Atg17 recruits Atg9 to organize the pre-autophagosomal structure

Authors


  • Communicated by: Akihiko Nakano

* Correspondence: yohsumi@nibb.ac.jp

Abstract

Autophagy is a degradation system of cytoplasmic proteins and organelles via formation of double-membrane vesicles called autophagosomes. In the yeast Saccharomyces cerevisiae, autophagosomes are formed via the pre-autophagosomal structure (PAS) in a manner dependent on Atg proteins. Under nutrient-rich condition, Atg9 is recruited to the PAS by binding to Atg11 for the Cvt pathway. However, because Atg9 is recruited to the PAS in atg11Δ cells in starved condition and autophagy is induced, autophagy-specific mechanism for the Atg9 recruitment to the PAS has been assumed. Here, we demonstrate that, in autophagy-inducing condition, Atg9 is recruited to the PAS in a manner dependent on Atg17. Atg9 physically interacts with Atg17 in the presence of rapamycin. This interaction requires Atg1, a protein kinase essential for autophagy. Consistently, the Atg17-dependent PAS localization of Atg9 requires Atg1. However, its kinase activity is dispensable for this process. It rather regulates the equilibrium of assembly and disassembly of Atg9 at the PAS.

Introduction

In starved conditions, cells need to synthesize macromolecules, such as proteins, for their survival. Because external materials are limited under starvation, intracellular materials are recycled via autophagy. First, double-membrane vesicles termed autophagosomes sequester the cytoplasmic materials nonselectively. Then, the outer membrane of autophagosomes fuses with the vacuolar/lysosomal membrane, and the inner single-membrane vesicles are released into the vacuole/lysosome lumen as autophagic bodies (Takeshige et al. 1992; Baba et al. 1994). Autophagic bodies are then disrupted and the digested materials are transported back to the cytoplasm.

The molecular mechanism of autophagosome formation has been studied primarily by using budding yeast, Saccharomyces cerevisiae. Genetical screens of autophagy mutants revealed an overlap with genes involved in the cytoplasm to vacuole targeting (Cvt) pathway that delivers the vacuolar resident enzymes, aminopeptidase I (Ape1) and α-mannosidase (Ams1), during nutrient-rich condition (Tsukada & Ohsumi 1993; Thumm et al. 1994; Scott et al. 1996). Both cargo proteins are delivered to the vacuole by autophagy in autophagy-inducing conditions, such as nitrogen-starvation or the presence of rapamycin, a Tor kinase inhibitor. Reflecting these genetical and functional overlaps, the Cvt pathway shares common mechanistic features with autophagy, including the formation of double-membrane vesicles, termed the Cvt vesicles, which are smaller (140~160 nm diameter) than autophagosomes (300~900 nm diameter), and the breakdown of single-membrane vesicles in the vacuolar lumen (Baba et al. 1997). Identified genes involved in either autophagy or the Cvt pathway are given a generic name, ATG (autophagy-related).

In S. cerevisiae, autophagosomes and the Cvt vesicles are formed via a peri-vacuolar structure designated as the pre-autophagosomal structure (PAS), which is organized by dynamic assembly and disassembly processes of various Atg proteins (Suzuki et al. 2001, 2007; Reggiori et al. 2004; Kawamata et al. 2008). Shintani & Klionsky (2004) suggested that Atg11, a specific factor for the Cvt pathway, acts as a scaffold for the PAS organization in nutrient-rich condition. In addition to Atg11, comprehensive microscopic analysis of GFP-fused Atg proteins demonstrated that Atg17 is another scaffold for the PAS organization (Suzuki et al. 2007). In contrast to Atg11, Atg17 is specifically required for autophagy and functions as the PAS scaffold in autophagy-inducing conditions (Cheong et al. 2008; Kawamata et al. 2008). Both scaffold proteins physically interact with Atg1-Atg13 complex, which plays a pivotal role between the Cvt pathway and autophagy in response to nutritional condition change or to rapamycin treatment (Kamada et al. 2000). Recent studies revealed that Atg1-Atg13 complex and autophagy-specific components, Atg29 and Atg31, are required for the PAS organization in a mutually dependent manner with Atg17 in autophagy-inducing condition (Cheong et al. 2008; Kawamata et al. 2008).

Atg9 is an integral membrane protein essential for both autophagy and the Cvt pathway. GFP-fused Atg9 locates at multiple punctate structures in cytoplasm, one of which corresponds to the PAS (Noda et al. 2000; Kim et al. 2002). It has been suggested that Atg9 does not steadily reside at the PAS, but rather dynamically repeats association and dissociation and that this dynamics plays a role for supplying the lipids to create autophagosomes and the Cvt vesicles (Reggiori et al. 2004). Recent studies revealed the molecular mechanism for the recruitment of Atg9 to the PAS. The PAS localization of Atg9 is completely abolished in atg11Δatg17Δ cells, whereas the PAS localization of both Atg11 and Atg17 is normal in atg9Δ cells (Suzuki et al. 2007), showing that Atg9 is recruited to the PAS at downstream of these scaffold proteins. On the other hand, atg9Δ cells show defects in the PAS localization of many other Atg proteins, including Atg2-Atg18 complex, Atg8 and Atg14. These observations placed Atg9 at right downstream of Atg11 and Atg17 in the hierarchy scheme for the PAS organization (Suzuki et al. 2007). In this study, we addressed the molecular mechanism of the PAS localization of Atg9 for autophagy.

