additional proteins in cargo recognition and packaging. After synthesis in the cytosol, prApe1 forms an oligomer, which is targeted through a receptor, Atg19, and an adaptor, Atg11 (Kim et al., 1997; Scott et al., 2001; Shintani et al., 2002; Shintani and Klionsky, 2004; Yorimitsu and Klionsky, 2005). There is no evidence that the Cvt pathway exists outside of fungi. However, a similar recognition mechanism might be used in mammalian examples of selective autophagy (Iwata et al., 2006).

Most of the Atg proteins at least transiently colocalize at the phagophore assembly site (PAS; the presumed nucleating site of the autophagosome) in both the Cvt pathway and nonselective autophagy; however, this site is poorly defined and characterized. Knowing how Atg proteins are targeted to the PAS will help us dissect the vesicle formation process and understand how these proteins interact and function. The cargo complex including Atg11 is the initial component for PAS assembly in vegetative conditions (Shintani et al., 2002), whereas Atg17 has been shown to be an initial factor for PAS assembly in starvation (Suzuki et al., 2007). Components in the Atg17 complex, such as Atg1 and Atg13, are also involved in organizing the PAS during starvation (Cheong et al., 2008; Kawamata et al., 2008). However, many questions regarding the temporal order of protein assembly at the PAS still remain.

Among the conserved genes, eight participate in two separate but related conjugation systems involving two ubiquitin-like proteins, Atg8 and Atg12. Atg8 is conjugated to phosphatidylethanolamine (PE) through the serial action of Atg4, Atg7, and Atg3 (Ichimura et al., 2000). The C-terminal arginine residue of Atg8 is first removed by the cysteine protease Atg4, exposing a glycine residue that is now accessible to the E1-like activating enzyme Atg7. The activated Atg8 is then transferred to the E2-like conjugating enzyme Atg3 and finally conjugated to PE. Atg8 can be deconjugated from Atg8–PE through a second cleavage by Atg4, although the mechanism that regulates the timing of Atg4 cleavage is not known. The second ubiquitin-like protein, Atg12, is covalently linked to Atg5 through the action of the E1-like Atg7 and an E2-like enzyme, Atg10 (Mizushima et al., 1998). The Atg12–Atg5 conjugate also binds another protein, Atg16, whose oligomerization leads to a tetrameric complex of Atg12–Atg5-Atg16 (Mizushima et al., 1999; Kuma et al., 2002). The Atg8–PE level is reduced in the absence of Atg12, Atg5, or Atg16, and the Atg12–Atg5 conjugate may act as an E3-like enzyme in the Atg8–PE conjugation system (Suzuki et al., 2001; Hanada et al., 2007). However, the efficiency of Atg12–Atg5 conjugation is apparently not affected by the Atg8–PE system (Kuma et al., 2002). Although in vitro reconstitution of both Atg8–PE and Atg12–Atg5 conjugation has been described previously (Ichimura et al., 2004; Hanada et al., 2007; Shao et al., 2007; Fujioka et al., 2008), in vivo reconstitution of the two reactions in eukaryotic cells is not available. Thus, the physiological significance of some of the in vitro observations has not been verified.

Many steps of autophagy have been analyzed in vivo using standard molecular genetics in single- and double-deletion mutant strains, and a smaller subset of the reactions have been reconstituted in vitro. It is unlikely that the entire process will be reconstituted in vitro in the near term because of the complexity of the system and the lack of knowledge regarding basic issues, such as the origin of the sequestering membrane. In addition, in vitro studies inherently lack many critical elements such as the cytoskeleton and the regulatory controls that play important roles in vivo. To provide a physiological reconstitution system, we deleted 24 ATG genes directly required for autophagy and/or the Cvt pathway using a multiple knockout (MKO) strategy. We used this MKO strain to determine the minimum requirements for cargo packaging, initial assembly of the starvation-specific PAS, and conjugation of the ubiquitin-like Atg proteins. Our results with the MKO strain verify the role of particular Atg proteins in cargo packaging, and extend previous studies on PAS assembly and protein conjugation that were based on work with individual or double-deletion strains and/or in vitro studies. With this work, we present a new method to tackle complex cellular pathways, particularly those with extensive regulatory networks such as autophagy that cannot be easily replicated in vitro.

