The Autophagy-related Protein Kinase Atg1 Interacts with the Ubiquitin-like Protein Atg8 via the Atg8 Family Interacting Motif to Facilitate Autophagosome Formation*

Background: Atg1 is a protein kinase essential for the initiation of autophagosome formation. Results: Atg1 interacts with Atg8 to associate with forming autophagosomal membranes; specific disruption of this interaction causes a significant defect in autophagy. Conclusion: Atg1 on forming autophagosomal membranes promotes autophagosome formation, distinct from its role for triggering the process. Significance: This study provides novel insights into the regulatory mechanisms of autophagy. In autophagy, a cup-shaped membrane called the isolation membrane is formed, expanded, and sealed to complete a double membrane-bound vesicle called the autophagosome that encapsulates cellular constituents to be transported to and degraded in the lysosome/vacuole. The formation of the autophagosome requires autophagy-related (Atg) proteins. Atg8 is a ubiquitin-like protein that localizes to the isolation membrane; a subpopulation of this protein remains inside the autophagosome and is transported to the lysosome/vacuole. In the budding yeast Saccharomyces cerevisiae, Atg1 is a serine/threonine kinase that functions in the initial step of autophagosome formation and is also efficiently transported to the vacuole via autophagy. Here, we explore the mechanism and significance of this autophagic transport of Atg1. In selective types of autophagy, receptor proteins recognize degradation targets and also interact with Atg8, via the Atg8 family interacting motif (AIM), to link the targets to the isolation membrane. We find that Atg1 contains an AIM and directly interacts with Atg8. Mutations in the AIM disrupt this interaction and abolish vacuolar transport of Atg1. These results suggest that Atg1 associates with the isolation membrane by binding to Atg8, resulting in its incorporation into the autophagosome. We also show that mutations in the Atg1 AIM cause a significant defect in autophagy, without affecting the functions of Atg1 implicated in triggering autophagosome formation. We propose that in addition to its essential function in the initial stage, Atg1 also associates with the isolation membrane to promote its maturation into the autophagosome.

In autophagy, a cup-shaped membrane called the isolation membrane is formed, expanded, and sealed to complete a double membrane-bound vesicle called the autophagosome that encapsulates cellular constituents to be transported to and degraded in the lysosome/vacuole. The formation of the autophagosome requires autophagy-related (Atg) proteins. Atg8 is a ubiquitin-like protein that localizes to the isolation membrane; a subpopulation of this protein remains inside the autophagosome and is transported to the lysosome/vacuole. In the budding yeast Saccharomyces cerevisiae, Atg1 is a serine/ threonine kinase that functions in the initial step of autophagosome formation and is also efficiently transported to the vacuole via autophagy. Here, we explore the mechanism and significance of this autophagic transport of Atg1. In selective types of autophagy, receptor proteins recognize degradation targets and also interact with Atg8, via the Atg8 family interacting motif (AIM), to link the targets to the isolation membrane. We find that Atg1 contains an AIM and directly interacts with Atg8. Mutations in the AIM disrupt this interaction and abolish vac-uolar transport of Atg1. These results suggest that Atg1 associates with the isolation membrane by binding to Atg8, resulting in its incorporation into the autophagosome. We also show that mutations in the Atg1 AIM cause a significant defect in autophagy, without affecting the functions of Atg1 implicated in triggering autophagosome formation. We propose that in addition to its essential function in the initial stage, Atg1 also associates with the isolation membrane to promote its maturation into the autophagosome.
Autophagy (macroautophagy) is a pathway for membrane transport to the lysosome/vacuole (1). During this process, a cup-shaped membrane called the isolation membrane is formed and expanded into a double membrane-bound vesicle called the autophagosome. The autophagosome eventually fuses with those lytic organelles, allowing degradation of its contents. Under starvation conditions, autophagosomes randomly sequester portions of the cytoplasm. In contrast, autophagosomes also selectively engulf degradation targets, such as cytotoxic protein aggregates, and damaged or surplus organelles (2,3). These processes require autophagy-related (Atg) 2 proteins. The core Atgs mediate the formation of the autophagosomal membrane, whereas receptor and adaptor Atgs recognize degradation targets and recruit the core Atgs to allow membrane formation along the surfaces of targets. The ubiquitin-like protein Atg8 (microtubule-associated protein 1 light chain 3 (LC3) or gamma-aminobutyric acid A receptorassociated protein (GABARAP) in mammals) is conjugated to phosphatidylethanolamine (PE) and thereby anchored to the isolation membrane, where it is involved in the expansion of the membrane (4 -7). A proportion of Atg8 is retained inside the complete autophagosomes and is transported to the lysosome/vacuole. In addition, Atg8 also plays an important role in the incorporation of degradation targets into autophagosomes; it interacts with receptor proteins that contain the Atg8 family interacting motif (AIM) (or the LC3 interacting region), which binds to highly conserved, hydrophobic pockets in Atg8 (2,3,8,9). Atg8-PE conjugates on the isolation membrane may link the target-receptor complex to the membrane to facilitate its engulfment.