Results

Subcellular localization of Atg9-GFP

In yeast S. cerevisiae, during the process of autophagy and the Cvt pathway, autophagosomes and the Cvt vesicles are generated from the PAS, respectively (Suzuki et al. 2001). The PAS is a dynamic structure organized by continuous assembly and disassembly of many Atg proteins (Reggiori et al. 2004; Kawamata et al. 2008). Thus, understanding the dynamics at the PAS for each Atg protein is possible to uncover the molecular mechanism of autophagosome formation. Atg9 is essential for both autophagy and the Cvt pathway and it is the only integral membrane protein involved in the formation of autophagosomes and Cvt vesicles (Noda et al. 2000). Microscopic study for Atg9-GFP revealed a multiple cytoplasmic punctate pattern (Noda et al. 2000). One of them colocalized with mRFP-Ape1, which is a PAS marker, representing a population of Atg9 that is targeted to the PAS (arrow in Fig. 1A). In time-lapse image, Atg9-GFP co-stained with mRFP-Ape1 (arrow in Fig. 1A) was visible as a less motile and brighter punctate structure than the other punctate structures stained only by Atg9-GFP (Fig. 1A and Supporting Video S1). To clearly visualize the motile punctate structures, NaN3 treatment was tested for cells expressing Atg9-GFP. As shown in Fig. 1B, this treatment fixed the cytoplasmic punctate structures, enabling us to clearly observe the punctate structures in the cytoplasm. Like Atg9, Atg23, an essential factor for the Cvt pathway, also displays a multiple punctate pattern in cytoplasm (Tucker et al. 2003). The NaN3 treatment clearly demonstrated that the colocalization of Atg23-GFP with Atg9-mCherry (Fig. 1C). We also examined the localization of Atg9 on mitochondria in NaN3-treated cells. Although it has been reported that some of the cytoplasmic Atg9 punctate structures are adjacent to mitochondria (Reggiori et al., 2005a), we were not able to observe the significant localization of Atg9-GFP on mitochondria (Fig. 1D). Also in the streaming image of nonfixed cells, Atg9-GFP was not localized on mitochondria (see Supporting Video S2).

Figure 1.

Atg9 localizes to multiple puncta in cytoplasm. (A) STY1530 cells were grown in synthetic rich medium to OD600 of 1.0–1.5. Localization of GFP and mRFP was analyzed under a fluorescent microscope. Images were captured in the streaming mode at the rate of 200 milliseconds per frame. (B) ORY0900 cells were incubated with 1 mM NaN3, as described in Experimental procedures, and observed by fluorescent microscope. (C) STY2446 cells carrying Atg9-mCherry plasmid (pST107) were observed as in (B). (D) ORY0900 cells carrying mito-RFP plasmid (Kondo-Okamoto et al. 2006) were observed as in (B).

Atg9 localizes to the PAS in a manner dependent on Atg17 in autophagy-inducing condition

Recent studies revealed that Atg17, an autophagy-specific factor, functions as a scaffold for the PAS organization in autophagy-inducing condition (Cheong et al. 2008; Kawamata et al. 2008). However, in atg17Δ cells, the PAS is still organized in a manner dependent on Atg11, a Cvt pathway-specific factor. In this case, although the Cvt pathway is carried out, autophagy is blocked. Consistently, Suzuki et al. (2007) demonstrated that in the presence of rapamycin, which inhibits Tor kinase, Atg9-GFP successfully localizes to the PAS in either atg11Δ or atg17Δ cells, but does not in atg11Δatg17Δ cells. These imply that Atg9 is recruited to the PAS in a manner dependent on Atg17 in autophagy-inducing condition.

For clear assessment of the PAS localization of Atg9, TAKA assay (the transport of Atg9 after knocking-out Atg1 assay) has been performed (Shintani & Klionsky 2004). This assay is based on the accumulation of Atg9-GFP at the PAS in atg1Δ cells (Fig. 2). Recent study, however, demonstrated that, like Atg17, Atg1 is essential for the PAS organization in autophagy-inducing condition. To initiate the PAS organization, Atg1 and Atg17 are targeted to the perivacuolar site in an interdependent manner (Cheong et al. 2008; Kawamata et al. 2008). Consistently, as shown in Fig. 2, atg1Δatg11Δ cells showed no accumulation of Atg9-GFP at particular dot structure even after rapamycin treatment. Thus, the TAKA assay is unable to assess the Atg17-dependent recruitment of Atg9 to the PAS.

Figure 2.

Accumulation of Atg9-GFP at the PAS in atg1Δ and atg2Δ cells. Wild-type, atg1Δ, atg1Δatg11Δ, atg2Δatg11Δ, atg2Δatg11Δatg17Δ and atg2Δatg17Δ cells that carry chromosomally integrated ATG9-GFP were grown in SD + CAS medium supplemented with uracil and tryptophan to OD600 of 1.0–1.5 (nutrient-rich; left panels) and then treated with rapamycin for 2.5 h (right panels). The localization of Atg9-GFP was observed by fluorescent microscopy.

To clearly assess the Atg17-dependent PAS localization of Atg9, we used the mutant, which accumulates Atg9 at the PAS without inhibiting the PAS organization by Atg17. Like atg1Δ cells, atg2Δ cells accumulate Atg9 at the PAS (Fig. 2; Reggiori et al. 2004). Atg2 is the most downstream Atg protein in the PAS organization process (Suzuki et al. 2007), and Atg17-GFP normally localizes to the PAS in atg2Δatg11Δ cells (Kawamata et al. 2008). As expected, in contrast to atg1Δatg11Δ cells, atg2Δatg11Δ cells accumulated Atg9-GFP at the PAS in response to rapamycin treatment (Fig. 2). This rapamycin-dependent accumulation at the PAS was abolished by further deleting ATG17 (atg2Δatg11Δatg17Δ, Fig. 2), showing its dependency on Atg17. Thus, microscopic assay using atg2Δ background allows us to clearly demonstrate the Atg17-dependent PAS localization of Atg9.

Atg9 physically interacts with Atg17 in a manner dependent on Atg1

Next, we tested whether Atg9 physically interacts with Atg17 by coimmunoprecipitation assay. Cells expressing either Atg17 or Myc3-tagged Atg17 (Myc3-Atg17) were incubated in the presence or absence of rapamycin after conversion to the spheroplasts and subjected to coimmunoprecipitation with anti-Myc antibody. As shown in Fig. 3A, from rapamycin-treated spheroplasts, significant amount of Atg9 was coprecipitated with Myc3-Atg17, whereas only the basal level of Atg9 was detected with untagged Atg17, indicating that Atg9 specifically interacts with Atg17 (Fig. 3A, compare lane 4 with lane 5). On the other hand, in the absence of rapamycin, coprecipitated Atg9 with Myc3-Atg17 was significantly less than that in the presence of rapamycin (Fig. 3A, compare lane 1 with lane 4). This indicates that the Atg9-Atg17 interaction is enhanced upon rapamycin treatment. It is noted that the Atg9-Atg17 interaction is not affected in atg11Δ cells (Fig. 3A, lane 6), showing that it is an independent event from the Atg9-Atg11 interaction reported previously (He et al. 2006). In contrast, as shown in Fig. 3B, Atg9 was not coprecipitated with Myc3-Atg17 in atg1Δ cells (lane 7), indicating that Atg1 is required for the Atg9-Atg17 interaction. In atg2Δ cells, Atg9 was coprecipitated with Myc3-Atg17 as in wild-type cells (Fig. 3B, lane 8). These are totally consistent with our microscopic results showing that Atg1 but not Atg2 is required for the Atg17-dependent PAS localization of Atg9 (Fig. 2).

