A role for Atg8–PE deconjugation in autophagosome biogenesis

Abstract

Formation of the autophagosome is likely the most complex step of macroautophagy, and indeed it is the morphological and functional hallmark of this process; accordingly, it is critical to understand the corresponding molecular mechanism. Atg8 is the only known autophagy-related (Atg) protein required for autophagosome formation that remains associated with the completed sequestering vesicle. Approximately one-fourth of all of the characterized Atg proteins that participate in autophagosome biogenesis affect Atg8, regulating its conjugation to phosphatidylethanolamine (PE), localization to the phagophore assembly site and/or subsequent deconjugation. An unanswered question in the field regards the physiological role of the deconjugation of Atg8–PE. Using an Atg8 mutant that bypasses the initial Atg4-dependent processing, we demonstrate that Atg8 deconjugation is an important step required to facilitate multiple events during macroautophagy. The inability to deconjugate Atg8–PE results in the mislocalization of this protein to the vacuolar membrane. We also show that the deconjugation of Atg8–PE is required for efficient autophagosome biogenesis, the assembly of Atg9-containing tubulovesicular clusters into phagophores/autophagosomes, and for the disassembly of PAS-associated Atg components.

Introduction

Macroautophagy, hereafter referred to as autophagy, is a highly conserved, vesicular trafficking pathway for eliminating cytoplasmic contents ranging from soluble proteins to entire organelles. Autophagy, and autophagy-related processes, also function in biosynthetic modes including the delivery of resident vacuolar hydrolases¹ and unconventional protein secretion.^2-4 The degradative autophagic pathway involves the sequestration of cytoplasm into double-membrane vesicles, termed autophagosomes, which fuse with the vacuole, releasing the inner membrane vesicles and cargo into the vacuolar lumen for degradation.⁵ Autophagy was first identified in yeast as a starvation response essential for overcoming conditions of nutrient deprivation; however, in multicellular organisms, aside from its role in cellular homeostasis, autophagy is required for processes as diverse as the removal of protein aggregates and damaged or superfluous organelles, development and cellular remodeling, and immunity.^6-8 Indeed, defects in autophagy are associated with various diseases, such as cancer, gastrointestinal disorders, and neurodegeneration.^9-11 Interestingly, even though the role of autophagy seems to have significantly diverged in multicellular organisms, the core components of the autophagy machinery are evolutionarily conserved from yeast to human.

The formation process and the itinerary of autophagosomes in yeast can be divided into several conceptual steps. First, the cargo and autophagy-related proteins arrive at the site of autophagosome formation, termed the phagophore assembly site (PAS, also called the pre-autophagosomal structure^12,13); second, the Atg proteins act in concert with lipids and with one another, resulting in the expansion of the phagophore membrane and sequestration of cargo; third, the expanding phagophore membranes join together generating a complete autophagic vesicle; fourth, almost all the Atg proteins are recovered from the completed autophagosome, presumably for reuse in subsequent rounds of autophagosome formation; and, finally, the completed autophagosome fuses with the vacuole.

Central to autophagosome formation and phagophore expansion in yeast is a highly conserved ubiquitin-like protein, Atg8,^14,15 which has at least seven different mammalian orthologs belonging to the LC3 and GABARAP subfamilies.^16-20 In order to function in the autophagy pathway, Atg8 has to be conjugated to the lipid phosphatidylethanolamine (PE). Atg8–PE conjugation is initiated by the cysteine protease Atg4-mediated cleavage of the C-terminal arginine residue of precursor Atg8 to expose a glycine residue.^12,13 This glycine is then conjugated to PE through the sequential actions of Atg7, a ubiquitin E1-like activating enzyme homolog, and Atg3, a ubiquitin E2-like conjugating enzyme analog.¹⁴ Atg8–PE conjugation is facilitated by a dimer of the Atg12–Atg5-Atg16 protein complex, which presumably exerts an E3-like function in Atg8 lipidation,²¹ and perhaps in part dictates the site of conjugation.²²

In yeast, under nutrient-rich conditions, Atg8 mostly exists in the unconjugated form; however, upon nitrogen starvation its expression level significantly increases and most of the Atg8 is converted to the PE-conjugated form.^23,24 Atg8–PE is recruited to the PAS; this localization not only depends on the Atg8–PE conjugation machinery and on the Atg12–Atg5-Atg16 complex, but also on Atg9, a transmembrane protein proposed to function as a membrane carrier for autophagosome formation, and the autophagy-specific phosphatidylinositol 3-kinase (PtdIns3K) complex I.^12,13,25-27 Analogous to yeast Atg8, the mammalian Atg8 homologs are also conjugated to PE and localize to the phagophore and autophagosome membranes.^18,19,28-31 Immunoelectron microscopy (IEM) experiments show that Atg8–PE is localized on both the inner and outer phagophore membranes, but not on the surface of completed autophagosomes.¹³ Upon autophagosome completion, Atg8–PE that lines the inner autophagosome membrane is delivered to the vacuole where it is degraded;^17,23,24 whereas, Atg8 that is present on the outer membrane of the autophagosome¹³ is released, via the deconjugation of Atg8–PE by a second Atg4-dependent cleavage step. At this point, however, the mechanism that determines the timing of this cleavage, possibly preventing premature autophagosome membrane closure, is not known. The deconjugation of Atg8–PE is clearly important for autophagy, as mutations that specifically block Atg8–PE deconjugation result in a significant reduction in the magnitude of autophagy,¹³ although the reason for this decrease has not been known. In this work, we investigated the significance of Atg8–PE deconjugation in autophagy. Our results revealed that a block in deconjugation affects several steps of autophagy; we present evidence that Atg8–PE deconjugation is necessary for autophagosome biogenesis, affects optimal phagophore expansion, and plays a role in determining autophagosome size and number. Finally, we propose that the deconjugation of Atg8–PE is a prerequisite for the disassembly of proteins including Atg14 from completed autophagosomes.

