|Home | About | Journals | Submit | Contact Us | Français|
Autophagy is the degradation of a cell's own components within lysosomes (or the analogous yeast vacuole), and its malfunction contributes to a variety of human diseases. Atg9 is the sole integral membrane protein required in formation of the initial sequestering compartment, the phagophore, and is proposed to play a key role in membrane transport; the phagophore presumably expands by vesicular addition to form a complete autophagosome. It is not clear through what mechanism Atg9 functions at the phagophore assembly site (PAS). Here we report that Atg9 molecules self-associate independently of other known autophagy proteins in both nutrient-rich and starvation conditions. Mutational analyses reveal that self-interaction is critical for anterograde transport of Atg9 to the PAS. The ability of Atg9 to self-interact is required for both selective and nonselective autophagy at the step of phagophore expansion at the PAS. Our results support a model in which Atg9 multimerization facilitates membrane flow to the PAS for phagophore formation.
Autophagic degradation of unneeded or damaged cellular components is essential for various cellular functions including proper homeostasis. Along these lines, the malfunction of autophagy is implicated in a variety of diseases, including cancer, neurodegeneration, cardiac disorders, and pathogen infection (Shintani and Klionsky, 2004a ). During autophagy, cytosolic proteins and organelles are engulfed into a double-membrane vesicle, the autophagosome, which then fuses with a lysosome (or the vacuole in fungi and plants) where its cargos are degraded. The autophagy-related (Atg) protein Atg9 plays a central role in the nucleation step during autophagosome formation in eukaryotes ranging from yeast to mammals (Noda et al., 2000 ; Young et al., 2006 ). Being the only identified integral membrane protein that is absolutely required in autophagosome formation, Atg9 is proposed to be the “carrier” of lipids to the phagophore assembly site (PAS, also known as the preautophagosomal structure; Kim et al., 2002 ). Atg9 is absent from the completed autophagosomes, suggesting that the protein is retrieved upon vesicle completion. Previous work done in the yeast Saccharomyces cerevisiae has shown that Atg9 interacts with multiple autophagy-related proteins via its two cytosol-facing termini (Reggiori et al., 2005 ; He et al., 2006 ; Legakis et al., 2007 ; Yen et al., 2007 ). However, the physiological role of such a multisubunit complex in autophagy and the function of Atg9 in this complex are still unclear.
Selective autophagy, such as pexophagy (degradation of excess peroxisomes), mitophagy (clearance of damaged mitochondria), and the cytoplasm-to-vacuole targeting (Cvt) pathway, targets specific cargos (Xie and Klionsky, 2007 ). The Cvt pathway occurs during vegetative growth in yeast, in which two vacuolar hydrolases, α-mannosidase and the precursor form of aminopeptidase I (Ape1 [prApe1]), are transported to the vacuole where prApe1 is processed into mature Ape1. Nonselective, bulk autophagy occurs at a basal level and is induced by developmental signals and/or stress conditions (Levine and Klionsky, 2004 ). For instance, during starvation, nutrients are provided to ensure cell survival through elevated autophagic degradation of bulk cytoplasm and subsequent release of the breakdown products from the lysosome. Previous studies in yeast show a cycling route of Atg9 transport between the PAS and some peripheral compartments including mitochondria, during vegetative growth: the anterograde transport of Atg9 to the PAS facilitated by Atg9 binding partners may deliver lipids to the PAS, and the retrieval of Atg9 from the PAS may recycle the protein back to the membrane origin for the next round of delivery (Reggiori et al., 2004a ; He et al., 2006 ). Nevertheless, to date little is known about the mechanism targeting Atg9 to the PAS during starvation-induced bulk autophagy.
Unlike other intracellular trafficking vesicles that usually bud from the surface of a preexisting organelle, an autophagosome is thought to assemble by fusion of new membrane fragments with the phagophore, the initial sequestering compartment, at the PAS. However, the membranous structure at the PAS, or the expanding phagophore, has not been clearly elucidated; how small membranes are incorporated into this structure remains unknown. In this article, we present data that Atg9 interacts with itself in both nutrient-rich and starvation conditions independent of other Atg proteins. The self-interaction, which is mediated by the C terminus of the protein, promotes the trafficking of Atg9 from its origins to the PAS and is required for both selective and nonselective starvation-induced autophagy. Through examining the expansion of Atg9-containing phagophores by fluorescence and immunoelectron microscopy, we found that the ability of Atg9 to multimerize is an essential function during formation of a normal phagophore at the PAS. Our data provide new conceptual insights to the molecular mechanism governing Atg9 anterograde transport and assembly of the PAS during bulk autophagy and hence refine the cycling model of Atg9 transport.