Figure 3.

Atg9 interacts with Atg17 and Atg11 in vivo. Immunoprecipitates and total lysates were subjected to immunoblot analysis with the indicated antibodies. (A) Coimmunoprecipitation of Atg9 with Myc3-Atg17. Myc3-Atg17 was immunoprecipitated from total lysates prepared from atg17Δ (STY2034; lanes 1 and 4) and atg11Δatg17Δ (STY1818; lanes 3 and 6) strains carrying plasmid encoding Myc3-Atg17. atg17Δ strain carrying plasmid encoding Atg17 was used as control (lanes 2 and 5). (B) Coimmunoprecipitation of Atg9 with Myc3-Atg17. Myc3-Atg17 was immunoprecipitated from total lysates prepared from atg17Δ (STY2034; lanes 1 and 5), atg1Δatg17Δ (STY1817; lanes 3 and 7), and atg2Δatg17Δ (STY2343; lanes 4 and 8) strains carrying plasmid encoding Myc3-Atg17. atg17Δ strain carrying plasmid encoding Atg17 was used as control (lanes 2 and 6). (C) Coimmunoprecipitation of Atg9 with Myc3-Atg11. Myc3-Atg11 was immunoprecipitated from total lysates prepared from atg11Δ (STY1458; lanes 1 and 5), atg1Δatg11Δ (STY1854; lanes 3 and 7), and atg2Δatg11Δ (STY2345; lanes 4 and 8) strains carrying plasmid encoding Myc3-Atg11. atg11Δ strain carrying plasmid encoding Atg11 was used as control (lanes 2 and 6). (D) The Atg9-Atg17 interaction is normal without the PAS localization. Myc3-Atg17 was immunoprecipitated from total lysates prepared from atg17Δ (STY2034; lane 1) and atg11Δatg17Δatg29Δ (STY3609; lane 3) strains carrying Myc3-Atg17 plasmid, which were treated with rapamycin. Cells carrying plasmid encoding Atg17 was used as control (lanes 2 and 4).

In atg1Δ cells, although Atg17-dependent PAS localization of Atg9 is blocked, Atg9-GFP was highly accumulated at the PAS irrespective of rapamycin treatment (Fig. 2). This accumulation seems to be mediated by Atg11 because it was abolished by further deleting ATG11 (Fig. 2). Consistently, Myc3-Atg11 in atg1Δ cells successfully coprecipitated Atg9 in both presence and absence of rapamycin (Fig. 3C, lanes 3 and 7). Therefore, in contrast to the Atg9-Atg17 interaction, the Atg9-Atg11 interaction does not require Atg1.

Next, we tested whether Atg9 interacts with Atg17 without those PAS localization. Atg29 has been shown to be required for the scaffold function of Atg17. In atg11Δatg29Δ cells, the PAS is not organized in the presence as well as in the absence of rapamycin (Kawamata et al. 2008). Consistently, Atg9-GFP was not accumulated by rapamycin treatment in atg2Δatg11Δatg29Δ cells (data not shown). We transformed atg11Δatg17Δatg29Δ strain with Myc3-Atg17 plasmid and tested the interaction of Atg9 with Myc3-Atg17 by coimmunoprecipitation assay. As shown in Fig. 3D, Myc3-Atg17 from rapamycin-treated atg11Δatg17Δatg29Δ cells coprecipitated Atg9 to the same extent as that from wild-type cells, showing that Atg9 normally interacts with Atg17 irrespective of the PAS localization.

Atg17-dependent PAS localization of Atg9 is required for autophagy

Hydropathy plot analysis of Atg9 predicted that it contains six to eight transmembrane segments with ~35 and 25 kD hydrophilic regions at the N- and C-termini, respectively (Noda et al. 2000), both of which are exposed to the cytosol (He et al. 2006). To examine the region involved in the interaction with Atg17, we truncated the 200 amino acids from the N-terminal end of Atg9-GFP (Fig. 4A) and the resulting ΔN2-Atg9-GFP was tested for the interaction with Atg17. Cells coexpressing Myc3-Atg17 with either full-length of Atg9-GFP (FL-Atg9-GFP) or ΔN2-Atg9-GFP were converted to spheroplasts and incubated with or without rapamycin. As shown in Fig. 4B, with rapamycin, ΔN2-Atg9-GFP was not coprecipitated with Myc3-Atg17 (lane 4), whereas FL-Atg9-GFP was successfully coprecipitated (lane 2), indicating that the N-terminal region is required for the Atg9-Atg17 interaction. We further confirmed the interaction of the N-terminal hydrophilic region of Atg9 with Atg17 by two-hybrid assay. The N-terminal region of Atg9 corresponding to 2–302 amino acid residues successfully interacted with Atg17 (Fig. 4C). He et al. (2006) demonstrated that the single histidine to leucine substitution at position 192 (H192 L), which is deleted in ΔN2-Atg9-GFP, caused the loss of interaction with Atg11, resulting in the defect in the Cvt pathway but not in autophagy. Indeed, we confirmed that the ΔN2-Atg9-GFP also failed to interact with Atg11 by coimmunoprecipitation assay (data not shown). Thus, ΔN2-Atg9-GFP does not interact with either Atg11 or Atg17. Then, we expected that ΔN2-Atg9 fails to localize to the PAS not only in nutrient-rich condition but also in autophagy-inducing condition. As expected, microscopic observation of ΔN2-Atg9-GFP, showed that, although the cytoplasmic motile structures remained to exist (see Supporting Video S3), none of cytoplasmic dot structures colocalized with mRFP-Ape1 in the presence as well as in the absence of rapamycin (Fig. 5A). This indicates that ΔN2-Atg9-GFP fails to localize to the PAS in autophagy-inducing condition, and thus that the Atg17-dependent PAS localization is blocked. Next, we examined the autophagic activity of cells expressing the ΔN2-Atg9 by ALP assay, which monitors autophagy-dependent processing of the cytosolic form of alkaline phosphatase Pho8Δ60 (Noda et al. 1995). As shown in Fig. 5B, cells expressing ΔN2-Atg9 gave no increase of ALP activity in the presence of rapamycin, indicating that ΔN2-Atg9 is defective in autophagy. Supporting this, autophagic body was not observed in cells expressing ΔN2-Atg9 (data not shown). Taken together, our results suggest that the Atg17-dependent PAS localization of Atg9 is essential for autophagy.

