The C-terminal region of an Apg7p/Cvt2p is required for homodimerization and is essential for its E1 activity and E1-E2 complex formation.

Apg7p/Cvt2p, a protein-activating enzyme, is essential for both the Apg12p-Apg5p conjugation system and the Apg8p membrane targeting in autophagy and cytoplasm-to-vacuole targeting in the yeast Saccharomyces cerevisiae. Similar to the ubiquitin-conjugating system, both Apg12p and Apg8p are activated by Apg7p, an E1-like enzyme. Apg12p is then transferred to Apg10p, an E2-like enzyme, and conjugated with Apg5p, whereas Apg8p is transferred to Apg3p, another E2-like enzyme, followed by conjugation with phosphatidylethanolamine. Evidence is presented here that Apg7p forms a homodimer with two active-site cysteine residues via the C-terminal region. The dimerization of Apg7p is independent of the other Apg proteins and facilitated by overexpressed Apg12p. The C-terminal 123 amino acids of Apg7p (residues 508 to 630 out of 630 amino acids) are sufficient for its dimerization, where there is neither an ATP binding domain nor an active-site cysteine essential for its E1 activity. The deletion of its carboxyl 40 amino acids (residues 591-630 out of 630 amino acids) results in several defects of not only Apg7p dimerization but also interactions with two substrates, Apg12p and Apg8p and Apg12p-Apg5p conjugation, whereas the mutant Apg7p contains both an ATP binding domain and an active-site cysteine. Furthermore, the carboxyl 40 amino acids of Apg7p are also essential for the interaction of Apg7p with Apg3p to form the E1-E2 complex for Apg8p. These results suggest that Apg7p forms a homodimer via the C-terminal region and that the C-terminal region is essential for both the activity of the E1 enzyme for Apg12p and Apg8p as well as the formation of an E1-E2 complex for Apg8p.

Autophagy is responsible for the bulk of intracellular protein degradation in the lytic organelles, lysosome/vacuole (1,2). When cells exist under conditions of nutrient starvation, the cytoplasmic components are nonselectively sequestered into autophagosomes, double-membrane structures, and are subse-quently targeted to the lysosome/vacuole for degradation. The entire process is conserved through eukaryotes from yeast to mammals. Unique membrane dynamics are observed in the process of autophagy. In the case of the yeast, Saccharomyces cerevisiae, cytoplasmic components are nonselectively surrounded by membranes, which on expansion and completion, give rise to an autophagosome (3,4). Autophagosomal membranes are morphologically distinct from any other known organellar membranes (3). The outer membrane of the autophagosome fuses with the vacuolar membrane (5). The inner membrane structure, which is referred to as an autophagic body, is released into the lumen (3,4). Finally, the cytoplasmic components within an autophagic body are degraded in the vacuole (3). In the case of dynamic autophagosomal membraneformation and fusion with the vacuole (lysosome in mammals), the molecular mechanism for this process remains unknown.
Several autophagy-defective (apg and aut) mutants have been isolated via the application of yeast genetics (6, 7). These apg and aut mutants genetically overlap with most cvt mutants, which have defects in the cytoplasm-to-vacuole targeting (Cvt) pathway of aminopeptidase I (Refs. 8 -10; for a review, see Ref. 11), indicating that the mechanism for autophagy and the Cvt pathway share some common features. Recently, some of the characteristics of individual APG gene products have been elucidated. APG1/AUT3 encodes a protein kinase (12,13). Apg13p is phosphorylated and interacts with Apg1p and Vac8p (14,15). Apg6p/Vps30p forms a complex with Apg14p and is localized on as-yet unidentified membrane structures (16). Aut9p/Apg9p is an integral membrane protein that is required for both the Cvt and autophagic pathways and is localized on large perivacuolar punctate structures (17,18).
Of the APG gene products characterized thus far, the Apg12p modification system and the Apg8p/Aut7p 1 membrane-targeting system have been the subjects of considerable attention in that they function as protein modifiers similar to ubiquitin (for reviews, see Refs. 19 -22). In the case of autophagy, Apg12p binds covalently to Apg5p (23,24). In this conjugation system, Apg7p and Apg10p function as E1 and E2 enzymes, respectively (24 -26). After Apg12p-Apg5p conjugation, Apg16p is assembled with the conjugate, resulting in a high molecular weight Apg12p⅐Apg5p⅐Apg16p complex (27), which is essential for the subsequent formation of autophagosomes. A second modifier, Apg8p, is processed by a novel cysteine protease, Apg4p/Aut2p, and Apg7p and Apg3p/Aut1p are essential for Apg8p lipidation (conjugation with phosphatidylethanolamine), suggesting that Apg7p and Apg3p are, respectively, E1 and E2 enzymes for Apg8p (28 -30). The expression of Apg8p is enhanced by starvation, and Apg8p is associated with autophagosomal membranes under conditions of starvation and with Cvt vesicles under conditions of active growth (31,32). Apg8p also interacts with two ER-to-Golgi vesicular-soluble N-ethylmaleimide factor attachment protein receptors (v-SNAREs), Bet1p and Sec22p, in addition to a vacuolar t-SNARE (Vam3p) and a v-SNARE (Nyv1p) (33).