The S. cerevisiae strains used in this study are listed in Table 1. For disruption of ATG9, the entire coding region was replaced by the Kluyveromyces lactis LEU2 gene using PCR primers containing ~45 bases of identity to the regions flanking the open reading frame. For PCR-based integration of the green fluorescent protein (GFP) or tandem affinity purification (TAP) tag, pFA6a-GFP(S65T)-TRP1, or pBS1479 and pBS1539, was used as the template, respectively (Longtine et al., 1998 ; Puig et al., 2001 ). For integration of the Atg9-3GFP fusion, the integrative plasmid pATG9–3GFP(306) was linearized by digestion with StuI and integrated into the URA3 gene locus. For integration of the Atg9-3DsRed fusion, the DNA fragment containing the ATG9 gene and native promoter was released from pATG9–3GFP(306) and cloned into pTPIARP2–3DsRed(305) using XhoI and BamHI; the resulting integrative plasmid pAtg9-3DsRed(305) was linearized by digestion with AflII and integrated into the LEU2 gene locus.
Yeast cells were grown in rich medium (YPD; 1% yeast extract, 2% peptone, 2% glucose) or synthetic minimal medium (SMD; 0.67% yeast nitrogen base, 2% glucose, amino acids, and vitamins as needed). Starvation experiments were conducted in synthetic medium lacking nitrogen (SD-N; 0.17% yeast nitrogen base without amino acids and 2% glucose).
Plasmids expressing HA-Atg13 (pHAAtg13(315); Cheong et al., 2008 ), RFP-Ape1 (pRFPApe1(414); Stromhaug et al., 2004 ), GFP-Atg8 (pGFP-AUT7(414); Abeliovich et al., 2003 ); Atg9 (pAPG9(416), essentially constructed the same as pAPG9(414); Noda et al., 2000 ); Atg9-GFP (pAPG9GFP(416) and pCuAPG9GFP(416); Noda et al., 2000 ); Atg9-PA (pAtg9PA(314); He et al., 2006 ); HA-Atg11 (pCuHA-CVT9(414); Kim et al., 2001 ); cyan fluorescent protein (CFP)-Atg11 (pCuHACFPCVT9(414); Kim et al., 2002 ); GFP-Atg2 (pCuGFPAPG2(414); Wang et al., 2001 ); and Atg18-GFP (pCVT18GFP(414); Guan et al., 2001 ) have been described previously. Yeast two-hybrid plasmids expressing Atg11 (pAD-Atg11), Atg18 (pAD-Atg18), Atg23 (pAD-Atg23), Atg27 (pAD-Atg27), Atg9 (pAD-Atg9 and pBD-Atg9), and the Atg9 N terminus (pAD-Atg9N) or the C terminus (pAD-Atg9C) have been described previously (He et al., 2006 ).
The plasmids pAtg9Δ787-997(416) and pAtg9Δ870-997(416) were generated by amplifying the Atg9Δ787-997 and Atg9Δ870-997 fragments from pAPG9(416) and cloning them into the two AatII sites in pAPG9(416). To generate the plasmid expressing Atg9Δ766-997-GFP driven by the TPI1 promoter (pS1S2(416)), the N terminus of ATG9 (928 base pairs) was amplified and cloned in the vector pPEP416 (Reggiori et al., 2000 ) digested with EcoRI/SacII. The 3′ primer introduced an XbaI site and a SacII site preceded by a stop codon. The resulting pS1(416) plasmid was then cut with XbaI/SacII, and the central part of ATG9 (1404 base pairs) was amplified and inserted using the same enzymes. For internal deletions of Atg9 (Atg9Δ766-785-GFP, Atg9Δ766-770-GFP, Atg9Δ771-775-GFP, Atg9Δ776-780-GFP and Atg9Δ781-785-GFP), the truncated open reading frames were amplified by PCR and cloned into NotI and BamHI sites of pAPG9GFP(416). For generation of nontagged pAtg9Δ766-785(416), pAtg9Δ766-770(416), pAtg9Δ781-785(416), or pCuAtg9Δ766-770-GFP(416), and pAtg9Δ766-770–3HA(426), the fragment containing the indicated deletion was released from pAtg9Δ766-785-GFP(416), pAtg9Δ766-770-GFP(416), or pAtg9Δ781-785-GFP(416) by AgeI and SphI digestion and introduced into pAPG9(416), pCuAPG9GFP(416), or pAtg9-3HA(426), respectively. Point mutations in Atg9 amino acids 766–770 were introduced by site-directed mutagenesis. To construct the two-hybrid plasmid pBD-Atg9Δ766-770, the Atg9Δ766-770 fragment was amplified from pAtg9Δ766-770-GFP(416) and cloned into pGBDU-C1 using BamHI and SalI sites.