Results

The inability to deconjugate Atg8–PE results in its mislocalization and in defects in the Cvt and autophagy pathways

The expression of an Atg8 variant, Atg8ΔR, which lacks the C-terminal arginine, in an atg4Δ atg8Δ mutant strain allows the initial step of Atg4-mediated cleavage of Atg8 to be bypassed while still obtaining conjugation of Atg8. Thus, in this background, Atg8ΔR can be conjugated to PE, but once Atg8–PE is formed, the PE moiety cannot be deconjugated from Atg8 because of the absence of Atg4 (Fig. 1A). In the yeast S. cerevisiae, the core components required for autophagy are shared with the cytoplasm to vacuole targeting (Cvt) pathway, which delivers vacuolar hydrolases such as precursor aminopeptidase I (prApe1) to the vacuole under nutrient-rich conditions.^32,33 In wild-type cells most of the Ape1 was detected in the mature form (mApe1), whereas cells lacking ATG8 or ATG4 showed a complete block in prApe1 maturation. As seen previously,¹³ there was a partial block in prApe1 maturation in atg4Δ atg8Δ cells expressing Atg8ΔR (Fig. 1B). Thus, expression of Atg8ΔR was not able to completely rescue the Cvt pathway defect in this strain background.

Figure 1. The deconjugation of Atg8–PE is required for efficient Cvt and autophagy pathways, and for the wild-type localization pattern of Atg8. (A) The lipidation pattern of Atg8. Atg8 conjugation was analyzed in a wild-type (TN124) strain, and in atg8Δ, atg4Δ, or atg4Δ atg8Δ cells expressing chromosomally tagged Atg8ΔR. Cells were grown in nutrient-rich medium to mid-log phase, shifted to starvation medium for 1 h and collected. Protein extracts were subjected to SDS-PAGE followed by immunoblotting with anti-Atg8 antibody. The positions of Atg8 and Atg8–PE are indicated. (B) The Cvt pathway is partially blocked in cells that are unable to deconjugate Atg8–PE. Protein extracts from the cells in (A) were analyzed by immunoblotting using anti-Ape1 antiserum. The positions of precursor and mature Ape1 are indicated. (C) Nonselective autophagy is reduced in cells expressing the constitutively PE conjugated form of Atg8. The wild-type strain (WT; YZX247), atg1Δ strain transformed with an empty vector (UNY5), or atg4Δ atg8Δ cells expressing GFP-Atg8ΔR (UNY52) were grown in SMD medium and incubated for 4 h in SD-N to induce autophagy. Cell extracts were prepared from these strains, and Pho8Δ60 activity was measured. The Pho8Δ60 activity from the wild-type strain incubated in SD-N was set at 100%, and the Pho8Δ60 activities for all other data points were normalized to this value. The results represent the mean and standard deviation (SD) of three independent experiments. (D) The delipidation of Atg8 is critical for its wild-type localization pattern. Localization of GFP-Atg8 in wild-type (SEY6210) or atg4Δ strains, and GFP-Atg8ΔR localization in atg4Δ atg8Δ cells was determined by fluorescence microscopy under nitrogen starvation conditions. Cells were grown in synthetic medium until mid-log phase, stained with FM 4-64 and shifted to SD-N for 2 h before imaging. DIC, differential interference contrast. Scale bar, 2.5 μm. The percentages of cells showing puncta were quantified in (E). The error bars represent the standard deviation from three independent experiments; n = 411 for the wild-type strain, 246 for the atg4Δ strain, and 754 for the atg4Δ atg8Δ strain expressing GFP-Atg8ΔR.

Next, we quantitatively measured bulk autophagy under nitrogen starvation conditions by monitoring the activity of Pho8Δ60, a nonspecific autophagy protein marker that encodes an altered form of alkaline phosphatase that can only be delivered to the vacuole via autophagy, where it is processed to its mature, active form.^34,35 Expression of the chronically PE-conjugated form of GFP-Atg8 (GFP-Atg8ΔR) in atg4Δ atg8Δ cells resulted in approximately 30% Pho8Δ60 activity compared with the wild-type strain, indicating a significant defect in nonspecific autophagy (Fig. 1C). Complementation with a plasmid expressing Atg4 restored the Pho8Δ60 activity and prApe1 maturation to wild-type levels in an atg4Δ atg8Δ strain expressing Atg8ΔR (Fig. S1), suggesting that the deconjugation of Atg8–PE is required for normal autophagy.

To explore the mechanism by which the block in delipidation of Atg8–PE results in a defect in autophagy, we expressed GFP-Atg8ΔR in atg4Δ atg8Δ cells, and examined its localization in nitrogen-starvation conditions by fluorescence microscopy, after staining the vacuole membrane with the dye FM 4-64. Approximately 37% of wild-type cells showed perivacuolar GFP-Atg8 puncta (Fig. 1D and E), and several cells showed GFP-Atg8 fluorescence in the vacuolar lumen, consistent with the delivery of GFP-Atg8-containing autophagic bodies into the lumen following the fusion of completed autophagosomes with the vacuole.¹² In atg4Δ cells, the recruitment of GFP-Atg8 to the perivacuolar puncta was only detected at a low background level; these cells are defective in autophagy, and accordingly there was no vacuolar lumenal GFP-Atg8 fluorescence.¹³ Unlike the wild type, atg4Δ atg8Δ cells expressing GFP-Atg8ΔR showed distinct vacuolar rim staining colocalizing with FM 4-64 (Fig. 1D), although a similar percentage of cells displayed perivacuolar puncta compared with wild-type cells (Fig. 1E). Localization to the vacuolar rim likely reflects the fusion of the autophagosome outer membrane with the vacuole, coupled with the defect in deconjugation, resulting in mislocalization of Atg8. Essentially the same localization pattern was seen when atg4Δ atg8Δ cells expressing GFP-Atg8ΔR were grown in rich medium (Fig. S2). In both rich medium and under starvation conditions, the perivacuolar puncta observed in the atg4Δ atg8Δ cells expressing GFP-Atg8ΔR colocalized with a PAS marker, mCherry-prApe1 (Fig. S3). Taken together, these results suggest that the deconjugation of Atg8–PE by Atg4 is critical for the normal functioning of the Cvt and autophagy pathways, and in ensuring the proper localization pattern of Atg8 in both rich medium and starvation conditions.

