Atg8/LC3–PE Is Not Sufficient to Drive Membrane Hemifusion in Physiological Conditions
Previously, Atg8–PE was shown to drive the hemifusion of membranes, a function implicated in the expansion of the phagophore (Nakatogawa et al., 2007
). Those data, however, relied on in vitro
fusion reactions using liposomes containing 55% phosphatidylethanolamine. PE plays a critical role in membrane dynamics; it is a cone-shaped “non-bilayer preferring” lipid with a strongly negative spontaneous curvature that facilitates hemifusion (Chernomordik and Kozlov, 2003
). As such, liposomes having 50% or more PE are inherently unstable and prone to fusion by a variety of factors not associated with membrane dynamics or vesicle trafficking (Brugger et al., 2000
). Therefore, we sought to examine the putative role of Atg8–PE in membrane hemifusion using liposomes with physiological concentrations of PE.
The concentration of PE at the phagophore is not known; however, the concentration in organelles implicated as possible sources of the autophagosome membrane (Hailey et al., 2010
; Lynch-Day et al., 2010
; Ravikumar et al., 2010
; Yen et al., 2010
) ranges from 15-20% at the yeast ER and Golgi, to as much as 25% at the plasma membrane and mitochondria (van Meer et al., 2008
). In order to establish lipidation conditions at these PE surface densities, we expressed and purified Atg8G116
(hereafter Atg8ΔR), Atg7, and Atg3 and incubated all three with liposomes comprised of varying amounts of PE. When ATP was included to drive the reaction, Atg8ΔR was efficiently lipidated in a PE-concentration-dependent manner ().
Physiological Levels of Atg8–PE Are not Sufficient to Drive Membrane Hemifusion
As in the Nakatogawa et al
. paper, Atg8–PE could mediate tethering of liposomes. For example, reactions run with 55 mole % PE liposomes produced robust light scattering, consistent with liposome aggregation, while negative controls (-ATP, -Atg3, -Atg8) did not (Figure S1A
). Reduction of the PE mole % had very little impact on this aggregation all the way down to 30%. In fact, the ability to cluster liposomes is well correlated with the capacity for lipidation, and both readouts were close to maximal at physiologically-relevant PE concentrations (25-30%; ). Thus, tethering requires lipidated Atg8, but is not dependent upon unusually high PE surface densities.
Using a fluorescence dequenching assay, we tested whether these aggregated liposomes had fused. Atg8–PE mediated fusion in a protein (Atg8ΔR) concentration-dependent manner when liposomes contained 55% PE (). Negative controls—reactions lacking Atg7—also exhibited a modest level of membrane fusion independent of Atg8ΔR protein concentration, and illustrated the inherent fusogenicity of liposomes with high PE. As previously noted (Nakatogawa et al., 2007
), high concentrations of Atg7 and Atg3 inhibited fusion even as lipidation continued, confirming that we were measuring the same phenomenon previously described. When liposomes containing physiological levels of PE (30%) were tested for fusion, however, only a negligible increase in NBD fluorescence was recorded, compared to negative controls () although Atg8ΔR was conjugated to PE with efficiencies similar to those obtained using liposomes containing 55% PE (). Indeed, the rate of fusion of the “negative control” 55% PE liposomes (measurements in the absence of Atg7) was always higher than any experimental condition utilizing lower surface densities of PE regardless of coupling or Atg8ΔR levels (Figure S1B
The high concentration of PE (55%) used in the initial report (Nakatogawa et al., 2007
) was perhaps necessitated in part because the efficiency with which Atg8 couples to liposomes depends on the surface density of PE (). As a result, in this system it is very difficult to separate the contribution of PE to membrane fusion from the critical role of PE in coupling the putative fusogen. Furthermore, high concentrations of Atg7and Atg3 tend to inhibit fusion (; (Nakatogawa et al., 2007
). All told, factors that maximize conjugation (high concentrations of substrate lipid, and coupling enzymes) are each confounding factors in interpreting the fusogenicity of Atg8. Thus, we designed a novel lipidation strategy that is independent of PE surface densities and that does not require Atg7, Atg3 or ATP (Figure S1C
). We engineered Atg8 with a C-terminal cysteine that can be covalently bonded to a phospholipid carrying a maleimide reactive site on its headgroup. In this way, the total number of membrane reactive sites is a function of the maleimide density, and the liposome PE surface density can be varied independently.
