Early Steps in Autophagy Depend on Direct Phosphorylation of Atg9 by the Atg1 Kinase

1Macroautophagy, hereafter referred to as autophagy, is an important cellular mechanism for the bulk degradation of cytoplasmic material. Autophagy is a sequential process beginning with the sequestration of cytoplasmic cargo in an expanding membrane sac known as the phagophore, which forms at the preautophagosomal structure (PAS). The phagophore expands to form a double-membrane autophagosome that engulfs the cargo and subsequently fuses with the lysosome or vacuole where the cargo is degraded. This conserved process serves to ensure cellular homeostasis, and is integral for the cellular response to stress conditions such as nutrient starvation as well as in embryonic development and the defense against several human pathogens. Indeed, defects in autophagy pathways have been associated with numerous human pathologies including infectious diseases, neurodegenerative disorders, and cancer. Despite these fundamental functions, the molecular mechanisms underlying autophagy regulation remain poorly understood.

2Both nonselective “bulk” autophagy and selective autophagy of specific proteins or organelles have been described. A selective autophagy-related pathway in yeast is the cytoplasm-to-vacuole targeting (Cvt) pathway, which fulfills a biosynthetic function by delivering three resident enzymes, aminopeptidase 1 (Ape1), α-mannosidase (Ams1), and aspartyl aminopeptidase (Ape4), to the vacuole. Genetic analysis in yeast has identified 36 mostly conserved components (termed Atg1–Atg36) that are required for different steps of autophagy. Many of these components are common to both autophagy and the Cvt pathway, although autophagy- and Cvt-specific genes exist. Atg1 is a serine-threonine kinase required for both selective and bulk autophagy pathways, and it is highly conserved from yeast to mammals. A single homolog of Atg1 has been identified in Caenorhabditis elegans and Drosophila melanogaster, and the homologs Unc-51-like kinase 1 (ULK1) and ULK2 have been identified in mammals. The kinase activity of Atg1 and its homologs is needed for autophagy in mammalian cells, and for both autophagy and the Cvt pathway in yeast.

3Among proteins essential for autophagy is the conserved multimembrane-spanning protein Atg9. A role for Atg9 in the regulation of autophagy by providing membranes for autophagosome formation in yeast and possibly also in mammals has been suggested. Atg9 localizes to several vesicles throughout the cytoplasm, and shuttling of Atg9 to and from the PAS has been proposed to be important for autophagy function. Atg9 vesicles are 30–60 nm in diameter and contain approximately 30 molecules of Atg9. These vesicles fuse to form the preautophagosomal structure, and subsequently Atg9 becomes integrated into the outer autophagosomal membrane. After fusion of autophagosomes with the vacuole, Atg9 is recycled. Several factors including Atg1, Atg2, and Atg18 have been implicated in the regulation of Atg9.

4Despite the important role of Atg1 in autophagy, little is known about its physiological substrates for autophagy function. Although some Atg1 kinase targets have been described in D. melanogaster and mammalian cells, the mechanism by which Atg1 kinase regulates autophagy remains elusive. Understanding the detailed events of how Atg1 phosphorylation regulates autophagy and identification of its crucial substrates for autophagy function is key to deciphering the mechanisms of autophagy induction during nutrient starvation and other stress conditions.

5In this study, we set out to identify Atg1 kinase substrates in yeast. We used a combination of peptide array-based consensus screening and quantitative mass spectrometry to deduce Atg1 kinase substrates. Among 24 confirmed phosphorylation targets, this approach identified Atg9 as a direct kinase substrate of Atg1. Indeed, mutation of the Atg1-dependent phosphorylation sites on Atg9 resulted in severe defects in both selective and bulk autophagy. Although nonphosphorylatable Atg9 mutants shuttle normally in the cell, they fail to interact with and recruit the essential autophagy proteins Atg18 and Atg8 to the PAS, resulting in the inability to form autophagosomal isolation membranes. Together, these findings identify Atg9 as a direct Atg1 kinase target essential for autophagy function in yeast, and unravel a mechanistic role for Atg1 at the PAS early during autophagosome formation.