We were able to construct the MKO strain because all the ATG genes directly required for autophagy and the Cvt pathway are nonessential, and the loxP–Cre system allows the use and recycling of markers for MKOs (Fig. 1 A; Gueldener et al., 2002). Among the 31 ATG genes, 24 of them were deleted in strain YCY123 (Table I).ATG15, ATG22, ATG25, ATG26, ATG28, and ATG30 were not deleted because the first two function after autophagosomes fuse with the lysosomal/vacuolar membrane, and the rest are either not found in Saccharomyces cerevisiae or are only required for peroxisome degradation in other methylotrophic yeasts (Cao and Klionsky, 2007). ATG31 was not deleted because we finished generating the MKO strain and the experiments described in this manuscript before it was published (Kabeya et al., 2007). We refer to strain YCY123 as the MKO strain. Additional deletions or the presence of ATG genes are indicated in parentheses.

The MKO strain does not have any obvious unexpected defect other than a slight growth delay (Fig. 1 B). In the MKO strain, we observed one abnormally enlarged vacuole, which is similar to the atg18Δ phenotype (Fig. 1 C; Dove et al., 2004). The normal vacuole morphology was restored when we introduced Atg18 back into the MKO strain (Fig. 1 C). We also tested whether the MKO strain was defective for other vacuolar targeting pathways. Carboxypeptidase Y (Prc1) transits through the ER, Golgi complex, and late endosome–multivesicular body (MVB) for delivery to the vacuole. Carboxypeptidase S (Cps1) is a cargo for the MVB pathway and transits through the early secretory pathway similar to Prc1, but is packaged into lumenal vesicles of the MVB before vacuolar delivery. Vacuolar alkaline phosphatase (Pho8) is transported through the ER to the Golgi, but it bypasses the endosome–MVB (Bowers and Stevens, 2005). By pulse-chase analysis and immunoprecipitation, we followed Prc1, Cps1, and Pho8 processing in the MKO strain. As ATG6 is essential for the carboxypeptidase Y pathway, we transformed a plasmid expressing Atg6 into the MKO and the atg6Δ strains. As shown in Fig. 1 D, although there was a kinetic delay in the MKO strain expressing Atg6, most of the precursor Prc1, Cps1 (not depicted), and Pho8 became mature after 45 min in chase medium. Taking the growth delay of the MKO strain into consideration, these data suggest that the MKO strain expressing Atg6 is not defective for the carboxypeptidase Y, MVB, and alkaline phosphatase vacuolar protein delivery pathways, which are nonautophagic.

We next wanted to determine if the MKO strain faithfully reproduced different steps of autophagy when expressing the appropriate autophagy genes. Previous studies with standard deletion strains have only been able to demonstrate the requirement, but not the sufficiency, for various proteins involved in different steps of autophagy. In the next set of experiments, we reconstituted the cargo packaging step of the Cvt pathway. At the beginning of the cargo packaging process, prApe1 forms a large oligomer, and a bright single punctate structure representing the oligomer can be observed by fluorescence microscopy if the protein is tagged with a fluorophore (Kim et al., 1997; Shintani et al., 2002). The same pattern was also detected in the MKO (atg15Δ RFP-APE1) strain expressing prApe1 tagged with RFP (Fig. 2 A, left). Additional knockout of ATG15 in the MKO strain slightly reduced its growth rate but did not affect Cvt vesicle/autophagosome formation (not depicted). In ape1Δ cells, both the Atg19 receptor and the Atg11 adaptor are dispersed in the cytosol (Yorimitsu and Klionsky, 2005). To determine whether the MKO strain reproduced these phenotypes, we constructed a version lacking APE1 and transformed it with a plasmid expressing either YFP-Atg19 or GFP-Atg11. In the absence of prApe1 (the cargo), YFP-Atg19 was cytosolic (Fig. 2 A, middle), but GFP-Atg11 formed a punctate structure (Fig. 2 A, right). Atg11 is able to interact with itself and several other Atg proteins, and the coiled-coil domains through which Atg11 self-interacts overlap with the interaction sites used by Atg1, Atg17, and Atg20 (Yorimitsu and Klionsky, 2005). Therefore, it is possible that without competition from the other Atg proteins, Atg11 can form a large oligomer by self-interaction in the MKO strain. Analysis of the Atg11 protein by gradient fractionation suggests that it indeed forms a large complex in the MKO strain (unpublished data).

