In vivo and in vitro reconstitution of Atg8 conjugation essential for autophagy.

In an analogous manner to protein ubiquitination, The C terminus of Atg8p, a yeast protein essential for autophagy, conjugates to a head group of phosphatidylethanolamine via an amide bond. Though physiological role of this reaction is assigned to membrane organization during autophagy, its molecular details are still unknown. Here, we show that Escherichia coli cells coexpressed Atg8p, Atg7p (E1), and Atg3p (E2) allowed to form conjugate of Atg8p with endogenous PE. Further, we established an in vitro Atg8p-PE reconstitution system using purified Atg8pG116, Atg7p, Atg3p, and PE-containing liposomes, demonstrating that the Atg7p and the Atg3p are minimal catalysts for Atg8p-PE conjugate reaction. Efficiency of this lipidation reaction depends on the state of the substrate, PE (phospholipid bilayer and its lipid composition). It is also suggested that the lipidation induces a conformational change in the N-terminal region of Atg8p. In vitro system developed here will provide a powerful system for further understanding the precise role of lipidation and interaction of two ubiquitin-like systems essential for autophagy.

Autophagy, an intracellular bulk degradation system in the lysosomes/vacuoles of eukaryotic cells is necessary for the recycling of cytoplasmic components for survival during nutrient starvation conditions (1)(2)(3). Membrane dynamics in autophagy are distinct from classical membrane trafficking. In particular, the mechanism by which this new double membrane-bounded compartment, the autophagosome, is formed has remained poorly understood. Genetic screens of the budding yeast, Saccharomyces cerevisiae, have identified 16 genes potentially involved in this process, named ATG genes (formerly APG/AUT) (4). Dissection of autophagosome biogenesis through intensive characterization of these gene products Atgps (1) has revealed that the overall process of autophagy in yeast is similar to that in higher eukaryotes, exhibiting conservation of the molecular machinery involved (5).
We have observed a small number of autophagosome-like structures of abnormal morphology in the null mutant of ATG8 (20). Whereas Abeliovich et al. (21) reported that significantly smaller vesicles than normal autophagosome were observed in a similar ⌬atg8 strain. These observations imply that the Atg8p-PE may function in normal development of autophagosomal membrane. In fact, previous immunoelectorn microscopic analyses demonstrated the specific localization of Atg8p to the isolation membrane (intermediate structure of autophagosome) under the starvation conditions (20).
The precise roles of the two conjugates during autophagosome formation remain unclear. Atg8p-PE conjugation is severely reduced in the absence of the Atg12p-Atg5p conjugate.
Thus, it appears likely that the functions of these two conjugates, Atg8p-PE and Atg12p-Atg5p, are closely related. To identify the molecular events governing autophagosome formation, we need to address not only the function of each of the two conjugates, but also the interrelationship between Ag8p-PE and Atg12p-Atg5p. In this study, to address the molecular machinery involved in membrane dynamics, we focused on Atg8p lipidation, and developed in vitro reconstitution system of Atg8p-PE.
Protein Purification-To purify the recombinant Atg proteins, we constructed glutathione S-transferase (GST) fusions of the desired Atgs (GST-Atg7-Myc, GST-Atg3, GST-Atg3 C234S and GST-Atg8 G116 ) by insertion of PCR products into the cloning site (BamHI/SalI) of pGEX4-T-1 (Amersham Biosciences). Following introduction of the obtained plasmids into BL21 (DE3) E. coli, expression of each protein was induced by addition of 0.05 mM isopropyl ␤-D(-)-thiogalactopyranoside to the medium. After induction, cells were disrupted by sonication in TBST (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Triton X-100) containing 2 mM DTT, and 1 mM phenylmethylsulfonyl fluoride. After removal of cell debris by centrifugation at 10,000 rpm for 10 min, obtained cell lysates were incubated with 1 ml of glutathione-Sepharose 4B (50% slurry in TBST) (Amersham Biosciences) for 30 min at 4°C to collect recombinant proteins. The proteins bound to the Sepharose beads were washed with TBST, then suspended in 400 l of TBS containing 2 mM DTT. The GST-fused Atg proteins were treated with thrombin (10 units) at 20°C for 1 h; the glutathione-Sepharose 4B/GST complexes were then removed by centrifugation. Protein concentrations were determined by the Bradford method (Pierce) using bovine serum albumin as a standard. The purified Atg proteins were then diluted in TBS containing 50% glycerol for storage at Ϫ30°C.