Considering the fact that the function of Apg7p as a unique E1 enzyme for two substrates, Apg7p promises to be a key enzyme for the functional divergence or correlation between the Apg12p and Apg8p pathway. Furthermore, in view of the functional relationship between the Apg12p, Apg7p, and Apg12p-Apg5p conjugate, it would be of interest to examine the differential intracellular distribution of Apg7p, Apg12p, Apg8p, and Apg5p (24,31). Apg7p is mainly present in the cytoplasm (26,34). Apg12p is distributed in the cytoplasm as well as on a membrane compartment, whereas Apg5p, Apg8p, and Apg12p-Apg5p are localized only on the membrane compartment(s). These findings suggest that a mechanism for the targeting of Apg12p to the Apg5p-localized membrane during the Apg12p-Apg5p conjugation reaction might well be operative. The present study shows that Apg7p forms a homodimer via the Cterminal region with two active-site cysteines and that the C-terminal region is essential not only for interaction with Apg12p and Apg8p, an activity of the E1 enzyme, but also for interaction with Apg3p in the formation of an E1-E2 complex.

EXPERIMENTAL PROCEDURES
Strains, Media, Materials, and Molecular Biological Techniques-Escherichia coli strain DH5␣, the host for plasmids and protein expression, was grown in Luria Broth medium in the presence of the required antibiotics (35). The S. cerevisiae apg7⌬ mutants and PJ69-4A used in this study are listed in Table I. The apg mutant strains have been described previously (6). All yeast strains were cultured in a rich medium (YPD: 1% yeast extract, 2% polypeptone, 2% glucose, 20 mg/liter adenine, 20 mg/liter tryptophan, 20 mg/liter uracil, and 50 mM succinate/NaOH, pH 5.0), MVD medium (0.67% yeast nitrogen base without amino acids, 0.5% casamino acids, and 2% glucose), or SD medium (0.67% yeast nitrogen base without amino acids, 2% glucose, and appropriate amino acids) as described by Kaiser et al. (36). The nitrogen starvation medium contained 0.17% yeast nitrogen base without amino acids, ammonium sulfate, and 2% glucose. For the galactose-inducible expression of proteins in the yeast, SSG medium (0.67% yeast nitrogen base without amino acids, 0.2% sucrose, and 2% galactose) was employed. The solid medium contained 2% Bacto agar. Standard genetic and molecular biological techniques were performed as described by Kaiser et al. (36) and Ausubel et al. (35). The polymerase chain reaction was performed with a program temperature control system PC-701 (ASTEC, Fukuoka, Japan). The DNA sequence was determined using an ABI 373A DNA sequencer (PE Applied Biosystems, Foster City, CA). Restriction enzymes were purchased from TOYOBO (Osaka, Japan) and New England Biolabs (Beverly, MA). Oligonucleotides were synthesized by ESPEC oligo service (Ibaraki, Japan). Two-hybrid screening was performed using the system described by James et al. (37). The expression of a protein under the control of a galactose-inducible promoter was performed after the manufacturer's recommended protocol (Stratagene, La Jolla, CA). pRS series vectors were generous gifts from P. Hieter, and the pGAD-C1 vector, pGBD-C1 vector, and PJ69-4A strain were generous gifts from P. James (37,38). pGem-T vector was purchased from Promega (Madison, WI). A series of pESC vectors was purchased from Stratagene.
Plasmid Construction-Plasmids used in this study are also listed in Table I. For two-hybrid screening, the bait plasmid (pGBD-APG7),  which encodes the Apg7p fused in-frame to the C-terminal end of the GAL4-DNA binding domain, has been described previously (26). pGAD-APG7⌬N containing APG7, which encodes the C-terminal half (amino acids 262-630 out of 630 amino acids), was isolated in the two-hybrid screening. Parts of the APG7-encoding deletion of the C-and N-terminal regions were amplified by polymerase chain reaction and inserted into the BamHI-SalI sites of pGAD-C1. pGAD-APG7 contains the same insert sequences as pGBD-APG7. pGAD-APG12 has been described previously (26).
To express HA-Apg7p and Myc-Apg7p, the DNA sequence that encodes a repeated hemagglutinin (HA) epitope tag and a triplet-repeated Myc epitope tag was inserted just before the start codon of the APG7 gene, and the resulting DNA fragments were cloned into the pGem-T vector. To construct galactose-inducible expression plasmids for these epitope-tagged Apg7ps, the DNA fragments were introduced into the BamHI and SalI sites of pESC-LEU and pESC-URA vectors (Stratagene), resulting in pESCL-HAAPG7, pESCL-MycAPG7, pESCU-HAAPG7, and pESCU-MycAPG7, respectively. For the deletion of the C-terminal region of Apg7p, polymerase chain reaction was performed with appropriate primers, and the resulting product was inserted into the BamHI-SalI sites of pESC-LEU and pESC-URA vectors (Stratagene) to generate pESCL-HAAPG7⌬C40, pESCL-MycAPG7⌬C40, pESCU-HAAPG7⌬C40, and pESCU-MycAPG7⌬C40, respectively.
