Apg10p, a novel protein‐conjugating enzyme essential for autophagy in yeast

Takahiro Shintani, Noboru Mizushima, Yoko Ogawa, Akira Matsuura, Takeshi Noda, Yoshinori Ohsumi

Author Affiliations

  1. Takahiro Shintani1,
  2. Noboru Mizushima1,
  3. Yoko Ogawa2,
  4. Akira Matsuura3,
  5. Takeshi Noda1 and
  6. Yoshinori Ohsumi*,1
  1. 1 Department of Cell Biology, National Institute for Basic Biology, Okazaki, 444‐8585, Japan
  2. 2 Present address: Department of Neurobiology, Graduate School of Medicine, The University of Tokyo, Tokyo, 113‐0033, Japan
  3. 3 Present address: Department of Life Science, Tokyo Institute of Technology, Yokohama, 226‐8501, Japan
  1. *Corresponding author. E-mail: yohsumi{at}


Autophagy is a cellular process for bulk degradation of cytoplasmic components. The attachment of Apg12p, a modifier with no significant similarity to ubiquitin, to Apg5p is crucial for autophagy in yeast. This reaction proceeds in a ubiquitination‐like manner, and requires Apg7p and Apg10p. Apg7p exhibits a considerable similarity to ubiquitin‐activating enzyme (E1) and is found to activate Apg12p with ATP hydrolysis. Apg10p, on the other hand, shows no significant similarity to other proteins whose functions are known. Here, we show that after activation by Apg7p, Apg12p is transferred to the Cys‐133 residue of Apg10p to form an Apg12p–Apg10p thioester. Cells expressing Apg10pC133S do not generate the Apg12p–Apg5p conjugate, which leads to defects in autophagy and cytoplasm‐to‐vacuole targeting of aminopeptidase I. These findings indicate that Apg10p is a new type of protein‐conjugating enzyme that functions in the Apg12p–Apg5p conjugation pathway.


Proteins are often modified by a variety of moieties in cells: phosphates, carbohydrates, lipids or proteins, etc. Such chemical processes, called post‐translational modifications, modulate protein function by altering protein activity, stability, subcellular localization or interaction with other molecules. A well characterized type of protein modification is ubiquitination, which is involved in selective degradation of short‐lived proteins, quality control of endoplasmic reticulum proteins, turnover of plasma membrane proteins and so on (Varshavsky, 1997; Bonifacino and Weissman, 1998; Ciechanover, 1998; Hershko and Ciechanover, 1998). Ubiquitin, a highly conserved 76‐amino‐acid‐residue protein, is transferred to target proteins to form isopeptide bonds between the C‐terminal glycine of ubiquitin and the ϵ‐amino group of a lysine residue in the target protein. Ubiquitination involves sequential enzymatic reactions by several classes of enzymes. The ubiquitin‐activating enzyme, E1, hydrolyzes ATP and forms an intermediate of ubiquitin adenylate, followed by the binding of ubiquitin to its own active‐site cysteine residue in a thioester linkage. Activated ubiquitin is next transferred to an active‐site cysteine residue of a ubiquitin‐conjugating enzyme, E2, to form the ubiquitin thioester in a similar fashion. At the final step, which is often catalyzed by ubiquitin‐protein ligase, E3, ubiquitin is attached to a lysine residue of the target protein via an isopeptide bond (Varshavsky, 1997; Bonifacino and Weissman, 1998; Hershko and Ciechanover, 1998).