Extraction of Total Lipids and Liposome Preparation-W3011 and GN10 E. coli strains were grown in LB medium containing 50 mM MgCl 2 to stationary phase. SEY6210 yeast strain was grown in YEPD to stationary phase. Both types of cells were harvested (cells wet-weight 10 g), and total lipids were extracted by Bligh and Dyer's methods (25). The total lipids were dissolved in 10 ml of chloroform, and the concentration of total phospholipids was quantified by a phosphorous assay using the phosphomolybdate reaction (26). Individual phospholipids were separated by two-dimensional thin-layer chromatography, and the composition of phospholipids was determined by the phosphorous assay Total lipids from E. coli and yeast were used to generate E. coli and yeast total lipid liposomes, respectively. To prepare liposomes with various phospholipid compositions, phospholipids were mixed in the appropriate ratios from stocks dissolved in chloroform as described above. PE from E. coli, dioleoylphosphatidylethanolamine (DOPE), 1-palmitoyl-2-oleoylphoshatidylcholine (POPC), dioleoyl phosphatidylglycerol (DOPG), phosphatidylinositol (PI) from bovine liver, and dioleoylphosphatidic acid (DOPA), dioleoylphosphatidylserine (DOPS) were purchased from Avanti Polar Lipids. After transfer to a glass tube, the chloroform solvent was removed by rotary evaporation. Samples were dried further in a desiccator under vacuum for 12 h. The resulting lipid film was suspended in a buffer (25 mM Tris-HCl, pH 7.5, 137 mM NaCl, 2.7 mM KCl) at a final concentration of 1 mM phospholipids by vortexing at room temperature. Samples were then subjected to sonication for 5 min to obtain small unilamellar liposomes.
Antibodies and Immunoblotting-Polyclonal antibodies against fulllength Atg3p and against the Atg8p N-terminal peptide have been previously described (7). A polyclonal antibody against full-length Atg8p and a monoclonal antibody against Myc (9E10) were purchased from Rockland and BabCo, respectively. SDS-PAGE was performed according to Laemmli's method. SDS-PAGE separation in the presence of 6 M urea (urea-SDS-PAGE gel) was used to distinguish Atg8p and Atg8p-PE conjugates, as described (13). To visualize the protein bands within the gel, SDS-PAGE gels were stained using GelCode Blue (Pierce). For immunoblot analyses, proteins were transferred to polyvinylidene difluoride membranes (Immobilon-PSQ, Millipore). Immunoblotting analysis was performed using rabbit or mouse antibodies against the specified proteins, then visualized using either peroxidase-conjugated goat anti-rabbit IgG or rabbit anti-mouse IgG. The immunoreactive protein bands were detected using Western Lightning Chemiluminescence Reagent Plus (PerkinElmer Life Sciences) Immunoprecipitation-In vitro Atg8p-PE conjugation was performed as described in the above. The Atg8p-PE-bound liposomes were collected by centrifugation at 15,000 rpm for 10 min. The harvested Atg8p-PE liposomes were mixed with the nearly equal amount of Atg8p, and treated with 2% CHAPS for 1 h on ice. The solubilized Atg8p-PE and Atg8p were suspended in IP buffer (50 mM Tris-HCl, pH 7.5. 150 mM NaCl, 2% CHAPS, and 0.01% bovine serum albumin (w/w)), and subjected to immunoprecipitation analysis according to previously reported methods with the following modifications (7). Briefly, samples were incubated at 4°C for 2 h in the presence or absence of the anti-N-terminal 14 peptide of Atg8p antibody. Protein G-Sepharose 4 Fast Flow beads (Amersham Biosciences) were subsequently added to the samples and further incubated at 4°C for 1 h. Immune complexes were washed with TBS containing 2% CHAPS and suspended in SDS-PAGE sample buffer. The resulting immunoprecipitates were separated by urea-SDS-PAGE and visualized by SyproOrange staining (Amersham Biosciences).
Trypsin Digestion-In vitro Atg8-PE conjugation was performed as described in the above. Then, Atg8-PE-conjugated liposomes were recovered by centrifugation at 15,000 rpm for 10 min, and solubilized in TBS containing 1 mM DTT and 1% CHAPS for 1 h on ice. The mixture of Atg8p-PE (2.0 g) and Atg8p (1.5 g), or each of them was treated with trypsin (0.01 g) (Sigma T8802) in 20 l of TBS containing 1 mM DTT and 1% CHAPS at 30°C for the indicated time (0, 1, 3, 7, and 15 min). The digestion was stopped by boiling in SDS-PAGE sample buffer. Samples were subjected to urea-SDS-PAGE and stained by CBB.

