Phosphatidylserine in Addition to Phosphatidylethanolamine Is an in Vitro Target of the Mammalian Atg8 Modifiers, LC3, GABARAP, and GATE-16*

In yeast, phosphatidylethanolamine is a target of the Atg8 modifier in ubiquitylation-like reactions essential for autophagy. Three human Atg8 (hAtg8) homologs, LC3, GABARAP, and GATE-16, have been characterized as modifiers in reactions mediated by hAtg7 (an E1-like enzyme) and hAtg3 (an E2-like enzyme) as in yeast Atg8 lipidation, but their final targets have not been identified. The results of a recent study in which COS7 cells were incubated with [14C]ethanolamine for 48 h suggested that phosphatidylethanolamine is a target of LC3. However, these results were not conclusive because of the long incubation time. To identify the phospholipid targets of Atg8 homologs, we reconstituted conjugation systems for mammalian Atg8 homologs in vitro using purified recombinant Atg proteins and liposomes. Each purified mutant Atg8 homolog with an exposed C-terminal Gly formed an E1-substrate intermediate with hAtg7 via a thioester bond in an ATP-dependent manner and formed an E2-substrate intermediate with hAtg3 via a thioester bond dependent on ATP and hAtg7. A conjugated form of each Atg8 homolog was observed in the presence of hAtg7, hAtg3, ATP, and liposomes. In addition to phosphatidylethanolamine, in vitro conjugation experiments using synthetic phospholipid liposomes showed that phosphatidylserine is also a target of LC3, GABARAP, and GATE-16. In contrast, thin layer chromatography of phospholipids released on hAtg4B-digestion from endogenous LC3-phospholipid conjugate revealed that phosphatidylethanolamine, but not phosphatidylserine, is the predominant target phospholipid of LC3 in vivo. The discrepancy between in vitro and in vivo reactions suggested that there may be selective factor(s) involved in the endogenous LC3 conjugation system.

However, there are differences among the conjugated forms of the three Atg8 homologs. For example, the amount of LC3-II in rat tissues cannot be correlated with the amounts of the other conjugates. In HEK293 cells, conjugation of LC3, but not of GATE-16 or GABARAP, is facilitated by overexpression of hAtg7 and hAtg3. In HeLa cells, LC3-II and GABARAP-PL, but not GATE-16-II, accumulate in the presence of the protease inhibitors, E64d and pepstatin A, under nutrient-rich conditions. In the livers of mice lacking Atg7, the amounts of the unlipidated forms of all three Atg8 homologs increase, suggesting that conjugation of all three molecules is active in the mouse liver (21).
In the present study, we focused on whether hAtg7, hAtg3, and phospholipid(s) are the minimum essential requirements needed for conjugation of the three mammalian Atg8 homologs and on the nature of the targets of these three human Atg8 homologs. We hypothesized that PE and phosphatidylserine (PS) may be targets, because the hydrophilic heads of both PE and PS have amino groups, which are required in ubiquitylation-like reactions. We reconstituted the conjugation systems for the mammalian Atg8 homologs in vitro using purified recombinant hAtg7, hAtg3, LC3, GABARAP, and GATE-16, as well as synthetic phospholipid liposomes. Furthermore, we directly analyzed the in vivo target phospholipid of endogenous LC3-II purified from HeLa cells by TLC.
Expression of GST-tagged Recombinant Human Atg Proteins in E. coli and Their Purification-GST-tagged proteins were expressed in E. coli and purified using glutathione-Sepharose 4B and PreScission protease according to the manufacturer's protocol (Amersham Biosciences).
In Vitro Assay for E1-and E2-Modifier Intermediates via Thioester Bond-Purified recombinant proteins (0.1 M each) were mixed in TN buffer (50 mM Tris-HCl, pH 7.0, 150 mM NaCl), and 5 mM ATP, 10 mM MgCl 2 , and 1 mM DTT were added to the mixture where indicated. The mixture was incubated at 25°C for 1 h, the reaction was stopped by the addition of an equal volume of SDS solution (1% SDS, 50 mM Tris-HCl, pH 6.8, 15% glycerol) in the absence or presence of a reducing reagent, DTT (100 mM), and the mixtures were incubated for 10 min at 37°C (25,26).
Preparation of Total Lipids from HeLa Cells and Formation of Liposomes-Total lipids in HeLa cells were extracted as described (13,22,27) and stored as a chloroform solution. To prepare dried lipid films, the chloroform was evaporated with nitrogen gas, and the samples were placed in bell jars under vacuum at room temperature for 12 h. The resultant dried lipid films were hydrated to a final concentration of 1 mM phospholipids in a buffer consisting of 25 mM Tris-HCl (pH 7.5), 137 mM NaCl, 2.7 mM KCl, vortexed vigorously at room temperature, and sonicated for 5 min at 4°C. After centrifugation at 20,000 ϫ g for 20 min, the supernatant was used as small unilamellar liposomes (13,27,28).