Maleimide-coupled Atg8 also promoted lipid mixing on liposomes with a composition similar to that shown previously (Nakatogawa et al., 2007
); i.e., 55% PE, ). This suggests that Atg3 and Atg7 are not necessary to drive fusion, nor is a particular Atg8 conformation that might be brought on by the physiologically relevant coupling machinery. When the reaction was quenched, fusion was much lower and independent of protein concentration (), analogous to the “no Atg7” controls (). Elimination of PE or a reduction to 30% PE eliminated the fusogenic potential of this protein, despite the fact that coupling of Atg8 to maleimide was efficient across all liposome compositions. Thus, we can conclude that fusion mediated by Atg8 is only possible on membranes carrying a very high surface density of the cone-shaped lipid, PE.
Similar results were obtained when we examined the mammalian Atg8 homologue LC3. Using the maleimide-coupling strategy, only LC3-conjugated liposomes containing 55% PE were able to drive hemifusion (Figure S1D, S1E
), even though the level of chemically-conjugated LC3–malPE was essentially identical across all PE concentrations tested. Taken together, liposomes with physiological PE concentrations were not able to undergo Atg8–PE or LC3–PE-dependent hemifusion.
SNARE Proteins Are Required for Autophagy
Our data indicated that Atg8/LC3 was not a self-sufficient fusogen able to direct phagophore expansion; accordingly, we were interested in identifying another component(s) that might act in this capacity. We recently reported that two proteins required for exocytosis, Sec2 and Sec4, are essential for autophagy (Geng et al., 2010
). The docking and fusion of post-Golgi secretory vesicles at the cell surface is brought about by the Q/t-SNAREs Sso1/2 and Sec9, and the R/v-SNAREs Snc1/2 (Aalto et al., 1993
; Brennwald et al., 1994
; Protopopov et al., 1993
). Given the essential role of Sec2 and Sec4 in autophagy, we investigated whether Sso1/2 and Sec9 are also required for this pathway.
We began our analysis with the GFP-Atg8 processing assay, in which this fluorescent autophagy marker is degraded in the vacuole to yield free GFP. In wild-type cells, at both permissive (PT, 24°C) and nonpermissive (NPT, 34°C) temperatures, there was no detectable free GFP in rich conditions, but a clear GFP band after starvation (). At the PT in nitrogen-starvation medium, the cleavage of GFP-Atg8 in sso1
cells was similar to that in wild-type cells; however, when these cells were starved at the NPT, there was no detectable free GFP, suggesting a severe defect in autophagy (). When we shifted the sso1
mutants back to PT, free GFP bands were observed in the mutants (; recovery, R), demonstrating that they had not lost their viability and that the block of autophagy was reversible. The expression of SSO1
genes from CEN plasmids restored the autophagy activity at the NPT in the sso1
mutants, respectively (Figure S2A
Both Autophagy and the Cvt Pathway Are Defective in sec9-4 and sso1Δ/2ts Mutants
Next, we performed the Pho8Δ60 assay that quantitatively measures the magnitude of autophagy. Wild-type cells showed a clear induction in Pho8Δ60 activity after incubation in starvation medium at both 24°C and 34°C (). In contrast, in the atg1
Δ mutant where autophagy is completely blocked, the Pho8Δ60 activity did not increase. In the sso1
strains, the Pho8Δ60 activity was somewhat lower than that of the wild type at the PT, indicating some defect even under these conditions, whereas at the NPT both mutants displayed essentially a complete block in autophagy (). The autophagy activities of sso1
mutants again recovered when the cells were shifted back to PT (SD-N Recovery). Although we found that the exocytic Q/t-SNAREs were required for autophagy, deletion of the R/v-SNAREs Snc1/2 had no effect on this process, suggesting that a subset, but not all, of the exocytic SNARE complex is involved in autophagy (Figure S2B
In S. cerevisiae, the cytoplasm to vacuole targeting (Cvt) pathway is a constitutive biosynthetic route that shares most of its molecular components with nonspecific autophagy, and is used for the delivery of certain resident vacuolar hydrolases such as precursor aminopeptidase I (prApe1). We investigated whether the aforementioned SNARE proteins are involved in the Cvt pathway by examining the maturation kinetics of prApe1 by a pulse-chase experiment. When examined at the NPT, the SNARE mutants showed a significant delay in prApe1 maturation kinetics compared with wild-type cells; however, the block in prApe1 maturation was more severe in the sso1Δ/2ts mutant ().