In recent years, the ubiquitin‐related proteins have been discovered and some of them have been found to be covalently attached to other proteins (Johnson and Hochstrasser, 1997; Kamitani et al., 1997; Saitoh et al., 1997; Hochstrasser, 1998; Lammer et al., 1998). RanGAP1 (Ran‐specific GTPase‐activating protein) (Matunis et al., 1996, 1998; Mahajan et al., 1997, 1998), PML (Kamitani et al., 1998; Müller et al., 1998), Sp100 (Sternsdorf et al., 1997, 1999) and I‐κBα (Desterro et al., 1998) were found to be covalently modified by SUMO‐1 (small ubiquitin‐related modifier‐1). The ligation of SUMO‐1 to RanGAP1 is crucial for its association with RanBP2, a component of the nuclear pore complex. Modification of PML or Sp100 by SUMO‐1 is also important for the assembly of subnuclear structures termed nuclear dots or PML nuclear bodies (Duprez et al., 1999; Müller and Dejean, 1999). Smt3p, a yeast homolog of SUMO‐1, is essential for viability (Johnson and Blobel, 1997) and is likely to contribute to cell cycle regulation (Li and Hochstrasser, 1999). The enzyme system involved in conjugation of SUMO‐1/Smt3p to other proteins has been identified recently (Gong et al., 1997, 1999; Johnson et al., 1997; Schwarz et al., 1998; Desterro et al., 1999; Okuma et al., 1999). SUMO‐1/Smt3p is activated by an Aos1–Uba2 heterodimer and conjugated to proteins by Ubc9. NEDD8 (neural precursor cell‐expressed developmentally down‐regulated gene 8) and its yeast homolog Rub1p, showing ∼50% identity to ubiquitin, are conjugated to cullin/Cdc53p, a common subunit of the multifunctional SCF ubiquitin ligase, in a ubiquitination‐like manner. The NEDD8 conjugation pathway requires at least two enzymes: an E1‐like heterodimeric activating enzyme, APP‐BP–hUba3, and an E2‐like conjugating enzyme, hUbc12 (Osaka et al., 1998; Gong and Yeh, 1999). The activating and conjugating enzymes of the Rub1p conjugation pathway have also been identified as Ula1p–Uba3p and Ubc12p, respectively (Lammer et al., 1998; Liakopoulos et al., 1998). Although substrates of Ubc9 and Ubc12 are not ubiquitin, Ubc9 and Ubc12 are structurally similar to the ubiquitin E2s.

Recently, we found a novel modifier essential for autophagy (Mizushima et al., 1998a). Autophagy is a cellular process essential for bulk degradation of cytoplasmic components in eukaryotes. We demonstrated previously that cytoplasmic components including proteins, carbohydrates, lipids, nucleic acids and organelles are transported to the vacuole in yeast by macroautophagy in response to nutrient starvation (Takeshige et al., 1992; Baba et al., 1994, 1995). When cells face nutrient starvation, cytoplasmic components are sequestered non‐selectively by double‐membrane‐bound structures termed autophagosomes and then targeted to the lysosome/vacuole to be degraded. Autophagy plays a central role in protein turnover, which may be important for cellular remodeling during development and differentiation. Taking advantage of yeast genetics, autophagy‐deficient mutants of Saccharomyces cerevisiae (apg and aut) have been isolated independently by two laboratories (Tsukada and Ohsumi, 1993; Thumm et al., 1994). Surprisingly, the apg and aut mutants partially overlap with cvt mutants that are based on defects in cytoplasm‐to‐vacuole targeting (Cvt) of aminopeptidase I (Harding et al., 1996; Scott et al., 1996). The Cvt pathway uses a similar mechanism to the autophagic pathway (Baba et al., 1997). During analyses of APG gene products, we found a novel protein conjugation system that is indispensable for both autophagy and the Cvt pathway (Mizushima et al., 1998a). Apg12p is a hydrophilic 186‐amino‐acid‐residue protein with no similarity to ubiquitin or ubiquitin‐related modifiers, which is covalently attached to Apg5p. The Apg12p–Apg5p conjugate further associates with Apg16p to form the Apg12p–Apg5p–Apg16p multimeric complex (Mizushima et al., 1999). The manner of conjugation of Apg12p to Apg5p is ubiquitination‐like: a covalent linkage of the C‐terminal glycine of Apg12p to the lysine‐149 side chain of Apg5p via an isopeptide bond. This reaction requires at least two proteins: Apg7p and Apg10p. Apg7p is similar to ubiquitin‐activating enzyme in its amino acid sequence around the ATP binding site and the active‐site cysteine residue, and has been found to be an E1‐like Apg12p‐activating enzyme (Mizushima et al., 1998a; Tanida et al., 1999). Apg10p, on the other hand, shows no significant similarity with other proteins whose functions are known. Here, we show that Apg10p is a new type of protein‐conjugating enzyme that functions in the Apg12p conjugation pathway.


apg10‐1 shows a defect in autophagy

We have isolated and characterized an apg10‐1 mutant (Tsukada and Ohsumi, 1993). Wild‐type cells accumulated autophagic bodies (ABs) in their vacuoles in a nitrogen‐starvation medium containing 1 mM phenylmethylsulfonyl fluoride (PMSF), which were observed under a light microscope (Takeshige et al., 1992 and Figure 1A, wild type). On the other hand, the apg10‐1 cells did not accumulate ABs (Figure 1A). Moreover, the apg10‐1 mutation severely reduced the viability of the cells under the starvation condition (Figure 1B). It was found that Apg12p is covalently attached to Apg5p via an isopeptide linkage in a ubiquitination‐like manner and this conjugation is crucial for autophagy (Mizushima et al., 1998a). The apg10‐1 mutant completely lost the ability to generate the Apg12p–Apg5p conjugate (Mizushima et al., 1998a and Figure 1C). These data suggest that the defect of the apg10‐1 mutant in autophagy is due to loss of the Apg12p–Apg5p conjugate, and that Apg10p probably functions at the Apg12p‐conjugating step.