Figure 4.

The N-terminal hydrophilic region of Atg9 is required for the interaction with Atg17. (A) Schematic diagram of ΔN2-Atg9-GFP construct used in this study. (B) ΔN2-Atg9-GFP fails to interact with Atg17. Myc3-Atg17 was immunoprecipitated from total lysates made from STY2339 strain cotransformed with Myc3-Atg17 plasmid and either FL-Atg9-GFP plasmid encoding full-length of Atg9-GFP (lanes 1 and 2) or ΔN2-Atg9-GFP plasmid (lanes 3 and 4). (C) DNA-binding (BD) and activation (AD) domains were fused to full-length Atg17 and Atg9 N-terminal region (amino acids 2–302), respectively. AH109 strains cotransformed with indicated vectors were grown on either SC lacking leucine and tryptophan (-LW) or SC lacking leucine, tryptophan and adenine (-ALW).

Figure 5.

Atg17-dependent PAS localization of Atg9 is required for autophagy. (A) STY1548 cells carrying the plasmid bearing FL-Atg9-GFP or ΔN2-Atg9-GFP were grown in selective medium to OD600 of 1.0–1.5 (upper) and then further grown for 3 h in the presence of rapamycin (lower). Localization of GFP and mRFP was imaged under a fluorescent microscope. Arrows indicate colocalization of GFP and mRFP. (B) Full-length Atg9 and ΔN2-Atg9 were expressed from centromeric plasmid in STY0478 strain. Lysates were prepared from nontreated (open bars) or rapamycin-treated (gray bars) cells. ALP activity in each lysate was measured as described previously (Noda et al. 1995). Values shown are mean ± SD of results from triplicates in each case.

Kinase activity of Atg1 is not essential for the Atg17-dependent PAS localization of Atg9 but regulates the equilibrium of assembly and disassembly of Atg9 at the PAS

Atg1 is a protein kinase, whose activity increases in a manner dependent on the association with Atg13 and Atg17 to induce autophagy (Kamada et al. 2000). Recently, it has been reported that the kinase activity of Atg1 is dispensable for the PAS organization in autophagy-inducing condition (Cheong et al. 2008; Kawamata et al. 2008). As described above, we demonstrated that Atg1 is required for the Atg9-Atg17 interaction (Fig. 3B). To address the function of the kinase activity for autophagy, we examined the Atg9-Atg17 interaction in cells expressing kinase-deficient Atg1. As kinase-deficient Atg1, we tested Atg1K54A and Atg1D211A, both of which display no kinase activity in vitro, although these normally interact with Atg13 and Atg17 (Matsuura et al. 1997; Kamada et al. 2000; Kawamata et al. 2008). From rapamycin-treated atg1Δatg11Δatg17Δ cells co-expressing Myc3-Atg17 and either wild-type Atg1 (Atg1WT), Atg1K54A or Atg1D211A, Myc3-Atg17 was precipitated with anti-Myc antibody. As shown in Fig. 6, compared with Atg1WT (lane 1), the kinase-deficient Atg1, Atg1K54A and Atg1D211A (lanes 2 and 3) apparently reduced the Atg9-Atg17 interaction, implying that the kinase activity enhances the interaction. However, the amount of Atg9 coprecipitated with Myc3-Atg17 from cells expressing the kinase-deficient Atg1 (lanes 2 and 3) was significantly more than that from cells carrying empty plasmid (lane 4). This shows that a certain level of the Atg9 interacts with Atg17 without Atg1 kinase activity.

Figure 6.

Atg1 kinase activity is required for efficient interaction between Atg9 and Atg17. A centromeric plasmid with either ATG1WT (lanes 1 and 5), atg1K54A (lanes 2 and 6), atg1D211A (lanes 3 and 7), or no insert (lanes 4 and 8) was introduced into STY2256 strain carrying plasmid encoding Myc3-Atg17. Spheroplasts were treated with rapamycin and coimmunoprecipitation was performed as in Fig. 3.

Next, we examined whether the loss of Atg1 kinase activity affects the Atg17-dependent PAS localization of Atg9. We utilized atg2Δ background to assess the PAS localization of Atg9. The PAS localization of Atg9-GFP was observed in atg1Δatg2Δatg11Δ cells expressing either Atg1WT or Atg1D211A. As shown in Fig. 7A, upon rapamycin treatment (lower panels), Atg9-GFP accumulation was induced in Atg1WT cells, while no accumulation was induced in cells carrying empty plasmid, showing that the Atg17-dependent PAS localization is recovered by Atg1WT plasmid. Like Atg1WT, Atg1D211A accumulated Atg9-GFP at the PAS upon rapamycin treatment, showing that the Atg1 kinase activity is not essential for the Atg17-dependent PAS localization of Atg9. Also in the atg2Δ background, less Atg9 was coimmunoprecipitated with Myc3-Atg17 from Atg1D211A cells than from Atg1WT cells (data not shown). These indicate that, even if the Atg9-Atg17 interaction is largely impaired, Atg9 is able to accumulate at the PAS in cells harboring kinase-deficient Atg1. The accumulation of Atg9 at the PAS with Atg1D211A was also observed in the presence of Atg2. As shown in Fig. 7B, in atg1Δatg11Δ cells expressing Atg1D211A, Atg9-GFP clearly accumulated at the PAS in the presence of rapamycin. We also examined the effect of the loss of kinase activity in the presence of Atg11. In atg1Δ cells expressing Atg1D211A, Atg9-GFP accumulated at the PAS without rapamycin, as well as with rapamycin (data not shown). These indicate that the loss of kinase activity of Atg1 results in the accumulation of Atg9 at the PAS in either Atg11- or Atg17-dependent manner, suggesting that the kinase activity of Atg1 regulates the equilibrium of assembly and disassembly of Atg9 at the PAS regardless of nutritional condition. Taken together, Atg1 plays two independent roles on the Atg9 intracellular localization; the structural role to organize the PAS, resulting in the Atg9 recruitment, and the regulatory role to control the dynamics of Atg9 recruitment to the PAS, which are independent and dependent on the kinase activity, respectively.

Figure 7.