Smaller and fewer autophagosomes are produced in cells blocked in the delipidation of Atg8–PE

To elucidate the nature of the decrease in autophagy activity seen in the absence of Atg8–PE deoconjugation, we decided to examine the role of Atg8–PE delipidation in autophagosome biogenesis. Accordingly, we examined the ultrastructure of autophagic bodies that accumulated in atg4Δ atg8Δ cells constitutively expressing Atg8–PE, by transmission electron microscopy (TEM). For TEM analysis, we used an atg4Δ atg8Δ strain expressing Atg8ΔR that was additionally deleted for the PEP4 and VPS4 genes in order to prevent the vacuolar degradation of autophagic bodies and to eliminate vesicles targeted to the vacuole via the multivesicular body pathway, respectively.^36,37 After 4 h starvation, the average number of autophagic bodies per vacuole in the wild-type strain deleted for only PEP4 and VPS4, was 9 ± 0.13 (mean ± SEM; n = 100), while in an atg1Δ pep4Δ vps4Δ strain, no autophagic bodies were observed, consistent with Atg1 being absolutely required for autophagy (Fig. 2A and B). Under the same conditions, in the atg4Δ atg8Δ pep4Δ vps4Δ strain expressing Atg8ΔR the average number of autophagic bodies observed per vacuole was 2 ± 0.04 (mean ± SEM; n = 178; Fig. 2A and B).

Figure 2. The deconjugation of Atg8ΔR is required for normal autophagosome formation. (A) The wild-type (pep4Δ vps4Δ; UNY111), experimental (atg4Δ atg8Δ pep4Δ vps4Δ cells expressing Atg8ΔR; YCY41), and negative control (atg1Δ pep4Δ vps4Δ; UNY112) strains were grown in YPD to mid-log phase and shifted to SD-N for 4 h to induce autophagy. The cells were fixed in potassium permanganate, embedded in Spurr’s resin and examined by TEM. Scale bar, 500 nm. Autophagic bodies observed in the experimental strain are also shown in higher magnification. (B) Average number of autophagic bodies per vacuole accumulated in the wild-type and experimental strains. No autophagic bodies were formed in atg1Δ cells. The error bars represent standard error of the mean (SEM). For each strain at least 100 cells were analyzed to score the number of autophagic bodies. (C) Quantification of autophagic body size. The average diameter of cross-sections of autophagic bodies are shown; error bars represent SEM; n > 500. (D) Percentage of wild-type or mutant cells as a function of the number of autophagic bodies per vacuole organized into 5 groups: 0, 1 to 4, 5 to 8, 9 to 12 and > 12. The number of autophagic bodies accumulated in the wild-type (n = 100) or experimental (n = 178) strain was determined from cells containing similar-sized vacuoles.

To determine whether the constitutive expression of Atg8–PE affected the size of the autophagic bodies, we quantified their diameter. The average diameter of autophagic bodies in atg4Δ atg8Δ cells expressing only the lipidated form of Atg8 was 226 ± 1.8 nm (mean ± SEM; n = 827) compared with wild-type cells in which the average diameter was 394 ± 3.5 nm (mean ± SEM; n = 534; Fig. 2C). Finally, we also examined the distribution of autophagic bodies per vacuole and found that in cells expressing only Atg8–PE, approximately 60% of the cells had no autophagic bodies compared with 6% of the wild-type (pep4Δ vps4Δ) cells. Approximately 76% of wild-type cells and about 24% of the cells expressing Atg8ΔR had five or more autophagic bodies per vacuole, respectively (Fig. 2D). Note that there was no statistical difference in the size of the vacuoles between the wild-type (pep4Δ vps4Δ) and atg4Δ atg8Δ Atg8ΔR (pep4Δ vps4Δ) cells (data not shown). Taken together, these data show that the deletion of ATG4 that abolished the delipidation of Atg8ΔR resulted in highly reduced autophagy by affecting both the number and size of autophagosomes produced.

The inability to deconjugate Atg8–PE results in a defect in autophagosome biogenesis

Our TEM results suggested that the defect in autophagy could result from at least two possibilities; atg4Δ atg8Δ cells expressing Atg8–PE are either defective in enclosing cargo within autophagosomes, or are kinetically delayed in the fusion of completed autophagosomes with the vacuole. We performed an experiment in order to determine whether the GFP-Atg8ΔR puncta formed in the atg4Δ atg8Δ strain during starvation conditions are incomplete structures, or functional autophagosomes that are capable of fusion with the vacuole. To perform this experiment, we examined the ability of wild-type or deconjugation-defective atg4Δ atg8Δ cells to complete autophagy when shifted from starvation conditions to rich medium. Accordingly, we quantified the decrease in fluorescence intensity of GFP-Atg8 puncta over time when these cells were starved for 1 h in SD-N and shifted back to rich medium. Samples were collected and observed by fluorescence microscopy 5, 10, 15, 20 and 25 min after the shift. In the first 5 min after the cells were shifted to rich medium, the number of GFP-Atg8 puncta did not change significantly (data not shown). In wild-type cells, over two-thirds of GFP-Atg8 puncta that accumulated after 1 h in SD-N medium, disappeared after a 10 min shift to rich conditions; in contrast, in the deconjugation-defective cells, over 70% of the GFP-Atg8 puncta remained present even after 25 min of growth (Fig. 3A and B). As a negative control, we examined the disappearance of GFP-Atg8 puncta in the atg1Δ strain that is defective in autophagosome formation and found that, consistent with a defect in autophagosome formation, GFP-Atg8 puncta (which in this case correspond to a nonfunctional PAS) persisted after a 25 min shift to rich medium (Fig. S4A and S4B). These results suggest that the GFP-Atg8ΔR puncta formed in the deconjugation-defective cells are either incomplete autophagosomes incapable of cargo sequestration or abnormal autophagosomes that are defective in fusion with the vacuole.

Figure 3. A defect in the deconjugation of Atg8–PE affects autophagic vesicle biogenesis. (A and B) An inability to deconjugate Atg8–PE results in autophagosomes that do not mature. (A) Mid-log phase yeast cells, were starved for 1 h in nitrogen starvation medium, and then shifted back to rich medium to downregulate autophagosome formation. Samples were collected and observed by fluorescence microscopy at the indicated time points. DIC, differential interference contrast. Scale bar, 5 μm. (B) For each time point, at least 30 cells were quantified to obtain the average number of GFP-Atg8 puncta per cell. The experiment was repeated three times. The mean and SD of the average numbers of GFP-Atg8 puncta per cell are shown. For each strain, the mean value at 5 min was set to 100% and other values were normalized. (C–E) The deconjugation of Atg8–PE is important for autophagosome formation. Spheroplasts from wild-type (WT) pep4Δ cells (C; UNY117), atg4Δ atg8Δ pep4Δ cells expressing Atg8ΔR (D; UNY118), or atg1Δ pep4Δ cells (E; UNY119) were subjected to a protease-protection analysis as described in Materials and Methods using anti-Ape1 antiserum. The presence of Pgk1 was analyzed to determine the efficiency of spheroplast lysis and the extent of contamination of the pellet fraction with cytosol. The sensitivity of the precursor form of Prc1 to exogenous protease was examined by immunoblot to verify that the lysis method did not result in the disruption of organellar membranes in the P5 fractions. Ape1* and Prc1* indicate partially degraded forms of these proteins. T, total lysate; S5, lysate supernatant fraction; P5, pellet fraction; PK, proteinase K; TX; Triton X-100.