Figure 1.

Cloning of the APG10 gene and phenotype of Δapg10 cell. (A) Wild‐type (YW5‐1B), apg10‐1 mutant (MT91‐4‐2) and Δapg10 (TFD10‐L1) cells were incubated in SD(−N) medium containing 1 mM PMSF at 30°C. After incubation for 6 h, cells were observed under light microscopy (Nomarski images). (B) Wild‐type (closed circles), apg10‐1 mutant (closed triangles) and Δapg10 (open circles) cells were grown to 1 OD600/ml in YPD and then transferred to SD(−N) medium. After incubation at 30°C for the times indicated, their viability was determined by phloxine B staining (Tsukada and Ohsumi, 1993). (C) Cell lysates from wild‐type cells with empty vector pRS316 (lane 1) or HAAPG12 (lane 2), and also apg10‐1 (lane 3) and Δapg10 cells with HAAPG12 (lane 4) were subjected to Western blotting analysis with anti‐HA antibody (16B12). (D) Amino acid sequence of Apg10p. Cysteine residues are indicated in bold.

Isolation and sequence analysis of APG10

The APG10 gene was cloned by complementation of reduced viability of apg10‐1 mutant under starvation conditions as described in Materials and methods. A YCp50‐based yeast genomic library (a gift from Dr Wada, Osaka University) was introduced into apg10‐1 cells, and screening of ∼10 000 transformants yielded plasmid 5‐5C containing an 8.0 kb genomic fragment. Subcloning of the fragment of 5‐5C revealed that a 1.2 kb XbaI–HindIII fragment containing one open reading frame (ORF) YLL042c was sufficient for complementation. The YLL042c is a novel gene and encodes a hydrophilic protein of 167 amino acids with a predicted molecular mass of 19.8 kDa (Figure 1D).

To determine whether YLL042c was the authentic APG10 gene, this ORF was disrupted by replacing with a LEU2 gene. The disruptant exhibited the same phenotype as the apg10‐1 cell: (i) the autophagic bodies did not accumulate in the vacuole during starvation (Figure 1A); (ii) the viability of the disruptant decreased under the starvation condition (Figure 1B); and (iii) the Apg12p–Apg5p conjugate was not generated at all (Figure 1C). Sequence analysis of the apg10‐1 allele revealed that the 482nd nucleotide of the coding sequence was altered from G to A, which led to translation termination at the 160th amino acid residue. Furthermore, a diploid cell obtained by crossing apg10‐1 and Δyll042c cells is also defective in autophagy. We concluded therefore, that YLL042c was the authentic APG10 gene.

The amino acid sequence of Apg10p did not provide any insight into its function. A BLAST search, however, revealed that it was homologous to a Caenorhabditis elegans protein of unknown function (DDBJ/EMBL/GenBank accession No. z54282) but exhibited no significant similarity to ubiquitin‐conjugating enzymes or ubiquitin‐protein ligases.

Formation of an Apg12p–Apg10p thioester

To elucidate how Apg10p participates in the Apg12p–Apg5p conjugation pathway, we first examined whether Apg10p could interact with other components of this pathway. In a two‐hybrid assay (James et al., 1996), we observed that Apg10p physically interacts with Apg7p and Apg12p but not with Apg5p (Figure 2A). Next we performed coimmunoprecipitation to confirm the Apg10p–Apg12p interaction. Total cell lysates of Δapg5 cells with HAAPG10 and/or MycAPG12 (3× Myc‐tagged construct) on 2μ plasmids were immunoprecipitated with either anti‐HA or anti‐Myc antibody. The resulting precipitates were analyzed by Western blotting. When the cells expressed both HAApg10p and MycApg12p, MycApg12p was coprecipitated by pulling HAApg10p down with anti‐HA antibody (Figure 2B, lane 17), and HAApg10p was coprecipitated with MycApg12p (Figure 2B, lane 18). These results indicate that Apg10p interacts with Apg12p in vivo. Under non‐reducing SDS–PAGE conditions, an additional 56 kDa band was also detected by both anti‐HA and anti‐Myc antibodies (Figure 2B, lanes 8 and 9), whereas it was not detected in cells harboring either HAAPG10 or MycAPG12 plasmid. Furthermore, its molecular mass was shifted from 56 to 61 kDa by changing MycAPG12 plasmid to a 6× Myc‐tagged construct (data not shown). This difference in molecular mass just corresponds with a size of 3× Myc. These findings show that the 56 kDa band contains both HAApg10p and MycApg12p, and these molecules conjugate 1:1 via a reducing reagent‐sensitive bond. Apg12p has three cysteine residues. Because the substitutions of all three cysteine residues of Apg12p by alanine did not affect the formation of the Apg12p–Apg10p conjugate, the dithiothreitol (DTT)‐sensitive bond is probably a thioester bond between the C‐terminal glycine of Apg12p and a cysteine residue of Apg10p, but is not a disulfide bond. Another DTT‐sensitive 67 kDa band was also detected by anti‐Myc but not by anti‐HA antibody (Figure 2B, lanes 6, 8 and 9), suggesting that it contained Apg12p but not Apg10p. We concluded, however, from several lines of evidence (described below) that the 67 kDa complex is not involved in the Apg12p–Apg5p conjugation pathway. Similar results were obtained with the cells expressing both HAApg10p and MycApg12p from CEN plasmids (data not shown).