Atg1 kinase activity is not essential for the Atg17-dependent PAS localization of Atg9 but rather regulates the equilibrium of assembly and disassembly of Atg9 at the PAS. A centromeric plasmid with either no insert (Vector), ATG1WT or atg1D211Awas introduced into atg1Δatg2Δatg11Δ (A) and atg1Δatg11Δ (B) cells that carry chromosomally-integrated ATG9-GFP. Cells were grown in SD + CAS medium supplemented with tryptophan to OD600 of 1.0–1.5 (nutrient-rich) or further grown in the presence of rapamycin for 3 h (rapamycin 3 h), and then observed by fluorescent microscopy.

Kinase activity of Atg1 controls the dynamics of Atg9 recruitment to the PAS by mediating the PAS localization of Atg2

Like atg1Δ cells, atg2Δ cells accumulate Atg9 at the PAS (Fig. 2). The recruitment of Atg2 to the PAS requires several Atg proteins, including Atg1 and Atg9, in both nutrient-rich and nitrogen-starved conditions (Shintani et al. 2001; Suzuki et al. 2007). In cells expressing Atg2G83E, which lacks the ability to localize to the PAS (Shintani et al. 2001), Atg9-GFP highly accumulated at the PAS (Fig. 8A), showing that the PAS localization of Atg2 affects the equilibrium of Atg9 assembly and disassembly at the PAS. Then, we tested whether the Atg1 kinase activity is required for the PAS localization of Atg2. As shown in Fig. 8B, the PAS localization of Atg2-GFP was completely diminished in cells expressing Atg1D211A. This suggests that the kinase activity of Atg1 controls the dynamics of Atg9 recruitment to the PAS by mediating the recruitment of Atg2 to the PAS.

Figure 8.

The kinase activity of Atg1 controls the dynamics of Atg9 recruitment to the PAS by mediating the PAS localization of Atg2 (A) STY0634 strain was transformed with a plasmid carrying either ATG2, atg2G83E, or no insert (Vector). Cells were grown in SD + CAS medium supplemented with tryptophan to OD600 of 1.0–1.5 (nutrient-rich) or further grown in the presence of rapamycin for 2.5 h (rapamycin 2.5 h), and then observed by fluorescent microscopy. (B) A centromeric plasmid with either no insert (Vector), ATG1WT or atg1D211Awas introduced atg1Δatg11Δ cells that carry chromosomally-integrated ATG2-GFP. The GFP fluorescence in cells grown in the presence of rapamycin for 3 h was observed by fluorescent microscopy.

Discussion

As many other Atg proteins, Atg9 localizes to the PAS. In addition, it localizes to cytoplasmic punctate structures, which are highly motile (Fig. 1A) and are clearly visualized by NaN3 fixation (Fig. 1B). These motile punctate structures were not affected in atg mutants such as atg1Δ and atg11Δatg17Δ, in which the PAS localization of Atg9 is altered (data not shown). The truncated version of Atg9-GFP, ΔN2-Atg9-GFP, which fails to localize to the PAS, was also successfully targeted to these motile puncta (Supporting Video S3). A portion of the Atg9 cytoplasmic puncta was shown to localize on mitochondria, leading a model where mitochondria may provide the autophagosomal lipids (Reggiori et al. 2005a). However, our NaN3 fixation (Fig 1D), as well as the streaming capture of nonfixed cells (Supporting Video S2), was not able to demonstrate the significant localization of Atg9-GFP on mitochondria. Thus, at present, we are not able to clearly assess the functional relationship between these structures and autophagy. However, it is noted that Atg23, which contributes to full induction of autophagy (Tucker et al. 2003), colocalized with Atg9 at these motile punctate structures (Fig. 1C; Legakis et al. 2007), implying the involvement of these structures in autophagy.

In nutrient-rich condition, Atg11, Atg23, Atg27 and actin are required for the recruitment of Atg9 to the PAS (Reggiori et al. 2005b; He et al. 2006; Yen et al. 2007). However, in autophagy-inducing conditions, such as in nitrogen starvation or in the presence of rapamycin, the PAS localization of Atg9 is apparently normal in the absence of these proteins (He et al. 2006). Therefore, there must be specific molecule(s) to target Atg9 to the PAS in autophagy-inducing conditions. Here, we demonstrate that it is Atg17 to interact with and recruit Atg9 to the PAS in autophagy-inducing conditions.

Recent studies demonstrated that Atg17-Atg29-Atg31 autophagy-specific components and Atg1-Atg13 complex function as scaffold for the PAS organization (Cheong et al. 2008, Kawamata et al. 2008). This PAS organization is induced by nitrogen starvation or rapamycin treatment and has been suggested to be required for autophagy. In rapamycin-treated atg11Δ cells, Atg9-GFP localizes to the PAS in a manner dependent not only on Atg17 but also on Atg1, as well as Atg13, Atg29 and Atg31 (Fig. 2, data not shown). Therefore, TAKA assay, a method to assess the PAS localization of Atg9 by observing the accumulation of Atg9-GFP at the PAS in atg1Δ background, is not able to assess the Atg17-dependent PAS localization of Atg9 (Fig. 2). In other words, TAKA assay allows us to assess only the Atg11-dependent PAS localization, which does not require Atg1. In contrast, by using atg2Δ, we clearly visualized the Atg17-dependent PAS localization of Atg9, as well as the Atg11-dependent PAS localization (Fig. 2).

In this study, we demonstrated that Atg9 physically interacts with Atg17 in vivo (Fig. 3A). The Atg9-Atg17 interaction is not mediated by Atg11, which also interacts with Atg9 because the Atg9-Atg17 interaction is normal in atg11Δ cells (Fig. 3A). We found that ΔN2-Atg9, in which the N-terminal 200 amino acids of Atg9 was deleted, lacks the interaction with Atg17 (Fig. 4B). By two-hybrid assay, the N-terminal hydrophilic region of Atg9 interacts with Atg17 (Fig. 4C), but not with Atg1, Atg13 and Atg29 (data not shown). In addition, the interaction between the N-terminal region of Atg9 and Atg17 was not affected in two-hybrid assay using atg1Δ and atg11Δ strains (data not shown). These suggest that Atg9 directly interacts with Atg17, although we cannot exclude out the possibility that other unknown factor mediates this interaction. We found that ΔN2-Atg9-GFP fails to localize to the PAS in the presence of rapamycin (Fig. 5A), showing that the Atg17-dependent PAS localization is abolished. Finally, ΔN2-Atg9 is defective in autophagy (Fig. 5B). From these, we conclude that the Atg17-dependent PAS localization of Atg9 is essential for autophagy. Our results also imply that the Atg9-Atg17 interaction is involved in the Atg17-dependent PAS localization of Atg9, although it remains possible that the N-terminal region may function other than the interaction with Atg17 for the Atg9 recruitment to the PAS.