Cells were grown to OD600 = 0.8 in SMD; for rapamycin treatment, cells were cultured with 0.2 μg/ml rapamycin at 30°C for an additional 2 h. Fifty milliliters of cells was harvested and resuspended in lysis buffer (20 mM Tris-HCl, pH 7.5, 150 mM KCl, 5 mM MgCl2, 1% Triton X-100, 1 mM PMSF, and protease inhibitor cocktail). The detergent extracts were incubated with IgG-Sepharose beads overnight at 4°C. The beads were washed with lysis buffer six times and eluted in SDS-PAGE sample buffer by incubating at 37°C for 30 min. The eluates were resolved by SDS-PAGE and immunoblotted with anti-Atg9 antiserum or anti-HA or anti-PA antibody.
Fluorescence signals were visualized on an Olympus IX71 fluorescence microscope (Mellville, NY). The images were captured by a Photometrics CoolSNAP HQ camera (Roper Scientific, Tucson, AZ) and deconvolved using DeltaVision software (Applied Precision, Issaquah, WA). When necessary, a mild fixation procedure was applied as previously described to visualize Atg9 without affecting various fluorescent proteins (He et al., 2006 ).
Bis-Tris gels (3–12%; Invitrogen, Carlsbad, CA) were used for blue native gel electrophoresis. Spheroplasts were generated from 25 ml of cells at midlog phase as described previously (Tomashek et al., 1996 ), and cell lysates were prepared according to the Invitrogen NativePAGE Novex Bis-Tris Gel System manual.
Four hundred milliliters of cells were cultured to OD600 = 0.8 in YPD. Spheroplasts were prepared and TAP purification was done as previously reported (Tomashek et al., 1996 ; Puig et al., 2001 ). The eluates were resolved by SDS-PAGE, and silver staining of proteins in polyacrylamide gels was performed using the SilverSNAP Stain Kit II (Thermo Scientific, Rockford, IL).
Immunoelectron microscopy (IEM) was performed according to the procedures described previously (Baba et al., 1997 ) with the following modifications: The blocking solution contained 0.05% Tween 20 or 0.5% cold fish gelatin (Sigma-Aldrich, St. Louis, MO). The anti-GFP (JL-8, Clontech/Takara Bio Group, Mountain View, CA) antibody was preadsorbed with a protein extract prepared from atg9Δ cells to reduce nonspecific staining. The secondary antibody was conjugated to Ultra Small gold particles (Aurion, Wageningen, The Netherlands) and visualized with silver enhancement.
Previously we showed that Atg9 is located at the PAS and multiple peripheral punctate sites (Reggiori et al., 2004a ). To determine whether Atg9 directly self-interacts, we took advantage of a new reagent, a yeast strain lacking 24 known ATG genes that function in S. cerevisiae, referred to as the multiple-knockout (MKO) strain (Cao et al., 2008 ). Analyses by fluorescence microscopy revealed that Atg9-GFP fusion proteins were indeed organized in clusters (detected as large puncta) in this strain, as well as in the wild-type strain (Figure 1A), whereas expressing GFP alone did not lead to observable cluster formation (Supplemental Figure S1). These findings led us to propose that Atg9 may form clusters through direct self-association involving none of the other known Atg proteins. To test this hypothesis, we coexpressed Atg9 proteins fused with two different tags, GFP and DsRed, in MKO and wild-type strains. As shown in Figure 1B, the Atg9-GFP and Atg9-DsRed fusion proteins were colocalized in the cell as assessed by fluorescence microscopy.
Next, we used a biochemical coimmunoprecipitation approach to test the self-interaction of Atg9. As shown in Figure 1C, Atg9 was coprecipitated with Atg9-protein A (PA) in the MKO strain as well as in the wild-type strain in nutrient-rich conditions. As a negative control, we examined the ability of TAP-tagged Atg1 to coimmunoprecipitate Atg9-GFP; in this case, we could not examine endogenous Atg9 because of its migration at the same position as TAP-Atg1 during SDS-PAGE. In contrast to the result with Atg9-PA, Atg9-GFP was not recovered with TAP-tagged Atg1; Atg1 interacts with various other proteins including Atg13 but not Atg9, and we verified that TAP-Atg1 was able to coimmunoprecipitate Atg13 (Supplemental Figure S2A). Similarly, Atg9-PA was not able to coimmunoprecipitate Atg1 (Supplemental Figure S2B). Along with our previously published yeast two-hybrid results suggesting Atg9 self-interaction (Reggiori et al., 2005 ), these data demonstrated that multiple Atg9 molecules were able to form a complex without any other known Atg proteins.