Expression of the Components of Atg8 System in E. coli-
Previous studies of the Atg8 system revealed that the C-terminal glycine of Atg8p is covalently conjugated to an amino group of PE through ubiquitination-like reactions requiring Atg7p and Atg3p, functioning as E1-and E2-like enzymes, respectively (7). We have not, however, excluded the possibility that other factor(s), such as an E3, is required for conjugation.
To address this issue, we applied an E. coli expression system. For this purpose, expression plasmids encoding the components of the Atg8 system, Atg7p bearing a Myc epitope tag, Atg3p, and either the processed or unprocessed form of Atg8p, under the control of inducible arabinose promoter were constructed (Fig. 1A). E. coli BL21 (DE3) cells harboring these expression plasmids were cultured to logarithmic growth phase, then cultured for an additional 60 min in the presence of 0.2% arabinose. Total cell lysates from the E. coli were then subjected to SDS-PAGE containing 6 M urea (urea-SDS-PAGE). Subsequent immunoblot analyses with anti-Myc and anti-Atg3p antibodies detected Atg7p-Myc and Atg3p at the expected molecular masses of 78 and 36 kDa, respectively (Fig.  1B, lanes 1-3). Atg8p R117 , the nascent, unprocessed form of Atg8p, was detected as a single band of ϳ13.5 kDa, the predicted molecular mass; Atg8p G116 , the processed form of Atg8p, migrated similarly to Atg8p R117 (Fig. 1B, lanes 1 and 2).
Interestingly, only upon simultaneous expression of Atg7p-Myc, Atg3p, and Atg8p G116 , an additional Atg8p band appeared FIG. 1. Expression of the Atg8 system components in E. coli. A, diagram of expression plasmids. To construct expression plasmids for Atg7p-Myc, Atg3p, Atg8p G116 (the processed form of Atg8p), and Atg8p R117 (nascent form of Atg8p) in E. coli, each gene was introduced into pBAD (pBAD18) or pBADC (pBAD33) plasmids. The inserted DNA fragments contain a Shine-Dalgarno (SD) sequence. B, modified Atg8p in E. coli expression system. Expressed proteins in E. coli cells were detected by immunoblotting with anti-Myc, anti-Atg3p, and anti-Atg8p antibodies. The mobility of the Atg8p bands in E. coli cells concurrently expressing Atg7p-Myc, Atg3p, and Atg8p G116 was compared with Atg8p and Atg8p-PE forms derived from ⌬atg7, ⌬atg3 and wild-type (wt) yeast by urea-SDS-PAGE. Yeast cell lysates were prepared by glass bead disruption, as previously reported (20). R indicates Atg8p R117 . C, Atg8p intermediates in E. coli expression system. Atg7p-Myc, Atg3p, and Atg8p G116 were induced in E. coli cells as indicated; CS indicates Atg7p C507S or Atg3p C234S , whereas R designates Atg8p R117 expressed by addition of 0.2% arabinose to cultures for 1 h. Each product was analyzed by the Western blotting as described above. (Fig. 1B, lane 3). This band, migrating faster than the Atg8p G116 form, was not observed in E. coli co-expressing the unprocessed Atg8p R117 form of Atg8p (Fig. 1B, lane 2). The faster migrating band is not an Atg8p degradation product, but a modified form, since only one band for Atg8p was observed by conventional SDS-PAGE (data not shown) (13). Furthermore, the mobility of the modified Atg8p by urea-SDS-PAGE corresponded to that observed for Atg8p-PE derived from yeast cells (Fig. 1B, lanes 3 and 6).