Atg1 (unc-51-like kinase 1 (ULK1) in mammals) is a serine/ threonine kinase essential for autophagy. In yeast, Atg1 forms a complex with the regulatory proteins Atg13, Atg17, Atg29, and Atg31 in response to autophagy-inducing signals (10,11). This complex serves as a scaffold to organize the pre-autophagosomal structure (PAS), a dynamic assembly of Atg proteins, during the initiation of autophagosome formation (1,(11)(12)(13). The formation of the PAS-scaffolding complex also enhances the kinase activity of Atg1 (10), which regulates the PAS local-* This work was supported in part by the Funding Program for Next Generation World-Leading Researchers (to H. N.) and grants-in-aid for scientific research (to Y. O.) from the Ministry of Education, Culture, Sports, Science and Technology, Japan. □ S This article contains supplemental Methods, Tables S1 and S2, Figs. S1-S7, and Movie S1. 1 To whom correspondence should be addressed. ization of downstream Atgs, probably via phosphorylation of some of those proteins. In a previous study, we found an intriguing behavior of Atg1 during autophagy (12). Although Atg1 has been implicated in the early stage of autophagosome formation, a fraction of this kinase is incorporated into the autophagosome and transported to the vacuole. In addition, although the PAS localization of Atg1 requires complex formation with the regulatory proteins, only Atg1 among these proteins is significantly transported to the vacuole.
In this study, we investigated the mechanism of the incorporation of Atg1 into the autophagosome. We found that Atg1 contains an AIM, with which it directly binds to Atg8. Mutations that affected this interaction accordingly reduced the vacuolar transport of Atg1. Thus, interaction with Atg8 allows Atg1 to associate with the isolation membrane, resulting in Atg1 incorporation into the autophagosome. We also found that mutations in the AIM of Atg1 attenuated autophagic activity, although they did not affect Atg1 functions involved in the initiation of autophagosome formation. These results suggest that Atg1 on the isolation membrane plays an important role in the promotion of autophagosome formation, distinct from its role in the initial stage.

EXPERIMENTAL PROCEDURES
Yeast Strains, Plasmids, and Media-The yeast strains and oligonucleotides used in this study are listed in supplemental Tables S1 and S2, respectively. Construction of these strains and plasmids and the compositions of culture media are described in supplemental Methods.
Fluorescence Microscopy-Fluorescence microscopy was performed as described previously (14).
Yeast Two-hybrid Assay-The yeast two-hybrid assay was performed using Matchmaker Gal4 Two-hybrid System 3 (Clontech). The AH109 strain was transformed with different combinations of pGADT7-ATG1 and pGBD-C-ATG8 plasmids and their vectors, and the overnight cultures were spotted onto SC-LT (control), SC-LTH (ϪHis), and SC-LTA (ϪAde) agar plates followed by incubation at 30°C for 1, 2, and 3 days, respectively.
Immunoprecipitation-Yeast cells expressing Atg1-GFP and 3ϫFLAG-Atg8 were grown to mid-log phase, treated with rapamycin for 2 h, and disrupted in HSE buffer (25 mM HEPES-KOH (pH 7.2), 750 mM sorbitol, 5 mM EDTA) containing 0.5ϫ Complete protease inhibitor mixture (Roche Applied Science) using Multi-beads shocker and 0.5-mm YZB zirconia beads (Yasui Kikai). Sodium chloride and Triton X-100 were added to the lysates to concentrations of 50 mM and 0.5%, respectively. A monoclonal anti-FLAG M2 antibody (Sigma) and Dynabeads protein G (Invitrogen) were incubated with the lysates at 4°C for 3 h. After washing the beads, the bound proteins were eluted by incubating the beads in SDS sample buffer at 65°C for 5 min.
To examine the formation of the PAS-scaffolding complex, the cell lysates were prepared from rapamycin-treated yeast cells expressing Atg1-GFP by essentially the same method as described above, except that TSG buffer (50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 10% glycerol) was used instead of HSE buffer. The lysates were treated with 0.2% n-dodecyl-␤-Dmaltoside at 4°C for 30 min and incubated with GFP-Trap_M beads (ChromoTek) at 4°C for 2 h. After washing the beads, the bound proteins were eluted as described above.