6Atg1 kinase activity is essential for autophagy function; however, phosphorylation targets have remained sparse. To identify Atg1 kinase targets relevant for autophagy, we mapped the Atg1 kinase consensus phosphorylation site motif using a peptide array. Previous large-scale studies deciphering kinase consensus site motifs used overexpression of the kinase of interest. However, as Atg1 overexpression is toxic in yeast and Atg1 is found in a large protein complex together with several proteins (including its essential activator Atg13), we purified endogenously expressed Atg1 to preserve the native complex and interactions, a prerequisite for the identification of physiological substrates. Protein A-tagged Atg1 was shown to be fully functional as determined by cleavage of the propeptide from the preenzyme Ape1 (prApe1) to generate its active form in the vacuole (Figure 1A). After induction of autophagy by rapamycin treatment, the Atg1 complex was affinity purified from yeast cultures and released from the beads by tobacco etch virus (TEV) protease cleavage (Figure 1B). The identity of Atg1, Atg11, and Atg13 was verified by mass spectrometric analysis of the respective silver-stained gel bands. The activity of the resulting soluble complex was then analyzed in an in vitro autophosphorylation reaction (Figure 1B; Figure S1A available online). Whereas the wild-type protein was active, the kinase-dead Atg1 D211A mutant did not show in vitro autophosphorylation activity, indicating the absence of copurified kinase activity. To identify the preferred phosphorylation motif, this soluble kinase was assayed in peptide arrays as described previously. The spot intensities from the peptide array were quantified, background corrected, and normalized. These values were used to build heat maps indicating the preference for each amino acid type at the respective position relative to the serine phosphorylation acceptor (Figure 1C). Atg1 had a unique phosphorylation motif characterized by preferences for hydrophobic residues at multiple positions. For example, Atg1 was most selective for aliphatic residues at position −3 and for both aliphatic and aromatic residues at positions +1 and +2, and there was strong negative selection against small and charged residues at these positions. In addition, Atg1 showed a clear preference for serine over threonine as the phosphoacceptor residue.

7A position-specific scoring matrix of selectivity values was used to search the Saccharomyces cerevisiae protein sequence database for sites closely matching the Atg1 kinase consensus sequence. Among these potential Atg1 kinase substrates, the autophagy protein Atg2 was identified (Figure 1D). We furthermore noted that Atg1 and Atg9 also contain Atg1 consensus sites, and included these in the subsequent analysis. To verify that the identified candidate substrates are bona fide Atg1 kinase targets, we synthesized 32 peptides spanning the consensus regions of the candidate proteins and incubated them with soluble Atg1 kinase in the presence of ATP. The resulting phosphorylated peptides were then analyzed by shotgun mass spectrometry. As in vitro kinase reactions can generate nonspecific phosphorylation artifacts, we controlled for both Atg1 phosphorylation specificity and for incorrect phosphopeptide identifications by sampling the reaction at 5, 30, and 60 min and monitoring the accumulation of phosphorylation over time by label-free quantification using the precursor ion intensities of the respective peptides. For most of the identified phosphorylation sites, accumulation of precursor ion intensities could be detected over time (Figure 1D, marked with asterisks; Figure S5), further indicating that the in vitro phosphorylation events were specifically associated with the Atg1 complex. In addition, experiments performed with the kinase-dead mutant of Atg1 did not generate any phosphorylated peptides. This confirms that the kinase activity detected is dependent only on Atg1 and not any other copurifying enzyme. Out of 32 candidates tested, 25 were verified in vitro as direct Atg1 kinase substrates, validating our approach. A refined consensus sequence was generated by the comparison of the 25 substrates, illustrated by a WebLogo cartoon (Figure 1E). Alignment of the 25 verified Atg1 kinase targets with the known mammalian ULK1 target Beclin-1 showed that the consensus preference is indeed conserved from yeast to mammals (Figure 1F). We focused our follow-up analysis on the known autophagy players Atg1, Atg2, and Atg9, as in budding yeast no Atg1 kinase substrate has been identified to date.

8To determine whether Atg1, Atg2, and Atg9 are phosphorylated in an Atg1-dependent manner in vivo, we analyzed the phosphorylation sites of these proteins by mass spectrometric stable isotope labeling by amino acids in cell culture (SILAC) quantification. We purified Atg1, the kinase-dead Atg1 D211A mutant, Atg2, and Atg9 from rapamycin-treated wild-type and atg1Δ cells under denaturing conditions using a C-terminally integrated His-biotin (HTB) tag. These constructs are fully functional, as judged by prApe1 processing (Figure S1B).