Atg9 clustering occurs not only in nutrient-rich conditions but also during nitrogen starvation (Figure 1A). To explore whether the same Atg9 complex forms during bulk autophagy, the MKO cells were subjected to nitrogen starvation or treatment with rapamycin, a drug that partly mimics starvation conditions and induces bulk autophagy. Atg9 was coprecipitated with Atg9-PA in both conditions in the MKO strain (Figure 1D). As before, Atg9-GFP was not coisolated with TAP-tagged Atg1. Similar results were obtained using wild-type cells (Supplemental Figure S2C). In addition, we found that Atg9 self-interaction is enhanced during starvation, based on quantification of the relative Atg9 amount precipitated by Atg9-PA (Supplemental Figure S2D), implicating an important role of self-interaction during autophagosome formation. Collectively, we concluded that Atg9 self-interacts independent of any other known Atg proteins in both nutrient-rich and starvation conditions, which may correlate with a role for Atg9 in both selective and nonselective bulk autophagy.
Next, we wanted to study the physiological function of Atg9 self-interaction in autophagy. Accordingly, we first mapped the interaction domain by the yeast two-hybrid approach. In the presence of full-length Atg9, the Atg9 C-terminal domain supported the growth of two-hybrid cells on selective plates lacking histidine as well as full-length Atg9, whereas the N terminus was not able to do so (Figure 2A), indicating that Atg9 self-interaction is mediated through the C terminus. To narrow down the critical region, we further constructed a series of Atg9 C-terminal deletion mutants (Figure 2B) and analyzed them using the coimmunoprecipitation assay. Atg9, Atg9Δ870-997, and Atg9Δ787-997, but not Atg9Δ766-785 or Atg9Δ766-997, were pulled down by full-length Atg9-PA (Figure 2C and unpublished data), suggesting that amino acids included in the region of 766–786 were required for self-interaction. Atg9 and the PA vector were coexpressed as a control and no detectable Atg9 was coprecipitated by PA. On the basis of this result, we generated four additional mutants each lacking five consecutive amino acids within the sequence of amino acids 766–786. By coimmunoprecipitation assays we determined that Atg9Δ766-770 was unable to interact with Atg9, whereas the other mutants displayed a normal interaction (Figure 2D and unpublished data). We note that amino acids 766–770 are highly conserved through evolution (Figure 2E).
Because Atg9 interacts with several other known Atg proteins, we further tested the ability of Atg9Δ766-770 to bind these other binding partners by yeast two-hybrid assays. Although Atg9Δ766-770 had a significantly compromised interaction with Atg9, it was still able to bind to its other binding partners, including Atg23, Atg27, Atg18, and Atg11, with apparently normal affinity (Figure 2F). In addition, we verified that the interaction between Atg9Δ766-770 and Atg23 was not impaired based on coimmunoprecipitation; we used a vector expressing PA alone as a negative control (Figure 2G). Therefore, the C-terminal mutant Atg9Δ766-770 specifically disrupts the self-interaction of Atg9 and was used for the following functional analyses.
To study the function of Atg9 self-interaction in autophagy, we adopted several established assays using the Atg9Δ766–770 mutant. The processing of prApe1 into mature Ape1 results in a migration shift during SDS-PAGE and can be monitored as an indicator for selective autophagy. As shown in Figure 3A, Atg9Δ766-770, which impaired the self-interaction of Atg9, blocked the maturation of prApe1, whereas the other three deletion mutants, which had no effect on Atg9 self-interaction, retained the capacity of prApe1 maturation similar to wild-type Atg9, although all of the Atg9 mutants displayed a level of stability at least equivalent to that of the wild-type protein. This indicated that the self-interaction of Atg9 is indispensable for selective autophagy.