In wild-type cells, Atg19 binds the propeptide of, and colocalizes with, prApe1. Atg11 interacts with Atg19 and targets the complex to the PAS (Scott et al., 2001; Shintani et al., 2002; Yorimitsu and Klionsky, 2005). Thus, there is a temporal order of protein interaction, whereby Atg11 cannot bind the prApe1 complex in the absence of Atg19. To test whether this temporal organization is faithfully retained in the MKO strain, we transformed the MKO (atg15Δ RFP-APE1) strain with plasmids expressing YFP-Atg19, CFP-Atg11, or both. In agreement with the previous model, RFP-Ape1 and YFP-Atg19 colocalized with each other in the absence of other Atg proteins (Fig. 2 B, top). In contrast, without Atg19, RFP-Ape1 did not colocalize with CFP-Atg11 (Fig. 2 B, middle). The three proteins colocalized in a single punctum when all of them were coexpressed (Fig. 2 B, bottom). These data suggest that prApe1, Atg19, and Atg11 are necessary and sufficient for cargo packaging in the Cvt pathway. We note, however, that only ∼50% of the puncta representing the cargo complex (comprised of prApe1, Atg19, and Atg11) were at the PAS (i.e., perivacuolar), which suggests that additional Atg proteins are required to facilitate the targeting of the cargo complex to this site.

The analysis of cargo packaging verified that the MKO strain faithfully replicated the temporal order of Atg protein interactions that have been deduced from previous studies. Next, we decided to use the MKO strain to address a question that would otherwise be extremely difficult to approach. In particular, we wanted to determine the order of assembly of Atg proteins at the PAS. In theory, it might be possible to order the proteins through an extensive analysis of multiple deletion strains, generating a series of epistatic relationships. However, a definitive study is best performed in the complete absence of the other (i.e., those not being examined) Atg proteins. Recent studies suggest that Atg11 and Atg17 are the two initial factors that establish the assembly sequence for Atg proteins at the PAS (Shintani et al., 2002; Suzuki et al., 2007). In particular, Atg11 is critical for PAS assembly during vegetative growth, whereas Atg17 is proposed to act as a scaffold for recruiting other Atg proteins during starvation-specific PAS formation.

To test whether Atg17 itself is sufficient to assemble the initial PAS, we observed the localization of Atg17 tagged with GFP in the MKO strain by fluorescence microscopy. In wild-type cells, Atg17-GFP displayed a clear PAS punctum as well as a diffuse cytosolic signal in both vegetative and starvation conditions; in some cells, we could detect more than one punctum, particularly under starvation conditions (Fig. 3 A). In contrast, Atg17-GFP was largely cytosolic in the MKO (ATG17-GFP) strain (Fig. 3 A, vector), which suggests that Atg17 is not sufficient for the initial PAS assembly even during starvation. Therefore, we decided to determine what additional factors were needed to allow correct localization of Atg17 during autophagy.