In Vitro Assay for LC3, GABARAP, and GATE-16 Conjugation-Each mutant hAtg8 homolog with exposed carboxyl-terminal Gly residues (0.1 M LC3 TFG , GABARAP VYG , and GATE-16 TFG ), hAtg7 (0.1 M), hAtg3 (0.1 M), and liposomes (30 M) were mixed in TN buffer in the presence of 5 mM ATP, 10 mM MgCl 2 , and 1 mM DTT and incubated at 25°C for 1 h. To stop the reaction, an equal volume of SDS solution containing 5% 2-mercaptoethanol was added, and the mixture was boiled for 5 min.
Identification of Phospholipid Conjugated to LC3-I-Endogenous LC3-II (LC3-phospholipid conjugate) in HeLa cells was accumulated in the presence of the protease inhibitors, E64d and pepstatin A, as described (18). After preparation of an LC3-II-enriched membrane fraction from the inhibitor-treated HeLa cells (about 1 ϫ 10 8 cells/ analysis), LC3-II was solubilized from LC3-II-rich total membrane fraction in TX solution (2% Triton X-100, 20 mM Tris-HCl, pH 7.5, 150 mM NaCl) (18) and purified by affinity chromatography on an anti-LC3 antibody-immobilized Sepharose column. Purified LC3-II was mixed with recombinant hAtg4B, a delipidating protease (17,18) from E. coli, and the mixture was incubated at 25°C for 120 min. Released phospholipids were extracted from reaction mixtures using Bligh-Dyer's method (22) and subsequently separated by TLC on Kieselgel 60 plates (Merck) using chloroform/methanol/deionized water (65:25:4, v/v/v) as a solvent. The TLC plates were dried in a fume hood (5 min) and sprayed with primuline reagent (29) to detect phospholipids under UV light (main wavelength, 365 nm).

In Vitro E1-Substrate Formation between hAtg7 and Each Mutant
Atg8 Homolog, LC3 TFG , GABARAP VYG , and GATE-16 TFG -In HEK293, COS7, and HeLa cells, almost all of the C termini of LC3, GABARAP, and GATE-16 are posttranslationally cleaved to expose a C-terminal Gly residue, which is essential for ubiquitylation-like reactions. Therefore, for in vitro E1-substrate reactions, hAtg7 and a series of mutant Atg8 homologs were expressed as GST fusion proteins in E. coli and purified using glutathione-Sepharose 4B and PreScission protease (Fig.  1). Coomassie Brilliant Blue staining of the resultant proteins on SDS-PAGE showed that all three had been purified almost to homogeneity (Fig. 1C).
We first focused on the formation of the E1-substrate intermediate between hAtg7 and LC3 TFG by mixing hAtg7 (0.1 M) and LC3 TFG (0.1 M) and incubating them at 25°C for 1 h in the presence or absence of ATP (5 mM). After stopping the reaction, proteins were separated by SDS-PAGE (4 -12% linear gradient) under nonreducing conditions, and LC3 TFG was recognized by immunoblotting using anti-LC3 antibody. In the presence of ATP, a band of about 100 kDa, corresponding to the hAtg7-LC3 TFG (E1-substrate) intermediate was recognized (Fig. 2A). Similar results were obtained in the assays for E1-substrate intermediates of GABARAP VYG and GATE-16 TFG (Fig. 2, B and C). Bands of about 100 kDa, corresponding to hAtg7-GABARAP VYG and hAtg7-GATE-16 TFG (E1-substrate) intermediates were recognized by their respective antibodies only in the presence of ATP, in a manner sensitive to the reducing reagent, DTT (100 mM). There was little intermediate formation when GABARAP VY and GATE-16 TF were substituted for GABARAP VYG and GATE-16 TFG , respectively. These results indicated that hAtg7, each of the Atg8 homologs, and ATP are the minimum materials essential for the formation of E1-substrate intermediates via thioester bonds.
