Autophagy is a conserved degradative pathway that is induced in response to various stress and developmental conditions in eukaryotic cells. It allows the elimination of cytosolic proteins and organelles in the lysosome/vacuole. In the yeast Saccharomyces cerevisiae, the integral membrane protein Atg9 (autophagy-related protein 9) cycles between mitochondria and the preautophagosomal structure (PAS), the nucleating site for formation of the sequestering vesicle, suggesting a role in supplying membrane for vesicle formation and/or expansion during autophagy. To better understand the mechanisms involved in Atg9 cycling, we performed a yeast two-hybrid–based screen and identified a peripheral membrane protein, Atg11, that interacts with Atg9. We show that Atg11 governs Atg9 cycling through the PAS during specific autophagy. We also demonstrate that the integrity of the actin cytoskeleton is essential for correct targeting of Atg11 to the PAS. We propose that a pool of Atg11 mediates the anterograde transport of Atg9 to the PAS that is dependent on the actin cytoskeleton during yeast vegetative growth.
As one of the major degradative mechanisms conserved among eukaryotic cells, autophagy mediates the turnover and recycling of long-lived cytosolic proteins and excess or damaged organelles (Klionsky, 2005). The cargo destined for autophagic degradation is sequestered in a double-membrane vesicle called an autophagosome, which fuses with the lysosome in mammalian cells or the vacuole in yeast. Eventually, the cargo is degraded by lysosomal/vacuolar resident hydrolases. Autophagy occurs in response to physiological stress or developmental signals (Levine and Klionsky, 2004). Recently, autophagy has been implicated in a variety of human diseases, including cancer, neurodegeneration, and pathogen infection (Shintani and Klionsky, 2004a). The initial identification of >20 autophagy-related (ATG) genes in the budding yeast Saccharomyces cerevisiae has highlighted this single-cell organism as a perfect model to study the molecular mechanism of autophagy, although orthologues of some yeast ATG genes have been found in higher eukaryotes.
In yeast, autophagy can be induced under starvation conditions to reuse nutrients for essential cellular activities and proper cellular remodeling; this starvation-induced bulk autophagy is considered nonspecific. Studies in yeast have also revealed that S. cerevisiae has selective autophagic pathways that target specific cargos (Scott et al., 1996; Hutchins et al., 1999). These pathways mechanistically and genetically resemble bulk autophagy. One such route is the cytoplasm to vacuole targeting (Cvt) pathway (Nair and Klionsky, 2005). In this pathway, two vacuolar hydrolases, the precursor form of aminopeptidase I (Ape1 [prApe1]) and α-mannosidase, are transported to the vacuole in a double-membrane vesicle called a Cvt vesicle, with the former subsequently being processed into mature Ape1. Compared with starvation-induced autophagy, the Cvt pathway occurs constitutively in growing conditions. Although some Atg proteins appear to be pathway specific, most are involved in both the specific and nonspecific pathways. However, it is not fully understood how these proteins coordinate and function at the molecular level in either bulk or selective autophagy.
Most yeast Atg components localize at a perivacuolar punctate structure called the preautophagosomal structure (PAS) or phagophore assembly site, which is proposed to be the site of autophagosome and Cvt vesicle formation (Kim et al., 2002; Noda et al., 2002). In most endomembrane trafficking systems, such as the early secretory pathway, vesicles form by budding from the surface of a preexisting organelle. However, in autophagy-related processes, the double-membrane sequestering vesicles appear to form de novo; that is, they expand by membrane addition during the formation process rather than being generated from a single piece of contiguous membrane (Reggiori and Klionsky, 2005). One of the major current challenges is to unveil where the membrane materials for autophagosomes or Cvt vesicles come from and how the lipids are transported to the assembly site. Among all Atg proteins, Atg9 is the best candidate that can help us understand this pivotal issue. Atg9 is the only characterized integral membrane protein required for both autophagosome and Cvt vesicle formation (Noda et al., 2000). However, this protein is absent from the completed vesicles, suggesting that it is retrieved before the vesicle sealing/completion step. Atg9 localizes to multiple punctate sites, with one of them corresponding to the PAS and others to mitochondria in addition to unidentified structures (Reggiori et al., 2005b). Recent studies reveal that Atg9 cycles between mitochondria and the PAS vesicle assembly site (Reggiori et al., 2004, 2005b). These characteristics make Atg9 a potential membrane carrier for vesicle formation.