In order to specifically determine whether the atg4Δ atg8Δ cells expressing Atg8ΔR were impaired in cargo sequestration, we performed a protease-protection assay under nitrogen starvation conditions using prApe1 as a marker, and examined whether this protein could be protected from exogenously added protease. In this assay pep4Δ cells, atg4Δ atg8Δ pep4Δ cells expressing Atg8ΔR, or atg1Δ pep4Δ cells were grown to mid-log phase and converted to spheroplasts. To induce autophagosome formation, the spheroplasts were incubated for 1 h in osmotically-supplemented nitrogen-starvation medium. They were then collected by centrifugation, osmotically lysed and fractionated into low speed supernatant and pellet fractions. The prApe1complex associates with the membrane fraction in a MgCl₂-dependent manner even when it is not enclosed in a completed autophagosome, and can therefore be recovered in a low speed pellet fraction. The prApe1-containing pellet fraction was divided into four parts and treated with or without detergent in the absence or presence of proteinase K. In a mutant that is defective in vesicle completion, the prApe1 complex is sensitive to exogenously added protease; however, in a strain that is normal for autophagosome formation, the prApe1 is protected within autophagic vesicles from proteinase K, and becomes protease sensitive only when autophagosome membranes are disrupted by the addition of detergent. In the pep4Δ strain, prApe1 was protected from proteinase K alone, indicating efficient autophagosome completion (Fig. 3C). In contrast, the pellet fraction of the atg4Δ atg8Δ pep4Δ strain expressing Atg8ΔR was largely sensitive to exogenously added proteinase K in the absence of detergent, indicating that the delipidation of Atg8 is necessary for vesicle completion during autophagy (Fig. 3D). The pellet fraction of the atg1Δ pep4Δ strain was accessible to proteinase K both in the absence or presence of detergent (Fig. 3E), consistent with a complete block in autophagosome formation in an atg1Δ mutant. Under our experimental conditions, the cytosolic marker Pgk1 was predominantly detected in the total lysate and supernatant fractions, confirming efficient spheroplast lysis (Fig. 3C–E). Finally, to confirm that the osmotic lysis procedure did not result in the disruption of membranes, such as lysis of the autophagosome or vacuolar membranes, we examined the protease sensitivity of carboxypeptidase Y (Prc1). Prc1 is a vacuolar exopeptidase that is synthesized as a precursor, which transits through the ER, Golgi complex, and late endosome/multivesicular body (MVB), and is delivered to the vacuole where it is converted to the mature, active form by removal of a propeptide.³⁸ Treatment of the pellet fraction with proteinase K alone did not result in the mature form of Prc1 in any of these strains; however, the addition of both Triton X-100 and proteinase K resulted in processing of precursor Prc1 to its mature form (Fig. 3C–E). These results confirm that the vacuolar, and presumably autophagosome, membranes were not disrupted in our pellet fractions.

Finally we considered the possibility that the mislocalization of Atg8–PE to the vacuolar membrane could not only result in defects in autophagy-related pathways, but could also more generally affect nonautophagic, vacuolar protein delivery pathways by interfering with the fusion of nonautophagic vesicles with the vacuole. By pulse-chase analysis and immunoprecipitation, we followed Prc1 processing in the atg4Δ atg8Δ Atg8ΔR mutant or in a wild-type strain. There was no difference in the kinetics of precursor Prc1 maturation between the wild-type strain and the atg4Δ atg8Δ strain expressing Atg8ΔR (Fig. S4C). These data suggest that the atg4Δ atg8Δ strain constitutively expressing Atg8–PE was specifically defective in autophagy-related pathways, and not in the nonautophagic, carboxypeptidase Y pathway.

The overexpression of Atg8ΔR does not rescue the defect in autophagy

The amount of Atg8 at the PAS regulates the level of autophagy by modulating the size of autophagosomes.¹⁵ Because Atg8ΔR is mislocalized to the vacuolar rim, we took into account the possibility that the autophagy defect in an atg4Δ atg8Δ mutant expressing Atg8ΔR could be due to the inadequate availability of Atg8–PE for autophagosome formation. We hypothesized that the overexpression of Atg8ΔR could supply sufficient Atg8–PE for the formation of autophagosomes, and thereby overcome the defect in autophagy. Therefore, we overexpressed GFP-Atg8ΔR under the control of the copper-inducible, CUP1 promoter. Although the majority of Atg8 was conjugated with PE, we were able to detect a small amount of nonlipidated Atg8 consistent with the overexpression of Atg8ΔR overwhelming the Atg8-lipidation machinery (Fig. S5A). When we overexpressed GFP-Atg8ΔR, we found that even though the vacuolar rim staining persisted, there was an approximately 2-fold increase in the percentage of atg4Δ atg8Δ cells showing recruitment of GFP-Atg8ΔR to the PAS when this chimeric protein was expressed under the control of the CUP1 promoter as compared with the endogenous ATG8 promoter (Fig. S5B and S5C). Furthermore, the size of the puncta formed in the strain overexpressing GFP-Atg8ΔR appeared larger than in the wild-type strain expressing GFP-Atg8. By measuring Pho8Δ60 activity after 4 h nitrogen starvation, we found that the overexpression of Atg8–PE did not result in an increase in autophagy activity; atg4Δ atg8Δ cells expressing GFP-Atg8ΔR under the endogenous ATG8 promoter or the CUP1 promoter exhibited ~38% and 41%, respectively, of the Pho8Δ60 activity seen in wild-type cells (Fig. 4A) even when copper was added to further induce CUP1 promoter-dependent expression (Fig. S5D).