The APG3 and APG8 open reading frames in which the stop codon was excluded were amplified by polymerase chain reaction using a yeast genomic library and appropriate oligonucleotides that incorporated a BamHI site on the 5Ј-primer and a SalI site on the 3Ј-primer (5Ј-CGGATCCATTATCATGATTAGATCTAC-3Ј, 5Ј-AGTCGACCCAAC-CTTCCATGGTATAGT-3Ј for APG3, 5Ј-AGGATCCAGAGACATGAAG-TCTACATT-3Ј, 5Ј-AGTCGACCCTGCCAA ATGTATTTTCTC-3Ј for APG8), and the amplified DNA fragments were cloned into pGem-T vector (Promega), resulting in pGEM-APG3 and pGEM-APG8, respectively. The BamHI-SalI fragments of pGEM-APG3 and pGEM-APG8 were introduced into the BamHI-SalI sites of pGAD-C1, thus generating pGAD-APG3 and pGAD-APG8, respectively. For the expression of Myc-Apg3p, the BamHI-SalI fragments of pGEM-APG3 were inserted into the BamHI-SalI f site of pESC-URA vector (Stratagene), resulting in pESCU-APG3Myc. For the expression of the N-terminal FLAG-Apg8p, the DNA sequence, which encodes for a FLAG epitope tag, was inserted just before the start codon of the APG8 gene, and the resulting sequence was introduced into the BamHI and SalI sites of pESC-URA vector (Stratagene), resulting in pESCU-FLAGAPG8.
Two-hybrid Experiment-The improved two-hybrid system was performed as described previously (36). A host strain, PJ69-4A (MAT a trp1 leu2 ura3 his 3 gal4⌬ gal80⌬ GAL2-ADE2 LYS2::GAL1-HIS3 met2::GAL7-lacZ), has been created that contains three easily assayed reporter genes, each under the control of a different inducible promoter. The ADE2 gene is under the control of the GAL2 promoter, and HIS3 gene is under the control of the GAL1 promoter. As a result, this strain is extremely sensitive to weak interactions and eliminates nearly all false positives using simple plate assays, i.e. a strong interaction of the examined gene proteins suppresses both Ade Ϫ and His Ϫ auxotrophy of PJ69-4A, whereas a weak interaction suppresses only His Ϫ auxotrophy.
Chemical Cross-linking-Yeast cells that express Myc-Apg7p were harvested and converted to spheroplasts in 1 M sorbitol with 10 mg/ml zymolyase 100T. The spheroplasts were lysed with ice-cold PNE buffer (10 mM potassium phosphate, pH 7.4, 1% Nonidet P-40, 150 mM NaCl, 1 mM EDTA, and protease inhibitors) and sonicated for a short time. The lysate was then centrifuged at 10,000 ϫ g for 5 min at 4°C to remove debris. The supernatant was incubated with 5 mM disuccinimidyl suberate for 30 min at 30°C. Free reactive groups of disuccinimidyl suberate were quenched by incubation with 50 mM Tris-Cl, pH 7.5, for 15 min at 30°C.
Glycerol Gradient Centrifugation-Yeast spheroplasts were prepared in the presence of 1 M sorbitol, lysed with ice-cold PNE buffer, and sonicated for a short time. The lysates were centrifuged at 10,000 ϫ g for 5 min at 4°C to remove cell debris. The supernatant was subjected to glycerol density centrifugation in 10 -40% glycerol in 10 mM potassium phosphate, pH 7.4, 1% Nonidet P-40, and 150 mM NaCl. After centrifugation at 200,000 ϫ g for 18 h at 4°C in a Beckman SW-41 rotor, the gradient was separated into 18 fractions of 600 l. Each fraction was analyzed by immunoblotting with an anti-Myc antibody (9E10) or an anti-HA antibody (16B12). Authentic thyroglobulin (670 kDa, 19 S), catalase (220 kDa, 11.2 S), aldolase (158 kDa, 7.4 S), and bovine serum albumin (67 kDa, 4.3 S) were used as internal S-value standards.
Immunoblotting-Immunoblots were performed using a previously described protocol (26). Briefly, yeast cells harboring the appropriate plasmids were suspended in 100 l of 0.2 N NaOH that contained 0.5% 2-mercaptoethanol. After incubation for 15 min on ice, 1 ml of acetone at Ϫ20°C was added, and the incubation was continued for an additional 30 min at Ϫ20°C. After centrifugation at 10,000 ϫ g for 5 min, the resulting pellets were resuspended in the appropriate volume of SDS sample buffer and boiled for 5 min. Lysates equivalent to 0.5-A 600 cells were separated by SDS-PAGE and transferred to a polyvinylidene difluoride membrane (Millipore). Mouse monoclonal anti-HA antibody (16B12), anti-Myc antibody (9E10), or rabbit anti-API antibody (a gift from Prof. Klionsky) was used for immunodetection. Development was performed by the ECL Plus detection methods (Amersham Pharmacia Biotech).