9Eleven autophosphorylation sites on Atg1 and four Atg1-regulated phosphorylation events on Atg2 were detected (Figures S1C and S1E). The predicted phosphorylation sites S517 on Atg1 and S249 on Atg2 were found phosphorylated in vivo, strongly supporting that these are bona fide Atg1 phosphorylation sites. In addition, S356 and S390 on Atg1 matched the consensus sequence, further indicating that these are also autophosphorylation sites of Atg1 (Figure S1D).

10In addition to S249, S1086 on Atg2 also conforms to the Atg1 consensus (Figure S1F). Thus, we mutated S249 and S1086 on Atg2 to nonphosphorylatable alanine or phospho-mimicking aspartate and tested their functionality by measuring prApe1 processing. Neither mutation had an effect on prApe1 maturation (Figure S1G), suggesting that these phosphorylation sites act in a more specialized autophagy pathway and/or serve redundant functions.

11SILAC analysis furthermore revealed a total of 35 serine and threonine phosphorylation sites on Atg9, which are all located in the cytosolic domains of the protein (Figure 2A; Figure S2A; Table S1). Six of these sites match the Atg1 consensus (Figure 2A; Figure S2B). Interestingly, phosphorylation of five out of the six sites matching the Atg1 kinase motif was absent or strongly reduced in Atg9 protein isolated from atg1Δ cells, showing that these sites are indeed phosphorylated in an Atg1-dependent manner in vivo (Figure 2B). The remaining predicted phosphorylation site, serine 657, could not be identified by mass spectrometry, most likely due to technical difficulties resulting from the arginine residue preceding serine 657.

12To determine whether Atg1 directly phosphorylates Atg9 on additional serine residues, we synthesized 54 peptides spanning the Atg9 protein sequence and incubated them with soluble Atg1 kinase. The resulting phosphorylated peptides were then analyzed as in Figure 1D. Three out of 54 peptides spanning Atg9 were phosphorylated in vitro by Atg1 on a serine residue (Figure 2C). Two of these phosphorylation sites, serine 657 and 948, were among the six consensus motif sites. The lack of identification of the remaining four consensus sites is likely due to stochastic sampling of the mass spectrometry method used or to the poor flying nature of the peptides themselves, as their sequences were not chosen specifically for their ability to be detected but rather for their biological significance. Indeed, for several peptides, neither the naked nor the phosphorylated form was detected. We therefore cannot exclude the possibility that these peptides were phosphorylated in vitro but not detected by mass spectrometry. The third phosphorylated site at S602 also matched the Atg1 kinase consensus motif; however, it is located in a luminal exposed loop of Atg9 and is most likely not accessible to the Atg1 kinase in vivo. Taken together, the peptide approach shows that Atg1 selectively and directly phosphorylates at least two consensus sites in vitro.

13To confirm these results, we performed additional in vitro phosphorylation experiments using as a substrate the C terminus of Atg9 (amino acids 757–997) purified from Escherichia coli. This fragment contains four of the six consensus phosphorylation sites (S802, S831, S948, and S969). Indeed, whereas soluble Atg1 complex purified from yeast readily phosphorylated the C-terminal Atg9757–997 fragment, phosphorylation was strongly decreased upon mutation of all four consensus phosphorylation sites (Figure 2D).

14In summary, these findings suggest that Atg9 is a bona fide target of Atg1, which directly phosphorylates Atg9 on several sites in vitro and in vivo.

15To investigate whether these phosphorylation events are important for Atg9 function in vivo, we generated nonphosphorylatable and phospho-mimicking mutants of Atg9. All highly regulated serine residues were mutated to alanine (A) or aspartate/glutamate (D/E; S19A/D, S657A/D, S802A/D, S831A/E, S948A/D, S969A/D) and analyzed for their effect on the Cvt pathway or autophagy when expressed in atg9Δ cells. S657, S831, and S948 were analyzed individually, in all possible double-mutant permutations, as well as the triple mutant (3A, 3D; Figure S2C). In addition, all six sites were mutated simultaneously (6A, 6D) or lacking S657, which is located next to a transmembrane region (5A, 5D). All mutants were expressed at similar levels as the wild-type protein and cofractionated with membranes (Figure 3A; Figure S2D). First, we monitored Cvt activity by measuring Ape1 processing. Whereas the individual S-to-A mutations had little effect on the Cvt pathway (Figure S2C), Atg9 containing all six mutated serines (6A) near completely negated Cvt activity (Figure 3A). Various combinations of multiple S-to-A mutations behaved as intermediates between the two extremes. This suggests that phosphorylation of these residues is indeed required for Cvt activity, even though no single site is essential for Atg9 function. Equal expression levels, membrane association, and increasing severity of the mutant phenotype further support that the defect monitored is due to the lack of phosphorylation rather than misfolding of the mutant protein. Interestingly, the aspartate/glutamate phospho-mimicking mutations showed similar though less severe defects in Ape1 processing compared to the alanine mutations (Figure 3A), indicating that the mutations into charged residues are not sufficient to mimic the phosphorylated state of Atg9. Alternatively, cycles of phosphorylation and dephosphorylation may be required for Atg9 function in the Cvt pathway.