We also observed that Atg9 self-interaction occurs under nitrogen-starvation conditions, which suggests that a protein complex containing multiple Atg9 proteins may be involved in bulk autophagy. To test this hypothesis, we carried out two assays to measure bulk autophagy activity when Atg9 self-interaction was altered. Atg8 is conjugated to phosphatidylethanolamine (PE) and remains associated with the completed autophagosome and thus is a marker for autophagy progression (Kirisako et al., 1999 ; Huang et al., 2000 ). During autophagy, the GFP-tagged Atg8 is transported to the vacuole where Atg8 is rapidly degraded, whereas the GFP moiety remains relatively stable. Thus, the accumulation of free GFP detected by Western blot reflects autophagy activity (Shintani and Klionsky, 2004b ). We assayed GFP-Atg8 processing with the mutants mentioned above and found that only with Atg9Δ766-770 and Atg9Δ766-785, both of which affected Atg9 self-interaction (Figure 2D and unpublished data), GFP-Atg8 processing was blocked. With two other mutants Atg9Δ781-785 and Atg9Δ787-997 that did not affect Atg9 self-interaction, GFP-Atg8 was processed similar to cells expressing wild-type Atg9 (Figure 3B). These data demonstrated that loss of Atg9 self-interaction caused a defect in bulk autophagy. To quantitatively confirm this result, we measured the Pho8Δ60 enzymatic activity. Pho8Δ60 is a truncated form of alkaline phosphatase that can be delivered to the vacuole only via autophagy (Noda et al., 1995 ). Approximately fourfold induction of Pho8Δ60 activity by starvation was observed with wild-type Atg9 and Atg9Δ781-785, whereas the empty vector and Atg9Δ766-770 showed only the basal level of Pho8Δ60 activity (Figure 3C). We further introduced alanine mutations at each conserved residue (Figure 2E) in the region 766–770, and found that these mutations also caused defects in the Cvt pathway and bulk autophagy (Table 2). Thus, even point mutations in the highly conserved interaction domain interfere with Atg9 function. Taken together, these results indicate that Atg9 self-interaction is also functionally essential for bulk autophagy. Thus our data revealed a previously unknown mechanism of Atg9 shared in both selective and bulk autophagy, involving formation of a multiple Atg9-containing complex that depends on interaction between Atg9 proteins.
We decided to investigate the underlying mechanisms of the functional defects seen with the Atg9Δ766-770 mutant. In the absence of Atg9, a number of Atg proteins are not correctly localized to the PAS, including Atg2, Atg14, and Atg18 (Suzuki et al., 2001 ). Therefore it is possible that formation of a multimeric Atg9 complex is required for recruitment of these proteins to the PAS. To examine this possibility, we visualized the localization of Atg2, Atg18, and Atg14 in cells expressing wild-type Atg9 or Atg9Δ766-770. As shown in Figure 4, in the presence of either wild-type Atg9 or Atg9Δ766-770, Atg2, Atg18, and Atg14 localized to a primary perivacuolar punctum, which colocalized with RFP-Ape1 and corresponded to the PAS. These data suggested that PAS recruitment of Atg proteins by Atg9Δ766-770 was not affected and thus was not the causal factor for the autophagy deficiencies.
We then hypothesized that the self-interaction may be involved in the trafficking of Atg9. According to the “cycling” model of Atg9 transport, when the retrograde transport of Atg9 is impaired, Atg9 will accumulate at the PAS as one primary punctum (Reggiori et al., 2004a ), but this accumulation was not observed with Atg9Δ766-770 in otherwise wild-type yeast cells; besides, Atg9Δ766-770 partially colocalized with mitochondria similar to wild-type Atg9 (unpublished data). Thus, Atg9 self-interaction is unlikely to be involved in the retrograde trafficking of Atg9 back to the peripheral (i.e., non-PAS) sites. Accordingly, to study whether self-interaction is involved in Atg9 anterograde transport, we used the TAKA (transport of Atg9 after knocking out ATG1) assay (Cheong et al., 2005 ). Atg1 is a key regulator that activates the retrieval of Atg9 from the PAS to the peripheral sites and deleting ATG1 restricts Atg9 to the PAS. The TAKA assay examines the epistasis of a second mutation relative to atg1Δ with regard to Atg9 localization at the PAS. Using fluorescence microscopy, we imaged the localization of Atg9Δ766-770-GFP and wild-type Atg9-GFP in atg1Δ cells in nutrient-rich and starvation conditions.