Immunoprecipitation-The spheroplasts in 1 M sorbitol were lysed by sonication. 800 l of PNE buffer was added to the lysate, and the mixture was centrifuged at 10,000 ϫ g for 5 min at 4°C to remove debris. The supernatant was precleared with 50 l of protein A-agarose (20% slurry, Santa Cruz Biotechnology, Santa Cruz, CA). 50 l of protein A-agarose and 1 g of anti-HA antibody (F7, Santa Cruz Biotechnology, Santa Cruz, CA) was then added to the lysate, and the mixture was rotated for 12 h at 4°C. The immunoprecipitate-bead complex was washed five times with ice-cold PNE buffer. The complex was then boiled for 5 min in SDS sample buffer in the presence of ␤-mercaptoethanol to elute proteins and centrifuged at 10,000 ϫ g for 5 min at 4°C. The supernatant was subjected to SDS-PAGE, transferred to a polyvinylidene difluoride membrane, and analyzed by immunoblotting with anti-HA (F7), anti-Myc (9E10), or anti-FLAG antibody (M2, Sigma).

Apg12p-activating Enzyme, Apg7p, Forms a Homodimer-
Considering the divergence in the localization of Apg7p, Apg12p, and the Apg12p-Apg5p conjugate, the possibility of targeting machinery for Apg12p to an Apg5p-associated membrane becomes an important issue. Previous studies indicated that Apg7p interacts more potently with Apg12p than with Apg10p (25,26,39). Therefore, it is possible that Apg7p-interacting protein(s) functions as targeting machinery in an Apg12pdependent manner. To investigate the Apg12p-dependent Apg7p-interacting protein, a yeast two-hybrid screening was carried out using a tester strain that overexpressed HA-Apg12p on a 2-m plasmid with Apg7p as the bait. Of 1 ϫ 10 6 independent clones, two positive candidates were isolated. DNA-sequencing analysis indicated that both inserts encode the C-terminal region of Apg7p (residues 262-630 out of 630 amino acids, Apg7p⌬N261) ( Fig. 1, APG7⌬N261, PJ69-4A [ APG12 2 ]). An interaction between Apg7p⌬N261 and Apg7p was observed without Apg12p being expressed ( Fig. 1, APG7⌬N261, PJ69-4A), although overexpressed Apg12p significantly enhanced the interaction. Similarly, the interaction of wild-type Apg7p with itself was also observed in the absence of overexpressed Apg12p (Fig. 1, APG7 wild type).
To confirm the interaction of Apg7p homooligomer, the coimmunoprecipitation with Myc-and HA-tagged Apg7p proteins was employed. Myc-Apg7p and HA-Apg7p were coexpressed in the apg7⌬ mutant under the control of a galactose-inducible promoter, which suppresses both Apg Ϫ and Cvt Ϫ phenotypes of the apg7⌬ mutant (data not shown). Cell lysates of the transformant were prepared, and HA-Apg7p was immunoprecipitated with an anti-HA antibody. When galactose was added to the medium, HA-Apg7p was immunoprecipitated with the anti-HA antibody ( Fig. 2A, lane 4). At the same time, Myc-Apg7p was also coimmunoprecipitated ( Fig. 2A, lanes 2). Essentially the same results were obtained by coimmunoprecipitation with an anti-Myc antibody (data not shown).
We next employed a cross-linking experiment using a chemical cross-linker, disuccinimidyl suberate. The lysate of the apg7⌬ cells, which express Myc-Apg7p, was prepared in the presence or absence of disuccinimidyl suberate and separated by SDS-PAGE, and the Myc-Apg7p was subsequently identified by immunoblot with an anti-Myc antibody. In the absence of the cross-linker, a 78-kDa band was observed that corre-Apg7p Dimerization and E1-E2 Complex for Apg8p sponds to monomeric Apg7p. When the cell lysate was treated with the cross-linker, the intensity of the monomer band decreased significantly, and a higher molecular mass band (about 160 kDa), the size of which corresponded to a dimer, appeared (Fig. 2B).
Glycerol density gradient ultracentrifugation also indicates that Apg7p forms a homodimer. Total cell lysates of the apg7⌬ cells overexpressing Myc-Apg7p were subjected to a 10 -40% glycerol density gradient centrifugation. The resulting fractions were analyzed by SDS-PAGE, and the Myc-Apg7p was subsequently identified by immunoblotting with an anti-Myc antibody. Myc-Apg7p was collected in fractions 12-15 and mainly sedimented with a sedimentation coefficient of ϳ7.4 S (Fig. 2C, fraction 13). The overexpression of Apg12p along with the Apg7p resulted in a higher concentration of Apg7p in fraction 13 (data not shown).
