**2. Recognition of vacuolar enzymes by Atg19 and Atg34**

#### **2.1 Selective transport of vacuolar enzymes by autophagic pathways**

In the budding yeast *S. cerevisiae*, α-mannosidase (Ams1) and a precursor form of aminopeptidase I (prApe1) are selectively delivered into the vacuole through the cytoplasm to vacuole targeting (Cvt) pathway under vegetative conditions, and via autophagy under starvation conditions. The Cvt pathway is topologically and mechanistically similar to autophagy (Lynch-Day and Klionsky, 2010); therefore, studies on the molecular mechanisms of cargo recognition in the Cvt pathway will provide insight into the basic mechanism of selective autophagy. prApe1, the primary Cvt cargo, is synthesized in the cytosol as a precursor form with a cleavable propeptide consisting of 45 amino acid residues at the N terminus (Klionsky et al., 1992) and assembles into a dodecamer. The prApe1 dodecamer further self-assembles into a higher order structure called the Ape1 complex. The existence of a specific receptor for prApe1 was proposed when it was observed that prApe1 transport to the vacuole by the Cvt pathway is both specific and saturable.

Two groups simultaneously discovered that Atg19 has all of the characteristics needed to be a receptor for prApe1 in Cvt transport (Leber et al., 2001; Scott et al., 2001). Characterization of the protein revealed that Atg19 is needed for the stabilization of prApe1 binding to the Cvt vesicle membrane, and that in *atg19*Δ cells, prApe1 maturation is inhibited while autophagy is not affected (Suzuki et al., 2002). In addition, Atg19 binds to prApe1 in a propeptide-dependent manner, suggesting that the propeptide region is responsible for the recognition of prApe1 by the Cvt pathway machinery (Shintani et al., 2002). A secondarystructure prediction suggested that the prApe1 propeptide forms a helix-turn-helix structure and that the first helix exhibits the characteristics of an amphipathic α helix (Martinez et al., 1997). Our previous study revealed that the region containing the first helix of the prApe1 propeptide (residues 1-20) is sufficient for interaction with Atg19 (Watanabe et al., 2010). This is consistent with a previous report showing that the first helix of the prApe1 propeptide is critical for prApe1processing (Oda et al., 1996). *In vitro* pull-down assays showed that the coiled coil domain of Atg19 (residues 124-253), which contains a predicted coiled coil between amino acids 160 and 187, directly interacts with the prApe1 propeptide. This is consistent with a previous report showing that the prApe1-binding site of Atg19 is located in the region between amino acid residues 153 and 191 (Shintani et al., 2002).

Ams1, another Cvt cargo, oligomerizes after synthesis and associates with the Ape1 complex through the action of Atg19. Atg19 has two stable domains, the N-terminal

selective transport of vacuolar enzymes to the vacuole through autophagy (Leber et al., 2001; Scott et al., 2001; Suzuki et al., 2010), p62 and neighbor of BRCA1 gene 1 (NBR1) in the autophagic degradation of ubiquitinated protein aggregates (Bjorkoy et al., 2005; Kirkin et al., 2009), PpAtg30 in pexophagy (autophagic degradation of peroxisome) (Farre et al., 2008), and Atg32 and Nix1 in mitophagy (autophagic degradation of mitochondria) (Kanki et al., 2009; Novak et al., 2010; Okamoto et al., 2009). Most of these receptors interact directly with

We have been studying the mechanisms of specific cargo recognition during autophagy, especially those of the selective delivery of vacuolar enzymes into the vacuole in yeast. We summarize here the current knowledge of such mechanisms as revealed by biochemical and

In the budding yeast *S. cerevisiae*, α-mannosidase (Ams1) and a precursor form of aminopeptidase I (prApe1) are selectively delivered into the vacuole through the cytoplasm to vacuole targeting (Cvt) pathway under vegetative conditions, and via autophagy under starvation conditions. The Cvt pathway is topologically and mechanistically similar to autophagy (Lynch-Day and Klionsky, 2010); therefore, studies on the molecular mechanisms of cargo recognition in the Cvt pathway will provide insight into the basic mechanism of selective autophagy. prApe1, the primary Cvt cargo, is synthesized in the cytosol as a precursor form with a cleavable propeptide consisting of 45 amino acid residues at the N terminus (Klionsky et al., 1992) and assembles into a dodecamer. The prApe1 dodecamer further self-assembles into a higher order structure called the Ape1 complex. The existence of a specific receptor for prApe1 was proposed when it was observed that prApe1 transport