Recently, it was found that Atg1 and Atg13 are required along with Atg17 in assembly of the starvation-specific PAS (Cheong et al., 2008). Accordingly, we hypothesized that Atg1, Atg13, and Atg17 are all required for the initial PAS assembly during nonspecific autophagy. To examine this, we transformed plasmids expressing Atg1, Atg13, or both into the MKO (ATG17-GFP) strain and observed localization of Atg17-GFP by fluorescence microscopy. Atg17-GFP showed clear punctate structures under starvation conditions only when Atg1 and Atg13 were expressed together (Fig. 3 A, Atg1 + Atg13). In contrast, expression of either Atg1 or Atg13 alone or in combination with Atg11 did not facilitate Atg17-GFP puncta formation (Fig. 3 A). The same localization pattern was detected with GFP-Atg1; this protein was cytosolic when expressed by itself, and formed punctate structures in starvation conditions only when Atg13 and Atg17 were both present (unpublished data). Furthermore, ∼75% of Atg17-GFP or GFP-Atg1 puncta were perivacuolar (detected with the lipophilic dye FM 4-64), which indicates that Atg1, Atg13, and Atg17 are sufficient for the initial PAS assembly during starvation (Fig. 3 B). All together, these data suggest that Atg17 is not sufficient for starvation-specific PAS formation; rather, Atg1, Atg13, and Atg17 all appear to have an initial role in PAS assembly during starvation.

Finally, we quantified the percentage of cells containing Atg17-GFP puncta in the MKO strain expressing different Atg proteins. Approximately 40% of the cells coexpressing Atg1 and Atg13 showed Atg17-GFP puncta under starvation conditions, whereas <5% of the cells displayed any punctate structure when Atg17-GFP was expressed by itself (Fig. 3 C). Additional expression of other components in the Atg17 complex further increased the percentage of cells displaying Atg17-GFP puncta. For example, Atg29 by itself did not significantly affect Atg17-GFP puncta formation, but expression of Atg29 in the presence of Atg1 and Atg13 increased the percentage of cells with puncta to ∼53% during starvation. Atg11, when expressed together with Atg1 and Atg13, resulted in an even greater increase to ∼80% (Fig. 3 C). Collectively, these results indicate that starvation-specific PAS formation initially requires a minimal set of Atg proteins consisting of Atg1, Atg13, and Atg17; other components in the Atg1 complex, such as Atg29 and Atg11, further enhance assembly.

Having established that the MKO strain reproduces the temporal order of action of components involved in cargo packaging in growing conditions, and determined the initial factors for PAS assembly during starvation, we decided to extend our analysis of the utility of this system. In particular, we chose to examine the Atg12–Atg5 conjugation system because this part of the autophagy process has been reconstituted in vitro. Therefore, components have been identified that are both necessary and sufficient for the reaction. The question we posed was whether the information gained from these previous experiments accurately reflected the complete in vivo situation. Previous data show that Atg5, Atg7, Atg10, and Atg12 are essential for the conjugation (Mizushima et al., 1998), and Atg16 is required for the efficiency of this reaction (Mizushima et al., 1999), but Atg3, Atg4, and Atg8 from the Atg8–PE system are not involved in Atg12–Atg5 conjugation (Kuma et al., 2002). Accordingly, we expressed various combinations of these proteins in the MKO strain. We used 3× HA-tagged Atg12 to allow detection of the Atg12–Atg5 conjugate. When expressed in an atg12Δ strain, we detected both free HA-Atg12 and HA-Atg12 conjugated to Atg5 (Fig. 4, lane 2). As expected, when only Atg5 and HA-Atg12 were expressed in the MKO strain, we did not detect the conjugated species (Fig. 4, lane 4). In contrast, if Atg7 and Atg10 were also present, we detected a faint band corresponding to the HA-Atg12–Atg5 conjugate; however, the majority of the HA-Atg12 was in the free form (Fig. 4, lane 5). When we added Atg16, we found that a substantially greater amount of conjugate was formed (Fig. 4, lane 6).

Next, we examined possible contributions from the Atg8 conjugation system. When we expressed Atg4 and Atg8 together with Atg5, Atg7, Atg10, Atg12, and Atg16, the Atg12–Atg5 conjugation efficiency was further enhanced (Fig. 4, lane 7). Atg8–PE was not generated in this strain because of the absence of Atg3 (not depicted). Thus, it seems that even without the formation of Atg8–PE, the presence of some proteins in the Atg8–PE system can facilitate the conjugation of Atg12 to Atg5. Furthermore, the level of the Atg12–Atg5 conjugate was even greater when Atg3 was added (Fig. 4, lane 8). Thus, our data suggest that Atg5, Atg7, Atg10, and Atg12 are minimum requirements for Atg12–Atg5 conjugate formation, and Atg16 facilitates the conjugation, in agreement with published data. We further found, however, that components in the Atg8–PE system also enhance the conjugation reaction. The same conclusions were reached in either vegetative (Fig. 4) or starvation (not depicted) conditions.