ALP Assay-To quantify autophagic activity of yeast cells, an alkaline phosphatase (ALP) assay was performed as described previously (15).

Atg1 Is Transported to and Degraded in the Vacuole via
Autophagy-In systematic analysis of the PAS localization of Atg proteins (12), Atg1 was found to be transported to the vacuole via autophagy. When yeast cells expressing Atg1 tagged with GFP were treated with the autophagy inducer rapamycin, GFP fluorescence in the vacuole increased in a wild-type strain, but not in atg mutants such as atg14⌬ (Fig. 1A). Immunoblotting analysis using antibodies against GFP showed that wildtype cells produce GFP-containing fragments under autophagyinducing conditions, but atg14⌬ cells do not (Fig. 1B). These fragments were likely to have been produced by proteolysis of Atg1-GFP in the vacuole because their accumulation was blocked by the addition of phenylmethylsulfonyl fluoride (PMSF), which inactivates the vacuolar protease Prb1 (supplemental Fig. S1A). In contrast, specific fragments were not detected by immunoblotting using antibodies against Atg1 (supplemental Fig. S1B), suggesting that the Atg1 portion of the fusion protein is degraded in the vacuole, whereas the GFP portion is resistant to vacuolar proteases.
Atg1 Directly Binds to Atg8 via the AIM-We next addressed the mechanism by which Atg1 is efficiently incorporated into the autophagosome. First, we examined the involvement of factors known to function in selective types of autophagy in yeast (16 -18). Deletion of genes encoding adaptor (Atg11) and receptor proteins (Atg19, Atg34, and Atg32) did not significantly affect the vacuolar transport of Atg1 (supplemental Fig.  S2). However, a double mutation in Atg8 (P52A/R67A), which decreases its binding to the AIM but not its function in autophagosome formation (8,19), retarded Atg1 transport to the vacuole (Fig. 1C). Although an as yet unidentified receptor containing the AIM might mediate Atg1 incorporation into the autophagosome, previous studies showed that not only receptors but also some core Atgs use the AIM to interact with Atg8 (20). Therefore, we examined the possibility that Atg1 directly interacts with Atg8 via the AIM. Atg1 consists of the N-terminal kinase domain, the C-terminal region responsible for binding to Atg13, and the intervening middle region (Fig. 1D). The yeast two-hybrid assay clearly showed that Atg1 binds to Atg8; an indicator strain coexpressing Atg1 fused to the Gal4 transcription activation domain and Atg8 fused to the Gal4 DNAbinding domain could grow on agar medium lacking histidine ( Fig. 1E and supplemental Fig. S3). Furthermore, the Atg1 middle region alone interacted with Atg8, whereas Atg1 lacking this region did not, suggesting that Atg1 interacts with Atg8 via its middle region (Fig. 1E and supplemental Fig. S3). When compared with the full-length protein, the Atg1 middle region exhibited a higher affinity to Atg8; yeast strains coexpressing this Atg1 variant could also grow in the absence of adenine (Fig.  1E). We found that the P52A/R67A double mutation in Atg8 negatively affected the interaction of Atg8 with the Atg1 middle region (Fig. 1E), suggesting that the AIM binding ability of Atg8 is important for the interaction with Atg1.
Based on common features of known AIMs (20), we identified a putative AIM with the amino acid sequence Tyr 429 -Val 430 -Val 431 -Val 432 in the middle region of Atg1 (Fig. 1D). When Tyr 429 and Val 432 were simultaneously replaced with alanine residues, the interaction between Atg1 and Atg8 was severely impaired (Fig. 1E). This was also confirmed by coimmunoprecipitation experiments (Fig. 1F). Although immunoprecipitation of Atg8 tagged with the 3ϫFLAG sequence coprecipitated wild-type Atg1-GFP, the efficiency was significantly reduced by the Y429A/V432A double mutation. Thus, we conclude that this sequence in Atg1 functions as the AIM.
Association of Atg1 with the Isolation Membrane Facilitates Autophagosome Formation-We showed that the mutations in the AIM abolished the vacuolar transport of Atg1 (Fig. 2, A and  B). Given that Atg8 localizes to the isolation membrane and is retained on the complete autophagosomal membrane, these results suggest that Atg1 interacts with Atg8 using the AIM, thereby associates with the isolation membrane, and is eventually encapsulated into the autophagosome.