We decided to investigate the molecular regulatory mechanisms underlying Atg9 cycling and, in particular, what factors regulate the anterograde transport of Atg9 to the site of vesicle formation. In this study, we discovered that a peripheral membrane protein, Atg11 (Kim et al., 2001), is an interaction partner of Atg9. The interaction requires the second coiled-coil (CC) domain of Atg11 and the Atg9 N-terminal cytosolic domain. A missense mutation (H192L) in the Atg9 N-terminal domain that disrupts its interaction with Atg11 results in the impaired cycling of Atg9 and a defect in selective autophagy. In addition, we found that in actin mutant cells, Atg11 colocalized with Atg9 and was retained on mitochondria, indicating that Atg11 is not able to direct Atg9 to the PAS in the absence of an intact cytoskeletal network. These data support a model in which a pool of Atg11 links Atg9 to the PAS along the actin cable under vegetative growth conditions.
Atg11 is an interaction partner of Atg9
Atg9 is the only known transmembrane protein required for both bulk autophagy and selective autophagic processes (e.g., the Cvt pathway; Noda et al., 2000). Unlike most other Atg proteins, which are restricted to the perivacuolar PAS, Atg9 localizes to several punctate structures; one of them is at the PAS, whereas the others are primarily confined to mitochondria (Reggiori et al., 2005b). Atg9 cycles between the two compartments, suggesting that it plays a role in providing lipids to the forming autophagosomes or Cvt vesicles (Reggiori et al., 2004). However, other than actin (Reggiori et al., 2005a), the factors that regulate the anterograde transport of Atg9 to the PAS have not been identified. Therefore, we performed a yeast two-hybrid–based screen of Atg proteins to identify potential Atg9-interacting proteins. Yeast two-hybrid cells harboring Atg9 and a peripheral membrane protein, Atg11, showed robust growth on plates lacking histidine, indicating that Atg11 could interact with Atg9 (Fig. 1 A). The same result was obtained on plates lacking adenine (unpublished data).
Atg11 functions in selective types of autophagy (i.e., the Cvt pathway and pexophagy) but is not essential for bulk autophagy (Kim et al., 2001). Atg11 plays a role in organizing the PAS and linking cargo to the vesicle-forming machinery at the PAS (Shintani et al., 2002; Yorimitsu and Klionsky, 2005). This protein is a component of at least two complexes in yeast. One is the Atg1–Atg11 complex, which is involved in the induction of bulk and selective autophagy (Kamada et al., 2000). The other is the Atg19–Atg11 complex, which recognizes and delivers prApe1 and α-mannosidase to the PAS (Shintani et al., 2002; Yorimitsu and Klionsky, 2005). We decided to determine whether these complexes are involved in Atg11 and Atg9 interaction. To address this issue, we used a biochemical approach to examine whether Atg11 was able to form a complex with Atg9. We tagged Atg11 with GFP and tagged Atg9 with protein A (PA) at the chromosomal locus. Wild-type, atg1Δ, and atg19Δ cells expressing the integrated Atg9-PA and Atg11-GFP fusions were lysed, and the PA-tagged protein was isolated with IgG–Sepharose beads. Atg11 was coprecipitated with Atg9-PA in all three strains (Fig. 1 B), which verifies that these two proteins are present in a complex, although we do not know whether they interact directly. Thus, the absence of either Atg1 or Atg19 did not affect the formation of a complex between Atg9 and Atg11. This finding suggests that there might be multiple populations of Atg11 within the cell that interact with different sets of Atg proteins.
Our recent data show that two other Atg proteins, Atg23 and Atg27, interact with Atg9 and are required for Atg9 cycling. The interaction between Atg9 and either Atg23 or Atg27 is not mediated through Atg11 (Tucker et al., 2003; and unpublished data). Thus, we extended the analysis by determining whether the Atg9–Atg11 interaction was dependent on these other Atg9-interacting proteins. We found that Atg11 was coprecipitated with Atg9 in atg23Δ and atg23Δ atg27Δ cells despite a lower overall efficiency of recovery (Fig. 1 C). This suggests that Atg9 and Atg11 were able to form a complex in the absence of Atg23 and Atg27, although these two proteins may facilitate the interaction. The interaction between Atg9 and Atg11 in the atg23Δ or atg27Δ strains was confirmed by yeast two-hybrid analyses; the atg23Δ or atg27Δ two-hybrid cells expressing Atg9 and Atg11 were able to grow on −histidine (−His) selective plates, which is comparable with the wild-type cells (unpublished data).