16Next, we quantified bulk autophagy using the Pho8Δ60 assay that measures the autophagic delivery of the Pho8Δ60 phosphatase into the vacuole. As expected, Pho8Δ60 activity in wild-type cells was increased after 5 hr of nitrogen starvation, whereas no such increase was measured in atg9Δ cells. Interestingly, in Atg9 nonphosphorylatable mutant cells (3A, 5A, 6A), the starvation-induced increase in Pho8Δ60 activity amounted to only 10%–30% of that in wild-type cells, indicating that Atg9 phosphorylation is required for bulk autophagy in vivo (Figure 3B). Similarly, the phospho-mimicking mutants showed up to a 50% loss in autophagy activity. These results were confirmed by a GFP-Atg8 cleavage assay, in which processing of GFP-Atg8 was strongly inhibited in the Atg9 mutant cells (Figure S2E). In summary, these results demonstrate that Atg9 phosphorylation by Atg1 is required for the Cvt pathway and autophagy function in vivo.

17To determine whether the defect in autophagy observed in Atg9 mutant cells is due to a failure in early events such as the formation of the autophagosomal membrane or at a later stage of the process as the targeting to or fusion with the vacuole, we examined autophagosome closure using protease protection assays. Whereas in wild-type cells about 50% of Ape1 is protected from cleavage by proteinase K, indicating that Ape1 is present in autophagosomes, all was cleaved in the Atg9 mutants (Figure 3C; Figure S2F). These findings suggest that Atg9 phosphorylation is required at an early step of autophagy either to initiate autophagosome formation or to ensure the proper growth or closure of these double-membrane vesicles.

18As Atg1 phosphorylates Atg9 and both proteins are part of the PAS, these proteins are predicted to at least transiently interact. To test Atg1-Atg9 binding, we used an in vivo proximity assay based on fusing a histone lysine methyltransferase domain to a bait protein while fusing the substrate, histone H3, to the prey protein. Upon binding, the prey is stably methylated in vivo, which can subsequently be detected by immunoblotting using methylation-specific antibodies. This assay revealed that Atg1 and Atg9 indeed interact in vivo, and that this interaction was not affected by the 6A or 6D mutations (Figure 3D). Similarly, soluble Atg1 was able to bind to the wild-type and nonphosphorylatable (4A) mutant of the recombinant Atg9 C-terminal fragment (Figure 3E). As expected, binding of Atg9 to Atg1 is not required for Atg1 kinase function in vivo, as Atg1 kinase purified from atg9Δ cells was found to be active (Figure 3F).

19Together, these results show that Atg1 kinase activity and association occur independent of Atg9 phosphorylation and suggest that unphosphorylated Atg9 is a substrate of Atg1.

20As Atg9 shuttling to and from the PAS depends on Atg1, we examined whether Atg9 phosphorylation may regulate this process. As previously reported, both atg1Δ and kinase-dead Atg1-D211A mutant cells show accumulation of Atg9 at the PAS (Figure S3A), demonstrating that the kinase activity of Atg1 is indeed required for Atg9 dynamics. Furthermore, Atg1 contains an LC3-interaction region (LIR), which allows binding to Atg8/LC3 and thereby targets Atg1 to autophagosomes, where it might regulate autophagosome formation. Mutation of the LIR motif in Atg1 (Atg1-V432A/E433A) results in normal PAS localization, but autophagosomal association is reduced. Interestingly, this LIR mutant Atg1 did not alter Atg9 shuttling, implying that Atg1 regulates Atg9 retrieval specifically at the PAS rather than on autophagosomes or the vacuolar membrane. Mutation of all six phosphorylation sites on Atg9, which causes defective Cvt and autophagy, did not affect its PAS localization or its distribution between the PAS and the cytosol (Figures 4A and 4B). Also, the Atg9-6A phospho mutant accumulated at the PAS in atg1Δ cells similar to the Atg9 wild-type (Figure S3B), demonstrating that the mutations do not interfere with its PAS recruitment. Taken together, these results show that the phenotype of the Atg9 phospho mutants does not simply mimic that of an Atg1 deletion. Our findings suggest that Atg1-dependent phosphorylation of Atg9 does not affect its shuttling to and from the PAS, but regulates a different step in the autophagy pathway.