Figure 2.

In vivo interaction of Apg10p with Apg12p. (A) Two‐hybrid assay for interactions of Apg10p with Apg5p, Apg7p and Apg12p. Yeast cells (PJ69‐4A) harboring the indicated plasmids were streaked out on plates with medium lacking adenine to assay for interaction‐dependent activation of the ADE2 gene. (B) Total lysates from Δapg5 cells with HAAPG10 and/or MycAPG12 on 2μ plasmid were immunoprecipitated with anti‐HA (mAb 16B12) or anti‐Myc (mAb 9E10) antibody. Precipitates were analyzed by Western blotting using anti‐HA or anti‐Myc antibody. Apg12p adduct of Apg10p is indicated by ● and an unidentified complex of 67 kDa by ●●. The positions of cross‐reacting IgG heavy chain (H) and light chain (L) are indicated by arrowheads.

Next we examined the effects of Apg7p and Apg5p on formation of the Apg12p–Apg10p conjugate. Apg7p has already been found to function as an E1‐like Apg12p‐activating enzyme by Mizushima et al. (1998a) and Tanida et al. (1999). Apg7p probably activates Apg12p with hydrolysis of ATP and subsequent transfer to its own thiol group of the Cys‐507. By analogy with ubiquitination, an Apg12p covalently bound to the Cys‐507 residue of Apg7p can be transferred to a cysteine residue of a hypothetical E2‐like enzyme. Figure 3 shows that the Apg12p–Apg10p conjugate was not formed in Δapg7 cells, suggesting that Apg12p is tranferred to Apg10p after Apg7p functions (Figure 3, lanes 8 and 9). The non‐covalent interaction between Apg10p and Apg12p was also lost completely in Δapg7 cells. Considering that Apg7p should activate the C‐terminus of Apg12p, the Apg7p‐dependent formation of the Apg12p–Apg10p conjugate suggests that Apg12p and Apg10p are linked via a thioester bond. As stated above, the interaction between Apg7p and Apg10p was confirmed in the two‐hybrid assay. Taken together, these results suggest that an activated Apg12p is transferred directly from Apg7p to an active‐site cysteine residue of Apg10p.

Figure 3.

Effects of the deletion of APG5 or APG7 on Apg12p–Apg10p thioester formation. Coimmunoprecipitation was performed with wild‐type (KA311B), Δapg5 (YNM122) or Δapg7 (YTS12) cells harboring HAAPG10 and MycAPG12 on 2μ plasmids. The resulting precipitates were analyzed by Western blotting. The DTT‐resistant Apg12p–Apg5p conjugate is indicated by ●.

When coimmunoprecipitation was carried out using wild‐type cells, Apg12p–Apg10p thioester was generated as in the Δapg5 cells, i.e. the deletion of Apg5p, a target molecule of the Apg12p conjugation, did not promote accumulation of the Apg12p–Apg10p conjugate (Figure 3, lanes 2 and 3). Overexpression of Apg5p also had no effect on the formation of the Apg12p–Apg10p thioester (data not shown). These results indicate that the amount of Apg5p would not change the kinetics of Apg12p–Apg10p thioester formation. Moreover, in wild‐type cells, the Apg12p–Apg5p conjugate was precipitated with anti‐Myc but not with anti‐HA antibody (Figure 3B, lanes 3 and 12) providing evidence that Apg10p no longer associates with Apg12p after Apg12p is transferred to Apg5p. The interaction between Apg10p and Apg5p was not detected in the two‐hybrid assay (Figure 2A). These data suggest that Apg12p is not transferred directly from Apg10p to Apg5p, and other factor(s) may be necessary for the Apg10p‐to‐Apg5p transfer of Apg12p.