We extended our analysis with the MKO strain to the reconstitution of Atg8–PE conjugation. Although there have been extensive studies on the in vitro reconstitution of Atg8–PE, in vivo reconstitution has only been described in Escherichia coli where autophagy does not occur (Ichimura et al., 2004). The C-terminal arginine of nascent Atg8 is normally removed by Atg4, which can also release Atg8 from PE. When the C-terminal arginine is removed, as in Atg8ΔR, the initial processing step is bypassed. Previous experiments using purified Atg8ΔR, Atg3, Atg7, and PE-containing liposomes suggest that Atg8–PE formation is at its peak when the PE content reaches 70% of the total membrane, but with a concentration close to that of yeast organelle membranes, the efficiency is very low (Ichimura et al., 2004). However, the efficiency of conjugation is boosted by the presence of the Atg12–Atg5 conjugate, and further addition of Atg16 only slightly increases Atg8–PE formation (Hanada et al., 2007). Our data support this current view of Atg8 lipidation. Moreover, they provided additional insights into both the conjugation and deconjugation reactions.

We first tested if Atg8, Atg4, Atg7, and Atg3 are sufficient for Atg8–PE formation. When we transformed a plasmid expressing these proteins into the MKO (ATG3) strain, Atg8–PE was essentially not detected in both growing and starvation conditions (Fig. 5 A, lane 3); even when we expressed all the components for the Atg12–Atg5 conjugation system, we were still not able to detect a signal for Atg8–PE (Fig. 5 A, lane 4). These data do not agree with the in vitro reconstitution data. One of the differences between these previous studies and our experiments is that the previous analyses used the modified Atg8ΔR to bypass the need for the initial Atg4 cleavage, and omitted Atg4 from the subsequent reactions. However, the absence of Atg4 also blocks the deconjugation of Atg8–PE and thus favors the accumulation of the conjugated species. Considering the role of Atg4 in determining the balance between Atg8 and Atg8–PE, we introduced a plasmid expressing Atg8ΔR, Atg7, and Atg10, into the MKO (ATG3) strain; in this case, a significant amount of Atg8–PE was detected (Fig. 5 A, lane 5). When we additionally expressed Atg5, HA-Atg12, and Atg16, almost all of the Atg8 was in its conjugated form (Fig. 5 A, lane 6). Thus, when Atg4 is present, as is the case in vivo, Atg8–PE levels are extremely low in the MKO strain, even in the presence of the Atg12–Atg5 conjugation system. These results suggest that the in vitro system does not faithfully reflect the contribution of other factors that are required to control the activity of Atg4.

Finally, we attempted to further define the reason for the low level of Atg8–PE in this system. In wild-type cells, a population of Atg8 is constantly degraded inside the vacuole, so the total amount of Atg8 is lower than that in the MKO strain. Therefore, we used a pep4Δ strain, which blocks Atg8 degradation, as a control (Fig. 5 B, lane 11). When deconjugation occurs, an Atg8–PE band was not detected (Fig. 5 B, lane 3), whereas expression of Atg8ΔR in an atg4Δ atg8Δ strain defective in deconjugation causes all of the protein to accumulate in the lipidated form (Fig. 5 B, lane 10). When we expressed Atg8ΔR, Atg7, and Atg10 along with Atg5, Atg12, or Atg16 separately, Atg8–PE conjugation was not increased (Fig. 5 B, lane 4–7). However, the presence of the Atg12–Atg5 conjugate facilitated Atg8–PE formation (Fig. 5 B, lane 8; and not depicted). In this in vivo system, Atg16 also facilitated the formation and/or enhanced the stability of Atg8–PE (Fig. 5 B, lane 9), which could be indirect through its action on the Atg12–Atg5 conjugate (Fig. 4, lane 6).