During Atg8p lipidation, the C-terminal glycine of Atg8p first links to Cys 507 of Atg7p, then is transferred to Cys 234 of Atg3p through generation of a thioester bond (7). The mutant proteins Atg7p C507S -Myc and Atg3p C234S possess serine residues at the active site cysteines. Expression of the Atg7p C507S -Myc or Atg3p C234S mutants in E. coli in place of the wild-type Atg7p-Myc or Atg3p, respectively, abolished the appearance of the modified Atg8p form (Fig. 1C, lanes 5 and 6). Instead, expression of Atg3p C234S generated a novel 50-kDa band, detectable with both anti-Atg8p and anti-Atg3p antibodies (data not shown and Fig. 1C, lane 6). Appearance of the band was resistant to high concentrations of dithiothreitol (100 mM DTT) and dependent upon the Atg8p C-terminal glycine and active Atg7p (Fig. 1C, lanes 7 and 8). These results suggest that this species is an Atg8p-Atg3p ester conjugate formed between the C-terminal glycine (Gly 116 ) of Atg8p and the Atg3p C234S serine residue (Fig. 1C, lane 6). We concluded that sequential enzymatic reactions occur in E. coli cells in a manner similar to that observed in yeast cells. Thus, these co-expression experiments indicate that the E1 and E2 enzymes, Atg7p and Atg3p, are sufficient for the modification of the processed form of Atg8p.
PE Is the Target of Atg8p in E. coli-The above results prompted us to attempt the complete in vitro reconstitution of Atg8p conjugation using purified Atg proteins and liposomes. Plasmids encoding GST fusions of each of the Atg proteins were constructed using the pGEX4T-1 vector, to produce GST-Atg7-Myc, GST-Atg3, and GST-Atg8 G116 , then introduced into the BL21 (DE3) E. coli strain. The recombinant proteins were induced by IPTG and purified as described under "Experimental Procedures." Purified Atg proteins exhibited high purity, as demonstrated by Coomassie Brilliant Blue (CBB) staining in SDS-PAGE ( Fig. 2A, lanes 3-5). Liposomes were prepared using E. coli total phospholipids, as described under "Experimental Procedures." After mixing the purified Atg7p-Myc, Atg3p, and Atg8p G116 with E. coli phospholipid-containing liposomes, the mixture was incubated at 30°C in the presence of an ATP regeneration system. The reaction products were then subjected to urea-SDS-PAGE and visualized by CBB staining. The modified form of Atg8p was generated in this in vitro reaction in an ATP-dependent manner, as determined by the presence of the faster migrating, Atg8p-specific band ( Fig. 2A, lanes 6  and 7). E. coli membrane phospholipids consist primarily of PE, phosphatidylglycerol (PG), cardiolipin (CL), and phosphatidic acid (PA). PE is the major glycerophospholipid (ϳ75% of total lipids). E. coli PE is produced exclusively from phosphatidylserine (PS) via decarboxylation. An E. coli strain, GN10, possesses a null mutant of the pssA gene that encodes PS synthase; thus, this strain completely lacks PE (22). When this in vitro reaction was performed using liposomes made from the total lipids of the GN10 strain, the modified Atg8p could not be detected (Fig. 2A, lane 8). Liposomes were generated from a mixture of 30% GN10 total lipids and 70% purified E. coli PE; when these liposomes were subjected to in vitro reaction with purified Atg proteins, we observed the successful modification of Atg8p at levels similar to those seen with wild-type E. coli liposomes ( Fig. 2A, lanes 6 and 9). These results clearly dem-onstrate that PE is the target of Atg8p modification. We also investigated Atg8p-PE formation using liposomes made from pure phospholipids. Atg8p-PE conjugation could be reproduced with the liposomes composed of 70% dioleoylphosphatidylethanolamine (DOPE) and 30% 1-palmitoyl-2-oleoylphoshatidylcholine (POPC) (Fig. 2B). Taking together, we concluded that the PE-containing liposome is sufficient for conjugation to Atg8p; biological membranes containing additional components do not appear to be necessary.
Characterization of in Vitro Atg8p Lipidation-While in vitro conjugation studies of ubiquitin-like proteins, such as SUMO have been performed, conjugation reactions in the absence of E3 have always demonstrated low efficiencies. Such reactions were improved by the addition of excess amounts of E1 and E2 to SUMO (27). To examine the efficiency of in vitro Atg8p-PE conjugation, we titrated the amounts of Atg8p G116 and PE against fixed amounts of Atg7p-Myc and Atg3p. The results clearly indicated that the amount of Atg8p-PE was augmented with increasing Atg8p and PE ranging from a 5-fold to a 20-fold molar excess of these targets to Atg7p-Myc and Atg3p (Fig. 2B). These results suggest that Atg7p and Atg3p induce in vitro Atg8p-PE conjugation in a catalytic manner. Hence, we conclude that the minimal essential factors necessary for Atg8p-PE conjugation are Atg7p, Atg3p, Atg8p G116 , PE-containing membranes, and ATP.