Two groups simultaneously discovered that Atg19 has all of the characteristics needed to be a receptor for prApe1 in Cvt transport (Leber et al., 2001; Scott et al., 2001). Characterization of the protein revealed that Atg19 is needed for the stabilization of prApe1 binding to the Cvt vesicle membrane, and that in *atg19*Δ cells, prApe1 maturation is inhibited while autophagy is not affected (Suzuki et al., 2002). In addition, Atg19 binds to prApe1 in a propeptide-dependent manner, suggesting that the propeptide region is responsible for the recognition of prApe1 by the Cvt pathway machinery (Shintani et al., 2002). A secondarystructure prediction suggested that the prApe1 propeptide forms a helix-turn-helix structure and that the first helix exhibits the characteristics of an amphipathic α helix (Martinez et al., 1997). Our previous study revealed that the region containing the first helix of the prApe1 propeptide (residues 1-20) is sufficient for interaction with Atg19 (Watanabe et al., 2010). This is consistent with a previous report showing that the first helix of the prApe1 propeptide is critical for prApe1processing (Oda et al., 1996). *In vitro* pull-down assays showed that the coiled coil domain of Atg19 (residues 124-253), which contains a predicted coiled coil between amino acids 160 and 187, directly interacts with the prApe1 propeptide. This is consistent with a previous report showing that the prApe1-binding site of Atg19 is

located in the region between amino acid residues 153 and 191 (Shintani et al., 2002).

Ams1, another Cvt cargo, oligomerizes after synthesis and associates with the Ape1 complex through the action of Atg19. Atg19 has two stable domains, the N-terminal

Atg8-family proteins, which are crucial factors in autophagosome biogenesis.

**2. Recognition of vacuolar enzymes by Atg19 and Atg34** 

to the vacuole by the Cvt pathway is both specific and saturable.

**2.1 Selective transport of vacuolar enzymes by autophagic pathways** 

structural studies.

domain (residues 1-123) and the Ams1 binding domain (ABD; residues 254-367, see below for further details). Ams1 associates with Atg19 via the ABD that is distinct from the prApe1 binding site and therefore Atg19 can simultaneously interact with both prApe1 and Ams1. prApe1, Ams1, and Atg19 assemble into a large complex called the Cvt complex, which was identified as an electron-dense structure localized close to the vacuole by electron microscopy (Baba et al., 1997). Atg11 interacts with Atg19 to recruit the Cvt complex to the preautophagosomal structure (PAS), which plays a central role in autophagosome formation near the vacuole (Shintani et al., 2002; Suzuki and Ohsumi, 2010). Atg19 further interacts with Atg8, which is localized at the PAS and involved in the elongation of autophagosomes, using the Atg8 family-interacting motif (AIM; 412-WEEL-415) to induce formation of the Cvt vesicle (Noda et al., 2008). Atg8 is conjugated to phosphatidylethanolamine (PE) and associates with autophagosomes or the Cvt vesicle (Ichimura et al., 2000). This explains why the vesicle selectively surrounds only the cargo. After transport to the vacuole, the prApe1 propeptide is removed via a proteinase Bdependent reaction to generate mature Ape1 (mApe1), and the Ape1 complex disassembles back into dodecamers. Atg34, an Atg19 paralog, functions as an additional receptor protein for Ams1 but not prApe1 only under starvation conditions (Suzuki et al., 2010). Although Atg34, similar to Atg19, has the predicted coiled coil (residues 130-157), Atg34 is not capable of interacting with prApe1.

Recently, two cargoes that are selectively delivered to the vacuole have been identified: leucine aminopeptidase III (Lap3) (Kageyama et al., 2009) and aspartyl aminopeptidase (Ape4) (Yuga et al., 2011). Lap3 is transported to the vacuole for degradation only when it is overproduced under nitrogen starvation conditions. Lap3 forms a homohexameric complex of ~220 kDa, which further forms an aggregate independently of prApe1. Although this transport is partially mediated by Atg19, it remains to be determined whether Lap3 can interact with Atg19. Ape4 is the third Cvt cargo, which is similar in primary structure and subunit organization to Ape1. Ape4 lacks the N-terminal propeptide that is used by prApe1 for binding to Atg19. As the Ape4-binding site in Atg19 is located between the prApe1- and Ams1-binding sites (residues 204-247), these enzymes are unlikely to compete with each other for binding to Atg19. As Atg34 did not interact with Ape4, it might not be involved in Ape4 transport. More recently, Suzuki *et al*. elucidated that selective autophagy downregulates Ty1 transposition by eliminating Ty1 virus-like particles (VLPs) from the cytoplasm under nutrient-limited conditions (Suzuki et al., 2011). Although Ty1 VLPs are not vacuolar enzymes, they are targeted to autophagosomes by an interaction with Atg19. The N-terminal domain of Atg19 is specifically required for selective transport of Ty1 VLPs to the vacuole, though Atg19 is able to interact with Ty1 Gag without the N-terminal domain. Selective autophagy might safeguard genome integrity against excessive insertional mutagenesis caused during nutrient starvation by transposable elements in eukaryotic cells.