Figure 4. The overexpression of GFP-Atg8ΔR does not rescue the defect in autophagy. (A) The defect in bulk autophagy is not rescued in cells overexpressing the constitutively PE-conjugated form of Atg8. The Pho8Δ60 assay was used to monitor autophagy activity from the following: An atg8Δ strain (YZX231), a wild-type strain expressing GFP-Atg8 (YZX247), or an atg4Δ atg8Δ strain expressing GFP-Atg8ΔR under the control of the endogenous ATG8 promoter (UNY52) or the CUP1 promoter (YZX334). Where necessary, the strains were transformed with empty vectors, such that all strains being assayed had the same auxotrophic markers. Alkaline phosphatase activity was monitored from protein extracts prepared from cells grown in SMD or after a 4 h shift to SD-N. Specific activity (nmoles phosphate/mg/min) was normalized to protein concentration. The results represent the mean and SD of three independent experiments. (B–D) The deconjugation of Atg8–PE is important for autophagosome formation. Spheroplasts prepared from atg8Δ pep4Δ cells overexpressing GFP-Atg8 (B; UNY125), atg4Δ atg8Δ pep4Δ cells overexpressing GFP-Atg8ΔR (C; UNY124), or atg8Δ pep4Δ cells bearing an empty vector (D; YZX214) were examined by a protease-protection analysis as described in Figure 3 and Materials and Methods.

ImmunoEM (IEM) experiments to detect GFP-Atg8 revealed that in the wild-type atg8Δ pep4Δ strain overexpressing GFP-Atg8 under the control of the copper inducible promoter, we could detect cells showing robust autophagosome formation and autophagic bodies within the vacuole lumen (Fig. S6A). Gold particles decorating GFP-Atg8 were visible at the phagophore, on completed autophagosomes and within autophagic bodies. We found that the frequency of autophagosome/autophagic body formation was reduced in an Atg8–PE deconjugation-defective mutant; atg4Δ atg8Δ pep4Δ cells overexpressing GFP-Atg8ΔR, showed fewer autophagosomes and autophagic bodies compared with the wild-type strain (Fig. S6B–D). On average, wild-type cells overexpressing GFP-Atg8 showed nine autophagic bodies per cell, whereas in the mutant strain overexpressing GFP-Atg8ΔR there were approximately four autophagic bodies per cell (Fig. S6D). We also confirmed that the puncta formed in the atg4Δ atg8Δ cells overexpressing GFP-Atg8ΔR were primarily unclosed structures that were unable to efficiently protect prApe1 from proteinase K digestion in a protease-protection assay (Fig. 4B–D), and that did not mature based on the persistence of the punctate chimeric GFP signal in the cytosol (Fig. S6E). Therefore, the overexpression of Atg8ΔR was not sufficient for double-membrane vesicle biogenesis and/or closure. Taken together, these results suggested that the decrease in autophagy observed in atg4Δ atg8Δ cells expressing GFP-Atg8ΔR was not merely due to the inadequate amount of Atg8–PE availability at the PAS, but rather that a block in Atg8–PE deconjugation interfered with a critical step in the autophagosome biogenesis process.

A defect in Atg8-PE deconjugation results in impaired autophagosome formation

In order to examine the ultrastructure of the phagophore and phagophore membrane dynamics in the Atg8 deconjugation mutant strain, we performed IEM to analyze the membrane structures decorated with Atg9-GFP as described previously.^39,40 Wild-type, atg8Δ, atg4Δ atg8Δ, or atg4Δ atg8Δ cells expressing CUP1 promoter-driven Atg8ΔR were grown in YPD medium and shifted to SD-N for 1 h before being embedded for IEM and analyzed at the ultrastructural level after immunogold labeling with anti-GFP antibodies. In wild-type starved cells, although the Atg9-positive compartments were detected at a lower frequency than in growing cells, two types of Atg9-positive structures could be seen. The first were small tubulovesicular clusters decorated with a few gold particles, and one to three of these profiles were detected in the cell sections displaying clear labeling. Under starvation conditions, the rate of double-membrane vesicle formation is higher than in the presence of nutrients, and fluorescence microscopy analyses showed that Atg9 was dispersed in small and more numerous punctate structures (not shown), probably reflecting the increased dynamics of autophagosomal membranes. As a result, the observed structures are very likely Atg9 reservoirs, and the difficulty in detecting them is probably due to their reduced dimensions, e.g., fewer epitopes.

The second type of observed structures were also clusters of tubulovesicular membranes, but they were more prominently labeled and their overall surface was more spread out than that of the Atg9-positive clusters observed in growing wild-type cells (Fig. 5A and B). When detected, there was no more than one of these structures per cell section, and often they were close to the vacuole. All together, these observations suggest that this compartment could represent the PAS or a nascent autophagosome (Fig. 5A and B). In the atg8Δ strain, in contrast, the frequency of appearance of the Atg9-positive structures was much higher than in the wild type, consistent with a block in autophagy displayed by these cells. These compartments consisted of more linear and/or circular tubules than the Atg9-positive structures observed in wild-type cells and they were often adjacent to the vacuole, indicating they could also be autophagosomal intermediates (Fig. 5C and D). This more complex ultrastructure may indicate that some fusion events were taking place, but that they were not sufficient to complete the autophagosomes. The atg4Δ atg8Δ strain displayed a frequency and morphology of Atg9-positive tubulovesicular clusters very similar to those seen in the atg8Δ strain (Fig. 5E and F). In the atg4Δ atg8Δ cells expressing CUP1 promoter-driven Atg8ΔR, in contrast, the frequency of appearance of Atg9-positive membranes was similar to that detected in wild-type cells, in agreement with the observation that autophagy was partially operative in this background. With cells expressing CUP1 promoter-driven Atg8ΔR we observed three different membranous profiles: (1) There were Atg9 reservoirs similar to the ones observed in starved wild-type cells (not shown). (2) Most commonly, we observed rosette-like structures composed of enlarged tubules and vesicles (Fig. 5G and H). These Atg9-containing rosette-like structures suggest that the phagophore membrane expands much more than in the atg4Δ atg8Δ double mutant; however, its biogenesis cannot be completed to generate an autophagosome. (3) We also detected circular mono-cisternae that probably represented expanded phagophores (Fig. 5I and J).

Figure 5. Ultrastructure of the Atg9-containing compartments by IEM. Wild-type (WT, MMY067; A and B), atg8Δ Atg9-GFP (MMY085) (C and D), atg4Δ atg8Δ Atg9-GFP (UNY184; E and F), and atg4Δ atg8Δ Atg9-GFP CUP1p-ATG8ΔR (UNY185; G–J) cells were grown in rich medium to exponential phase, shifted to nitrogen starvation medium for 1 h and processed for IEM. Cryosections obtained from each sample were immunogold-labeled with an anti-GFP antibody to localize Atg9-GFP. Scale bars, 200 nm. The superscript denotes 10-nm gold particles.