In this study, we created and validated an MKO strain in which 24 ATG genes directly required for autophagy and the Cvt pathway were deleted (Fig. 1). Using this in vivo system, we determined minimum requirements for cargo packaging (Fig. 2), initial starvation-specific PAS assembly (Fig. 3), Atg12–Atg5 conjugation (Fig. 4), and Atg8–PE formation (Fig. 5). Although single- and double-deletion analyses and in vitro reconstitution have provided us with important insights into mechanisms, in general the minimum requirements for, and regulation of, these steps have not been explored extensively. The MKO strain serves as a valuable and powerful tool for these purposes. Moreover, it also has an obvious advantage over an in vitro reconstitution system, more accurately recapitulating what occurs in vivo and making studies of complex steps/aspects of autophagy possible and more efficient.

Our in vivo reconstitution of the Atg12–Atg5 and Atg8–PE conjugation systems demonstrated the usefulness of the MKO strain, as it provided us with information missing from the in vitro studies. For example, our data suggest that components in the Atg8–PE system contribute to the efficiency of Atg12–Atg5 conjugation (Fig. 4). With regard to the Atg8–PE conjugation system, the results are even more intriguing. We find that when Atg4 is present to cleave the C-terminal arginine of Atg8 and to deconjugate Atg8 from PE, a situation that is more physiologically relevant than the conditions used for the in vitro reconstitution studies; even all the proteins in the two conjugation systems cannot sustain a detectable amount of Atg8–PE (Fig. 5 A), in contrast to wild-type cells. This means that either deconjugation occurs too quickly, conjugation efficiency is still somehow compromised in the MKO strain, or both. One possibility is that the plasmid-based Atg4 in the MKO strain results in overexpression and elevated Atg4 activity that favors deconjugation of Atg8–PE. Because we do not have antibodies against the Atg4 protein, we cannot directly determine the Atg4 levels. To address this possibility, however, we examined the functionality of our Atg4-expressing plasmid in an atg4Δ strain. We found that expression of Atg4 in the atg4Δ strain complemented prApe1 maturation, and this strain did not accumulate unconjugated Atg8 (unpublished data), which suggests that our plasmid expressing Atg4 is functional, and that additional factors are needed to regulate Atg4's protease activity or deconjugation function in the MKO strain. Regulation of Atg4 is an important topic for autophagy. For example, why is Atg8 synthesized with residues after the C-terminal glycine when these residues are immediately removed by Atg4? Also, how is Atg4 regulated to prevent premature deconjugation of Atg8–PE? One recent study suggests that Atg4 activity is subject to redox regulation (Scherz-Shouval et al., 2007), although the actual mechanism through which this might occur in vivo is not known. Follow-up analyses will yield more insights into the regulation of deconjugation and conjugation.

With the MKO strain, we can study the order of protein assembly at the PAS more directly and efficiently compared with analyses with single- or double-deletion strains. Previous studies suggest that Atg11 and Atg17 are the initial factors for PAS assembly in vegetative and starvation conditions, respectively (Shintani et al., 2002; Suzuki et al., 2007). However, it is possible that some proteins in the autophagy pathway may act together or be able to substitute for one another in order to target Atg11 and Atg17 to the PAS. Our data suggest that in vegetative conditions, Atg11 is not sufficient to localize itself and the cargo complex to the PAS even though the cargo complex forms, which means that an additional protein is needed for efficient assembly of the Cvt-specific PAS. We also provided direct evidence that not only Atg17 but also Atg1 and Atg13 are important for the initial starvation-specific PAS assembly (Fig. 3 A). Approximately 75% of the Atg17-GFP puncta observed in the presence of Atg1 and Atg13 were perivacuolar (Fig. 3 B), which suggests that these three proteins are sufficient for this process. We also noted that MKO cells expressing Atg1, Atg13, and Atg17-GFP tend to have more punctate structures than the wild-type cells. Some of the puncta might be precursor or intermediate structures of the initial PAS. In addition to Atg1, Atg13, and Atg17, other proteins such as Atg29 and Atg11 also contribute to the efficiency of the assembly process (Fig. 3 C). During the course of our studies, a paper was published verifying a role for Atg29 in the assembly of the starvation-specific PAS (Kawamata et al., 2008). However, there are some discrepancies between the two sets of studies. For example, Kawamata et al. (2008) found that Atg29 is essential for assembly of the starvation-specific PAS, whereas our present studies indicate that this protein enhances PAS assembly but is not essential. Further analyses will be needed to resolve this and other questions pertaining to this topic. Although Atg11 is a Cvt-specific factor, it contributed to PAS assembly along with Atg1, Atg13, and Atg17 during starvation conditions (Fig. 3 C). Generation of the Cvt-specific PAS that is dependent on Atg11 during vegetative growth may help to recruit components in the Atg17 complex, resulting in enhanced PAS assembly when cells are subsequently shifted to starvation conditions (Cheong et al., 2008).