In Vitro E2-substrate Formation between hAtg3 and Each Mutant Atg8 Homolog-We next investigated whether each of the Atg8 homologs could form an E2-substrate intermediate with hAtg3 in vitro. Similar to the other proteins above, hAtg3 was expressed as a GST fusion protein in E. coli and purified with glutathione-Sepharose 4B and PreScission protease (Fig. 1C). To form an E2-substrate intermediate between hAtg3 and LC3 TFG , we incubated hAtg7 (0.1 M), hAtg3 (0.1   FEBRUARY 10, 2006 • VOLUME 281 • NUMBER 6 JOURNAL OF BIOLOGICAL CHEMISTRY 3019 M), and LC3 TFG (0.1 M) at 25°C for 1 h in the presence or absence of ATP (5 mM). After stopping the reaction, the proteins were separated on SDS-polyacrylamide gels (4 -12% linear gradient or 12%) under nonreducing conditions, and LC3 TFG was recognized by immunoblotting with anti-LC3 antibody (Fig. 3A). In the presence of ATP, a band of about 60 kDa, corresponding to an hAtg3-LC3 TFG (E2-substrate) intermediate, was observed. This band was not observed in the absence of hAtg7 or ATP and was sensitive to the reducing reagent, DTT (100 mM). When LC3 TF was employed instead of LC3 TFG , little E2-substrate intermediate was detected, indicating that the C-terminal Gly of LC3 is essential for this reaction in vitro.

Three Human Atg8 Homologs Conjugate to PE and PS
Similar results were obtained in the assays for E2-substrate intermediates of GABARAP VYG and GATE-16 TFG (Fig. 3, B and C). When hAtg7 (0.1 M), hAtg3 (0.1 M), ATP (5 mM), and GABARAP VYG or GATE-16 TFG (0.1 M) were mixed, an E2-substrate intermediate between hAtg3 and GABARAP VYG or GATE-16 TFG was detected by immunoblotting with the respective antibodies. These intermediates were sensitive to 100 mM DTT and were dependent on the presence of hAtg7 and ATP. When GABARAP VY and GATE-16 TF were substituted for GABARAP VYG and GATE-16 TFG , respectively, little intermediate was observed, indicating that the C-terminal Gly of GABARAP and GATE-16 is required for this reaction in vitro.
In Vitro Conjugation Reactions of the Three Atg8 Homologs, Mediated by hAtg7 and hAtg3-Yeast Atg8 is conjugated to PE in vitro and in vivo. It is unclear whether hAtg7 and hAtg3 are the minimum enzymes required for the conjugation reactions and whether GATE-16 is conju-  Fig. 2, A-C, respectively. Total proteins were separated by nonreducing SDS-PAGE (4 -12% linear gradient or 12%), and hAtg3 was recognized by immunoblotting with anti-hAtg3 antibody (WB:␣-hAtg3). hAtg7 (WB:␣-hAtg7) and the three Atg8 homologs (WB:␣-LC3, ␣-GABARAP, and ␣-GATE-16) were recognized by immunoblotting, as described in the legend to Fig. 2. gated to a phospholipid(s). To clarify these problems, we investigated whether the conjugated forms of the three Atg8 homologs are generated by hAtg7 and hAtg3 after E1 and E2 reactions in vitro. As a phospholipid source, we used liposomes generated from total lipids isolated from HeLa cells, because LC3-II (an LC3-phospholipid conjugate) and GABARAP-PL (a GABARAP-phospholipid conjugate) are present in HeLa cells under nutrient-rich conditions.
We found that LC3-II was generated when hAtg7 (0.1 M), hAtg3 (0.1 M), liposomes (30 M), and LC3 TFG (0.1 M) were incubated in the presence of ATP (5 mM) at 25°C for 1 h, as determined by immunoblotting with anti-LC3 antibody (Fig. 4A). In the absence of hAtg7, hAtg3, ATP, or liposomes, or when LC3 TF was substituted for LC3 TFG , little conjugate was observed. Similar results were obtained for GABARAP and GATE-16 conjugation reactions (Fig. 4, B and C); i.e. GABARAP-PL was detected by immunoblotting with anti-GABARAP antibody only when all components (hAtg7, hAtg3, ATP, GABARAP VYG , and liposomes) were present in the reaction mixture, whereas little GABARAP-PL was formed when GABARAP VY was substituted for GABARAP VYG (Fig. 4B). GATE-16-II, a conjugated form of GATE-16, was detected by immunoblotting with anti-GATE-16 antibody only when all components (hAtg7, hAtg3, ATP, GATE-16 TFG , and liposomes) were present in the reaction mixture, but little GATE-16-II was formed when GATE-16 TF was substituted for GATE-16 TFG (Fig. 4C). These results indicated that the minimum essential requirements for in vitro conjugation of the hAtg8 homologs include hAtg7, hAtg3, ATP, liposomes, and each hAtg8 homolog.