The deconjugation of Atg8–PE by Atg4 is required for the disassembly of Atg14 from the PAS

Among the known Atg proteins, Atg8 is the last protein to be recruited to the PAS.²⁷ Therefore, a defect in the delipidation of Atg8–PE is unlikely to affect the recruitment of other Atg proteins to this site. Instead, we hypothesized that the deconjugation of Atg8–PE may be an important step in autophagosome completion, which in turn serves as a cue for the disassembly of other Atg components. To test this hypothesis, we developed a disassembly assay where we examined the localization pattern of a marker protein, Atg14, which is a component of the PtdIns3K complex I, in an atg4Δ atg8Δ strain overexpressing Atg8ΔR under nitrogen starvation conditions; the PtdIns3K complex I is localized to the PAS where it generates PtdIns3P that is needed for the recruitment of other Atg proteins. As controls for this experiment we examined the localization of Atg14-GFP in an atg4Δ atg8Δ strain expressing Atg4 and CUP1 promoter-driven Atg8 (referred to here as the wild type). The quantification of Atg14-GFP puncta revealed that in the atg4Δ atg8Δ strain overexpressing Atg8ΔR, ~73% of the cells had perivacuolar Atg14-GFP puncta, compared with the wild-type strain where ~27% of the cells showed Atg14-GFP puncta (Fig. 6). In order to confirm that the defects in the localization of the Atg14-GFP marker protein was specifically due to the absence of Atg4 activity resulting in the inability to deconjugate Atg8–PE, we examined Atg14-GFP localization in the atg4Δ atg8Δ strain expressing CUP1p-Atg8ΔR and the catalytic site mutant of Atg4, Atg4^C147S. This mutant of Atg4 lacks proteolytic activity and is defective in autophagy.¹³ We found that in the presence of this mutant, ~76% of cells showed Atg14-GFP puncta (Fig. 6). Finally, to rule out the possibility that the increase in Atg14-GFP puncta was simply due to an accumulation of autophagosomal intermediates resulting from a general block in autophagy, we examined the localization of the Atg14-GFP chimera in an atg8Δ strain after a 1 h shift to nitrogen starvation medium. We found that in this strain only ~33% of the cells showed Atg14-GFP puncta. These results are consistent with the delipidation of Atg8–PE controlling the timing of dissociation of Atg14, and presumably other Atg proteins, upon autophagosome completion.

Figure 6. Atg4-mediated deconjugation of Atg8–PE is necessary for the disassembly of Atg14 from the PAS. (A) Atg14-GFP localization was examined in an atg8Δ strain (UNY177), the wild-type strain (atg4Δ atg8Δ Atg14-GFP CUP1p-Atg8 Atg4), Atg8–PE deconjugation-defective strain (atg4Δ atg8Δ Atg14-GFP CUP1p-Atg8ΔR), or an Atg8–PE deconjugation-defective strain bearing a catalytic site mutant of Atg4 (atg4Δ atg8Δ Atg14-GFP CUP1p-Atg8ΔR Atg4^C147S). Cells were grown to mid-log phase in synthetic medium and shifted to SD-N for 2 h before imaging. (B) Quantification of Atg14-GFP puncta. DIC, differential interference contrast. Scale bar, 5 µm. The number of cells (n) counted to determine PAS localization for each strain is shown.

Discussion

In this work, we have addressed the question of the mechanistic significance of Atg8–PE deconjugation in autophagosome biogenesis. It has been previously reported that a defect in this deconjugation process results in a reduction in autophagy;¹³ however, the molecular mechanisms that caused an attenuation in autophagy activity were not clear. Our results show that in wild-type cells the proteolytic activity of Atg4 was involved in preventing the nonspecific localization of Atg8–PE to compartments such as the vacuole limiting membrane (Fig. 1C). Thus, in wild-type cells, not only is nonspecifically localized Atg8–PE cleaved by Atg4, but in addition there have to be mechanisms that protect Atg8–PE at bona fide sites such as the PAS from unregulated cleavage by this protease during the process of autophagosome formation. At this point we do not know at what step of autophagosome biogenesis the cleavage of Atg8–PE takes place—whether it occurs only upon autophagosome completion, or whether Atg8–PE attached to the autophagosome membrane is cleaved while the phagophore membrane is expanding. We also note that in the atg4Δ mutant GFP-Atg8ΔR accumulates on the vacuole rim, even in the absence of Atg1 (Fig. S3), indicating that this localization is not due solely to the fusion of the vacuole with autophagosomes containing Atg8–PE on the outer surface. This observation supports a hypothesis that Atg8 conjugation to PE occurs not just at the PAS or phagophore, but at other sites in the cell; subsequent deconjugation by Atg4 removes Atg8–PE from all sites except the PAS/phagophore, although the mechanism restricting Atg4 activity at the latter is not known.

The steady-state level of Atg8 regulates the size of autophagosomes; the artificial reduction of the cellular level of Atg8 results in smaller autophagosomes compared with a wild-type strain, but the number of autophagosomes is the same as that of the wild type.¹⁵ Currently it is not known how different amounts of Atg8 modify the curvature of the phagophore to yield autophagosomes of varying sizes, but it appears that delipidation of Atg8–PE has to be a part of that mechanism. Indeed, in an atg4Δ atg8Δ strain expressing Atg8ΔR, the size of the autophagic bodies produced was about half that of the wild-type strain (Fig. 2). Even the overexpression of Atg8–PE in the atg4Δ atg8Δ strain was not sufficient to overcome the defect in autophagosome size. Additionally, an Atg8–PE delipidation defect resulted in an 80% reduction in the number of autophagic bodies produced, compared with a wild-type strain. Therefore, optimal autophagosome formation seems to require a balance between Atg8 lipidation and delipidation reactions.