It is an important question what physical state of lipid molecules is required for lipidation reaction. To investigate this problem, we pretreated 70% DOPE/30% POPC liposomes with an amphoteric detergent, 20 mM CHAPS, on ice for 30 min. We then subjected these lipids to the in vitro reconstitution assay in the presence of 10 mM CHAPS, the critical micelle concentration (CMC). As assessed by subsequent urea-SDS-PAGE analysis, the detergent completely prevented Atg8p lipidation (Fig. 2C, lane 2). Since the Atg8p-Atg3p C234S ester conjugate could be formed in the presence of detergent (Fig. 2C, lanes  3-5), activation of Atg8p and the thioester transferring of Atg8p from E1 to E2 occur normally in the presence of detergent. The formation of the amide bond between Atg8p and PE, however, requires PE in the lipid bilayer phase.
Lipid Composition of Target Membrane-We prepared liposomes with differential PE contents by combination of GN10 total lipids and purified E. coli PE. Interestingly, the yield of Atg8p-PE increased in a dose-dependent manner with increasing PE content up to 70% of the total lipids, despite an equal amount of total PE (70 nmol) present in each reaction (Fig. 3A). Liposomes containing more than 90% PE, however, resulted in a remarkably diminished quantity of Atg8p-PE conjugates (Fig. 3A). PE has a small hydrophilic headgroup responsible for its tendency to form non-bilayer structures (28). Thus, liposomes with excess PE may possess an unfavorable structure for Atg8p lipidation. We confirmed these results using two synthetic phospholipids, DOPE and POPC. When DOPE content was increased in DOPE/POPC liposomes, the yield of Atg8p-PE conjugate peaked at ϳ70% DOPE liposome content (Fig. 3B). By Western blotting analysis with an anti-Atg8p antibody, we could detect the Atg8p-PE species, even when the reaction was performed with 20 or 100% PE liposomes (Fig. 3B). The fast migrating band is Atg8p-PE, as it could be detected with 100% DOPE liposomes in the absence of additional lipids (Fig. 3B). From these results, we concluded that the PE content of membranes is an important factor governing the efficiency of Atg8p-PE formation.
The lipid composition of E. coli membranes is quite different from that of yeast. E. coli membranes contain more than 70% PE, while yeast organelle membranes contain ϳ15-30% PE (29). We examined in vitro Atg8p-PE formation using yeast total lipid liposomes, containing a ratio of PE/PC/PS/PI ϭ 16.6/ 51.7/13.7/18.0. Atg8p-PE conjugation in yeast liposomes was slightly more efficient than that seen in 20% DOPE/80% POPC liposomes, despite a lower PE content (Fig. 4). To determine the effect of other phospholipids on reaction efficiency, we mixed negatively charged phospholipids at final concentrations of 10 or 20% to the 20% DOPE/80% POPC liposomes. The level of Atg8p-PE formed in the 20% DOPE liposomes was enhanced by increases in either PI or PG (Fig. 4). A similar enhancement was also obtained by the addition of either PS or PA (data not shown). The Atg8p-PE conjugation reaction thus appears to be sensitive to membrane lipid composition.
High Sensitivity of Atg8p-PE in Western Blotting-We have two anti-Atg8p polyclonal antibodies, one generated by immunization with the N-terminal 14 amino acids (Met 1 -Arg 14 ) of Atg8p (hereafter anti-N-peptide antibody) (20) and the other a commercially available antibody against full-length Atg8p (Rockland). In both the E. coli and in vitro reconstitution experiments with Atg8p, the signal intensity of Atg8p-PE by immunoblot analysis was much higher than that of Atg8p G116 alone when using the anti-N-peptide antibody (see Fig. 3C). We quantified the differences in the signal intensities of Atg8p G116 and Atg8p-PE by immunoblot analysis with the anti-N-peptide antibody following the in vitro reaction at 30°C for 1 h in the absence or presence of ATP. By CBB staining, Atg8p G116 was significantly converted to Atg8p-PE in the presence (lane 1), but not the absence (lane 5), of ATP (Fig. 5A). By CBB staining, the two bands representing Atg8p-PE (lane 1) and Atg8p G116 (lane 5) were of similar intensities (Fig. 5A). Immunoblot analysis, however, with the anti-N-peptide antibody detected the Atg8p-PE band with a greater intensity than that of Atg8p G116 (Fig. 5A). Even when the mixture was prepared with Atg8p G116 and Atg8p-PE at a ratio 19:1, the signal for Atg8p-PE exhibited higher in intensity than that for Atg8p by the anti-N-peptide antibody immunoblotting (Fig. 5B). In contrast, Western analysis using the antibody against the whole Atg8 protein gave signals corresponding faithfully to the amounts of Atg8p and Atg8p-PE detected by CBB staining (Fig. 5B). These results indicate that the Atg8p-PE conjugate exhibits a higher affinity for the anti-N-peptide antibody.