Atg11 is predicted to contain four CC domains (Fig. 2 A; Yorimitsu and Klionsky, 2005). Each CC domain mediates interactions of Atg11 with different Atg proteins. To test whether these CC domains are responsible for the interaction between Atg11 and Atg9, we used CC domain deletion mutants in a series of yeast two-hybrid assays. As shown in Fig. 2 B, the two-hybrid mutant activation domain (AD)–Atg11 C-terminal truncation (11N; Δ627–1,178 and lacking CC3-4) allowed the cells to grow in the presence of binding domain (BD)–Atg9 as well as the full-length AD-Atg11. In contrast, cells expressing an AD-Atg11 N-terminal truncation (11C; Δ1–817 and lacking CC1-3) could not grow on selective −His plates (Fig. 2 B). Thus, the Atg11 N terminus is sufficient for Atg9–Atg11 interaction. In addition, Atg11ΔCC2 (Δ536–576) abolished the interaction with Atg9, whereas cells containing mutants Atg11ΔCC1 (Δ272–321), Atg11ΔCC3 (Δ627–858), or Atg11ΔCC4 (Δ859–1,178) grew well on selective plates (Fig. 2 C). These data indicated that Atg11 CC domain 2 is required for the Atg9–Atg11 interaction.
From a hydropathy plot analysis, Atg9 contains hydrophilic N and C termini flanking six to eight transmembrane domains (Noda et al., 2000). To define the topology of Atg9, we investigated the protease sensitivity of the Atg9 N and C termini. Spheroplasts derived from pep4Δ cells expressing either the N-terminal–tagged PA-Atg9 fusion or the C-terminal–tagged Atg9-PA fusion were osmotically lysed and centrifuged at 13,000 g. In agreement with previous studies, approximately two thirds of the total Atg9 was present in the S13 supernatant fraction that contained the PAS, and one third was present in the P13 pellet fraction (Fig. 2 D; Noda et al., 2000; Reggiori et al., 2005b). When the P13 fraction was treated with exogenous proteinase K, both the N- and C-terminal PA tags were cleaved in the absence or presence of detergent, and no bands of smaller molecular mass were detected (Fig. 2 D), indicating that both the N and C termini of Atg9 were accessible to protease on the cytosolic side of the membrane. To verify that intracellular membranous structures were intact after osmotic lysis, we simultaneously monitored the protease sensitivity of an endogenous vacuole membrane protein, Pho8. The precursor form of Pho8 that accumulated in the pep4Δ background contains a small cytosolic tail and a lumenally oriented propeptide (Fig. 2 D). In the absence of detergent, only the Pho8 cytosolic tail but not the lumenal propeptide was accessible to proteinase K. Upon the addition of both detergent and proteinase K, the Pho8 lumenal propeptide was removed as a result of the disruption of all membranous compartments, which is shown as a further shift of the molecular mass (Fig. 2 D). Thus, these data verified the integrity of the relatively fragile vacuole and presumably other intracellular membranous compartments after osmotic lysis, suggesting that both the N and C termini of Atg9 are exposed to the cytosol (Fig. 2 D). Accordingly, the topology of Atg9 appears to be conserved between yeast and mammalian cells (Young et al., 2006).
So far, no known functional domains have been identified in the Atg9 N- or C- terminal regions. To further analyze the Atg9–Atg11 interaction, we generated truncated Atg9 mutants containing the N terminus, C terminus, or transmembrane region. As shown in Fig. 2 E, in the presence of Atg11, the Atg9 N-terminal domain supported the growth of two-hybrid cells as well as the full-length Atg9, whereas neither the C-terminal nor the transmembrane domains of Atg9 were able to do so (Fig. 2 E). This result showed that Atg11 interacts with the N-terminal region of Atg9. We further constructed a series of Atg9 N-terminal truncation mutants and analyzed them for interaction with Atg11 by yeast two-hybrid analysis. Atg9 N-terminal amino acids 159–255 appeared to be the minimal region that mediates the interaction with Atg11 (Fig. 2 F). Collectively, we concluded that Atg11 and Atg9 interact through the Atg11 CC domain 2 and the Atg9 N terminus.
Atg11 recruits Atg9 to the PAS
In wild-type cells, Atg11 localizes at the PAS (Kim et al., 2001), and Atg9 cycles between the PAS and mitochondria (Reggiori et al., 2004, 2005b). To examine the role of the interaction between Atg9 and Atg11, we analyzed the cycling of Atg9 in the presence of overexpressed Atg11 or in the absence of this protein. The chromosomally tagged Atg9-YFP chimera displayed a multiple punctate distribution in wild-type cells, with one of the puncta colocalizing with the PAS marker blue fluorescent protein (BFP)–Ape1 (Fig. 3 A). In contrast, in the atg1Δ background, Atg9 was restricted to the PAS, which is in agreement with previous observations that indicated a role for Atg1 in the retrograde transport of Atg9 from the PAS to mitochondria (Reggiori et al., 2004, 2005a,b). In 91% (109/120) of the cells overexpressing Atg11, Atg9-YFP localized solely to the perivacuolar PAS (represented by CFP-Atg11), which is similar to the situation observed in atg1Δ cells (Fig. 3 A; Kim et al., 2002), and did not localize to mitochondria (not depicted). Thus, excess Atg11 was able to restrict Atg9 to the PAS. Overexpressed Atg11 displays a dominant-negative phenotype, interfering with the vacuolar import of prApe1 through the Cvt pathway (unpublished data). The dominant-negative phenotype presumably reflects the defect in Atg9 cycling.