The sso1Δ/sso2ts Mutant Is Defective in Autophagosome Formation
To understand the nature of the autophagic defect in the SNARE mutants, we examined the ultrastructure of autophagic bodies that are formed in the sso1Δ/2ts mutant by electron microscopy. In a wild-type strain carrying a deletion of PEP4 (the gene encoding vacuolar proteinase A) and VPS4 (required for the multivesicular body pathway), after 1.5 h starvation at NPT, there were 7.25 ± 0.49 autophagic bodies per vacuole (mean ± SEM, n = 48; ). In contrast, under the same conditions, autophagic bodies were hardly visible in the sso1Δ/2ts pep4Δ vps4Δ strain; the average number of autophagic bodies was 0.69 ± 0.22 (mean ± SEM, n = 53; ).
Next, we examined whether Sso1/2 and Sec9 function at the stage of autophagosome biogenesis by examining the ability of the sso1
mutants to sequester GFP-Atg8 in completed autophagosomes, rendering this chimeric protein insensitive to exogenously added proteinase K. Spheroplasts generated from cells shifted to starvation conditions for 1 h at NPT, were osmotically lysed, and treated with proteinase K either in the presence or absence of detergent. The vam3
Δ strain accumulates completed autophagosomes; the population of GFP-Atg8 that is enclosed within autophagosomes was insensitive to treatment with proteinase K alone in this strain (). Consistent with a block in autophagosome formation in the atg1
Δ mutant, all of the GFP-Atg8 was accessible to proteinase K in the absence or presence of detergent. The protease protection pattern in the sec9-4
strain was similar to that seen in atg1
Δ cells, indicating that the Q/t-SNAREs Sso1/2 and Sec9 are required for sequestering vesicle completion (). We independently confirmed this result by examining the ability of the sso1
mutant to sequester prApe1 at the NPT by monitoring its sensitivity to exogenously added proteinase K; the data again suggested that Sso1/2 are required for sequestering vesicle completion (Figure S2C
Sso1/2 and Sec9 Are Required for Efficient Atg9 Cycling and Atg8 Recruitment
Atg9 is a critical component that affects autophagosome biogenesis and localizes to several peripheral cytoplasmic sites and the PAS (Mari et al., 2010
; Noda et al., 2000
; Reggiori et al., 2005
). The anterograde movement of Atg9 from peripheral sites to the PAS is diminished in a sec2
mutant (Geng et al., 2010
). Thus, we hypothesized that Atg9 localization would be affected in the sso1
mutants. Wild-type or sso1
cells expressing Atg9-3xGFP, and the PAS marker RFP-Ape1, were incubated in nitrogen starvation medium at NPT, then observed by fluorescence microscopy. In wild-type cells, Atg9-3xGFP localized as multiple puncta; in approximately 56% of the cells, one of the Atg9-3xGFP puncta colocalized with RFP-Ape1 (). In contrast, in the sso1
mutant, Atg9-3xGFP was observed at the PAS in only 13% of the cells (), suggesting that efficient anterograde movement of peripheral pools of Atg9-containing vesicles to the PAS may require a SNARE-mediated membrane fusion event. Similar results were seen in the sec9-4
mutant; Atg9 colocalized with the PAS marker in approximately 20% of the cells (Figure S3A and S3B
Atg9 Anterograde Movement and Atg8 Localization Are Impaired in the sso1Δ/2ts Mutant
To confirm these results, we used an epistasis assay in which we deleted ATG1
in the sso1
mutant. In an atg1
Δ strain, the retrograde transport of Atg9-3xGFP from the PAS is blocked (Reggiori et al., 2004a
), and the fluorescent protein is restricted to the PAS unless a secondary mutation interferes with its anterograde transport. In the atg1
Δ mutant, at the NPT under starvation conditions, Atg9-3xGFP was present as a single PAS punctum (). Under the same conditions, in the atg1
mutant, Atg9 was present in multiple puncta, consistent with a defect in the anterograde movement of Atg9. Furthermore, colocalization between Atg9-3xGFP and RFP-Ape1 could only be observed in approximately 25% of these cells ().