In the absence of Atg4 in yeast, Atg8 is not recruited to the PAS, but several of the Atg components such as the Atg12–Atg5-Atg16, Atg1-Atg13-Atg17, and Atg2-Atg18 complexes, Atg9 and Atg14 accumulate at the PAS.²⁷ These results suggest that the Atg4-mediated initial cleavage of the Atg8 arginine tail and its subsequent conjugation to PE is specifically required for the PAS recruitment of Atg8. Two studies examined the effects of preventing PE conjugation of the mammalian Atg8 homologs on autophagosome formation. In the first study, the authors generated ATG3-deficient mice,⁴¹ and in the second, the authors overexpressed a protease-inactive form of the mammalian Atg4 homolog in a cell culture system.²² Both approaches showed essentially similar phenotypes including incomplete autophagosomes that are defective in membrane closure, strongly suggesting that the PE-conjugated form of LC3 (LC3-II) is required for phagophore closure. A more recent study showed that the mammalian LC3 and GABARAP orthologs of Atg8 are both required for autophagy, but have different roles in autophagosome biogenesis: The LC3 subfamily of proteins are required for phagophore membrane elongation, whereas the GABARAP subfamily is required for autophagosome membrane closure.²⁰ The situation in mammalian cells appears to be different from that in yeast, as abnormally elongated phagophores have not been observed in the latter in the absence of Atg components. Also, because the complementary experiment of exclusively examining the effect of a block in deconjugation of PE from Atg8 orthologs has not been performed in mammalian cells, we do not know whether closure of autophagosomes requires a balance between Atg8 lipidation and delipidation, as is the case in yeast.

Based on the observation that the Atg8–PE conjugate drives the tethering and hemifusion of liposomes in vitro, it has been proposed that Atg8–PE acts as a “fusogen” and plays a role in the expansion of autophagosome membranes.⁴² Our prApe1 protease protection (Figs. 3 and 4) and IEM results (Fig. 5), suggest that in vivo the chronic expression of Atg8–PE is in fact a limiting factor for phagophore expansion and autophagosome sealing. One reason for this could be that phagophores developing in Atg8–PE deconjugation-defective cells strongly bind other Atg components (Fig. 6) and prevent them from participating in their normal roles in autophagosome biogenesis. Furthermore, recent data suggest that the demonstration of a fusogenic role for Atg8 may be an artifact resulting from the in vitro use of liposomes containing nonphysiological levels of PE.⁴⁰

Using the Atg8–PE deconjugation-defective mutant, we have been able to dissect the temporal order of the release of Atg14 relative to that of Atg8. While there are several studies focusing on the order of recruitment of Atg proteins in PAS organization, we know very little about the order of exit of these proteins upon autophagosome completion. Our work shows that the deconjugation of Atg8–PE serves as a signal for the release of at least some of the Atg components upon autophagosome completion (Fig. 6). The steady-state level of Atg8 is upregulated in response to starvation conditions, perhaps to compensate for the Atg8 molecules that are trapped inside the inner membrane of completed autophagosomes; however, the expression level of most other Atg proteins remains relatively unchanged. Thus, in this scenario the delipidation of Atg8–PE may be critical for the disassembly of Atg components after autophagosome completion in order to facilitate successive rounds of autophagosome formation. Nonetheless, atg4Δ atg8Δ cells that express Atg8ΔR display a frequency and morphology of Atg9-containing structures that are similar to the wild type, and this strain is partly functional for autophagy. Furthermore, in the presence of Atg8ΔR the Atg9-containing membranes are more expanded than is seen in the atg4Δ atg8Δ double-mutant strain lacking any Atg8, which is completely defective for autophagy. These findings suggest that Atg8 (or Atg8ΔR)–PE also contributes to the expansion of the phagophore membrane. Additional work will be needed to determine the precise functions of Atg8 in autophagosome formation.

Materials and Methods

Yeast strains, media, reagents and antisera

The yeast strains used in this study are listed in Table 1. The loxP-Cre system was used to construct knockout strains.⁴⁴ Integration of the GFP tag at the 3′ end of the ATG14 open reading frame was performed by a PCR-based procedure.⁴⁵ Strains were grown in YPD (1% yeast extract, 2% peptone, and 2% glucose), SMD (2% glucose and 0.67% yeast nitrogen base without amino acids, but supplemented with vitamins and appropriate auxotrophic amino acids), or SD-N (2% glucose, and 0.17% yeast nitrogen base without amino acids and ammonium sulfate). In experiments where SD-N was supplemented with Cu²⁺, 2 μM CuSO₄ was added. FM 4-64 was obtained from Invitrogen (T-3166). Antisera against Ape1³³ and Atg8²³ have been described previously. Anti-Pgk1 antiserum was a gift from Dr. Jeremy Thorner (University of California, Berkeley). Antibodies to carboxypeptidase Y (Prc1) were obtained from Invitrogen (A-6428, no longer available), and monoclonal anti-YFP antibody clone JL-8 was from Clontech/Takara Bio Group (632381).

Table 1. Yeast strains used in this study.