To further examine the role of Atg11 in recruiting Atg9 to the PAS, we used the TAKA (transport of Atg9 after knocking out ATG1) assay (Cheong et al., 2005). This assay examines the epistasis of a second mutation relative to atg1Δ with regard to Atg9 localization at the PAS. We visualized chromosomally tagged Atg9-YFP in atg11Δ single and atg1Δ atg11Δ double deletion cells and simultaneously labeled the vacuolar membrane with FM 4-64. As shown previously, in atg1Δ cells, Atg9-YFP localized to a single perivacuolar punctum, which corresponds to the PAS (Fig. 3 B). In contrast, in nearly 90% of the atg11Δ or atg1Δ atg11Δ mutants, Atg9-YFP fluorescence showed multiple puncta and did not localize at a single perivacuolar structure, suggesting that its anterograde transport was blocked because of the ATG11 deletion (Fig. 3 B). These results suggest that Atg11 is involved in the anterograde transport of Atg9.
Our mapping of the Atg9 and Atg11 interaction domains indicated that Atg11ΔCC2 but not Atg11 lacking its other CC domains was defective in forming a complex with Atg9 (Fig. 2). Accordingly, we examined the effect of Atg11 CC deletions on Atg9 subcellular distribution. As shown in Fig. 3 C, in cells expressing Atg11ΔCC1 in the atg1Δ background, Atg9 localized to the PAS, which is similar to the result seen in atg1Δ cells expressing wild-type Atg11. In contrast, in cells expressing Atg11ΔCC2, which lacks the Atg9-interacting domain, Atg9 displayed a multiple punctate distribution (Fig. 3 C) resembling that observed in atg1Δ atg11Δ cells even though the mutant protein was expressed at a level similar to the wild-type protein (Yorimitsu and Klionsky, 2005). Thus, we concluded that the interaction of Atg9 with Atg11 CC domain 2 is required to direct Atg9 to the PAS.
The Atg9H192L mutant is defective for the Cvt pathway
To further clarify the physiological functions of the Atg9–Atg11 interaction, we decided to test whether autophagic processes were affected by its disruption. It has been reported that Atg11 CC domain 2, the Atg9-interacting domain, interacts with multiple Atg proteins, including at least one (Atg1) that is involved in Atg9 retrograde transport (Yorimitsu and Klionsky, 2005). Thus, Atg11ΔCC2 may cause pleiotropic effects when used in functional studies. To bypass this problem, we decided to use native Atg11 and instead to isolate mutations within the Atg9 N-terminal region that disrupt the interaction with Atg11. We performed a PCR-based random mutagenesis on the N-terminal region of Atg9 followed by a yeast two-hybrid screen for loss-of-interaction mutants. A missense mutation was identified with a single histidine to leucine substitution at position 192 (H192L), which is located in the minimal region (amino acids 159–255) needed for the Atg9–Atg11 interaction (Fig. 2 E). Two-hybrid cells harboring this mutation completely lost the capacity for growth on selective −adenine (−Ade) plates in the presence of Atg11 (Fig. 4 A), whereas the expression level of the mutant protein was comparable with that of the wild type (not depicted). This indicates that the interaction between Atg9 and Atg11 was abolished by the H192L mutation. The loss of Atg9–Atg11 interaction was confirmed by a coimmunoprecipitation assay. As shown in Fig. 4 B, endogenous Atg11 was recovered only with wild-type Atg9 but not with the Atg9H192L mutant. In contrast, yeast two-hybrid data indicated that the interaction between Atg9 and either Atg2 or Atg18, two proteins involved in the retrograde transport of Atg9, was unaffected (unpublished data).
Because both ATG9 and ATG11 are essential genes for the Cvt pathway, which is a type of selective autophagy, we monitored the processing of prApe1 as a marker protein for this transport route. As shown in Fig. 4 C, cells expressing Atg9H192L did not generate the mature form of Ape1 compared with cells expressing wild-type Atg9, indicating that the Cvt pathway was impaired by the Atg9 H192L mutation.