The DTT‐sensitive 67 kDa band was even detected in the Δapg7 cells when immunoprecipitation was carried out by anti‐Myc antibody (Figure 3, lane 9), suggesting that the 67 kDa complex could be generated without activation of Apg12p. Moreover, the band appeared even when an APG12ΔG (C‐terminal Gly deletion) construct, which could conjugate to Apg5p no longer, was used for immunoprecipitation instead of a wild‐type APG12 construct (data not shown). Therefore, we concluded that the 67 kDa complex did not participate in the Apg12p–Apg5p conjugating reaction.

Apg12p is attached to the Cys‐133 residue of Apg10p

There are three cysteine residues in Apg10p (Figure 1D). To determine which cysteine residue contributes to thioester formation, each cysteine residue of Apg10p was substituted by serine using site‐directed mutagenesis. First, we examined the abilities of the substitutes to generate the Apg12p–Apg5p conjugate by Western blotting. The Apg12p–Apg5p conjugate of 70 kDa was formed in Δapg10 cells harboring HAAPG10 and HAAPG12 on CEN plasmids (Figure 4A, lane 3). The substitution of Cys‐26 or Cys‐137 by Ser (C26S or C137S, respectively) in Apg10p did not affect the generation of the conjugate at all, whereas the substitution of Cys‐133 by Ser (C133S) resulted in a complete loss of the conjugation, although Apg10pC133S was detected in an amount comparable to that of wild‐type Apg10p (Figure 4A).

Figure 4.

Apg12p is transferred to the Cys‐133 residue of Apg10p. (A) Cell lysates from Δapg10 cells (TFD10‐L1) harboring HAAPG12 and each mutant form of HAAPG10 on CEN plasmids were subjected to Western blotting analysis with anti‐HA antibody (mAb 16B12). (B) Cell extracts from the wild‐type cells (KA311B) harboring MycAPG12 and mutated HAAPG10 were subjected to immunoprecipitation and the resulting precipitates were analyzed by Western blotting with the rabbit anti‐HA antibody.

To investigate whether Apg12p was covalently attached to Cys‐133 of Apg10p by a thioester bond, coimmunoprecipitation was carried out using the C133S or C133A (the substitution of Cys‐133 by Ala) mutants (Figure 4B). As compared with the positive control (Figure 4B, lanes 2 and 3), quite a small amount of Apg12p–Apg10pC133S conjugate was detected (Figure 4B, lanes 5 and 6). This conjugate was found to be resistant to the reducing reagent (Figure 4B, lanes 14 and 15), which suggested that an ester bond is formed between Apg12p and Apg10pC133S instead of a thioester bond. The C133A mutation caused a complete loss of the Apg12p–Apg10p conjugate even in the absence of the reducing reagent (Figure 4B, lanes 8 and 9). These results indicate that Cys‐133 is an active center of Apg10p and contributes to thioester formation. Both mutations prevented the interaction between the free forms of Apg12p and Apg10p (Figure 4B, lanes 6 and 9).

Next we asked whether the Cys‐133 mutation of Apg10p affected autophagy and Cvt of proaminopeptidase I (proAPI). Autophagic activities were measured using a biochemical assay monitoring autophagy‐dependent processing of alkaline phosphatase Pho8Δ60 (Noda et al., 1995). In wild‐type cells, Pho8Δ60 was activated by transferring the cells from a rich medium to a nitrogen‐depleted medium. The Δapg10 cells expressing Apg10pC133S did not activate Pho8Δ60 even in the starvation medium, while those expressing Apg10pC26S or Apg10pC137S showed normal autophagic activity (Figure 5A). These data indicate that the C133S mutation in Apg10p causes a defect in autophagy. ProAPI is synthesized in the cytoplasm and imported directly to the vacuole by a mechanism closely related to the autophagic pathway, and then processed to the mature form by proteinase B. As expected, proAPI was not processed in the Δapg10 cells expressing Apg10pC133S (Figure 5B), implying that the formation of a thioester intermediate is also essential for API transport.

Figure 5.

The cysteine‐133 residue of Apg10p is essential for autophagy and the Cvt pathway. (A) Autophagic activities were measured by an ALP assay (Noda and Ohsumi, 1998). YTS3 cells (Δapg10 PHO8::pho8Δ60) harboring each mutant APG10 on CEN plasmids were grown to 1 OD600/ml in SC medium lacking tryptophan and then transferred to SD(−N) medium. Lysates from the cells after incubation for 0 and 4 h were used for assay. Error bars indicate the SD of three independent experiments. (B) Cell lysates from Δapg10 cells (TFD10‐L1) harboring empty vector, pRS316 (lane 1), wild‐type HAAPG10 (lane 2), HAAPG10C26S (lane 3), HAAPG10C133S (lane 4) or HAAPG10C137S (lane 5) on CEN plasmids were subjected to Western blotting analysis with anti‐API antiserum.