To examine the autophagic activity of cells expressing the AIM mutant of Atg1, we performed an ALP assay ( Fig. 2C and supplemental Fig. S4A) (15). The results revealed that the Atg1 AIM mutant cells exhibited a partial but significant defect in autophagy. In accordance with this, the vacuolar transport of GFP-Atg8 also decreased in the mutant cells (supplemental Fig.  S4, B and C). We also examined accumulation of autophagic bodies, inner membrane vesicles released into the vacuolar lumen upon fusion between autophagosomal outer membranes and the vacuolar membrane, using vacuolar protease-deficient strains (supplemental Movie S1). The Atg1 AIM mutant cells accumulated autophagic bodies less than wild-type cells. These results suggest that the mutations in the AIM of Atg1 decrease the efficiency of autophagosome formation. Coimmunoprecipitation analysis showed that the Atg1 AIM mutant interacted normally with Atg13 and Atg17, suggesting that the mutations in the AIM did not affect the formation of the PASscaffolding complex (Fig. 2D). In wild-type cells, Atg1 kinase activity is up-regulated under autophagy-inducing conditions (10), which can be assessed by the appearance of a slower migrating band of autophosphorylated Atg1 in immunoblotting analysis (21). That band was detected in the AIM mutant at a level similar to wild-type Atg1 and disappeared upon deletion of ATG13, which is essential for activation of Atg1 (Fig. 2E). Thus, the stimulation of Atg1 kinase activity occurred normally in the AIM mutant. Furthermore, the Atg1 AIM mutant localized to the PAS as efficiently as the wild-type protein (Fig. 2F  and supplemental Fig. S5). It was also shown that the mutations in the AIM of Atg1 did not reduce the PAS localization of Atg8; it rather accumulated at the PAS, probably due to delayed autophagosome formation (Fig. 2G and supplemental Fig. S5).

FIGURE 1. Atg1 interacts with Atg8 via the AIM.
A, wild-type (YNH512) and atg14⌬ (YNH731) cells expressing Atg1-GFP were grown to mid-log phase, treated with rapamycin for 6 h, and observed under a fluorescence microscope. Arrowheads indicate vacuoles. Images obtained using differential interference contrast (DIC) optics are also shown, with scale bars representing 5 m. B, the same strains as in A were treated with rapamycin for the indicated time periods and subjected to immunoblotting analysis using monoclonal antibodies against GFP. GFPЈ represents GFP-containing fragments generated by proteolysis of Atg1-GFP in the vacuole. C, ATG8 WT (YNH695) and ATG8 P52A/R67A (YNH696) cells expressing Atg1-GFP were analyzed as described in B. D, schematic diagram of Atg1. Atg1 consists of three regions: the N-terminal kinase domain (N), the middle region (M), and the C-terminal Atg13-binding region (C). The AIM mutant (AIM mut) of Atg1 contains two alanine substitutions at Tyr 429 and Val 432 . E, the indicator strain AH109 was transformed with plasmids expressing a transcription activation domain (AD) fused with full-length Atg1 (F) or the middle region of Atg1 (M), with or without mutations at the AIM (AIM mut), in combination with plasmids expressing a DNA-binding domain (BD) fused with either Atg8 WT or the Atg8 P52A/R67A mutant. These strains were grown on SC-LTH (ϪHis), SC-LTA (ϪAde), and SC-LT (control) agar plates. F, ATG8 (YNH740) and 3ϫFLAG-ATG8 (YNH741) cells expressing Atg1 WT -GFP or Atg1 AIM mut -GFP from single-copy plasmids were treated with rapamycin for 2 h and disrupted to prepare cell lysates (input) for immunoprecipitation (IP) using a monoclonal antibody against the FLAG sequence. The immunoprecipitates were analyzed by immunoblotting with antibodies against GFP and Atg8.
Taken together, these results suggest that mutations in the AIM of Atg1 cause a defect in autophagy by abolishing Atg1 association with the isolation membrane rather than by affecting the functions of Atg1 involved in the initiation of autophagosome formation.
A double mutation in the C-terminal region of Atg1, Y878A/ R885A, impairs the interaction with Atg13, resulting in reduced PAS localization of Atg1 (13). This Atg1 mutant also exhibited a defect in PAS localization when it was coexpressed with wildtype Atg1 (supplemental Fig. S6). We showed that this mutant was less efficiently transported to the vacuole than wild-type Atg1 (Fig. 2, H and I), although autophagy occurred normally in these cells. These results suggest that PAS localization is a prerequisite for Atg1 to associate with autophagosomal membranes.