Anterograde movement of Atg9H192L is blocked in growing conditions
We were interested in determining whether the impairment of the Cvt pathway seen with Atg9H192L (Fig. 4 B) was caused by defects in Atg9 cycling, particularly anterograde movement to the PAS. Accordingly, we used the TAKA assay to visualize the cycling of Atg9H192L-GFP in growing conditions and concurrently stained mitochondria with the dye MitoFluor red. In the atg1Δ background, wild-type Atg9-GFP was restricted to the PAS, as marked with BFP-Ape1 in 94% (116/123) of the cells examined (Fig. 5). In contrast, Atg9H192L-GFP distributed to multiple punctate structures in 84% (131/156) of the atg1Δ cells, which colocalized, in part, with the mitochondrial labeling but not with the PAS. Thus, the Atg9H192L mutation acted epistatically to atg1Δ, suggesting that the anterograde movement of Atg9H192L from mitochondria to the PAS was defective.
Bulk autophagy is not affected by disruption of the Atg9–Atg11 interaction
Atg9 is essential for both bulk autophagy and the Cvt pathway, whereas Atg11 is required solely in the Cvt pathway (Noda et al., 2000; Kim et al., 2001). To clarify whether Atg9 anterograde transport via Atg11 is involved in bulk autophagy, we used the Atg9H192L mutant to analyze the progression of bulk autophagy by several established assays.
Atg8 conjugated to phosphatidylethanolamine remains associated with the completed autophagosome and is a marker for autophagic delivery to the vacuole (Kirisako et al., 1999, 2000; Huang et al., 2000). After delivery of the GFP-tagged Atg8 chimera, the GFP moiety is cleaved and remains relatively stable in the vacuole, whereas Atg8 is rapidly degraded. Thus, the accumulation of free GFP reflects the progression of bulk autophagy, which can be readily detected by Western blotting (Shintani and Klionsky, 2004b; Cheong et al., 2005). In atg9Δ cells, essentially no free GFP was detected, indicating that bulk autophagy was blocked by ATG9 deletion (Fig. 6 A). In cells expressing wild-type Atg9 or Atg9H192L, free GFP was detectable starting 2 h after cells were shifted to starvation conditions (synthetic medium lacking nitrogen [SD-N]) to induce bulk autophagy, although GFP-Atg8 processing showed a delay in Atg9H192L cells compared with wild-type cells. This result demonstrated that bulk autophagy retained similar activity even when Atg9 failed to interact with the Cvt-specific component Atg11.
To confirm the aforementioned result, we quantitatively measured bulk autophagy activity using another marker protein, Pho8Δ60, which encodes an altered form of alkaline phosphatase that is only delivered to the vacuole via autophagy (Noda et al., 1995). The Pho8Δ60 enzymatic activity was measured in wild-type and atg1Δ cells and in atg9Δ cells transformed with a plasmid expressing wild-type Atg9 or Atg9H192L or an empty vector in growing (synthetic minimal medium [SMD]) and starvation (SD-N) conditions (Fig. 6 B). The atg1Δ or atg9Δ cells transformed with an empty vector showed only the basal level of Pho8Δ60 activity after autophagy induction (SD-N), indicating that bulk autophagy was defective after deleting either gene. In contrast, there was an increase of Pho8Δ60 activity in atg9Δ cells expressing either wild-type Atg9 or Atg9H192L, indicating that Atg9H192L rescued the autophagy defect in atg9Δ cells, which is comparable with the wild-type Atg9 protein. Collectively, these data suggested that bulk autophagy does not depend on the interaction between Atg9 and an Atg protein that is needed for anterograde movement during specific autophagy (Atg11).
Atg9H192L cycling is normal during bulk autophagy
Because bulk autophagy activity was not affected by the Atg9 H192L mutation, it was tempting to speculate that Atg9H192L cycled normally under starvation conditions even though it was not capable of forming a complex with Atg11. To test this hypothesis, we visualized the movement of Atg9H192L by the TAKA assay after autophagy induction. Cells were treated with the drug rapamycin, which mimics starvation conditions and induces bulk autophagy. As shown in Fig. 7, in the atg1Δ background without rapamycin treatment (−rap), wild-type Atg9-GFP was restricted to the PAS, whereas Atg9H192L-GFP could not move to the PAS and displayed a multiple punctate localization. After treatment with rapamycin (Fig. 7, +rap), in 85% (41/48) of the cells, Atg9H192L-GFP colocalized with the PAS marker BFP-Ape1 similarly to wild-type Atg9, indicating that Atg9 recruitment to the PAS was normal. Therefore, this result demonstrated that during bulk autophagy, the Atg9H192L mutation did not interfere with the cycling of Atg9.