21As both Atg1 and Atg9 are required to recruit the autophagy-essential Atg2 and Atg18 proteins to the PAS, we next analyzed whether their recruitment depends on Atg9 phosphorylation. As previously reported, both Atg2 and Atg18 formed a punctate structure representing PAS localization in wild-type cells upon rapamycin treatment. This PAS localization was lost in atg9Δ cells. Whereas Atg2 localized normally to the PAS in the Atg9 mutants (Figures S4A and S4B), Atg18 PAS localization was significantly decreased (Figures 4C and 4D). Also, Atg18 failed to localize to the PAS in the absence of active Atg1 (Figure S4C). These results demonstrate that Atg18 recruitment is to a large part dependent on Atg9 phosphorylation and suggest that the phospho-Atg9-independent Atg18 recruitment is not sufficient to rescue autophagy defects (compare Figures 3B and 4D).

22In line with this notion, it has previously been reported that Atg9 interacts with Atg18 in an Atg1-dependent manner. We reasoned that the decreased recruitment of Atg18 to the PAS might be due to a loss of interaction with the Atg9 phosphorylation site mutants. Indeed, endogenously expressed Atg18-TAP (tandem affinity purification) was able to coprecipitate with wild-type Atg9-GFP, and this interaction was dependent on the presence of active Atg1 (Figure 4E; Figure S4D). Consistently, the interaction of Atg18 with Atg9 was decreased after phosphatase treatment (Figure 4F), and Atg18 binding to the Atg9-6A and Atg9-6D mutants was hardly detectable (Figure 4E). Taken together, our results show that Atg1-dependent phosphorylation of Atg9 is essential for its interaction with Atg18.

23Atg8 localizes to the PAS during autophagosome formation. Phosphatidylethanolamine conjugation, the Atg12-Atg5-Atg16 complex, and Atg9 are all required for its proper recruitment. We therefore asked whether Atg1-dependent Atg9 phosphorylation is required for Atg8 recruitment to the PAS. As previously reported, Atg8 failed to be recruited to the PAS in atg9Δ cells (Figures 5A and 5B). Similarly, the Atg9-6A and Atg9-6D mutants showed a strong defect in Atg8 puncta formation, strongly suggesting that Atg1-dependent phosphorylation of Atg9 is required for this recruitment process. This failure in recruitment was neither due to a defect in Atg5-Atg12 conjugation nor to a failure in the lipidation of Atg8, as both events took place in the Atg9-6A and Atg9-6D mutants to the same degree as in wild-type cells (Figures S4E and S4F). To test whether the small number of Atg8 dots in Atg9 mutant cells represents functional autophagosomes, we analyzed GFP-Atg8 structures in cells expressing giant Ape1 oligomers. Ape1 overexpression results in the formation of a giant Ape1 oligomer, which is too large to be fully enwrapped by the isolation membrane. These oligomers become partially enwrapped by Atg8-containing membranes in wild-type cells and are visible as cup-shaped structures by fluorescence microscopy, allowing the visualization and analysis of elongated isolation membrane intermediates. Only 20% of Atg9 mutant cells showed Atg8 recruitment to the giant Ape1 oligomers, whereas in 80% of wild-type cells Atg8 colocalized with Ape1, suggesting that Atg9 phosphorylation is required for this recruitment step (Figures 5C and 5D). As previously reported, wild-type cells formed elongated cup-shaped Atg8-positive structures, although Atg9 mutant cells only formed small patches (Figures 5C and 5E). Together, these findings show that Atg9 phosphorylation is required at a very early step of autophagosome formation to promote productive PAS assembly competent of isolation membrane formation.