We extended our analysis of autophagosome formation to include GFP-Atg8. In wild-type cells, after incubation at the NPT in nitrogen starvation medium, two different categories of GFP-Atg8 fluorescence signals could be visualized: In approximately 35% of the cells, GFP-Atg8 clearly colocalized with RFP-Ape1 puncta at the PAS, and in approximately 60% of the cells, GFP-Atg8 was present in the vacuole (). In the sso1Δ/2ts mutant, the targeting of GFP-Atg8 to the PAS was drastically affected; approximately 85% of the cells showed both GFP-Atg8 and RFP-Ape1 puncta, with no colocalization between the two marker proteins. Consistent with a complete block in autophagy (), vacuolar GFP-Atg8 staining was not observed in the sso1Δ/2ts mutant (). Finally, there was a much higher level of RFP-Ape1 detected in the vacuole in the wild type compared to the sso1Δ/2ts strain () in agreement with the defect in prApe1 import in the mutant strain ().
To determine the stage at which vesicle formation is defective, we examined the localization of two early markers of vesicle formation, Atg11 or Atg1, in the sso1
mutant. In wild-type or sso1
cells incubated in rich or starvation media at NPT, there was no difference in the localization pattern of GFP-Atg11 or GFP-Atg1 respectively, which colocalized with RFP-Ape1 (Figure S3C, S3D, S3E and S3F
). Taken together, these results underscore the requirement of Sso1 and Sso2 in the movement of Atg9 containing reservoirs to the PAS, but not for the colocalization between early Atg markers and RFP-Ape1.
The Formation of the Atg9-Containing Tubulovesicular Clusters Is Disrupted in the sso1Δ/2ts Mutant
Next, in order to gain mechanistic insight into the role of the exocytic Q/t-SNAREs in autophagosome formation, in particular on the membrane dynamics during this event, we analyzed the ultrastructure of the Atg9-containing compartments by immuno-EM as described previously (Mari et al., 2010
). Wild-type (pep4
Δ) or pep4
cells expressing Atg9-GFP were grown at 24°C in YPD medium and then shifted to 34°C in either YPD or SD-N and examined by immuno-EM. In a wild-type strain grown in rich medium or starved, Atg9-GFP-positive structures appear as a cluster of vesicles and tubules (Mari et al., 2010
). Similar clusters could infrequently be observed in the sso1/2ts
mutant at the PT (). As previously published, the sso1
cells accumulated clusters of secretory vesicles even at the PT (Jantti et al., 2002
), and Atg9 labeling was more prominent and principally localized to these structures (), which often were also associated with one electron-dense circular structure (, asterisk). Certain secretory mutants such as sec4
accumulate two different populations of vesicles, both 100 nm in diameter: One population carries cell wall and plasma membrane components, and the second carries periplasmic enzymes (Harsay and Bretscher, 1995
). In the sso1
mutant, we observed at least two different sizes of vesicles; large (80-100 nm) and small (30-40 nm), and Atg9 appeared to be preferentially associated with the small vesicles. After 1.5 h at the NPT in rich medium, the Atg9-positive vesicles were not clustered but were more dispersed, being located among the huge number of secretory vesicles accumulated by the sso1
mutant (), often with an electron dense circular structure close by (, asterisk). Secretion is reduced under starvation conditions (Geng et al., 2010
) and accordingly the sso1
mutant did not accumulate large numbers of vesicles after starvation at the NPT; instead, smaller and more circumscribed clusters of secretory vesicles were observed (). In this case also, Atg9 was frequently found to be associated with the smaller vesicles, but neither the tubules nor the tubulovesicular clusters positive for Atg9 were observed.