Atg11 localization to the PAS is dependent on the actin cytoskeleton
Recently, we have shown that Atg9 anterograde traffic to the PAS is blocked when the actin cytoskeleton is disrupted by either treatment with the drug latrunculin A or point mutations in ACT1, the gene encoding actin (Reggiori et al., 2005a). In particular, the impairment of actin function leads to a defect in the Cvt pathway, whereas bulk autophagy is normal; a similar phenotype was also observed with the Atg9H192L mutant. Because Atg9 anterograde transport was also dependent on its interaction with Atg11 (Fig. 3), we wondered whether there was a functional connection between actin and Atg11 in Atg9 cycling and autophagic processes. Previous data show that Atg11 is needed to recruit prApe1 to the PAS (Yorimitsu and Klionsky, 2005). The atg9-null mutant does not affect the recruitment of prApe1 to the PAS or the localization of Atg11 (unpublished data), suggesting that Atg11 localization at the PAS is not dependent on the Atg9–Atg11 interaction. Thus, we propose that Atg11 mediates actin-dependent Atg9 cycling in the Cvt pathway.
To test this hypothesis, we used an actin mutant, act1-159, which has defects in actin cable depolymerization, Atg9 cycling, and the Cvt pathway (Reggiori et al., 2005a). As shown in Fig. 8 A, in wild-type cells, GFP-Atg11 localized to the PAS as a single punctum; chromosomally tagged Atg9-RFP displayed a multiple punctate localization pattern. In the act1-159 mutant, however, GFP-Atg11 redistributed to cytoplasmic patches, which partially colocalized with Atg9-RFP. To reveal the identity of these multiple compartments, we stained cells with the mitochondrial dye MitoFluor red and the vacuolar membrane dye FM 4-64. As shown in Fig. 8 B, in 81% (81/100) of the cells, the GFP-Atg11 patches were absent from the perivacuolar PAS position; instead, they colocalized with the mitochondrial labeling, indicating that Atg11 localized at mitochondria when the actin network was defective. Therefore, the proper localization of Atg11 to the PAS is dependent on the actin cytoskeleton and may underlie the actin-dependent Atg9 cycling during the Cvt pathway and other types of specific autophagy.
Atg9 is the only characterized transmembrane protein involved in the formation of the sequestering vesicles that form during the Cvt pathway, pexophagy, and autophagy. Accordingly, it is the best candidate to mark the source of the vesicle membrane. Recently, we have shown that Atg9 localizes to mitochondria in addition to the PAS, implicating this organelle in supplying membrane during autophagy-related processes (Reggiori et al., 2005b). Only Atg19, which is a receptor for biosynthetic cargos, and Atg8 remain associated with the completed vesicles; most of the soluble Atg proteins involved in vesicle formation presumably dissociate from the membrane before or upon vesicle completion. In contrast, a specific retrieval mechanism operates in the cycling of Atg9 and the associated protein Atg23 (Reggiori et al., 2004). The retrograde movement of Atg9 from the PAS to mitochondria requires Atg1–Atg13, Atg2, Atg18, and the PtdIns3–kinase complex. The transit of Atg9 to the PAS involves Atg23, Atg27, and actin (Reggiori et al., 2005a; Yen et al., 2006; and unpublished data); however, the mechanism by which actin mediates Atg9 movement is not known.
To identify other Atg components involved in the anterograde movement of Atg9, we performed a yeast two-hybrid screen for proteins that interact with Atg9 and identified Atg11 (Fig. 1). This result is intriguing because Atg11 has previously been shown to interact with Atg19 and Atg1, which are components involved in distinct steps of a specific autophagic process. In conjunction with the present result, we propose that Atg11 acts as a scaffold to coordinate the delivery of multiple components, including the cargo–receptor complex, components involved in vesicle formation, and proteins involved in supplying membrane, to the site of vesicle formation, the PAS (Fig. 9). The Atg11 CC domain2 interacts with the N terminus of Atg9 (Fig. 2), and the interaction occurs in the absence of Atg1 or Atg19 (Fig. 1), suggesting that there are distinct and multiple populations of Atg11 within the cell. Atg11 self-interaction (Yorimitsu and Klionsky, 2005) may then allow these various populations of Atg proteins to be delivered to the PAS in a coordinated manner. Consistent with this model, the overexpression of Atg11 restricted Atg9 to the PAS, presumably as a result of enhanced delivery (Fig. 3). In contrast, a mutation of H192L that disrupts interaction between Atg9 and the specific autophagy component Atg11 resulted in a defect in transporting Atg9 to the PAS (Fig. 5). Furthermore, the absence of Atg9 at the PAS caused by this point mutation led to a block in the Cvt pathway (Fig. 4).