Apg12p is a modifier protein with no significant similarity to ubiquitin and ubiquitin‐related modifiers. The covalent binding of Apg12p to Apg5p is essential for autophagy in yeast. The Apg12p–Apg5p conjugating reaction is initiated by Apg7p, an Apg12p‐activating enzyme, with ATP hydrolysis (Mizushima et al., 1998a; Tanida et al., 1999). Here, we have characterized Apg10p as an Apg12p‐conjugating enzyme acting after Apg7p in the Apg12p–Apg5p conjugation pathway. Coimmunoprecipitation experiments revealed that a reducing reagent‐sensitive covalent adduct was formed between Apg12p and Apg10p in an Apg7p‐dependent manner (Figures 2B and 3). This adduct is likely to be a thioester linked between the C‐terminal carboxy group of Apg12p and the active‐site cysteine‐133 of Apg10p (Figure 4). These findings suggest that the high‐energy thioester bond between Apg12p and Apg7p is transferred to Apg10p through transacylation. The Apg12p transfer from Apg7p to Apg10p certainly occurs directly because the interaction between Apg7p and Apg10p was observed in the yeast two‐hybrid assay (Figure 2A). As shown above, Apg10p functions as an ‘E2’ in the Apg12p–Apg5p conjugation system.

On the other hand, we speculate that the transfer of Apg12p from Apg10p to Apg5p, a final target molecule of Apg12p, requires other protein(s) on the grounds of several lines of evidence: (i) the interaction between Apg10p and Apg5p was not detected by a two‐hybrid assay (Figure 2A); (ii) Apg12p did not interact with Apg5p in the Δapg7 or Δapg10 cells with a two‐hybrid assay but did in wild‐type cells (data not shown); (iii) Apg10p no longer associated to the Apg12p–Apg5p conjugate (Figure 3, lanes 2, 3, 11 and 12), which suggests that neither Apg10p nor Apg12p contributes to the recognition of Apg5p; and (iv) moreover, the changes in expression level of Apg5p did not affect the kinetics of the Apg12p–Apg10p thioester formation (Figure 3). The other factor(s) that recognizes Apg5p may be essential for the transfer of Apg12p from Apg10p to Apg5p. Our model for an enzyme system of the Apg12p–Apg5p conjugation pathway is shown in Figure 6. Apg7p hydrolyzes ATP and forms the high‐energy thioester bond between its active‐site cysteine and the C‐terminus of Apg12p. The activated Apg12p is transferred directly from Apg7p to the active‐site cysteine‐133 of Apg10p to form the thioester bond. Then Apg12p is transferred to the target protein Apg5p, which may require the additional protein(s).

Figure 6.

Model of the enzymatic pathway for the Apg12p–Apg5p conjugation system. Apg12p is activated by Apg7p with ATP hydrolysis; the carboxy group of the C‐terminal glycine of Apg12 is conjugated to the thiol group of the cysteine‐507 residue of Apg7p via a high‐energy thioester bond (∼). Subsequent transfer of Apg12p to the thiol group of the cysteine‐133 residue of Apg10p results in the formation of the Apg12p–Apg10p thioester. Finally, the C‐terminus of Apg12p is covalently attached to the lysine‐149 residue of Apg5p via an isopeptide bond, which may require hypothetical Apg5p‐recognizing protein(s).

In the coimmunoprecipitation experiment, the non‐covalent interaction between Apg12p and Apg10p was also observed, which required both Apg7p and Apg10p activity (Figures 3 and 4B). This suggests that the non‐covalent association is dependent on Apg12p–Apg10p thioester formation. Therefore, the non‐covalent complex may actually exist as one of the intermediate forms of the reaction during transfer of Apg12p from Apg10p to Apg5p. Alternatively, the non‐covalent complex may be due to the unexpected cleavage of the thioester bond during the preparation of samples.