LC3, GABARAP, and GATE-16 Are Conjugated to both PE and PS in
Vitro-Conjugates of LC3 and GABARAP have been reported to be sensitive to phospholipase D, suggesting that these hAtg8 homologs target a phospholipid(s). Ubiquitin and other modifiers are conjugated to targets via amide bonds between the carboxyl group in their C-terminal Gly residue and the side chain amino group in a Lys residue or in the hydrophilic head of PE. Since PE and PS have amino groups in their hydrophilic heads, these phospholipids are potential targets of the human Atg8 homologs. To determine whether phospholipids are conjugated to the Atg8 homologs, we prepared liposomes composed of pure synthetic phospholipids (POPC, a molecular species of PC; DOPE, a molecular species of PE; and DOPS, a molecular species of PS), and reconstituted the conjugation reactions in vitro (Fig. 5).
We first focused on the effects of synthetic phospholipid-based liposomes on the conjugation of LC3. When liposomes consisting of 100% DOPS were used in the in vitro reconstitution of LC3 conjugation, LC3-II was observed by immunoblotting with anti-LC3 antibody (Fig.  5A, lane 2). In contrast, when liposomes containing 100% DOPE and 100% POPC were employed, no LC3-II was observed (Fig. 5A, lanes 1  and 3). We further investigated the optimum concentration of PS (DOPS) in the liposomes for in vitro conjugation by varying the concentration of DOPS in the liposomes from 0 to 100%. LC3-II was observed when 10 -100% DOPS liposomes were used (Fig. 5A, lanes 8 -15), with the optimum DOPS concentration for in vitro LC3 conjugation being 55-85%. These results indicated that PS is a target of LC3.
We next focused on the effects of synthetic phospholipid-based liposomes on in vitro GABARAP conjugation. Since there are differences in the amounts of the modified forms of LC3, GABARAP, and GATE-16 in rat tissues, there may be divergence among these hAtg8 homologs regarding their target phospholipids or their optimum concentrations. However, we observed little difference between LC3 and GABARAP conjugation reactions in vitro (Fig. 5B). GABARAP-PL was observed using 100% DOPS liposomes (Fig. 5B, lane 2) but not using liposomes composed of 100% DOPE or 100% POPC. Use of a series of DOPE-and DOPS-containing liposomes in the reaction mixture (Fig. 5B) indicated that their optimum concentrations for in vitro GABARAP conjugation were similar to those observed for LC3 conjugation (Fig. 5B, lanes 8 -23). These results indicated that both PS and PE are targets of GABARAP conjugation in vitro.
Previous biochemical studies of GATE-16-II and the results shown in Fig. 4C suggested that PS and PE are potential targets of GATE-16 conjugation. Therefore, we used synthetic phospholipid liposomes to investigate GATE-16 conjugation. When 100% DOPS and a series of DOPE liposomes were employed in the in vitro reaction, GATE-16-II was observed by immunoblotting with anti-GATE-16 antibody (Fig.  5C). The optimum concentrations of DOPS and DOPE liposomes for in vitro GATE-16 conjugation were similar to those observed for LC3 and GABARAP (Fig. 5C, lanes 8 -23). These findings indicated that GATE-16-II is a protein-phospholipid conjugate and that the targets of GATE-16 are also PS and PE. In conclusion, both PS and PE are targets of all three Atg8 homologs (LC3, GABARAP, and GATE-16) in these protein-phospholipid conjugation reactions.
The in Vivo Target of Endogenous LC3 Is Phosphatidylethanolamine but Not Phosphatidylserine-In the in vitro conjugation reaction, LC3 could conjugate with both PS and PE. However, it is not known whether this is also the case in vivo. The results of a recent study involving the incubation of [ 14 C]ethanolamine in COS7 cells for 48 h suggested that  Fig. 4, A-C, respectively. LC3-I and LC3-II were recognized by immunoblotting with anti-LC3 antibody; GABARAP and GABARAP-PL were recognized by immunoblotting with anti-GABARAP antibody; and GATE-16-I and GATE-16-II were recognized by immunoblotting with anti-GATE-16 antibody.  In the in vitro reaction (upper panel), hAtg3 directly recognizes PS and PE, and LC3 is conjugated to PE and PS via amide bonds between the C-terminal Gly in LC3 and an amino group of each phospholipid. Since endogenous LC3 is preferentially conjugated to PE, there will be a mechanism(s) to select a target phospholipid after transferring to hAtg3 in vivo (lower panel). One hypothetical factor, X, is an E3-like enzyme to select PE as a target for LC3. Another hypothetical factor, Y, is a regulatory factor that inhibits the association of LC3 (or hAtg3) with PS. FEBRUARY 10, 2006 • VOLUME 281 • NUMBER 6 phosphatidylethanolamine is a target of LC3 (17). However, the experiment could not exclude the possibility that phosphatidylserine may be another endogenous target of LC3 in the ATG conjugation reaction, as in the case of in vitro conjugation.