It is known that Atg11 is needed for specific types of autophagy such as the Cvt pathway but is not essential for nonspecific autophagy (Kim et al., 2001). We found that the Atg9 point mutant that disrupts the interaction with Atg11 imposed little effect on the bulk autophagy induced during starvation (Fig. 6). The essentially normal autophagy function was correlated with the normal localization/transport of the binding-defective Atg9 mutant in the presence of the autophagy inducer rapamycin (Fig. 7). This finding suggests that the anterograde transport of Atg9 during bulk autophagy may be mediated by a different mechanism that is at least relatively independent of Atg11. Thus, other proteins that may interact with Atg9 deserve further investigation to reveal the Atg9 cycling machinery that operates during bulk autophagy; however, we cannot rule out the possibility that a low level of interaction between Atg9H192L and Atg11 allows the anterograde transport of Atg9H192L after autophagy induction.
Finally, we found that actin is required for the localization of Atg11 to the PAS (Fig. 8). This observation, coupled with the role of Atg11 in Atg9 anterograde movement but not vice versa, suggests that Atg11 mediates the connection between actin and Atg9 delivery from the mitochondria to the PAS. Actin is not needed for bulk autophagy in yeast (Reggiori et al., 2005a), which is in agreement with our findings in the present paper that the Atg9H192L mutant is not defective for nonspecific autophagy. It is not known how Atg9 might move along actin cables. The third Atg11 CC domain displays some similarity with that of Myo2; however, Atg11 lacks the N-terminal motor domain that functions in Myo2 movement (Monastyrska et al., 2006). Thus, it is not clear how Atg11 might actually mediate the anterograde movement of Atg9. Continued analysis of Atg9 cycling and its interactions with Atg11 and other Atg proteins may provide insight into the underlying mechanisms of membrane delivery during Cvt vesicle and autophagosome formation.
Materials And Methods
Strains, plasmids, and media
The S. cerevisiae strains used in this study are listed in Table I. For gene disruption, the entire coding region was replaced by the S. cerevisiae TRP1, the Kluyveromyces lactis LEU2 or URA3, the Saccharomyces kluyveri HIS3, or the Escherichia coli kanr gene using PCR primers containing ∼50 bases of identity to the regions flanking the open reading frame. For PCR-based integrations of the PA and GFP tags at the 3′ end of the ATG9 and ATG11 genes, pHAB102 and pFA6a-GFP-HIS3 were used as templates to generate strains expressing fusion proteins under the control of their native promoters (Longtine et al., 1998; Abeliovich et al., 2003).
Yeast cells were grown in rich medium (YPD; 1% yeast extract, 2% peptone, and 2% glucose) or SMD (0.67% yeast nitrogen base, 2% glucose, amino acids, and vitamins as needed). Starvation experiments were conducted in SD-N (0.17% yeast nitrogen base without amino acids and 2% glucose).
Plasmids expressing Atg11 truncations (Yorimitsu and Klionsky, 2005), GFP-Atg8 (pGFP-AUT7(414); Abeliovich et al., 2003), Atg9 (pAPG9(416); essentially constructed the same as pAPG9(414); Noda et al., 2000), Atg9-GFP (pAPG9GFP(416); Noda et al., 2000), and CFP-Atg11 (pCuHACFPCVT9(414); Kim et al., 2002) have been described previously. BFP-Ape1 (pBFPApe1(414)) was constructed by introducing BFP (Qbiogene) on a BamHI PCR-generated fragment into a unique BglII site situated after the start codon of APE1 (Shintani et al., 2002). GFP-Atg11 (pTS495) was constructed in a similar manner starting with ATG11 cloned into the pRS416 vector. PA-Atg9 (pCuPAAtg9(416)) was generated by cloning ATG9 into the pRS416-CuProtA vector (Kim et al., 2002). Full-length Atg9 (pBD-Atg9) and the Atg9 N-terminal (pBD-Atg9N), C-terminal (pBD-Atg9C), and transmembrane (pBD-Atg9TM) regions were amplified by PCR from pAPG9(416) and introduced into the two-hybrid vector pGBDU-C1 using BamHI and SalI sites. A total of 19 other ATG genes were amplified and cloned into the two-hybrid vector pGAD-C1 using BamHI and SalI sites. The serial deletion mutants of the Atg9 N terminus were generated by PCR amplification from pBD-Atg9N and ligated into the EcoRI and BamHI sites of pGBDU-C1. The plasmid expressing the Atg9H192L mutant (pAtg9H192L(416)) was constructed by releasing the ATG9 fragment containing H192L from pAtg9H192LGFP(416) (see Random mutagenesis screen…mutant) and was cloned into pAPG9(416) using SacI and AgeI. pAtg9PA(314) and pAtg9H192LPA(314) were constructed by PCR amplifying ATG9 and ATG9H192L together with its native promoter from pAPG9GFP(416) and pAtg9H192LGFP(416) and incorporating them into pNopPA(314) using XhoI and XmaI sites. pNopPA(314) was a gift from F. Reggiori (University of Utrecht, Utrecht, Netherlands).