Next, we examined the affinity of the two Atg8p forms for these antibodies in the absence of SDS denaturation using a direct dot-blot assay. The in vitro reaction samples (examined in Fig. 5A) were added to a dilution buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.2 mM DTT, 2% glycerol, 10 mM CHAPS, and 0.5 mg/ml bovine serum albumin) and spotted onto poly-vinylidene difluoride membranes after serial dilution (27.5-0.05 ng of Atg8p). Subsequently, each spot was washed by spotting with TBS and subjected to the immunoblot analysis with both the anti-N-peptide and anti-whole Atg8p antibodies. The resulting immunoblot signals demonstrated the high affinity of Atg8p-PE for the anti-N-peptide antibody, at least 12-fold higher than that of Atg8p G116 for this antibody by quantification using NIH image (Fig. 5C). In contrast, the affinity of Atg8p and Atg8p-PE for the antibody against whole Atg8p was equivalent (Fig. 5C).
We confirmed the high affinity of Atg8p-PE for the anti-Npeptide antibody by immunoprecipitation experiments. A mixture of equal amounts of Atg8p and Atg8p-PE was subjected to immunoprecipitation using the anti-N-peptide antibody in the presence of 2% CHAPS (Fig. 5D, lane 1). Following harvest using protein G Sepharose, eluted molecules were separated by urea-SDS-PAGE. By SyproOrange staining, no precipitates were detectable in control samples lacking antibody. The resulting immunoprecipitates using the anti-N-peptide antibody contained Atg8p-PE, but only minimal Atg8p (Fig. 5D, lane 3). Thus, the lipidated form of Atg8p preferentially binds the anti-N-peptide antibody (Fig. 5D, lanes 2 and 3). These results strongly suggest that conformation of lipidated Atg8p at the N-terminal is distinct from that of free Atg8p.
Further, to ascertain the conformation change of Atg8p-PE, Atg8p, and Atg8p-PE were subjected to limited proteolysis by trypsin. First, the mixture of Atg8p and Atg8p-PE was treated with trypsin at 30°C for the indicated time (Fig. 5E, lanes 1-5). The resulting products were separated by SDS-PAGE, and visualized by CBB staining. As shown in Fig. 5E 5 g; Fig.  5E, lane 1) was relatively resistant to trypsin digestion (Fig.  5E, lane 4) and still detectable at 15 min (Fig. 5E, lane 5). The remaining Atg8p (Fig. 5E, lanes 1-5) was not derived from cleavage of Atg8p-PE, since Atg8p and Atg8p-PE were digested in the same time course when treated individually (Fig. 5E,  lanes 6 -11). These results suggested that Atg8p becomes highly sensitive to trypsin digestion after lipidation. Hence, we concluded that PE conjugation may induce a conformation change in the Atg8p N terminus. DISCUSSION E. coli cell provides a suitable system for co-expression of multiple proteins under strict control. Here, we applied this expression system to the ubiquitin-like modification of Atg8p, successfully accomplishing the sequential reactions in vivo necessary for Atg8p conjugation. Co-expression of multiple components involved in a concerted mechanism provides a fruitful FIG. 3. In vitro assay of Atg8p-PE conjugation with variable PE membrane content. A, in vitro Atg8p-PE with E. coli liposome. In vitro reconstitution of Atg8 system was performed at 30°C for 20 min as described in Fig. 2 in the presence (ϩ) or absence (Ϫ) of the ATP regeneration system. The products were then analyzed by CBB-stained urea-SDS-PAGE gel. Liposomes were prepared from W3011 total lipids (wild-type; wt), GN10 total lipids (0% PE), a lipid mixture of GN10/ E. coli PE with the indicated PE content (10 -90%), and E. coli PE alone (100% PE). Each reaction mixture (100 l) contained the equal amount of PE (7 nmol), with the exception of PE-negative liposomes. B, in vitro Atg8p-PE generation from pure phospholipid liposomes. In vitro reactions were performed as described above using DOPE/POPC liposomes with the indicated PE content. The in vitro samples (20 l) were then subjected to urea-SDS-PAGE, and visualized by CBB staining. C, in addition, these samples (2 l) were also analyzed by Western blotting with an antibody against N-terminal peptide of Atg8p. abstraction facilitating the elucidation of the mechanisms of complicated reaction systems. We have also succeeded making Atg12p-Atg5p conjugate in E. coli cells. 2 As the final target of Atg8p modification is phosphatidylethanolamine, we expected that this lipid acceptor endogenously resided in E. coli membranes. The E. coli inner membrane turned out to be an unexpectedly efficient target membrane, which contains ϳ70% PE.