The Apg7p-Apg7p interaction was first found in the presence of excess Apg12p. Thus, it is possible that an endogenous Apg12p mediates the interaction of Apg7p. To investigate this possibility, we examined the cross-linking experiment of Myc-Apg7p in the apg12⌬ mutant. Even in the apg12⌬ mutant, a higher molecular mass band corresponding to Apg7p homodimer (about 160 kDa) appeared in the presence of a chemical cross-linker as in wild-type (Fig. 2D). The dimerization of Apg7p was further investigated in other apg mutants. As in wild-type and the apg12⌬ cells, a higher molecular weight band of Apg7p (about 160 kDa) appeared depending on the chemical cross-linker in other apg mutants (apg1c apg2, apg3, apg4,   FIG. 1. Two hybrid analysis of self interaction of Apg7p. The domain structure of wild-type Apg7p is schematically represented on the top. The black box shows ATP binding domain (residues 331-336 out of 630 amino acids), and Cys-507 is the active-site cysteine. In a two-hybrid assay, a tester strain, PJ69-4A, contains two reporter genes under the control of a different inducible promoter, and its growth phenotype is shown according to the auxotrophy: ϩϩϩ, cells grew well on the SD-Ade and SD-His plates (colony size was about 1.5 mm after incubation at 30°C for 3 days on an SD-Ade plate); ϩϩ, cells grew on SD-Ade and SD-His plate (colony-size was about 0.5 mm after incubation at 30°C for 3 days on an SD-Ade plate and about 1.5 mm after incubation at 30°C for 3 days on an SD-His plate); ϩ, cells grew on SD-His plate (colony size is about 0.5 mm after incubation at 30°C for 7 days on SD-His plate); Ϫ, cells did not grow on SD-His plate after incubation at 30°C for 7 days. As host strains, a tester PJ69- apg5, apg6, apg8, apg9, apg10, apg13, and apg14 mutants) (representative data are presented in Fig. 2D, and the data on the other mutants are not shown). These results indicate that Apg7p forms a homodimer with itself without the participation of other APG gene products.
The C-terminal Region of Apg7p Is Essential for Apg7p Dimerization-The issue arises as to the nature of the essential domain for the formation of Apg7p-homodimer. Motif analysis of the amino acid sequence of Apg7p showed that no potential dimerization motifs such as a coiled-coil or a leucine zipper exist on the molecule. To determine the region of Apg7p that is essential for dimer formation, systematic deletion analyses were performed. The original clone isolated (Apg7p⌬N261) lacks N-terminal 261 amino acid residues, suggesting that a region that is proximal to the C terminus may be important for dimerization. Since Apg7p⌬N261 still possesses both an ATPbinding site and an active-site cysteine residue, we first deleted the catalytic domain and examined the resultant construct (Apg7p⌬N507) to determine whether it is capable of binding to full-length Apg7p. Apg7p⌬N507 interacts with full-length Apg7p, indicating that the C-terminal portion (residues 508 -630 out of 630 amino acids), which contains neither an ATP binding domain nor an active-site cysteine, is sufficient for interaction with the full-length Apg7p (Fig. 1, APG7⌬N507).
The location of the essential domain within the C-terminal 123 amino acids (residues 508 -630) is also an open question. ClustalW analysis revealed that a C-terminal region containing 40 amino acids (residues 591-630) has a significant homology with the equivalent region of mammalian Apg7p homologs and that it shows a weak homology with the equivalent region of Uba1p (Fig. 3A) (40). The relevance of this region (C40 region) to Apg7p-Apg7p interaction was thus investigated. In a two-hybrid analysis, the deletion of the C40 region from Apg7p results in the complete loss of its ability to interact with fulllength Apg7p (Fig. 1, APG7⌬C40), suggesting that the C40 region is essential for Apg7p dimerization. An attempt was made to determine whether the C40 region is sufficient for interaction with full-length Apg7p in a two-hybrid system. However, it was found that cells expressing only GBD-C40 were able to grow in selection plates. Therefore, further analyses on the C40 region itself were not pursued, and efforts were concentrated on Apg7p⌬C40. A coimmunoprecipitation assay  (24,39,40,42). B, a mutant protein, Myc-Apg7p⌬C40 was expressed at the same level of wild-type Myc-Apg7p. The apg7⌬ cells were transformed with both pESCL-MycAPG7 and pESCU-HAAPG7 (lane 2) or both pESCL-MycAPG7⌬C40 and pESCU-HAAPG7 (lane 4), respectively. After galactose induction, total lysates from the transformants were prepared and analyzed by immunoblotting (IB) using an anti-Myc antibody (9E10). C, total lysates of the apg7⌬ cells, which express a set of wild-type Myc-Apg7p and wild-type HA-Apg7p (lanes 1-4) or a set of mutant Myc-Apg7p⌬C40 and wild-type HA-Apg7p (lanes 5-8) were prepared, and HA-protein was immunoprecipitated with an anti-HA antibody (IP: Anti-HA). The resulting sediment was analyzed by SDS-PAGE, and epitope-tagged proteins were detected by immunoblot using anti-HA or anti-Myc antibodies (IB, Anti-Myc or Anti-HA).
confirmed the loss of interaction of Apg7p with Apg7p⌬C40. Both wild-type HA-Apg7p and Myc-Apg7p⌬C40 were coexpressed in the apg7⌬ cells. The cell lysate was immunoprecipitated with the anti-HA antibody, and Myc-Apg7p and Myc-Apg7p⌬C40 were examined by immunoblot with the anti-Myc antibody. Myc-Apg7p⌬C40 was expressed at a level similar to that of wild-type Myc-Apg7p in a galactose-dependent manner (Fig. 3B, lanes 4 and 2). Wild-type Myc-Apg7p was coimmunoprecipitated with HA-Apg7p (Fig. 3C, lane 2). In contrast, the mutant Apg7p⌬C40, was not coimmunoprecipitated with HA-Apg7p (Fig. 3C, lane 6). These results indicate that the C40 region of Apg7p is required for the formation of the Apg7p homodimer.