Ultrastructural Analysis of the Atg9-Containing Membranes and of the prApe1 Oligomer
To better understand the ultrastructure of the PAS in the sso1
mutant, we carried out immuno-labeling with anti-Ape1 antibody to localize prApe1 oligomers, which are typically associated with this site. In the wild-type strain grown in rich medium at 24°C, the prApe1 oligomer was rarely observed, but when detected, it was associated with a cluster of vesicles and tubules as expected, which represents the PAS or the Atg9 reservoirs (data not shown; (Mari et al., 2010
). In the same cells starved at 34°C, the cytoplasmic prApe1 complex was seldom observed, due to the increase of the autophagic flux that transports this complex to the vacuole more efficiently, and, as a result, the prApe1 oligomer was often found in the interior of autophagic bodies (). This proteinaceous structure was much more easily detected in the cytoplasm of the sso1
mutant shifted to NPT in either rich or starvation medium as a direct consequence of the block in the Cvt pathway and autophagy. In the presence of nutrients, the prApe1 oligomer was found embedded within the clusters of secretory vesicles, and it appeared to be preferentially associated with small vesicles, but again no tubules were observed (). The distinct morphological appearance of the prApe1-containing complex suggested that the circular electron dense structures observed in the labeling for Atg9-GFP (, asterisks) were prApe1 oligomers. Under starvation conditions, the prApe1 oligomer in the sso1
mutant was in most cases exclusively associated with 1 to 3 large vesicles, and not with small vesicles or tubules ().
To further visualize the nature of the phagophore membrane in the sso1
mutant, we extended this analysis by immuno-localizing GFP-Atg8. In contrast to the wild type, GFP-Atg8 appeared to be primarily dispersed throughout the cytosol, or associated with certain vesicular structures, although occasionally we could detect this protein decorating the phagophore membrane in the mutant strain (Figure S4
). Taken together, the results of our immuno-EM experiments suggest that the exocytic Q/t-SNARES appear to be primarily involved in the organization of the Atg9-positive membranes into tubulovesicular clusters, which are key in the generation of the PAS, and therefore autophagosome biogenesis and normal autophagy progression.
The Defect in Autophagy Is Not Due to the Accumulation of Exocytic Vesicles
We wanted to examine whether the autophagy defect in the SNARE mutants could be an indirect effect caused by a paucity of membrane for autophagosome biogenesis, due to the accumulation of exocytic vesicles. The overexpression of SNC1
in the sso1
mutant strongly suppresses the temperature sensitive growth defect (Jantti et al., 2002
). We found that the overexpression of Snc1 was indeed able to rescue the growth defect, but not the block in autophagy in the sso1
mutant at 34°C (). The overexpression of Sso1 was able to complement the defects in both growth and autophagy in the sso1
mutant, whereas transformation with a plasmid that overexpresses SEC9
poorly suppressed the growth defect, but was unable to rescue the block in autophagy. Similarly, we found that the overexpression of Snc1 or Sso1 in the sec9-4
mutant restored the growth but not the autophagy defect at 34°C (). These results suggest that the alleviation of the secretion defects in the sec9-4
mutants is not sufficient for restoring normal levels of autophagy. Furthermore, they also suggest that each of these SNARE proteins has a specific role in autophagy that cannot be compensated by simply overexpressing another SNARE protein.