Complete in vitro reconstitution of Atg8p lipidation with purified Atg8p G116 , Atg7p, Atg3p, and PE-containing liposome was established. Conjugation could proceed quite efficiently in multiple catalytic cycles in the presence of ATP, indicating that Atg8p, E1, E2 and liposomes are sufficient for chemical reactions of conjugate formation. PE-deficient liposome could not accept Atg8p, but exogenous supplementation with PE revealed an optimum PE content for the reaction at ϳ70% PE.
The most critical unknown remaining in this process is the site of Atg8p lipidation in yeast cells during autophagosome formation. We showed that lipid bilayer structure is indispensable for Atg8p lipidation. The most simple explanation from the above results is a membrane with high PE content is preferable for the Atg8p lipidation. Generally, yeast organelle membranes contain ϳ15-30% PE (29). However, phospholipids in biomembranes are distributed asymmetrically throughout lipid bilayers; PE is primarily exposed on the cytoplasmic surface of organelle membranes (30). Therefore, small regions with high PE content may not be unrealistic within yeast cells. Negatively charged phospholipids also preferentially modulate the efficiency of the lipidation reaction, suggesting that the site of Atg8p-PE modification in vivo may be membrane microdomains containing substantial PE and additional negatively charged lipids.
While we do not know which proteins recognize phospholip-2 T. Hanada and Y. Ohsumi, unpublished results.  4 and 6 -8). B, Atg8p-PE exhibits a high affinity for the N-terminal peptide antibody. Atg8p (95 ng) and Atg8p-PE (5 ng), derived from in vitro reaction samples, were mixed and subjected to urea-SDS-PAGE. Subsequently, Western blot was performed on these samples using antibodies against either the N-terminal peptide (N) or the whole (W) Atg8 protein. C, dot blot assay. The in vitro products (shown in panel a) were suspended in dilution buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.2 mM DTT, 2% glycerol, 10 mM CHAPS, and 0.5 mg/ml bovine serum albumin (Sigma 7030)) and spotted onto polyvinylidene difluoride membranes as indicated (dilution series of ϫ10-ϫ5,120 includes 27.6 -0.05 g of Atg8p (or Atg8p-PE)). The signals for Atg8p and Atg8p-PE in dot blot assays were detected according to the immunoblot procedures detailed above with antibodies against either the N-terminal peptide (P) or the whole protein (W) of Atg8p. The resulting signal intensities were evaluated by densitometry using NIH Image. D, immunoprecipitation with anti-N-peptide antibody. Equal amounts of Atg8p-PE and Atg8p from in vitro reaction were mixed in IP buffer (lane 1) prior to the addition of the anti-N-peptide antibody to the sample (lane 3). The immunoprecipitation was performed as described under "Experimental Procedures," and each sample was analyzed by urea-SDS-PAGE with SyproOrange staining. As a control, a sample lacking added antibody is shown in lane 2. E, trypsin digestion of Atg8p and Atg8p-PE. Mixture of Atg8p (1.5 g) and Atg8p-PE (2.0 g) was incubated at 30°C for the indicated time in the presence of trypsin (0.01 g) (1anes 1-5). Further, Atg8p (1.5 g) (lanes 6, 8, and 10) and Atg8p-PE (2.0 g) (lanes 7, 9, and 11) were treated individually with trypsin (0.005 g) at 30°C for the indicated time (lanes 6 -11). Samples were analyzed by urea-SDS-PAGE followed by CBB staining. Asterisk indicates proteolytic products. ids during Atg8p lipidation, previous studies have demonstrated that Atg7p and Atg3p form an E1/E2 complex in yeast (31). In the in vitro reaction mixture, Atg7p and Atg3p could be recovered from the liposome fraction (data not shown). Thus, the Atg7p/Atg3p complex may interact with the membrane surface, facilitating the transfer of Atg8p from Atg3p to PE. Further studies on lipid composition and its relevance for the membrane binding of the Atg7p/Atg3p complex will aid in our understanding of the site of lipidation.