As described above, the reaction mechanism for Apg12p–Apg5p conjugation is closely related to those for the ubiquitin or ubiquitin‐related modifier system, and Apg10p catalyzes the E2‐like reaction in the Apg12p–Apg5p conjugation pathway. A large number of E2s have been identified widely in eukaryotes and comprise the Ubc (ubiquitin‐conjugating enzyme) family that is characterized by conserved sequences, termed ‘E2 motifs’, containing the active‐site cysteine and a conserved His–Pro–Asn tripeptide (Haas and Siepmann, 1997). It has been found that some Ubcs are used for ubiquitin‐related modifier, e.g. Ubc9 and Ubc12 for SUMO‐1/Smt3p and NEDD8/Rub1p, respectively (Gong et al., 1997; Johnson and Blobel, 1997; Lee et al., 1998; Liakopoulos et al., 1998; Osaka et al., 1998; Schwarz et al., 1998). In spite of the similarity of the reaction mechanisms between the Apg12p and ubiquitin systems, the Apg12p system uses the protein with no significant sequence similarity to Ubcs as an E2. Because the structure of Apg12p does not resemble that of ubiquitin or ubiquitin‐related modifier at all, except for the C‐terminal glycine, the protein structurally unrelated to Ubcs, namely Apg10p, must be necessary for the recognition of Apg12p and/or Apg7p. This is the first example showing that the protein quite different from Ubcs functions as an E2‐like protein‐conjugating enzyme.

A BLAST search revealed that Apg10p has a potential C.elegans homolog of unknown function. In addition, the EST fragments that are closely related to Apg10p are also found in human (DDBJ/EMBL/GenBank accession No. AA448080) and mouse (DDBJ/EMBL/GenBank accession No. AA673136). The sequences around the cysteine‐133 residue are well conserved among these proteins, supporting the suggestion that the cysteine‐133 residue of Apg10p is the active site. Considering that the Apg12p–Apg5p conjugation system is conserved from yeast to human (Mizushima et al., 1998b), Apg10p should also be widely distributed in eukaryotes. Recently, it was reported that the Pichia pastoris homolog of Apg7p, Gsa7, is required for microautophagy of peroxisomes and is able to complement the defects in macroautophagy and API transport in S.cerevisiae (Kim et al., 1999; Yuan et al., 1999). These findings suggest that Apg10p may function in peroxisome degradation as well as macroautophagy and the Cvt pathway.

Materials and methods

Strains and media, genetic and molecular biological techniques

The S.cerevisiae strains used in this study are listed in Table I. Yeast cells were grown in yeast extract–peptone–dextrose (YPD) or synthetic complete (SC) media. SD(−N) medium (0.17% yeast nitrogen base without amino acids and ammonium sulfate and 2% glucose) was used for nitrogen starvation. Standard genetic manipulations were performed as described by Adams et al. (1998). DNA manipulations were performed using standard methods (Sambrook et al., 1989).

View this table:
Table 1. Yeast strains used in this study

Cloning, disruption and epitope tagging of APG10

APG10 was cloned by complementation of reduced viability of apg10‐1 in the SD(−N) medium as described previously (Funakoshi et al., 1997; Kametaka et al., 1998). A yeast mutant MT91‐4‐2 was transformed with a YCp50‐based yeast genomic library and Ura+ transformants were replica‐plated onto SD(−N) plates containing 10 μg/ml phloxine B, on which dead cells were stained red (Tsukada and Ohsumi, 1993). White colonies were picked up and subjected to a light microscopic observation to confirm the accumulation of autophagic bodies in their vacuoles.

Plasmids were rescued from the candidate cells and one plasmid, named 5‐5C, was used for further analysis. Subcloning identified a 1.2 kb XbaI–HindIII fragment as the minimal region with complementing activity.

The 2.4 kb XbaI–KpnI fragment was subcloned into pBluescript II KS+ (Stratagene) to generate pKSAPG10. The 2.0 kb SmaI–PstI LEU2 fragment from pJJ282 (Jones and Prakash, 1990) was cloned into PstI–EcoRV‐digested pKSAPG10 to generate pKSapg10::LEU2. The 4.24 kb NheI–NcoI fragment from pKSapg10::LEU2 was used for transformation of YW5‐1B and TN125. The disruption of the APG10 gene was confirmed by PCR.

C‐terminal epitope‐tagged APG10 (HAAPG10) was constructed by inserting the DNA sequence encoding a repeated hemagglutinin (HA)‐epitope tag (2× HA) just before the stop codon. The tagged HAAPG10 was confirmed to be functional by complementation of the apg phenotype of the apg10‐1 cell.