The Overexpression of SNAREs Involved in Secretion Cannot Suppress the Autophagy Defect in the sso1Δ/2ts or sec9-4 Mutants
Consistent with a recent mammalian study, temperature-sensitive mutations in several exocyst components in yeast result in a block in autophagy (Figure S5A
), presumably because their inactivation blocks SNARE assembly (Grote et al., 2000
). In mammalian cells, clathrin-mediated endocytosis of the plasma membrane is involved in the formation of pre-autophagic structures (Ravikumar et al., 2010
); however, we found that endocytosis is dispensible for autophagy in yeast (Figure S5B
; Reggiori et al., 2004b
), suggesting that in this organism if the plasma membrane is one of the sources for autophagosome formation, it is transported to the PAS by a mechanism that is independent of endocytosis.
Thus far, our results show that the exocytic SNAREs are involved in the formation of the tubulovesicular clusters of Atg9. One prediction from this result is that we would be able to observe colocalization between Sso1 and Atg9. To test this prediction, we transformed wild-type cells expressing Atg9-3xGFP with a plasmid bearing RFP-Sso1, and examined the localization of these chimeric proteins both under rich conditions and in starvation medium. Consistent with previous observations (Scott et al., 2004
), Sso1 predominantly showed patchy plasma membrane localization (). In addition, RFP-Sso1 showed cytosolic puncta; these puncta were also observed in experiments localizing endogenous Sso1 by immunofluorescence (Scott et al., 2004
). We observed colocalization between Atg9-3xGFP and cytosolic RFP-Sso1 puncta in cells exhibiting both fluorophores when grown in rich medium (), as well as in cells shifted to starvation medium (data not shown). In contrast, we rarely observed any colocalization between GFP-Sso1 and the PAS marker RFP-Ape1 (data not shown).
Tlg2 Is required for Atg9 Cycling and Forms a Novel Complex with Sso1 and Sec9
In order to understand how the function of the exocytic Q/t-SNARES in autophagosome formation is regulated, we examined the role of additional SNARE proteins in this process. The Q/t-SNARE Tlg2 localizes to the early endosome, and is required for efficient endocytosis and maintenance of normal levels of TGN proteins (Holthuis et al., 1998
). It was previously shown to be required for the Cvt pathway, but not for nonspecific autophagy (Abeliovich et al., 1999
). Tlg2 is not required for anterograde traffic through the Golgi complex or vacuolar protein delivery (Abeliovich et al., 1999
; Holthuis et al., 1998
); therefore, the defect in the Cvt pathway in the tlg2
Δ mutant is not likely to be due to a block in the trafficking of an essential Cvt component through the secretory pathway, or due to a block in the delivery of vacuolar proteases. When we examined the role of Tlg2 using the quantitative Pho8Δ60 assay, tlg2
Δ mutant cells displayed a significant reduction in activity (), although not as severe as the sso1
mutant () that showed an essentially complete block in autophagy.
Tlg2 Determines the Magnitude of Autophagy, Affects Atg9 Anterograde Transport, and Interacts with Sso1 and Sec9
Furthermore, the frequency of colocalization of Atg9-3xGFP with RFP-Ape1 dropped from approximately 55% in the wild-type strain to less than 30% in the tlg2Δ strain (). To confirm that the defect in Atg9-3xGFP localization to the PAS was due to a defect in its anterograde transport, we additionally deleted ATG1 in the tlg2Δ strain, and performed an epistasis analysis as described in . In the atg1Δ mutant under starvation conditions, in cells expressing both chimeric proteins, Atg9-3xGFP and RFP-Ape1 colocalized in 100% of the cells examined ( and data not shown); in contrast, in the atg1Δ tlg2Δ mutant, colocalization was approximately 84% ( and data not shown). Furthermore, in atg1Δ cells under starvation conditions, Atg9-3xGFP was predominantly present as a single PAS punctum, whereas in the atg1Δ tlg2Δ mutant, Atg9 was present in multiple puncta in approximately 43% of the cells examined, consistent with a defect in the anterograde movement of this protein ().