Microscopic results shown here are well correlated with the interaction between Atg9 and the PAS scaffold, Atg11 and Atg17. Atg17 but not Atg11 mediated the accumulation of Atg9 at the PAS in a manner dependent on Atg1 (Fig. 2). Consistently, in atg1Δ cells, the Atg9-Atg17 interaction was blocked (Fig. 3B), whereas the Atg9-Atg11 interaction was normal (Fig. 3C). In atg1Δ cells, as Atg9-GFP, Atg17-GFP was also highly accumulated at the PAS (data not shown). Thus, our results imply that the Atg9-Atg17 interaction is not resulted by those PAS localization but rather specifically regulated. In contrast to atg1Δ cells, atg2Δ cells are able to accumulate Atg9 at the PAS in an Atg17-dependent manner (Fig. 2). Consistently, the Atg9-Atg17 interaction in atg2Δ cells normally increased in response to rapamycin treatment (Fig. 3B). We also found that, in atg11Δatg29Δ cells, which fail to target Atg9 to the PAS in an Atg17-dependent manner (Kawamata et al. 2008), Atg9 successfully interacts with Atg17 (Fig. 3D), indicating that the Atg9-Atg17 interaction does not require the PAS localization. These results further support our notion that the physical interaction of Atg9 with Atg17 mediates the PAS localization of Atg9. This is also consistent with the hierarchy of recruitment of Atg proteins during the PAS organization (Suzuki et al. 2007), in which Atg9 is recruited to the PAS at right downstream of Atg17.

Recent studies revealed two Atg1 functions: recruitment of Atg proteins needed for assembly of the PAS and regulation of the kinetics of Atg protein movement in and out of the PAS, steps that are independent and dependent on Atg1 kinase activity, respectively (Cheong et al. 2008; Kawamata et al. 2008). We found that in cells expressing the kinase-deficient Atg1, Atg1D211A, Atg9-GFP highly accumulated at the PAS, indicating that the Atg1 kinase activity regulates the in/out equilibrium of Atg9 movement at the PAS (Fig. 7B). This is in contrast to the observation by Reggiori et al. (2004) that the kinase activity does not affect the localization of Atg9 at the PAS. This discrepancy is attributed to the possible residual kinase activity of Atg1K54A (Cheong et al. 2008), which has been used as a kinase-deficient Atg1. Indeed, we observed partial Atg9-GFP accumulation at the PAS in cells expressing Atg1K54A in both the presence and absence of Atg11 (data not shown). In contrast to Atg9, Atg2 failed to localize to the PAS in atg1D211A mutant (Fig. 8B). Thus, for Atg2, the kinase activity of Atg1 is necessary for the recruitment to the PAS. Taking account of the fact that Atg9 accumulates at the PAS in cells expressing Atg2G83E that is defective in the PAS localization (Fig. 7B), Atg1 kinase activity seems to act indirectly for the dynamics of Atg9 at the PAS by targeting Atg2 to the PAS.

Interestingly, in spite of the accumulation of both Atg9 and Atg17 at the PAS, the Atg9-Atg17 interaction was apparently reduced in the kinase-deficient Atg1 mutant (Fig. 6). Because atg2Δ cells did not reduce the Atg9-Atg17 interaction (Fig. 3B), it is not related to the altered equilibrium of assembly and disassembly of Atg9 at the PAS. Thus, the kinase activity of Atg1 may play another role for Atg9 function. Because a small number of autophagic bodies with reduced size were observed in cells expressing Atg1K54A in autophagy-inducing condition (Cheong et al. 2005), it is possible that the reduced Atg9-Atg17 interaction may result in a short supply of membrane source for autophagosome expansion. The accumulation of Atg9-GFP at the PAS in atg1Δ cells has been suggested to be due to block of Atg9 dissociation from the PAS in the cycling process between the PAS and the cytoplasm (Reggiori et al. 2004). If the dissociation of Atg9 from the PAS is blocked by the loss of Atg1 kinase activity, Atg9 accumulates at the PAS, even if less Atg9 is targeted to the PAS.

Finally, for the autophagy pathway and the Cvt pathway, specific components have been identified. Atg17 is specifically required for autophagy and Atg11 for the Cvt pathway. Although a candidate for the mammalian counterpart of Atg17, FIP200, as an interacting protein with ULK1/ULK2, mammalian Atg1 homologues, has been reported, it possesses no significant homology to Atg17 (Hara et al. 2008). Thus, so far, no apparent homologue of these specific Atg proteins has been found in mammalian and plant cells. Atg23 is essential for the Cvt pathway and required for efficient autophagy (Tucker et al. 2003; T. Sekito and Y. Ohsumi, unpublished results). As Atg11 and Atg17, no homologue of Atg23 has been found in mammalian and plant cells. We have also found that Atg23 is involved in the recruitment of Atg9 to the PAS by interacting with the C-terminal hydrophilic region of Atg9 (T. Sekito and Y. Ohsumi, unpublished results). Deletion of the C-terminal 200 amino acids results in the defect in the Cvt pathway and partial loss of autophagic activity, which is similar to atg23Δ phenotype. Together with the interaction with Atg11 and Atg17, the N- and C-terminal hydrophilic regions of Atg9 act as a binding region for these yeast-specific Atg proteins. These regions are highly variable among species, whereas the middle region containing transmembrane segments is conserved (Hanaoka et al. 2002; Yamada et al. 2005). Therefore, it is possible that the N- and C-terminal regions of S. cerevisiae Atg9 might adapt for yeast-specific type of autophagy, the Cvt pathway, and for regulation of the switch between the Cvt pathway and autophagy. It is noted that the region for self-interaction of Atg9 indispensable for both the Cvt pathway and autophagy is also located in the C-terminal region, which is adjacent to the predicted transmembrane segment (He et al. 2008). This region is not overlapped with the region required the interaction with Atg23 and well conserved among species. The diverse N- and C-terminal regions in the other organisms may be also involved in the specific type of autophagy, which is induced upon specific signal and delivers specific cargo for each case. Recently, the list of biological function of autophagy is growing. In addition to survival in starvation condition, studies using mammalian system revealed that this process is important for development, cellular differentiation, non-apoptotic programmed cell death and defense against diseases including cancer, neurodegenerative disorder and bacterial infection (Mizushima et al. 2008). For each, there might be a specific timing to be induced and a structural feature to enwrap the cargo. The N- and C-terminal hydrophilic region of Atg9 may play a role for various such functions of autophagy. Further analysis of the hydrophilic regions of Atg9 will provide an insight for the mechanism that regulate the size of forming vesicle, the induction and the specificity of the cargo in autophagy.