As shown in Figure 5A, in contrast to wild-type Atg9, which was restricted to the PAS (marked with RFP-Ape1) in 87% (52/60) of the cells, the Atg9Δ766–770 mutant distributed to multiple punctate or dispersed structures, in addition to the PAS location in 70% (52/74) of the cells. Such localization was observed regardless of nutrient supply (Figure 5B). During nitrogen starvation, in 53% (23/43) of the atg1Δ cells, Atg9Δ766-770 localized to multiple puncta, compared with only 11% (12/107) of the cells displaying peripheral localization with wild-type Atg9. Thus, these results demonstrated that loss of Atg9 self-interaction partially blocked the anterograde trafficking of Atg9 to the PAS during selective and bulk autophagy. Self-interacting and forming a complex could concentrate Atg9 molecules as clusters at the peripheral compartments, which may be important for efficient trafficking of Atg9 from these sites for subsequent delivery to the PAS.
On nutrient deprivation, formation of a continuous cup-shaped or ring-like structure by wild-type Atg9-GFP was visualized around the cargo prApe1 at the PAS and colabeled with Atg8 by fluorescence microscopy (Figure 5, B and C), suggesting that they were authentic phagophores. These structures were difficult to detect in wild-type cells, presumably because of their transient nature and the rapid dissociation of Atg9 from them upon autophagosome completion. Previous studies, however, show that autophagosomes emerge from Atg8-labeled intermediates in atg1 temperature-sensitive cells, indicating that in this mutant these are not dead-end structures (Suzuki et al., 2001 ). The cup-shaped structures were ~500 nm in diameter, which fits within the size range of completed autophagosomes in yeast (400–900 nm; Takeshige et al., 1992 ). Together, these data suggested that the expanding phagophore (or precursor membrane of the phagophore) is an Atg9-containing intermediate. Additionally, the Atg9-containing structures were seen with various sizes and curvatures (Supplemental Figure S3), which may represent different expansion stages of the phagophore. Nonetheless, we cannot definitively conclude that the structures accumulating in the atg1Δ mutant are authentic phagophores, and in particular the size may appear exaggerated by fluorescence microscopy because of the accumulation of Atg9-GFP as a result of the ATG1 deletion. With this caution in mind, we refer to the Atg9-containing structure as the phagophore for simplicity.
Interestingly, the phagophore structure was fragmented and/or failed to elongate normally when Atg9Δ766-770-GFP replaced the wild-type Atg9-GFP (Figure 5B). To further characterize the Atg9-containing structures at the PAS at a higher resolution, we applied IEM. Atg9-GFP or Atg9Δ766-770-GFP was expressed in an atg1Δ atg9Δ strain, and the Atg9 protein was detected with anti-GFP antibody. As previously reported (Baba et al., 1997 ), the Cvt complex formed by prApe1 and its receptor Atg19 was localized as a spherical electron-dense particle usually near the vacuole. Consistent with our result by fluorescence microscopy, wild-type Atg9 gold particles (arrows) were concentrated on the surface of membranous structures (arrowheads) surrounding the Cvt complex, whereas the Atg9Δ766-770 mutant that lost self-interaction appeared less clustered at the PAS (Figure 6, A and B). In sections prepared by freeze substitution but not immunostained, the membrane structures were more readily detected. In this case, it appeared that there were more membranes, potentially corresponding to the phagophore, surrounding the Cvt complex in wild-type cells relative to the Atg9Δ766-770 mutant. This result suggests that Atg9 self-interaction may be required to foster expansion of, or maintain the integrity of, the phagophore at the PAS.
To quantitatively analyze this difference, we further counted the average number of wild-type or mutant Atg9 molecules around the Cvt complex. As shown in Figure 6C, the molecule number of Atg9Δ766-770 per Cvt complex was markedly reduced compared with wild-type Atg9, suggesting that Atg9 self-association is essential for efficient delivery to the PAS. On the basis of these data and our previous Atg9 cycling model, we propose that an Atg9 complex is dependent on its self-interaction and that the complex formation may facilitate not only the flow of membrane to the PAS but also the fusion of small membrane fragments into a continuous larger phagophore to form an autophagosome.