Strain	Descriptive Name	Genotype	Source or reference
MMY067	WT	SEY6210 atg9Δ::KAN TPI1-ATG9-GFP::URA3	35
MMY085	atg8Δ Atg9-GFP	SEY6210 atg9Δ::KAN atg8Δ::ΔEU2 TPI1-ATG9-GFP::URA3	This study
SEY6210	WT	MATα ura3-52 leu2-3,112 his3-Δ200 trp1-Δ901 lys2–801 suc2-Δ9 GAL	43
TN124	Pho8Δ60 parent	MATα leu2–3,112 trp1 ura3–52 pho8::pho8Δ60 pho13Δ::LEU2	31
UNY5	atg1Δ	TN124 atg1Δ::KAN	This study
UNY45	atg4Δ atg8Δ Atg8ΔR	TN124 atg4Δ atg8Δ p1K-Atg8ΔR::TRP1 URA3	This study
UNY52	atg4Δ atg8Δ GFP-Atg8ΔR	TN124 atg4Δ atg8Δ p1K-GFP-Atg8ΔR::TRP1 URA3	This study
UNY111	pep4Δ vps4Δ	TN124 pep4Δ::URA3 vps4Δ::BLE	This study
UNY112	atg1Δ pep4Δ vps4Δ	TN124 atg1Δ::KAN pep4Δ::URA3 vps4Δ::BLE	This study
UNY113	atg4Δ	TN124 atg4Δ::URA3 TRP1	This study
UNY117	pep4Δ	TN124 pep4Δ::KAN TRP1 URA3	This study
UNY118	atg4Δ atg8Δ pep4Δ Atg8ΔR	TN124 atg4Δ atg8Δ pep4Δ::KAN Atg8ΔR::TRP1 URA3	This study
UNY119	atg1Δ pep4Δ	TN124 atg1Δ::KAN pep4Δ::URA3	This study
UNY126	atg4Δ atg8Δ Atg14-GFP	TN124 atg4Δ atg8Δ ATG14-GFP::KAN	This study
UNY177	atg8Δ Atg14-GFP	TN124 atg8Δ::URA3 Atg14-GFP::KAN	This study
UNY184	atg4Δ atg8Δ Atg9-GFP	SEY6210 atg9Δ::KAN atg8Δ::LEU2 atg4Δ::BLE TPI1-ATG9-GFP::URA3	This study
UNY185	atg4Δ atg8Δ Atg8ΔR Atg9-GFP	SEY6210 atg9Δ::KAN atg8Δ::LEU2 atg4Δ::BLE TPI1-ATG9-GFP::URA3 CUP1p-ATG8ΔR::TRP1	This study
YCY41	atg4Δ atg8Δ pep4Δ vps4Δ Atg8ΔR	TN124 atg4Δ atg8Δ ATG8ΔR::TRP1 pep4Δ::KAN vps4Δ::BLE	This study
YCY58	atg4Δ atg8Δ GFP-Atg8ΔR	TN124 atg4Δ atg8Δ GFP-ATG8ΔR::TRP1	This study
YCY59	atg4Δ atg8Δ atg1Δ GFP-Atg8ΔR	TN124 atg4Δ atg8Δ atg1Δ::URA3 GFP-ATG8ΔR::TRP1	This study
YZX206	atg4Δ atg8Δ	TN124 atg4Δ atg8Δ	This study
YZX214	atg8Δ pep4Δ	TN124 atg8Δ::KAN URA3 pep4Δ::TRP1	11
YZX231	atg8Δ	TN124 atg8Δ::KAN URA3 TRP1	11
YZX232	WT	TN124 URA3 TRP1	11
YZX247	atg8Δ GFP-Atg8	TN124 atg8Δ::KAN p1K-GFP-ATG8::URA3 TRP1	11
YZX293	atg4Δ atg8Δ Atg8ΔR	TN124 atg4Δ atg8Δ p1K-ATG8ΔR::TRP1	This study
YXZ334	atg4Δ atg8Δ GFP-Atg8ΔR CUP1p-GFP-Atg8ΔR	TN124 atg4Δ atg8Δ p1K-GFP-Atg8ΔR::TRP1 CUP1p-GFP-ATG8ΔR::URA3	This study
YZX396	atg4Δ atg8Δ GFP-Atg8ΔR	TN124 atg4Δ atg8Δ p1K-GFP-Atg8ΔR::TRP1 URA3	This study

Plasmids

Plasmids ATG8p-GFP-ATG8(406) containing ~1 kb of the ATG8 promoter region in front of the GFP-Atg8 open reading frame, ATG3p-ATG8(406) containing 380 base pairs of ATG3 5′ sequence, mCherry-prApe1(404) driven by the APE1 promoter, and ATG8p-GFP-ATG8ΔR(406) were described previously.¹⁵ For constructing pATG4(414), full-length ATG4 with 1 kb endogenous promoter and terminator were amplified from genomic DNA and cloned into the SacI site of pRS414. CUP1p-GFP-Atg8ΔR(406) and CUP1p-GFP-Atg8(406), were constructed by replacing the endogenous ATG8 promoter with the CUP1 promoter. Atg4^C147S was generated by PCR-based site-directed mutagenesis using full-length ATG4 (in pRS414) as a backbone.^‡

Electron microscopy

Two different procedures were performed to prepare samples for electron microscopy analysis. In the first, sample preparation was performed using a conventional embedding protocol as described previously.⁴³ Briefly, cells were fixed in potassium permanganate and embedded in Spurr’s resin. After resin polymerization, 65- to 75-nm sections were mounted on nickel grids and stained with 1% uranyl acetate followed by lead citrate. Ultrathin sections were imaged using a Philips CM-100 transmission electron microscope. Image acquisition and quantification of autophagic bodies have been described previously.¹⁵ For IEM, ultrathin sections were stained with anti-YFP antibody followed by 0.8-nm colloidal-gold-conjugated goat anti-mouse IgG (Aurion, 800.022), and then detected with silver enhancement. Samples were examined with an H-800 electron microscope (Hitachi High Technologies) at 125 kV. A film scanner was used to prepare images, and scale bars were added in Photoshop (Adobe). For IEM detection of Atg9-GFP, cells grown to early-log phase were shifted to nitrogen starvation medium for 1 h, fixed, embedded, cryo-sectioned and labeled as described previously.³⁹

Fluorescence microscopy

For imaging, yeast cells expressing proteins with fluorescent tags were grown to OD₆₀₀ = 0.6 to 0.8 in YPD or selective synthetic medium, or shifted to nitrogen starvation medium (SD-N) to induce nitrogen starvation. The conditions for labeling vacuolar membranes with FM 4-64 have been described previously.⁴⁶ Cells were visualized on a fluorescence microscope (IX71; Olympus) with DeltaVision (Spectris; Applied Precision), and images were captured with a 100× objective lens. A CCD camera (CoolSNAP HQ; Photometrics) fitted with differential interference contrast optics was used to capture images. Fifteen z-section images were collected and were deconvolved using softWoRx software (Applied Precision). All fluorescence microscopy images show a single focal z section.

Other assays

The protein extraction, western blotting, prApe1 maturation, prApe1 fractionation, alkaline phosphatase (Pho8Δ60) and protease-protection assays have been described previously.^35,47,48 For the protease-protection assay, spheroplasts were starved for 1 h in SD-N medium supplemented with 1.2 M sorbitol, collected by centrifugation and osmotically lysed. After a low speed centrifugation step to remove unbroken cells, the resulting total lysate was further separated into 5,000 × g lysate supernatant and pellet (P5) fractions. The efficiency of lysis and fractionation was determined by immunoblot analysis using anti-Pgk1 and anti-Ape1 antisera. The P5 fraction from each strain was divided into four parts and subjected to no treatment, treatment with detergent, or treatment with proteinase K in the presence or absence of detergent, and analyzed by immunoblot.

Glossary

Abbreviations:

AB

autophagic body

AP

autophagosome

Atg

autophagy-related

IEM

immunoelectron microscopy

mApe1

mature aminopeptidase I

PAS

phagophore assembly site

PE

phosphatidylethanolamine

prApe1

precursor aminopeptidase I

SD

standard deviation

SEM

standard error of the mean

TEM

transmission electron microscopy

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Supplemental Materials

Supplemental materials can be found at: www.landesbioscience.com/journals/autophagy/article/19385