Experimental procedures

Strains and media

The S. cerevisiae strains used in this study are listed in Table 1. Standard techniques were used for yeast manipulation (Kaiser et al. 1994). Cells were grown in either YEPD medium (1% yeast extract, 2% peptone and 2% glucose), SD + CAS medium (0.17% yeast nitrogen base w/o amino acids and ammonium sulfate, 0.5% ammonium sulfate, 0.5% casamino acids, 20 mg/L of adenine and 2% glucose), or SD medium (0.17% yeast nitrogen base w/o amino acids and ammonium sulfate, 0.5% ammonium sulfate and 2% glucose) supplemented with 100 mg/L of histidine (-ALW) or 20 mg/L of adenine and 100 mg/L of histidine (-LW). Specific methods based on polymerase-chain reactions (Longtine et al. 1998) were used to introduce C-terminally epitope-tagged versions of Atg9. These epitope-tagged forms replaced the endogenous copy of the respective genes at their natural chromosomal loci and were also expressed from the native promoter. Gene deletions were performed as described (Longtine et al. 1998; Noda et al. 2000). The STY1548 strain was generated by transforming STY0838 strain with pPS128 digested with AflII (plasmid is a gift from Dr Daniel J. Klionsky, University of Michigan, Ann Arbor, MI).

Table 1.  Yeast strains used in this study
StrainGenotypeSource
SEY6210MATα his3 leu2 lys2 trp1 ura3 suc2Robinson et al. (1988)
BJ2168MATa leu2 trp1 ura3 pep4–3 prb1–1122 prc1–407Yeast genetic stock center
AH109MATa trp1–901 leu 2–3 112 ura3–52 his3–200 gal4Δgal80Δ LYS2::GAL1UAS GAL1TATAHIS3 GAL2UASGAL2TATAADE2 URA3::MEL1UASMEL1TATAlacZ MEL1Clontech
BY4741MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0Giaever et al. (2002)
KVY55SEY6210 pho8::pho8Δ60Kirisako et al. (1999)
ORY0900SEY6210 ATG9::ATG9-GFP-kanMXSuzuki et al. (2007)
GYS616SEY6210 ATG9::ATG9-GFP-kanMX atg11Δ::LEU2 atg17Δ::HIS3Suzuki et al. (2007)
STY0111SEY6210 atg9Δ::TRP1This study
STY0838SEY6210 atg9Δ::kanMXThis study
STY1548SEY6210 atg9Δ::kanMX LEU2::mRFP1-APE1This study
STY2229SEY6210 ATG9::ATG9-GFP-natMX atg1Δ::kanMX atg2Δ::HIS3 atg11Δ::LEU2This study
STY2446SEY6210 atg9Δ::TRP1 ATG23::ATG23-GFP-kanMXThis study
STY0606ORY0900 atg1Δ::LEU2This study
STY0634ORY0900 atg2Δ::LEU2This study
STY2273ORY0900 atg1Δ::natMX atg11Δ::LEU2This study
STY2079ORY0900 atg2Δ::HIS3 atg11Δ::LEU2This study
STY1328ORY0900 atg2Δ::LEU2 atg11Δ::URA3 atg17Δ::spHIS5This study
STY1530ORY0900 LEU2::mRFP1-APE1Suzuki et al. (2007)
STY1458BJ2168 atg11Δ::LEU2This study
STY2034BJ2168 atg17Δ::kanMXThis study
STY1818BJ2168 atg11Δ::LEU2 atg17Δ::kanMXThis study
STY1817BJ2168 atg1Δ::LEU2 atg17Δ::kanMXThis study
STY2343BJ2168 atg2Δ::LEU2 atg17Δ::kanMXThis study
STY1854BJ2168 atg1Δ::kanMX atg11Δ::LEU2This study
STY2345BJ2168 atg2Δ::kanMX atg11Δ::LEU2This study
STY2256BJ2168 atg1Δ::natMX atg11Δ::LEU2 atg17Δ::kanMXThis study
STY2339BJ2168 atg9Δ::natMX atg17Δ::kanMXThis study
STY3609BJ2168 atg11Δ::LEU2, atg17Δ::kanMX, atg29Δ::natMXThis study
STY0478KVY55 atg9Δ::kanMXThis study
TMK721BY4741 ATG2::ATG2-GFP-kanMX atg1Δ::HIS3MX atg11Δ::LEU2This study

Plasmids

The plasmids used in this study are listed in Table 2. The HindIII-KpnI fragment containing entire ATG9 ORF was generated by PCR and cloned into yeast centromeric plasmid pRS316 to generate pST100. pST101 was constructed by cloning the 0.7-kb SmaI-XbaI fragment from pST100 and the 4.1-kb XbaI-SalI fragment from p8007 (Noda et al. 2000) tandemly into pJJ215 (Johns & Prakash 1990). The 4.8-kb SacI-SalI fragment from pST101 was cloned into pRS316 to generate pST102. For pST103 and pST104 construction, a BamHI site was generated at the N- and C-terminus, respectively, of the ATG9 ORF on pST102, using a QuikChangeTM Site-directed Mutagenesis Kit (Stratagene). The DNA fragment encoding GFP (S65T) with BamHI site and mCherry with BamHI site on both sides was then ligated to the BamHI site of pST104 to generate ATG9 plasmid that encoded Atg9 connected with GFP (pST106) and mCherry (pST107), respectively.