DISCUSSION
Previous studies have established the role of Atg1 kinase and its regulatory proteins in the initiation of autophagosome formation; in response to autophagy-inducing signals, Atg1 forms a large complex with Atg13 and the Atg17-Atg29-Atg31 ternary complex. This complex further forms a higher-order assembly that serves as a scaffold, which organizes the PAS and induces autophagosome formation. However, the autophagic transport of Atg1 to the vacuole suggested that the kinase remains associated with the autophagosomal membrane until a late stage of its formation. In contrast, few of the regulatory proteins are transported to the vacuole. These data indicated that Atg1 plays an additional role, distinct from its function in triggering autophagosome formation in association with the regulators. In agreement with these observations, it has recently been shown that Atg1, but not its regulators, localizes to the isolation membrane. 3 In the present study, our results suggest that the association of Atg1 with the isolation membrane is mediated by direct interaction with Atg8. Mutations that disrupt this interaction, and thereby prevent Atg1 association with the membrane, caused a significant defect in autophagy, although they did not impair Atg1 functions involved in the initiation of autophagosome formation. Thus, in accordance with the above notion, our results suggest that Atg1 plays an additional role in promoting autophagosome formation on the isolation membrane.
The interaction between Atg1 and Atg8 involves an AIM located in the middle region of Atg1. The AIM was first identified as a sequence commonly seen in receptors for selective autophagy and subsequently found in core Atg proteins such as yeast Atg3 and mammalian Atg4 (20). Although the AIM in Atg3 is specifically required for the cytoplasm-to-vacuole targeting pathway, which can be regarded as a type of selective autophagy, this study identified for the first time the AIM involved in starvation-induced, nonselective autophagy.
Most core Atg proteins localize to the PAS in a hierarchical manner; the PAS-scaffolding complex assembles first, whereas Atg8 is one of the most downstream factors (11)(12)(13). We showed that AIM mutations do not significantly affect the PAS localization of either Atg1 or Atg8. Therefore, it seems that the 3 K. Suzuki, M. Akioka, C. Kondo, and Y. Ohsumi, unpublished results. FIGURE 2. Characterization of an AIM mutant of Atg1. A and B, atg1⌬ cells (YNH204) expressing Atg1 WT -GFP or Atg1 AIM mut -GFP from single-copy plasmids were treated with rapamycin and analyzed by fluorescence microscopy (A) or immunoblotting with anti-GFP antibodies (B). Scale bars represent 5 m. GFPЈ represents GFP-containing fragments generated by proteolysis of Atg1-GFP in the vacuole. AIM mut, AIM mutant. C, ATG1 WT (YNH738), ATG1 AIM mut (YNH739), and atg1⌬ (YNH621) cells expressing a mutant form of ALP were incubated in nitrogen-deprived medium and subjected to the ALP assay. The same experiments were repeated three times; data are represented as means with standard deviations. AU, arbitrary units. D, YNH740 cells that carried either plasmids expressing Atg1 WT -GFP (W) or Atg1 AIM mut -GFP (A) or empty vector (Ϫ) were treated with rapamycin for 1 h and subjected to coimmunoprecipitation (IP) analysis using anti-GFP antibodies. The immunoprecipitates were analyzed by immunoblotting using antibodies against Atg1, Atg13, and Atg17. E, ATG1 WT (YNH736), ATG1 AIM mut (YNH737), ATG1 WT atg13⌬ (YNH746), and ATG1 AIM mut atg13⌬ (YNH747) cells were treated with rapamycin for 1 h and examined by immunoblotting using anti-Atg1 antibodies. F, ATG1 WT -GFP (YNH742) and ATG1 AIM mut -GFP (YNH743) cells treated with rapamycin for 1 h were observed under a fluorescence microscope (the numbers of cells examined were 272 and 221, respectively); the percentages of cells with GFP dots are shown. G, dot formation of GFP-Atg8 in ATG1 WT (YNH744) and ATG1 AIM mut (YNH745) cells was examined as described in F (the numbers of cells examined were 307 and 246, respectively). H, the vacuolar transport of Atg1 WT -GFP and Atg1 Y878A/R885A -GFP expressed from singlecopy plasmids in wild-type cells (BY4741) was examined by immunoblotting analysis as described in legend for Fig. 1. I, the same experiments as in H were repeated three times, and the intensities of the bands of Atg1-GFP and GFPЈ at the time point of 6 h were measured to estimate the efficiency of the production of processed GFP fragments (the intensities of GFPЈ were divided by the sum of those of Atg1-GFP and GFPЈ); results are represented as means with standard deviations.