Three Human Atg8 Homologs Conjugate to PE and PS
To identify the endogenous target phospholipid(s) of LC3 in vivo, we first purified endogenous LC3-II from HeLa cells in the presence of lysosomal protease inhibitors, E64d (10 g/ml) and pepstatin A (10 g/ml), as described (18), delipidated LC3-II with the delipidating enzyme, hAtg4B, and analyzed the phospholipids released on reaction with purified endogenous LC3-II. Endogenous LC3-II was solubilized from the total membrane fraction with TX solution and affinity-purified on an anti-LC3 antibody-immobilized Sepharose column. Purified LC3-II was incubated with either wild-type hAtg4B or inactive mutant hAtg4B C74A . Silver staining and immunoblotting analysis clearly showed that wild-type hAtg4B completely delipidated LC3-II to form LC3-I (Fig. 6, B and C). In contrast, incubation of the mutant hAtg4B C74A with LC3-II did not cause delipidation (Fig. 6, B and C). The phospholipids were extracted from the reaction products using the Bligh-Dyer method (22) and further analyzed by TLC. As clearly shown in Fig. 6D, treatment of LC3-II with wild-type hAtg4B released PE but not PS. No phospholipid spots were detected with the sample treated with the inactive mutant, hAtg4B C74A . These results indicated that PE, but not PS, is the target phospholipid conjugated with endogenous LC3 in vivo.

DISCUSSION
The results of the present study indicated that in vitro conjugation systems for the three Atg8 homologs, LC3, GABARAP, and GATE-16, can be reconstituted using purified hAtg7 (E1-like enzyme), hAtg3 (E2like enzyme), and synthetic phospholipid liposomes. All three hAtg7-Atg8 homolog intermediates (E1-substrate intermediates) formed via ATP-dependent thioester bonds, and all three hAtg3-Atg8 homolog intermediates (E2-substrate intermediates) formed via thioester bonds dependent on ATP and hAtg7. Finally, all three Atg8 homologs were conjugated to phospholipids, PS and PE. The findings presented here lead to three conclusions. First, the minimum components necessary for human Atg8-phospholipid conjugations were shown to be hAtg7, hAtg3, the respective Atg8 homolog (LC3, GABARAP, or GATE-16), phospholipid (DOPE or DOPS)-containing liposomes, and ATP. Second, the modified form of GATE-16, GATE-16-II, is a protein-phospholipid conjugate and therefore should be designated GATE-16-PL instead of GATE-16-II. Finally, PS and PE are targets of all human Atg8 modifiers in vitro.
We have further demonstrated that the in vivo target phospholipid of endogenous LC3-II is predominantly PE, as revealed by TLC analysis. This is the first direct evidence for the identity of the target phospholipid of the LC3 conjugation system in vivo.
These observations raise questions regarding the apparent discrepancy between in vitro and in vivo data. Our in vitro conjugation system requires only E1-like enzyme (Atg7) and E2-like enzyme (Atg3) for conjugation of the three Atg8 homologs with PE or PS as in yeast in vitro conjugation of Atg8 with PE (13). However, this does not exclude the possible involvement of some other as yet unidentified factors (Fig. 7) in promoting selective conjugation of LC3-I with PE in vivo. Both positive regulatory factor (X) and negative regulatory factor (Y) should be considered. One potential candidate for X is an E3-like enzyme, which specifically recognizes PE as the substrate of conjugation reaction of activated LC3-I on Atg3 (E2-like enzyme), making efficient and preferential transfer of LC3-I to PE possible.
We hypothesize a presumptive negative regulatory factor (Y), which inhibits LC3 conjugation to PS (Fig. 7). In our in vitro reconstitution experiments, all three Atg8 homologs (LC3, GABARAP, and GATE-16) could be conjugated with both PE and PS. However, cellular expression levels of the three homologs are quite different in native tissues and cultured cells (30). Therefore, the levels of the lipidated forms of the homologs must also be diverse. Hence, the preferential homologs to be lipidated may be different, depending on tissues and cells, and target phospholipid(s) may be also different depending on which of the three homologs is involved in the reaction. Thus, participation of negative regulatory factor (Y) is also a likely mechanism of in vivo lipidation of Atg8 homologs. Experiments are currently under way in our laboratory to screen for possible candidates of X and Y, since it is important to identify and characterize such regulatory factor(s) to improve our understanding of the mechanism of mammalian autophagy.