In vitro systems can be extremely useful and powerful tools for dissecting molecular mechanisms, and the present study does not suggest otherwise. Rather, we demonstrate that in vivo reconstitution can complement and extend in vitro analyses. Generation of a strain with 24 ATG genes knocked out is a unique and novel approach for studying complex pathways such as autophagy, and the strain will be a very useful tool for autophagy research. We can foresee potential uses of the strain in determining the temporal order of protein arrival at the PAS, in Atg protein retrieval after autophagosome formation, and many other aspects of autophagy. These experiments will reward us with new findings and open up areas for potential research.

The yeast S. cerevisiae strains used in this study are listed in Table I. Knockout strains were constructed using the loxP–Cre system (Gueldener et al., 2002). GFP tagging of Atg17 at the corresponding chromosomal locus was performed by a PCR-based procedure (Longtine et al., 1998). Yeast media including YPD (1% yeast extract, 2% peptone, and 2% dextrose) or synthetic minimal medium with casamino acids (SMD + CA) and SD-N for nitrogen starvation have been described previously (Shintani et al., 2002). FM 4-64 and antibodies to alkaline phosphatase (Pho8) and carboxypeptidase Y (Prc1) were obtained from Invitrogen. Antisera against Ape1 (Klionsky et al., 1992) and Atg8 (Huang et al., 2000) have been described previously. Antibody to the hemagglutinin epitope was obtained from Santa Cruz Biotechnology, Inc. Antisera against Cps1 and phosphoglycerate kinase (Pgk1) were gifts from S. Emr (Cornell University, Ithaca, NY) and J. Thorner (University of California, Berkeley, Berkeley, CA), respectively.

The MKO strains were derived from wild-type haploid strain SEY6210 (Robinson et al., 1988), and the loxP–Cre system (Gueldener et al., 2002) was modified for multiple gene knockouts (Fig. 1 A). The gene disruption cassette consisting of a selection marker flanked by two loxP sequences was amplified by PCR using primers containing 60 nucleotide stretches that are homologous to sequences upstream of the start codon or downstream of the stop codon of the target gene, followed by 19 or 22 nucleotides complementary to sequences flanking the loxP sites in the marker plasmid. The PCR products were transformed into yeast cells, and transformants were selected on plates lacking a particular amino acid or containing a specific antibiotic. Selected transformants were checked for correct integration by colony PCR using two sets of primers. ATG genes were deleted one at a time using markers including kanMX, ble^r, HIS5, URA3, and LEU2. A maximum of five ATG genes were deleted before Cre recombinase was expressed to loop out the markers. Replica plating was used to identify colonies that had lost all of the markers. The markers can thus be reused to delete other ATG genes. After several rounds of deletion and marker rescue, 24 ATG genes were deleted. The strain was named YCY123 (MKO).

For pATG4(414), pATG7(414), pATG10(414), pATG5(416), and pATG16(416), the individual gene with its endogenous promoter and terminator region was amplified from wild-type genomic DNA, digested with SacI, XmaI, KpnI, BamHI, and NotI, respectively, and cloned into the corresponding site of pRS414 or pRS416 (Sikorski and Hieter, 1989). SacI and BamHI sites were used to clone pATG6(414), NotI and NcoI for pATG8(414), SacI and PstI for pATG1(415), and XbaI and XmaI for pATG13(415). For pHA-ATG12(416), the insert was amplified from 3× HA-APG12 (Mizushima et al., 1998) and cloned into pRS416. The functionality of these plasmids was tested in the corresponding knockout strains by examining prApe1 maturation.