To extend our analysis, we examined the interaction between protein A (PA) tagged Sso1, GFP-Sec9 and GST-Tlg2 from cross-linked extracts prepared from wild-type cells that were starved for 1 h. Both GST-Tlg2 and GFP-Sec9 co-isolated with PA-Sso1, but not PA alone, indicating the existence of a complex in vivo
(). Furthermore, under the same conditions, PA-Sso1 could not co-isolate another t-SNARE, GST-Ufe1, suggesting a specific interaction between PA-Sso1 and GST-Tlg2. Previous work examining SNARE interactions in vitro
found that Sso1 and Sec9 functionally interacted with four R/v-SNARE proteins, Snc1, Snc2, Sec22 and Nyv1 (McNew et al., 2000
). However, the possibility that Sso1-Sec9 may interact with other SNAREs like Tlg2 was not explored. Considering that Tlg2 is found in a complex with PA-Sso1, we asked if Tlg2 was capable of driving fusion with Sso1-Sec9 in vitro. Recombinant Sso1-Sec9, Snc1 and Tlg2 were reconstituted into proteoliposomes and their ability to drive membrane fusion was determined. Whereas the cognate SNARE pair Sso1-Sec9-Snc2 displayed robust lipid mixing, Sso1-Sec9-Tlg2 was unable to do so, even when the regulatory Habc domain of Tlg2 was removed (H3-Tlg2) (Figure S6
). We also examined the possibility that Tlg2 requires activation, as it does with its known t-SNARE partners Tlg1 and Vti1 (Paumet et al., 2001
); nevertheless, no fusion was seen in the presence of the C-terminal Snc2 peptide. Thus, it is possible that an additional factor or some type of posttranslational modification is needed to reconstitute this fusion activity in vitro
The SM family protein Sec1, N-ethylmaleimide-sensitive fusion protein (NSF) Sec18, and the R/v-SNAREs Sec22 and Ykt6 are required for autophagy
Proteins of the Sec1/Munc18 family interact with SNAREs to regulate SNARE complex assembly and play complementary roles in fusion. There are four SM family proteins in yeast: Sec1, Vps45, Sly1, and Vps33. Of these proteins, Vps45, which associates with Tlg2, has previously been shown to be required for the Cvt pathway (Abeliovich et al., 1999
). Sec1 is required for exocytosis and forms a complex with the exocytic SNAREs (Carr et al., 1999
). Because of the role of Sso1/2 and Sec9 in autophagy, we asked if Sec1 was also required for this process. Indeed, the temperature sensitive sec1-1
mutant showed a block in autophagy ().
Autophagy Is Impaired in the Absence of Sec1, Sec18, and the R/v-SNAREs Sec22 and Ykt6
Previously, both the α
-SNAP protein Sec17 and the NSF protein Sec18, which mediate the disassembly of SNARE complexes (Mayer et al., 1996
), were shown to be required for autophagy; temperature sensitive sec17-1
mutants show no induction in Pho8Δ60 activity under starvation conditions at NPT (Isahara et al., 1999
; Ishihara et al., 2001
). However, that study reported that Sec18 is not required for autophagosome formation, but for the fusion of completed autophagosomes with the vacuole. We observed a strong block in autophagy as assayed by the GFP-Atg8 processing assay in both the SEC17
mutants () and a significant delay, but not a complete block, in prApe1 maturation kinetics compared with wild-type cells at NPT in rich conditions (Figure S7A
). In the previous study that concluded that sec18
was not required for autophagosome biogenesis, the authors examined the protection accessibility of steady state levels of membrane-associated prApe1 in sec18-1
cells shifted to starvation conditions at NPT. The results show that prApe1 is partially insensitive to proteinase K alone under those conditions; however, since prApe1 maturation is not completely blocked in this mutant at the time the assay was performed, a population of this oligomer was already transported to the vacuole, rendering it insensitive to exogenous proteinase K. We repeated the protease protection assay in which we examined the ability of the sec18-1
mutant to sequester GFP-Atg8 in completed autophagosomes by monitoring the sensitivity of this chimeric protein to exogenously added proteinase. In the vam3ts
mutant, a portion of GFP-Atg8 was insensitive to proteinase K treatment alone (). In contrast, the protease protection pattern in the sec18-1
mutant was similar to that seen in the atg1
Δ mutant, indicating that Sec18 is required for autophagosome biogenesis ().