In vitro lipidation reaction is highly efficient; the final molar ratio of Atg8p-PE to free PE could reach 1:14, when the reaction mixture contains 5.0 M Atg8p G116 and 70 M PE (shown in Fig. 2). Assuming that PE is evenly distributed across the inner and outer leaflets of the liposomes and that all liposomes are unilamellar, one out of every seven PE molecules in the outer surface must be modified by Atg8p. This result suggests that Atg8p molecules heavily cover the liposome surface. In the in vitro reaction, the reaction mixture became increasingly turbid through the formation of large liposome aggregates. The aggregate formation clearly correlated with the extent of Atg8p-PE. Treatment of the reaction mixture with the deconjugating enzyme Atg4p dissociated these aggregates, suggesting that Atg8p-PE causes liposome aggregation.
While we have not yet solved the tertiary structure of Atg8p, the structures of the mammalian Atg8p homologues, GABARAP, GATE-16, and LC3, have been determined. The structure of these Atg8p homologues contain a ubiquitin-fold and additional N-terminal two ␣-helices, which are not found in other members of the ubiquitin superfamily (32)(33)(34). We generated an antibody against N-terminal peptide of Atg8p, which covers the first ␣-helix of Atg8p (20). By Western blotting, we showed that the affinity of the anti-N-peptide antibody for Atg8p-PE was much higher than that for Atg8p. Whereas several factors can affect the efficiency of Western blotting of small protein like ubiquitin-like molecules, the reactivity of an additional anti-Atg8p did not appear to be affected by PE-modification, ruling out potential technical problems, such as transfer efficiency, as the reason for this differential recognition. Instead, it is more likely that the differences of reactivity of anti-N-terminal antibody could be accounted for by differential accessibility of the N-terminal region following PE-conjugation. We confirmed the induction of a conformation change in Atg8p by dot blot assay, immunoprecipitation experiments, and protease sensitivity without denaturing procedures. These results propose that the lipidation of Atg8p induces the conformation change of Atg8p under native conditions. Previously, we reported the presence of Atg8p in autophagic bodies by detection of HA-Atg8p with an anti-HA antibody, concluding that Atg8p-PE is cleaved off after autophagosome formation (20). Use of the anti-N-terminal antibody to detect Atg8p for immuno-EM analysis, however, could detect Atg8pspecific signals on the autophagosome and its intermediate structure, but not within the autophagic bodies. 3 This result may also reflect differences in Atg8p and the lipidated form.
Based on the crystal structure of Atg8p homologues, Atg8p is predicted to place the first ␣-helix (␣1) (Tyr 4 -Glu 8 ) in the N-terminal helical domain in the face of the central ␤ sheet (␤1) (Arg 28 -Lys 35 ) by hydrogen bonding between the backbone NH groups at the N terminus of helix ␣1 and Glu 34 , conserved throughout the Atg8p family. The crystal structure of GABARAP is proposed to adopt two distinct conformations, with the N-terminal ␣1 region in either a closed or open state, corresponding to the monomer and oligomer, respectively (35). This transition from the closed to the open conformation has been observed in high salt conditions. For Atg8p, a similar conformational change may be induced by Atg8p lipidation. As in the open form of GABARAP, movement of the N-terminal region of Atg8p-PE would expose the hydrophobic surface of the central ␤ sheets of ␤1 and ␤2. This change to an open structure of Atg8p-PE may facilitate the homodimeric interaction of conjugates with each other, may causing the membrane aggregation discussed above. Alternatively, the conformational change may induce interaction of the open form of Atg8p-PE with other molecules. Legesse-Miller et al. (36) has reported that Atg8p may be involved in fusion process mediated by SNARE (soluble N-ethylmaleimide sensitive factor attachment protein receptor) molecule. Thus, it is possible that Atg8p-PE participates in membrane fusion event by interacting with certain protein.
Another question that remains to be answered is the relationship between the Atg12p and Atg8p conjugation systems. We have previously reported only minimal formation of the Atg8p-PE conjugate in the absence of Atg12p-Atg5p (37). Preliminary experiments demonstrated an enhancement of the in vitro lipidation reaction by the presence of Atg12p-Atg5p conjugate. In vitro analysis of this interaction will give us critical information on the relationship between Atg8p lipidation and Atg12p-Atg5p conjugation systems and their roles in autophagy.