The promoter, open reading frame, and terminator inserts were subsequently removed by restriction digestion and cloned in various combinations to generate the multigene plasmids pATG1-ATG13(415), pATG7-ATG10(414), pATG5-HA-ATG12(416), pATG5-HA-ATG12-ATG16(416), and pATG8-ATG4-ATG7-ATG10(414). The key in the design was that the sites for cloning each subsequent gene were not present within the existing plasmid. A one-step site-directed mutagenesis protocol (Zheng et al., 2004) was used to remove the last arginine from the ATG8 open reading frame to generate pATG8ΔR(414). The NotI- and NcoI-digested Atg8ΔR fragment was exchanged with the Atg8 fragment to generate pATG8ΔR-ATG4-ATG7-ATG10(414), from which Atg4 was removed by BglII and SacII digestion followed by blunt-ending and religation to form pATG8ΔR-ATG7-ATG10(414). Plasmids pRFPAPE1(305) (Stromhaug et al., 2004), pGFPATG11(416) (pGFPCVT9(416); Kim et al., 2001), pYFPATG19(416) (pYFP-Cvt19; Shintani et al., 2002), pCuHACFPATG11(414) (Yorimitsu and Klionsky, 2005), and pATG18(415) (pCVT18(415); Guan et al., 2001) have been described previously.

5 ml of cells in mid-log phase grown in SMD + CA medium were collected and labeled in 500 μl SMD with 20 μCi ³⁵S for 5 min at 30°C. Chase medium (5 ml YPD containing 2 mM cysteine and 1 mM methionine) was added, and at each time point, 1 ml of cells were collected, precipitated with TCA, washed with acetone, dried, and resuspended in 100 μl MURB (50 mM sodium phosphate, 25 mM MES, pH 7.0, 1% SDS, 3 M urea, and 0.5% β-mercaptoethanol). Glass beads were added to break the cells, and then each sample was boiled and diluted in 800 μl TWIP (0.5% Tween 20, 50 mM Tris hydrochloride, pH 7.5, 150 mM NaC1, and 0.1 mM EDTA). The supernatant fraction was subjected to triple immunoprecipitation: the first precipitation was performed with 3 μl Prc1 antiserum, and the supernatant after removal of the antibody–antigen complex with protein A–Sepharose was immunoprecipitated with 1 μl Cps1 antiserum, and then with 6 μl Pho8 antiserum; the second and third immunoprecipitations were performed twice each to eliminate carryover of the previous antisera.

Yeast cells were grown at 30°C to mid-log phase in the appropriate medium, harvested, resuspended with 10% TCA, and held for 30 min on ice. Cell pellets were washed with 100% acetone, air-dried, and resuspended in MURB buffer. Cells were further broken by vortexing with glass beads and then heated at 75°C for 10 min. For each sample, 0.2 OD₆₀₀ units of cells were resolved by SDS-PAGE, transferred to a polyvinylidene fluoride membrane, and probed with appropriate antisera or antibodies.

Yeast cells were grown to mid-log phase in selective SMD medium. To label the vacuolar membrane, cells were stained with 0.8 mM FM 4-64 for 15 min, washed, and incubated in the same medium for 30 min. For starvation, cells were shifted to SD-N medium for 2 h. 1 ml of cells were pelleted and resuspended in 30 μl of the same medium before imaging. Microscopy was performed at room temperature on a microscope (IX71; Olympus) with a 100× Universal Plan Apochromat objective lens (Olympus) and a camera (CoolSnap HQ; Photometrics). The microscope was controlled by a workstation (DeltaVision Spectris; Applied Precision, LLC). Images were acquired and deconvolved with DeltaVision softWoRx software (Applied Precision, LLC).

This work was supported by National Institutes of Health Public Health Service grant GM53396 to D.J. Klionsky.