Because Atg9Δ766-770 partially affected the anterograde trafficking of Atg9 to the PAS, it is possible that the abnormal phagophore structure we observed with Atg9Δ766-770 is due to insufficient supply of this protein to the PAS. To test this possibility, we relied on the overexpression of Atg11, which increases the anterograde transport of Atg9 to the PAS (He et al., 2006 ). We imaged the GFP-tagged Atg9 and CFP-tagged Atg11 pairs with a yellow fluorescent protein (YFP)/CFP filter set by fluorescence microscopy, and, as reported previously (Kaksonen et al., 2003 ), no detectable bleed-through of the GFP signal from the CFP filter was seen. Because Atg9Δ766-770 did not affect its interaction with Atg11 (Figure 2F), we were able to overcome the inefficient anterograde transport of Atg9Δ766-770 to the PAS by overexpressing Atg11 in both atg1Δ and wild-type cells (Figure 7A and unpublished data), such that in nutrient-rich conditions the subcellular localization of Atg9Δ766-770 (68%; 36/53 cells) basically was indistinguishable from wild-type Atg9 (84%; 48/57 cells). The transport restoration was also observed in starvation conditions (Figure 7A; 73% of Atg9Δ766-770 cells [36/49] and 84% of wild-type Atg9 cells [46/55]). This phenotype indicates that the deletion of residues 766–770 did not absolutely block the ability of Atg9 to transit to the PAS and rules out accumulation in the endoplasmic reticulum due to protein misfolding as the cause of the functional defect. The increase in PAS localization of Atg9Δ766-770, however, did not result in the maturation of prApe1, indicating that this type of selective autophagy was still defective (Figure 7B). The Pho8Δ60 activity assay also showed that bulk autophagy was not rescued by enhancing the anterograde flow of the mutant Atg9 (Figure 7C). In either case, the overexpression of Atg11 did not interfere with the function of wild-type Atg9, indicating that this level of Atg11 did not cause a dominant negative phenotype. Therefore, these results clearly suggested that a normal multimeric state or the ability to multimerize is essential for Atg9 function after it reaches the PAS, which is a critical step during phagophore formation.
To further investigate the nature of the Atg9 complex, we performed a native gel analysis using the MKO strain expressing wild-type Atg9 or the mutant. We discovered that Atg9Δ766-770 migrated faster than wild-type Atg9 in native conditions (Figure 8A). As a control, we examined the dodecamer complex assembled by prApe1 (Kim et al., 1997 ) and found that it remained intact, suggesting that Atg9Δ766-770 was not able to form complexes as large (or as stable) as those seen with the wild-type protein. Consistent with the previous results showing that overexpression of Atg11 did not rescue the functional defect of the Atg9Δ766-770 mutant (Figure 7B), overexpression of Atg11 did not rescue the defect in complex formation of Atg9Δ766-770 in native conditions (Figure 8B). We further analyzed the Atg9 complex size in the wild-type background, by expressing wild-type or mutant Atg9 in the atg9Δ strain. As shown in Figure 8C, both wild-type and mutant Atg9 displayed similar migration patterns as in the MKO cells, suggesting that the known Atg9 interaction partners among Atg proteins may be dynamically associated with Atg9, rather than stably present in the core Atg9 complex. In addition, consistent with Figure 1 and Supplemental Figure S2, we also detected that the Atg9 complex remained intact upon rapamycin treatment (Figure 8C).
We noticed that Atg9Δ766-770 did not migrate as a monomer (~150 kDa on native gels, unpublished data), but ran at a position that might correspond to a tetramer. This might reflect the fact that the mutant only partially impaired Atg9 complex formation as suggested by a compromised, but not complete loss of, interaction seen by the yeast two-hybrid assay (Figure 2F), but it is also possible that the Atg9 complex contains unknown protein components as potential regulators of Atg9 function. To distinguish between these two possibilities, we performed a native purification of the Atg9 complex by the TAP method. We generated a strain expressing the chromosomally tagged Atg9-TAP fusion and another strain expressing the TAP tag alone integrated immediately after the ATG9 promoter as a control. The TAP tag was cleaved from Atg9-TAP after the first immunoprecipitation by IgG Sepharose, and the Atg9 complex was isolated by a second immunoprecipitation with calmodulin affinity resin. The proteins precipitated by Atg9 were analyzed on SDS-PAGE gels and detected by silver stain. Besides Atg9 (at the size of ~130 kDa), four major protein bands were detected at ~94, 54, 28, and 25 kDa, compared with the control (Figure 8D), suggesting that the Atg9 complex is actually composed of multiple proteins, although we cannot rule out at present that these represent degradation products. It will be helpful to further study the identity and function of these components in the Atg9 complex during phagophore formation and expansion.
Atg9 is the only known integral membrane protein that is required in the formation of sequestering vesicles during all types of autophagy, including the Cvt pathway, pexophagy, mitophagy (Kanki and Klionsky, 2008 ) and bulk autophagy, and it may play a key role in lipid delivery to the PAS. In addition to the PAS, Atg9 is localized at cytoplasmic punctate structures, which undergo dynamic movement revealed by time-lapse microscopy (Reggiori et al., 2005 ). Our data indicate that the clustered organization of Atg9 may be mediated by the ability to self-interact. We also found that self-interaction is important for efficient anterograde transport of Atg9 to the PAS (Figure 5). It is possible that Atg9 multimerization as clusters may function in promoting membrane “budding,” although the mechanism that triggers Atg9 movement from the peripheral compartments is not clear.