PA affinity isolation
Cells were grown to OD600 = 0.8 in SMD, and 100 ml were harvested and resuspended in lysis buffer (PBS, 200 mM sorbitol, 1 mM MgCl2, 0.1% Tween 20, 1 mM PMSF, and protease inhibitor cocktail). The detergent extracts were incubated with IgG–Sepharose beads overnight at 4°C. The beads were washed with lysis buffer eight times and eluted in SDS-PAGE sample buffer by incubating at 55°C for 15 min. The eluates were resolved by SDS-PAGE and immunoblotted with anti-YFP antibody.
Random mutagenesis screen and generation of the Atg9 mutant
The gap repair PCR mutagenesis method (Oda et al., 1996) was used to generate the Atg11 binding–defective Atg9 mutant. An AvrII site at 942 bp was introduced into the pAPG9GFP(416) plasmid by site-directed mutagenesis. The Atg9 N-terminal region containing the AvrII site was amplified and introduced into pGBDU-C1 with EcoRI and BamHI to generate pBD-Atg9N(AvrII). A gapped pGBDU-C1 plasmid was generated by digesting with EcoRI and BamHI. The PCR reaction was performed using Taq polymerase (New England Biolabs, Inc.) with pBD-Atg9N(AvrII) as the template. The dATP concentration was lowered to three fifths that of the other three deoxynucleoside triphosphates. The resulting mutagenized PCR product shares overlapping sequences of ∼100 bp at both 5′ and 3′ ends with the gapped pGBDU-C1 plasmid. The PCR product and the gapped plasmid were cotransformed into the two-hybrid strain PJ69-4A. The transformants were replicated on both −His and −Ade plates. The transformants that were not able to grow on either plate were selected and analyzed by Western blotting for protein stability using anti-Atg9 antiserum (Noda et al., 2000). Stable mutants were sequenced, and pBD-Atg9NH192L was identified. The Atg9 N-terminal fragment containing the H192L mutation was then released from pBD-Atg9NH192L by digestion at the introduced AvrII site and a natural NruI site. This fragment was cloned into pAPG9GFP(416) digested with the same restriction enzymes to generate pAtg9H192LGFP(416).
Cells expressing fusion proteins with fluorescence tags were grown in SMD media to OD600 = 0.8. For vacuolar membrane labeling, cells were pelleted, resuspended in fresh media at OD600 = 1.0, incubated with 2 μM FM 4-64 at 30°C for 15 min, and pelleted and cultured in the same media without FM 4-64 at 30°C for 30 min. For mitochondrial fluorescent labeling, MitoFluor red 589 (Invitrogen) was added to the growing culture at a final concentration of 1 μM, and the culture was incubated at 30°C for 30 min. Cells were then washed with the same culture medium before imaging to remove the excess dye. For rapamycin treatment, cells were cultured with 0.2 μg/ml rapamycin at 30°C for 1.5 h. When necessary, a mild fixation procedure was applied to visualize Atg9 without destroying various fluorescent proteins: cells were harvested, resuspended in a half volume of fixation buffer (50 mM KH2PO4, pH 8.0, 1.5% formaldehyde, and 1 μM MgCl2), and incubated at room temperature for 30 min with gentle shaking. Cells were then washed once with an equal amount of wash buffer (50 mM potassium phosphate, pH 8.0, and 1 μM MgCl2) and resuspended in 50 μl of wash buffer. Fluorescence signals were visualized on a fluorescence microscope (IX71; Olympus). The images were captured by a camera (Photometrics CoolSNAP HQ; Roper Scientific) and deconvolved using DeltaVision software (Applied Precision).
The authors thank Dr. Weibin Zhou and members of the Klionsky laboratory for comments and suggestions.
D.J. Klionsky is supported by National Institutes of Health Public Health Service grant GM53396.
Abbreviations used in this paper: AD, activation domain; Ape1, aminopeptidase I; Atg, autophagy related; BD, binding domain; BFP, blue fluorescent protein; CC, coiled coil; Cvt, cytoplasm to vacuole targeting; PA, protein A; PAS, preautophagosomal structure; prApe1, precursor form of Ape1; SMD, synthetic minimal medium.