24In summary, we have established a powerful screening method combining peptide arrays with quantitative mass spectrometry to identify kinase substrates. This method allowed us to determine the consensus phosphorylation sequence of the Atg1 kinase and the subsequent identification of 25 bona fide Atg1 kinase targets. Detailed analysis of one of these targets shows that (Figure 6, i) Atg9 phosphorylation by Atg1 facilitates efficient Atg8 and Atg18 recruitment to the PAS and is absolutely required for (Figure 6, ii) the interaction of Atg9 and Atg18 and (Figure 6, iii) isolation membrane expansion, which are necessary steps for autophagy and Cvt pathway function.

25Atg1 kinase activity is essential for autophagy, but its mechanism of action is largely unknown. Here we determined the consensus sequence of the Atg1 kinase complex and used a genomic search to identify potential substrates. Among the candidates validated by in vitro phosphorylation, we identified the conserved core autophagy protein Atg9 as a direct Atg1 target. Atg9 phosphorylation by Atg1 is essential for its interaction with Atg18 (WIPI2 in mammals) and efficient recruitment of this key autophagy factor and Atg8 to the PAS. This in turn is required for PAS assembly capable of Atg8-positive isolation membrane formation and elongation.

26Consensus motifs of a large number of yeast kinases have previously been analyzed using peptide arrays. However, these studies used overexpressed kinases, likely resulting in the loss of less abundant regulatory factors and possibly important secondary modifications on the kinase, both of which could influence substrate selectivity. Atg1 is a low-abundant protein in the cell, and its activity requires Atg13, another low-abundant protein. Therefore, we purified the native Atg1 kinase complex from yeast, preserving regulatory factors such as associated proteins and posttranslational modifications. Among other proteins, these isolated complexes contain the known regulators Atg13, Atg17, and Atg11, as verified by mass spectrometry. They show autophosphorylation activity as well as phosphorylation of myelin basic protein (MBP) as a model substrate, and both are increased approximately two-fold upon starvation or rapamycin treatment prior to complex isolation. Incubation of peptide arrays with native Atg1 kinase complexes revealed a selective motif distinct from known kinase target motifs, which tend to prefer either charged residues (protein kinase A and casein kinases) or proline (cyclin-dependent kinases and MAP kinases). By contrast, Atg1 clearly preferred hydrophobic residues at multiple positions. Among previously characterized kinases, NimA-related kinases (yeast Kin3 and mammalian NEK6) also select hydrophobic residues at the −3 position, although they prefer aromatic over aliphatic residues. The notion of this Atg1 kinase consensus allowed us to identify a set of bona fide Atg1 kinase targets, among them the autophagy-essential protein Atg9.

27Atg2 and Atg18 are essential for autophagy and the Cvt pathway. Atg18 contains a WD40 propeller and directly binds to Atg2, and both proteins require Atg9 for their proper PAS recruitment and function. Our results demonstrate that phosphorylation of Atg9 by Atg1 regulates its interaction with Atg18 and promotes Atg8 and Atg18 recruitment to the PAS and subsequent elongation of the isolation membrane, which is required for the Cvt pathway and autophagy to function (Figure 6). Interestingly, we show that when Atg1 cannot phosphorylate Atg9, Atg18 is still partially recruited to the PAS. However, this partial recruitment is apparently not sufficient to support autophagosome formation. Therefore, either the residual level of Atg18 recruitment lies under the threshold required at the PAS to promote these pathways, or it is the interaction of Atg9 with Atg18 rather than mere PAS recruitment that is essential for these pathways to function. As the latter is completely disrupted in the Atg9 mutants, the interaction of Atg9 with Atg18, possibly both in the cytoplasm and at the PAS, likely plays a central role in the promotion of autophagy. Additionally, Atg8 fails to be recruited efficiently to the PAS in the absence of Atg9 phosphorylation, and isolation membranes do not elongate, possibly due to the lack of Atg9-Atg18 interaction.

28Atg9 phosphorylation by Atg1 is required both under rich conditions and upon starvation, as the Cvt pathway and autophagy are nonfunctional in Atg9 mutant cells. The serine-to-aspartate mutations in Atg9 seem unable to mimic the phosphorylated state. Although this unfortunately does not allow analysis of Atg9 function during continuous phosphorylation, it is conceivable that cycles of phosphorylation and dephosphorylation are required for progressive autophagy and the Cvt pathway. Further analyses of the precise mechanism, temporal control, and involvement of Atg9 phosphorylation during Atg9 vesicle cycling and PAS formation will be important directions for future studies.