Of the four R/v-SNAREs that were shown to interact with Sso1 and Sec9 in vitro (McNew et al., 2000
); we found that the inactivation of Snc1, Snc2, and Nyv1 have no affect on autophagy (Figure S2B
and data not shown). However, two separate temperature sensitive alleles of sec22, sec22-1
showed severe blocks in autophagy and a kinetic delay in the Cvt pathway at the NPT in starvation conditions ( and Figure S7B
). Similarly, we found that Sec22 is required for autophagosome biogenesis as determined by a GFP-Atg8 protease protection assay (). We next tested whether a complex between Sec22, Sso1 and Sec9 could be detected in vivo
, and were able to isolate a complex between PA-Sec22, GST-Sec9 and GFP-Sso1 from cross-linked extracts prepared from wild-type cells that were starved for 1 h ().
Sec22 localizes to both the ER and the Golgi, and functions in ER-Golgi anterograde and retrograde transport (Liu and Barlowe, 2002
; Liu et al., 2004
). Both sec22-1
mutants showed defects in the anterograde transport of Atg9 to the PAS, when grown at the NPT in both rich medium and starvation conditions (Figure S7C
and data not shown). We were able to observe some colocalization between Atg9-3xGFP and RFP-Sec22 (Figure S7D
), although at present we do not know whether this is because of a disruption of the trafficking route of Atg9 through the ER, or whether a small amount of Sec22 colocalizes with Sso1/2 at the Atg9 peripheral sites to facilitate the formation of Atg9 tubulovesicular structures that then translocate to the PAS.
The R/v-SNARE Ykt6 is very versatile in that it participates in the secretory pathway at the early Golgi stage, and in three separate vacuole targeted vesicular routes, the ALP, CPY and Cvt pathways. Additionally, the overexpression of Ykt6 can overcome the growth defects of the sec22
temperature sensitive mutants at the NPT, and in ER-to-Golgi transport (Liu and Barlowe, 2002
; McNew et al., 1997
). Accordingly, we tested the role of the ykt6
temperature sensitive mutant in autophagy, and observed a severe block in GFP-Atg8 processing at the NPT under nitrogen-deprivation conditions (). Furthermore, GFP-Atg8 protease protection results suggested that Ykt6 is required for autophagosome formation (). This result was unexpected because Ykt6 is known to have a role in homotypic vacuole fusion where it forms a complex with the t-SNAREs Vam3 and Vam7 (Ungermann et al., 1999
), which are also required for the fusion of completed autophagosomes with the vacuole; consequently Ykt6 was assumed to function in autophagosome-vacuole fusion. Our results therefore suggest that Ykt6 may function at more than one step in autophagy. We found that RFP-Ykt6 colocalized with Atg9-3xGFP (Figure S7E
), and that PA-Ykt6, but not another v-SNARE, PA-Gos1, could bind GST-Sec9 and GFP-Sso1 in vivo
(). Previously it was shown that Ykt6 can function as a v-SNARE with Sso1 and Sec9 in vitro, when it is artificially attached to a membrane through a trans-membrane peptide anchor, but not when it is anchored to the membrane via its native C20
lipid chain (McNew et al., 2000
), leading to the conclusion that Ykt6 lipid-anchoring prevents it from functioning as a v-SNARE, thus imparting specificity in SNARE pairing. However, it is also possible that the membrane anchored Ykt6 may require other factors or may need to be palmitoylated (Fukasawa et al., 2004
) in order to drive liposome fusion in vitro. Taken together, we have found roles for two v-SNAREs, Sec22 and Ykt6, which can be found in a complex with Sso1-Sec9 in vivo
, in autophagosome formation.