The C-terminal Region Is Also Essential for Interactions of Apg7p with Two Substrates, Apg12p and Apg8p-What is the importance of the Apg7p-dimerization? Apg7p⌬C40 still contains both an ATP binding domain and an active-site cysteine residue but has a defect relative to the formation of Apg7p homodimer. We hypothesized that the dimerization of Apg7p may be somehow correlated with the E1 activity of Apg7p for the Apg12p and Apg8p. It would be interesting to determine whether Apg7p⌬C40 is able to bind to Apg12p and Apg8p. The interaction of Apg7p⌬C40 with Apg12p was first investigated using a two-hybrid analysis. Surprisingly, the interaction of Apg7p⌬C40 with Apg12p was completely abolished compared with the interaction of wild-type Apg7p and Apg12p (Fig. 4A,  Prey APG12). Furthermore, the loss of interaction of Apg7p⌬C40 resulted in a defect in the formation of the Apg12p-Apg5p conjugate. No Apg12p-Apg5p conjugate was present in the apg7⌬ mutant expressing Apg7p⌬C40, whereas the conjugate was present in the mutant expressing the wild-type Apg7p (Fig. 4B, lane 4 and 2). Similarly, a loss of interaction of Apg7p⌬C40 with Apg8p was also detected using a two-hybrid assay (Fig. 4A, Prey APG8). This was further confirmed via a coimmunoprecipitation experiment. A lysate of apg7⌬ cells coexpressing HA-Apg7p⌬C40 and FLAG-Apg8p was prepared, and HA-Apg7p was immunoprecipitated with an anti-HA antibody. FLAG-Apg8p in the sediment was identified by immunoblot with the anti-FLAG antibody. No FLAG-Apg8p was coimmunoprecipitated with HA-Apg7p⌬C40, whereas the FLAG-Apg8p was coimmunoprecipitated with wild-type HA-Apg7p (Fig. 4C, lanes 4 and 2). These results indicate that the C40 region of Apg7p is essential for the interaction of Apg7p with the two substrates.
These defects in the interactions of Apg7p⌬C40 with two substrates will result in pleiotropic defects of the autophagic FIG. 4. Effects of C-terminal deletion of Apg7p on the interaction of Apg7p with two substrates, Apg12p and Apg8p. A, summary of the interactions of deletion mutants of Apg7p with Apg12p and Apg8p by an improved two-hybrid system. Cells expressing both GBD-fused Apg7p variants and GAD-Apg12p (APG12) or GAD-Apg8p (APG8) were assayed for interaction-dependent activation of the ADE2 gene and HIS3 gene as described in Fig. 1. B, the formation of the Apg12p-Apg5p conjugate. The apg7⌬ cells harboring pAPG12HA-426 and either pESCL-MycAPG7 or pESCL-MycAPG7⌬C40 were cultured in the presence (ϩ) or absence (Ϫ) of galactose and lysed as described under "Experimental Procedures." The conjugate was recognized by immunoblot using anti-HA antibody. pESCL-MycAPG7/pAPG12HA-426: the lysate of the apg7⌬ cells, which express both wild-type Myc-Apg7p and HA-Apg12p; pESCL-MycAPG7⌬C40/pAPG12HA-426: the lysate of the apg7⌬ cells expressing both Myc-Apg7p⌬C40 and HA-Apg12p. C, coimmunoprecipitation of Apg8p with Apg7p. Total lysates from the apg7⌬ cells, which express FLAG-Apg8p and either HA-Apg7p or HA-Apg7p⌬C40 in the presence (ϩ) or absence (Ϫ) of galactose were immunoprecipitated (IP) with an anti-HA antibody. Precipitates were analyzed by immunoblotting (IB) using anti-HA or anti-FLAG antibody. Lanes 1 and 2, the apg7⌬ cells expressing FLAG-Apg8p and HA-Apg7p; lanes 3 and 4, the apg7⌬ cells expressing FLAG-Apg8p and HA-Apg7p⌬C40. IgG HC, heavy chain of IgG; IgG LC, light chain of IgG. D-E, effects of the deletion of the C-terminal region of Apg7p on the accumulation of autophagic bodies (D), the viability under nitrogen starvation conditions (E), and cytoplasm-to-vacuole targeting of aminopeptidase I (F). In D, cells grown to early logarithmic phase in MVD ϩ Ura medium were transferred to nitrogen starvation medium in the presence of phenylmethylsulfonyl fluoride and incubated for 8 h at 30°C. Representative Nomarski images of the cells are shown. In E, cells were plated on nitrogen starvation medium containing 10 g/ml phloxine B and incubated at 30°C for 3 days. Inviable cells were stained red (gray in monochrome), whereas viable cells were not stained (white in monochrome). apg7⌬, the apg7⌬ cells carrying control vector, pRS314; wild type, the apg7⌬ cells carrying pAPG7Myc-314; ⌬C40, the apg7⌬ cells carrying pAPG7⌬C40 -314. In F, cell lysates from the apg7⌬ cells harboring pRS314 (lane 1), pAPG7Myc-314 (lane 2), or pAPG7⌬C40 (lane 3) were subjected to immunoblotting analysis with anti-API antiserum. and Cvt pathways. In yeast, autophagy is induced by a variety of starvation conditions, and its progression is easily monitored by means of light microscopy; autophagic bodies accumulate in the vacuoles of wild-type cells under conditions of nitrogen starvation, and the detection of these autophagic bodies is facilitated by phenylmethylsulfonyl fluoride, a protease inhibitor that blocks their degradation (3). In the vacuoles of apg7⌬ cells expressing wild-type Apg7p, a significant accumulation of autophagic bodies was observed, whereas the apg7⌬ cells expressing Apg7p⌬C40 failed to accumulate autophagic bodies in the vacuole, as was the case of the apg7⌬ cells carrying a control vector (Fig. 4D, wild type, ⌬C40, and apg7⌬). A defect in autophagy results in a loss of viability under conditions of starvation (6). The colony color of the apg7⌬ mutant expressing Apg7p⌬C40 turned a pink color under nitrogen starvation conditions, as evidenced by phloxine B-staining, similar to the apg7⌬ mutant, whereas that of the apg7⌬ mutant expressing wild-type Apg7p was not stained (Fig. 4E, ⌬C40, apg7⌬, and wild type).