29Interestingly, although Atg1 kinase activity is required for Atg9 retrograde transport, direct phosphorylation of Atg9 by Atg1 was not required for this event, as fluorescence microscopy analyses revealed no significant alterations in the localization of nonphosphorylatable Atg9 mutants. Our data do not exclude the possibility that both antero- and retrograde trafficking kinetics are altered and the overall movement of Atg9 vesicles is slowed down. However, it seems more likely that Atg1 regulates Atg9 trafficking by other means than direct phosphorylation of Atg9. Thus, the critical Atg1 kinase substrate controlling Atg9 shuttling remains to be identified. Notably, the Atg1 LIR mutant, which is defective in binding to autophagosomes but still localizes to the PAS normally, shows no change in Atg9 trafficking behavior, implying that the function of Atg1 in regulating Atg9 transport might be restricted to the PAS.

30With the complex interplay of many autophagy proteins in PAS function and autophagosome formation, future studies are required to unravel the precise interplay and mechanism underlying Atg1 signaling in autophagy and Cvt pathway function. The identification of Atg9 as an important and direct Atg1 kinase target adds another piece to this still largely unassembled puzzle.

31Yeast strains and plasmids are listed in Tables S2 and S3.

32Yeast cells were grown in 2 l of YPD medium to an OD600 of 2 and treated with 220 nM rapamycin for 1 hr, harvested by centrifugation, and washed in PBS, 2% glucose. Cells were then resuspended in a pellet volume of lysis buffer (PBS, 10% glycerol, 0.5% Tween-20, 1 mM NaF, 1 mM phenylmethylsulfonylfluoride, 1 mM Na3VO4, 20 mM β-glycerophosphate, protease inhibitor cocktail [Roche]) and frozen in droplets in liquid nitrogen. After cell disruption with a freezer mill (6770; SPEX), the extract was thawed in lysis buffer and cleared by centrifugation. The cleared extract was incubated with 160 μl of IgG-coupled magnetic beads (Dynabeads; Invitrogen) for 1 hr at 4°C with rotation. The beads were washed six times for 5–10 min in lysis buffer with rotation and cleaved with lysis buffer containing 0.5 mM DTT and TEV protease for 1 hr at 16°C with slow shaking.

33Atg1 was genomically tagged at the C terminus with histone H3-HA (H3HA), and Atg9 with a C-terminal histone lysine methyltransferase domain (HKMT) was cloned in a CEN plasmid under the control of its endogenous promoter. Both proteins were tested for functionality by analyzing Ape1 processing. The M-Track methylation assay was performed as described. Methylation signals were enriched by immunoprecipitating Atg1-H3HA on HA agarose.

34All p values were calculated using an unpaired two-tailed Student’s t test. A p value less than 0.05 was considered to be significant. The intensities from the MS1 filtering in Figure 2C were normalized to the values from the 60 min reactions, and the error bars represent the normalized standard deviation of the replicate injections. In Figure 3B, the mean and standard deviation were calculated from four individual experiments. P3D: 0.0025; P5D: 0.0020; P6D: 0.00022; P3A: 0.00033; P5A: 0.00013; P6A: 0.0000091; Pdelta: 0.0000006. Figures 4B, 4D, and 5B show the mean and standard deviation calculated from two individual experiments. Figures 5D and 5E show the mean and standard deviation calculated from four (wild-type) or two (mutant) individual experiments. For Figure 4B, at least 14 cells of each mutant were analyzed per experiment. For Figure 4D, at least 310 cells of each mutant per experiment were counted blindly. For Figure 5B, at least 280 cells of each mutant were counted per experiment. For Figures 5D and 5E, at least 25 cells of each mutant were analyzed per experiment.

35Additional methods can be found in the Supplemental Experimental Procedures.

36D.P., M.S., W.R., L.W., C.A.B., A.M., I.H., T.P., J.H.L., and C.K. performed the experiments shown in the manuscript. M.K., I.S., S.S.L., and A.B. performed initial experiments. D.P., W.R., L.W., C.A.B., A.M., B.E.T., R.A., G.A., M.P., and C.K. participated in the experimental design. M.P. and C.K. wrote the paper.