Our microscopy analyses revealed for the first time that Atg9 is localized on the membrane structures enwrapping the Cvt complex at the PAS, and thus we suggest that Atg9 can be used as a marker for monitoring phagophore expansion, although the PAS in the atg1Δ mutant represents a relatively static structure compared with the authentic phagophore in wild-type cells. Also, we do note that we cannot unequivocally identify the apparent membrane fragments as belonging to the phagophore because we did not carry out dual immunolabeling with anti-Atg8, which is known to localize to the PAS in the atg1Δ strain. We speculate that the precise regulation of Atg9 self-interaction may facilitate tethering and fusion of small membranes at the PAS. Notably, no yeast SNAREs (N-ethylmaleimide-sensitive factor attachment protein receptor) have yet been localized to the PAS (Reggiori et al., 2004b ), implying that membrane fusion at, and closure of, the phagophore may adopt a mechanism different from the conventional SNARE-mediated manner, although it is possible that SNAREs are present at too low a level to detect by fluorescence microscopy. Recent studies suggest that lipidated Atg8–PE mediates membrane hemifusion of liposomes in vitro (Nakatogawa et al., 2007 ). Yet, how Atg8 leads to membrane hemifusion in vivo and how hemifusion of lipid bilayers contributes to phagophore expansion is not clear. A recent analysis of Atg8 function suggests that Atg8 may largely determine the autophagosome size, but not the biogenesis of the initial phagophore or its completion (Xie et al., 2008 ). As suggested in this study, it is possible that Atg9 plays a critical role in initiating the phagophore. Furthermore, Atg9 along with Atg8–PE, and other Atg proteins, actively participate in the fusion process that expands the phagophore into the autophagosome. However, as noted above, it is not yet known whether Atg8–PE is present on the Atg9-containing membrane segments that are involved in autophagosome biogenesis.
It should be noted that the cup-shaped phagophore is formed in cells lacking Atg1 (Figures 5B, B,6,6, A and B, and Supplemental Figure S3). Thus the Atg1 kinase, originally proposed to function in phagophore initiation, appears to be dispensable for this process. In addition, both Atg2 and Atg18 are absent from the PAS in atg1Δ cells (Suzuki et al., 2007 ). Therefore, the cup-shaped phagophore we observed around prApe1 seems to form independently of Atg1, Atg2, and Atg18. Given the previous reports showing that the absence of any of these three proteins causes accumulation of Atg9 at the PAS (Reggiori et al., 2004a ), it would be reasonable to hypothesize that Atg1, Atg2, and Atg18 function at the vesicle completion step and promote dissociation of Atg9 after vesicle sealing, and lacking any of them will arrest the phagophore as an intermediate structure. Notably, Atg1 also seems to function at a late stage and is not required for induction of micropexophagy in the methylotrophic yeast Pichia pastoris (Mukaiyama et al., 2002 ). Thus, the protein machinery responsible for autophagy induction, and the exact role of Atg1, remain to be further elucidated.
The MKO strain allowed us to investigate the formation of the Atg9 complex bypassing the complexity caused by multiple Atg9-binding partners among Atg proteins. Using Atg9 as a phagophore marker, the MKO strain will provide advantages for studying proteins required in distinct steps during autophagosome formation including nucleation, expansion, and completion of the sequestering vesicle. We found that additional proteins may be stably present in the Atg9 complex (Figure 8D), and it is possible that they function as regulatory units. It will be interesting to further characterize these members of the Atg9 complex for a better understanding of membrane dynamics during de novo autophagosome biogenesis.
The authors thank Dr. Fulvio Reggiori (University Medical Centre Utrecht) for providing the pS1S2(416) plasmid, Dr. Noriko Nagata (Japan Women's University) for the use of the electron microscopy facilities, members of the Klionsky lab for valuable comments and suggestions to this project, and Drs. Weibin Zhou and Clinton Bartholomew for critical reading of the manuscript. This work is supported by a Rackham Predoctoral Fellowship to C.H. and National Institutes of Health Public Health Service Grant GM53396 to D.J.K.
This article was published online ahead of print in MBC in Press (http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E08-05-0544) on October 1, 2008.