We next examined the effect of the C40 region of Apg7p on the Cvt pathway. Aminopeptidase I (API) is synthesized as the pro-form (proAPI) in the cytoplasm, transferred to the vacuole by a mechanism that is closely related to the autophagic pathway, and then processed to the mature form in the vacuole. In the case of the apg7⌬ cells expressing wild-type Apg7p, proAPI was processed to the mature form (Fig. 4F, lane 2). In contrast, in the case of apg7⌬ cells expressing mutant Apg7p⌬C40, the mature form of API was not detected, and proAPI accumulated, similar to that of the apg7⌬ cells (Fig. 4F, lane 3 and 1). Thus, the loss of function of Apg7p as the result of the deletion of the C40 region of Apg7p results in a defect in the Cvt pathway. These results indicated that the C40 region of Apg7p is essential for its E1 function in both the autophagic and Cvt pathways.
The C-terminal Region of Apg7p Is Necessary for the Formation of E1-E2 Complex for Apg8p-Recently, a comprehensive analysis of protein-protein interactions in the yeast S. cerevisiae by two-hybrid screening has indicated that Apg7p interacts not only with Apg12p and Apg8p but also with Apg3p ( Fig.  5A) (39). Recent findings revealed that Apg3p is an E2 enzyme for Apg8p. Thus far, no report has appeared on the formation of an E1-E2 complex in a protein modification system similar to ubiquitylation. Then we first investigated whether Apg3p is coimmunoprecipitated with Apg7p. Both HA-Apg7p and Apg3p-Myc were coexpressed under the control of a galactoseinducible promoter in the apg7⌬ cells, and the cell lysate was immunoprecipitated with an anti-HA antibody. Apg3p-Myc in the sediment was recognized by immunoblot with an anti-Myc antibody. Apg3p-Myc was coimmunoprecipitated with HA-Apg7p (Fig. 5B, lane 2). The coimmunoprecipitation of Apg3p with Apg7p occurs in the apg12⌬ mutant (data not shown). This result indicates that the Apg7p interacts with Apg3p to form an E1-E2 complex.
Since the C40 region of Apg7p is essential for its interaction with the two substrates, it is probable that this region is essential for the interaction of Apg7p with Apg3p, too. To investigate this possibility further, a coimmunoprecipitation experiment was performed in the apg7⌬ cells expressing both HA-Apg7p⌬C40 and Apg3p-Myc. No Apg3p-Myc was coimmunoprecipitated with HA-Apg7p⌬C40, whereas the Apg3p-Myc was coimmunoprecipitated with wild-type HA-Apg7p (Fig. 5B,  lanes 4 and 2). This finding was confirmed by a two-hybrid analysis (Fig. 5A). Furthermore, as is the case with the interaction of Apg12p and Apg8p, Apg3p does not interact with Apg7p⌬N507 (Fig. 5A). These results indicated that the Cterminal region of Apg7p is essential for the formation of the E1-E2 complex, Apg7p⅐Apg3p.