Table 2.  Plasmids used in this study
NameDescriptionSource
pRS314TRP1, CEN-ARS, amprSikorski & Hieter (1989)
pRS316URA3, CEN-ARS, amprSikorski & Hieter (1989)
pGADT7GAL4(768–881)AD, LEU2, 2µ, ampr, HA epitope tagClontech
pGBKT7GAL4(1–147)DNA-BD, TRP1, 2µ, kanr, Myc epitope tagClontech
pJJ215HIS3, amprJones & Prakash (1990)
p8007ATG9 in pRS316 (XbaI)Noda et al. (2000)
pST100ATG9 in pRS316 (HindIII-KpnI)This study
pST101ATG9 in pJJ215This study
pST102ATG9 in pRS316 (SacI-SalI)This study
pST103ATG9 (BamHI site immediately after the start codon) in pRS316This study
pST104ATG9 (BamHI site immediately before the stop codon) in pRS316This study
pST106ATG9-GFP in pRS316This study
pST107ATG9-mCherry in pRS316This study
pΔN2-GFPΔN2-ATG9-GFP in pRS316This study
pAD-Atg9NATG9(2–302) in pGADT7This study
pBD-Atg17ATG17 in pGBD-C2This study
pATG11ATG11 in pRS314This study
pMyc3-ATG11Myc3-ATG11 in pRS314This study
pATG17ATG17 in pRS314This study
pMyc3-ATG17Myc3-ATG17 in pRS314This study
pATG1ATG1 in pRS316This study
pATG1K54Aatg1K54Ain pRS316This study
pATG1D211Aatg1D211Ain pRS316This study
pTS119atg2G83E in YCplac33Shintani et al. (2001)
pYX142-mtRFPffNeurospora crassa ATP9(1–69)-fast folding DsRedKondo-Okamoto et al. (2006)

To construct the N-terminally truncated Atg9, pΔN2, N-terminal deletion fragment of ATG9 amplified by PCR were digested by BamHI and HpaI and substituted for the original BamHI-HpaI sequence containing whole ATG9 ORF and the downstream region in pST103. The oligonucleotide primer sequences used will be made available upon request. GFP fusion for the truncated Atg9, pΔN2-GFP, was constructed by replacement of the XbaI-SpeI fragment of pST106 with XbaI-SpeI fragments from pΔN2. ATG11 amplified from SEY6210 genomic DNA by PCR was cloned into the SacII and XhoI sites of pRS314 to generate pATG11. BamHI site was created immediately after the start codon of ATG11 ORF in pATG11 by using a QuikChangeTM Site-directed Mutagenesis Kit. Subsequently, Myc3 fragment with BamHI site at the both ends was inserted to construct pMyc3-ATG11. ATG17 amplified from SEY6210 genomic DNA by PCR was cloned into the KpnI and SacII sites of pRS314 to generate pATG17. BamHI site was created immediately after start codon of ATG17 ORF in pATG17 by using a QuikChangeTM Site-directed Mutagenesis Kit. Subsequently, Myc3 fragment with BamHI site at the both ends was inserted to construct pMyc3-ATG17. ATG1 amplified from genomic DNA by PCR was cloned into the KpnI and BamHI sites of pRS316 to generate pATG1. The Atg1 kinase-dead mutants, pATG1K54A and pATG1D211A, were made from pATG1 using a QuikChangeTM Site-directed Mutagenesis Kit. The functional complementation was confirmed for all plasmids constructed in this study by transforming corresponding atgΔ mutant and subsequent immunoblot analysis for Ape1 or ALP assay.

Two-hybrid assays

The two-hybrid assays were performed using the system purchased from Clontech. The N-terminal region of Atg9 (amino acids 2–302) amplified by PCR was cloned into SfiI-XhoI site of pGADT7 plasmid. The full-length Atg17 coding sequence amplified by PCR was cloned into pGBKT7 plasmid. Primer sequences to construct these plasmids will be provided upon request. AH109 transformants were selected on SD (-LW) plates and tested for growth on SD (-ALW) plates.

Fluorescence microscopy

Fluorescence signals were visualized with the use of an IX71 fluorescent microscope (Olympus) as described previously (Suzuki et al. 2007). Movie image was acquired by Metamorph software (Molecular Devices) in stream acquisition mode at the rate of 200 milliseconds per frame. For NaN3 treatment, cells grown in SD + CAS containing appropriate auxotrophic supplements until OD600 of 1.0–1.5, were suspended in 1 mL of ice-cold stop mix (0.9% NaCl, 1 mm NaN3, 10 mm EDTA, and 50 mm NaF). Cells were collected by centrifugation at 3000 g for 5 min and again suspended in 1 mL ice-cold stop mix. Cell suspension was subjected to fluorescence microscopy.

Co-immnoprecipitation experiments

Yeast cells were grown exponentially to OD600 = 1.0–2.0, and 50 OD600 units of cells were treated with zymolyase 100T (2 U/OD600 cell; Seikagaku Kogyo) in SD + CAS containing 20 mm Tris-HCl [pH 7.5] and 1.2 m sorbitol to generate spheroplasts. The spheroplasts were treated with or without 0.2 µg/mL rapamycin (Sigma, St. Louis, MO) for 30 min, and lysed in 1 mL ice-cold lysis buffer (PBS, 200 mm sorbitol, 1 mm MgCl2, 0.1% Tween 20) supplemented with 1 mm PMSF and a cocktail of protease inhibitors (complete EDTA-free; Roche Diagnostics). The extracts were centrifuged for 15 min at 20 000 g and the supernatants were incubated with anti-Myc antibody (9E10, BabCo) and Protein G-Sepharose at 4 °C for 3 h. The immunoprecipitated proteins were separated by SDS-PAGE and analyzed by immunoblotting with rabbit polyclonal antibodies against GFP (Molecular Probe), Atg1, Atg2, Atg9, Atg11 or Atg17.

Polyclonal antibodies to Atg1, Atg2, Atg9 and Atg17 have been preciously described (Kamada et al. 2000; Noda et al. 2000; Shintani et al. 2001; Kabeya et al. 2005). To prepare antiserum to Atg11, synthetic peptides corresponding to amino acids 356–382 and 534–547 were synthesized and conjugated individually to Keyhole limpet hemocyanin (KLH). Standard procedures were used to generate antiserum by Shibayagi (Gunma, Japan).

Other procedures

The alkaline phosphatase (ALP) assay was performed as described previously (Noda et al. 1995).

Acknowledgements

We thank Dr Koji Okamoto, Dr Roger Y. Tsien and Dr Daniel J. Klionsky for providing mito-RFP plasmid, mCherry plasmid and mRFP-Ape1 plasmid, respectively. We also thank the members of Ohsumi's laboratory for providing materials and constructive discussion, and the National Institute for Basic Biology Center for Analytical Instruments for technical assistance. This work was supported by Grants-in-Aids for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan.

Ancillary