Plasmid construction

HAAPG10 was cloned into the XbaI and EcoRI sites of pRS316 and pRS426 (Sikorski and Hieter, 1989) to generate pHA‐APG10‐316 and pHA‐APG10‐426, respectively. pHA‐APG12‐314 was described previously (Mizushima et al., 1998a) and pMyc‐APG12‐424 was generated by subcloning of the DNA fragment encoding 3× Myc‐tagged Apg12p (Mizushima et al., 1998a) into the SpeI and XhoI sites of pRS424. For the two‐hybrid assay, the entire APG10 ORF was amplified by PCR and inserted into the EcoRI site of pGBD‐C1 (James et al., 1996) to generate pGBD‐APG10 that expresses the DNA binding domain of Gal4p fused with Apg10p. pGAD‐APG5 was created by insertion of the entire APG5 ORF into the BamHI and PstI sites of pGAD‐C1 to fuse Apg5p to the C‐terminal end of the Gal4p activation domain. pGAD‐APG7 contains the same APG7 fragment as pGBD‐APG7 (Tanida et al., 1999). pGAD‐APG12 was described previously (Mizushima et al., 1999).

Site‐directed mutagenesis

Site‐directed mutagenesis of APG10 was performed by PCR (Ho et al., 1989). The following primers were used: C26S‐Fw, 5′‐CTGGGCCCTTTCCCAAGATGTCC‐3′; C26S‐Rv, 5′‐GACATCTTGGGAAAGGGCCCAGT‐3′; C133S‐Fw, 5′‐TTCCATCCATCCGATACATCATGTATA‐3′; C133S‐Rv, 5′‐CATGATGTATCGGATGGATGGAA‐3′; C133A‐Fw, 5′‐TTCCATCCAGCCGATACATCATGTATA‐3′; C133A‐Rv, 5′‐CATGATGTATCGGCTGGATGGAA‐3′; C137S‐Fw, 5′‐GATACATCATCTATAGTAGGTGAC‐3′; C137S‐Rv, 5′‐GTCACCTACTATAGATGATGTATC‐3′. The mutated APG10 fragments were inserted into the XbaI and EcoRI sites of pRS316 or pRS426. All mutations were verified by dideoxy sequencing.


Yeast cells harboring HAAPG10 and/or MycAPG12 on 2μ plasmid were broken with glass beads in IP buffer [50 mM Tris–HCl pH 7.5, 150 mM NaCl, 5 mM EDTA, 1 mM PMSF and 1× protease inhibitor mixture (Complete™, Boehringer Mannheim)]. Cell debris was removed by centrifugation at 5000 g for 5 min, and the protein concentration of the resultant lysate was determined by Bradford's method. After Nonidet P‐40 was added to the lysate to a final concentration of 1%, it was incubated with protein G–Sepharose 4 Fast Flow (Amersham Pharmacia Biotech) at 4°C for 1 h for preabsorption. The preabsorbed lysate containing 5 mg of protein was incubated with or without 1 μl of mouse monoclonal anti‐HA (16B12, BAbCo) or anti‐Myc (9E10) antibody at 4°C for 2 h and then 20 μl of protein G–Sepharose (50% slurry) were added to it. After incubation at 4°C for 2 h, unbound proteins were removed by three washes in 1 ml of IP buffer containing 1% Nonidet P‐40. The immunoprecipitated proteins were eluted by boiling in SDS gel‐loading buffer with or without 100 mM DTT. Proteins were separated by SDS–PAGE and analyzed by immunoblotting with anti‐HA (16B12), anti‐Myc antibody (9E10) or rabbit polyclonal anti‐HA antibody (BAbCo).

Western blotting analysis of whole yeast lysate

Cells were resuspended in 100 μl of 0.2 N NaOH, 0.5% 2‐mercaptoethanol. After incubation for 15 min on ice, 1 ml of ice‐cold acetone was added and further incubated for 30 min at −20°C. After centrifugation at 10 000 g for 5 min, the resulting pellets were resuspended in the appropriate volume of SDS loading buffer and boiled for 5 min. Lysates equivalent to 0.5 OD600 cells were separated by SDS–PAGE and electrotransferred to a polyvinylidene difluoride membrane (Millipore). Mouse monoclonal anti‐HA antibody (16B12) or rabbit anti‐aminopeptidase I antibody (a gift from Dr Klionsky) was used for immunodetection. Development was performed by the ECL detection methods (Amersham Pharmacia Biotech).

Other methods

For measurement of autophagic activity, the alkaline phosphatase (ALP) assay was performed as described previously (Noda and Ohsumi, 1998). The two‐hybrid assay was performed as described by James et al. (1996).


We would like to thank Dr Yoh Wada (Osaka University) for the yeast genomic library and Dr Daniel J.Klionsky (University of California, Davis) for anti‐API antibody. This work was supported in part by Grants‐in‐Aids for Scientific Research from the Ministry of Education, Science, Sports and Culture of Japan. T.S. is a Research Fellow of the Japan Society for the Promotion of Science.