DISCUSSION
The evidence presented herein indicates that Apg7p is a unique protein-activating enzyme that is capable of forming a homodimer and is essential for the two substrates (Apg12p and Apg8p). These characteristics have not been reported for other E1 enzymes. Furthermore, the Apg7p is able to form a stable E1-E2 complex. The dimerization occurs independently of other APG gene products examined thus far, supporting the possibility that Apg7p interacts with itself without the need for any other factors. The deletion of the C40 region of Apg7p results in the loss of Apg7p dimerization. It is surprising that the C40 region is also essential for the interaction of Apg7p with two substrates, Apg12p and Apg8p, even though the mutant Apg7p⌬C40 monomer still contains an ATP binding domain and an active site cysteine. The C40 region of Apg7p is also essential for the formation of the E1-E2 complex, Apg7p⅐Apg3p. The C-terminal 123 residues (residues 508 -630), which contain the C40 region, are sufficient for interaction with fulllength Apg7p. Combining these data, we conclude that homodimer formation via the C-terminal region is important for enzyme-substrate interaction and the formation of an E1-E2 complex. During starvation-induced autophagy, Apg8p became localized on the forming autophagosomal membranes (31). Our recent findings suggest that both Apg7p and Apg3p function as E1 and E2 enzymes, respectively, which is necessary for Apg8p to target the autophagosomal membranes (29,30). We therefore reason that multimer complexes that are formed during the two enzymatic reactions catalyzed by the Apg7p ho-modimer may play a key role in autophagy. An Apg7p-related scheme is shown in Fig. 6A. First, Apg7p undergoes homodimer formation (Fig. 6A, I). In one route of the next step, the Apg7p homodimer, which forms an enzyme-substrate conjugate with Apg12p via a thiol ester bond (Fig. 6A, IIa), functions as an E1 enzyme, which is essential for subsequent Apg12p-Apg5p conjugation (Fig. 6A, IIIa). Similarly, in the other route (Fig. 6A,   FIG. 6. Hypothetical scheme of the Apg7p homodimer and domain structure of the E1 enzymes are shown. A, the dimerization of Apg7p is essential for further reaction. Considering the fact that the C-terminal region of Apg7p, but neither Apg12p, Apg8p, nor Apg3p, interacts directly with wild-type Apg7p, the formation of Apg7p homodimer is necessary for interactions with substrates and E2 enzyme (I). After Apg7p dimerization, Apg12p(s) is activated by the Apg7p homodimer (IIa). Subsequent transfer of Apg12p to Apg10p results in the formation of the Apg12p-Apg10p intermediate. Finally, Apg12p is covalently attached to Apg5p via an isopeptide bond (IIIa). When the Apg7p homodimer interacts with Apg8p, Apg8p is activated by the Apg7p homodimer (IIb). Apg3p transiently interacts with Apg7p homodimer to effectively form Apg8p-Apg3p intermediate. Finally, Apg8p targets to an autophagosome (IIIb). PE, phosphatidylethanolamine. B, Box V is proposed as the essential domain for the homodimerization of Apg7p and Uba1p. The box V, which contains a cluster of acidic amino acids, is essential for homodimerization of Apg7p and the formation of E1-E2 complex. The C-terminal region of Uba1p has a similarity with the box V within Apg7p, and Uba1p will also then form a homodimer. UBI, ubiquitin. IIIb), Apg7p, which forms an enzyme-substrate conjugate with Apg8p, also functions as an E1 enzyme, which is essential for subsequent Apg8p-phophatidylethanoamine conjugation. The interaction of Apg3p (E2) with Apg7p (E1) may facilitate the effective targeting of Apg8p to autophagosomal membranes as well as to ER-to-Golgi vesicles (Fig. 6A, IIIb) (28 -32). How do these interactions correlate with the functions of these APG gene products? A key to revealing this question would be an interaction of Apg12p with Apg3p, which has been reported by a comprehensive two-hybrid screening (39). This suggests that Apg12p plays an important role in Apg8p targeting on autophagosomal membranes. At present, it is difficult to completely explain the functional correlation between Apg12p modification system and Apg8p-membrane targeting. Further analyses to clarify Apg12p versus Apg3p interaction will be necessary.
The human Apg7p/Cvt2p/Gsa7p homolog has been reported (41). The issue of whether or not the mammalian homolog forms a homodimer is of great interest. Several lines of observations suggest that the mammalian Apg7p homolog also is capable of entering into a homodimer formation. (i) The Apg12p conjugation system is conserved from yeast through mammalian cells (42). (ii) The human Apg7p homolog is an authentic protein-activating enzyme for human Apg12p (43). (iii) The C40 region of S. cerevisiae Apg7p is highly conserved among human Apg7p and mouse Apg7p (Fig. 4A). We are now investigating the possibility of homodimer formation in mammalian Apg7p homologs by means of cross-linking experiments and glycerol density gradient centrifugation.
Several E1 enzymes, Uba1p for ubiquitin, Aos1p-Uba2p for Smt3p, Ula1p-Uba3p for Rub1p, Apg7p for Apg12p, and Uba4p for Urm1p have been characterized in yeast (26, 44 -51). According to Johnson et al. (46) and Liakopoulos et al. (49), four similarity boxes (IϳIV) exist that are preserved in the E1 enzymes. Apg7p contains an ATP binding domain within the box I and an active-site cysteine within the box III, but no domains similar to the box II and the box IV are present within Apg7p (Fig. 6B) (26). We propose the C40 region of Apg7p as box V to be an essential domain for homodimer formation. Box V contains a cluster of acidic amino acids. Interestingly, the C-terminal region of Uba1p has some similarity with the C40 region of Apg7p. In an earlier study, Ciechanover et al. (52) report the purification of a mammalian homolog of Uba1p by "covalent affinity" chromatography and conclude that the purified enzyme is composed of two subunits. It is therefore possible that Uba1p also forms a homodimer via its C-terminal region, box V. The absence of a box V in two other E1 enzymes (Aos1p-Uba2p and Ula1p-Uba3p) suggests that they are functional when the exist as a heterodimer with only one activecenter cysteine. It is therefore important to understand how the divergence in molecular composition of protein-activating enzyme (E1) can be correlated with functional divergence in various cellular activities.