#### **2.2 Structural basis for Ams1 recognition by Ag19 and Atg34**

Scott *et al*. suggested that Ams1 is delivered to the vacuole in an Atg19-dependent manner (Scott et al., 2001). Ams1 was found to associate with Atg19, and a defect in the Ape1-Atg19 complex formation was shown to severely affect the import of Ams1 into the vacuole, whereas Ams1 was dispensable for transport of the Ape1-Atg19 complex. This suggests that Ams1 might exploit the prApe1 import system to achieve its own effective transport to the vacuole and that it is tethered to the Ape1-Atg19 complex through interaction with Atg19. In our recent study, we identified the Ams1 binding domain (ABD) in Atg19 and Atg34 by limited proteolysis of full-length Atg19, an *in vitro* pull-down assay as well as sequence alignment (Watanabe et al., 2010). In *atg19*Δ cells expressing Atg19ΔABD, Ams1 transport to the vacuole was inhibited, suggesting that the Atg19 ABD is required for Ams1 transport to the vacuole through the Cvt pathway. In such cells, prApe1 transport to the vacuole is the normal process. These results indicate that the Atg19 ABD is specifically responsible for the transport of Ams1, but not prApe1, to the vacuole through the Cvt pathway.

The Atg19 and Atg34 ABD structures were determined in solution using NMR spectroscopy (Figure 1A and B) (Watanabe et al., 2010). Both ABDs comprise eight β-strands (A-H), of which A, B, E, and H form an antiparallel β-sheet; the surface of this sheet faces a second antiparallel β-sheet comprising C, D, F, and G, thus forming a typical immunoglobulin-like β-sandwich fold. The Atg19 and Atg34 ABD structures are similar to each other with a root mean square difference of 2.1 Å for 102 residues (Z-score calculated by the Dalilite program (Holm and Park, 2000) is 12.8). There are relatively large structural differences between the Atg19 and Atg34 ABDs in the loops located at the bottom of the immunoglobulin fold (the loop connecting strands A and B (AB loop), CD, EF, and GH loops). In contrast, the loops located at the top of the immunoglobulin fold (the BC, DE, and FG loops) have a similar conformation. Furthermore, the residues comprising the top loops, especially those of the DE loop, are more strongly conserved between the Atg19 and Atg34 ABDs than those comprising the bottom loops (Figure 1). In the DE loop, His-310/296, Glu-311/297, Ile-314/300, and Lys-315/301 of Atg19/Atg34 are exposed. Among these exposed residues, His-310/296 and/or Glu-311/297 of the Atg19/Atg34 ABD are essential for Ams1 recognition. Further analysis showed that in *atg19*Δ*atg34*Δ cells expressing Atg19H310A (substitution of His-310 with alanine) but not Atg19E311A, transport of Ams1-GFP to the vacuole under autophagy-inducing conditions is inhibited. This indicates that the conserved His residue in the DE loop of the Atg19 ABD plays a critical role in Ams1 recognition and that Ams1 binding of the Atg19 ABD is essential for Ams1 transportation to the vacuole. Similar experiments using Atg34 mutants showed that His-296 of Atg34 ABD, which corresponds to His-310 of Atg19 ABD, also plays a critical role in Ams1 recognition.

The ABDs in Atg19 and Atg34 have a β-sandwich fold that is observed in a variety of immunoglobulins and immunoglobulin-like domains responsible for recognizing various proteins. Because antibodies generally recognize antigens using the hypervariable loops from both the VH and VL regions, their manner of antigen binding should differ from that of monomeric ABDs with Ams1. Interestingly, however, the ABD-Ams1 interaction resembles that observed between camelid antibody fragments and their antigens, as camelid antibodies lack a light chain and function as a monomer where hypervariable loops of the VH are responsible for antigen binding (Muyldermans, 2001). It also mimics the interaction of monobodies (artificially designed proteins that use a fibronectin type III domain as a scaffold) and their targets, as monobodies interact with their targets using similar loops in a monomeric immunoglobulin fold. Camelid antibody fragments and monobodies interact with their target proteins using loops clustered at one side of their immunoglobulin fold; these loops are topologically equivalent to the BC, DE, and FG loops of the Atg19 and Atg34 ABDs, one of which was shown to be crucial for Ams1 recognition as mentioned above. Therefore, they might recognize Ams1 using these loops in a similar manner with camelid antibodies and monobodies. In order to further elucidate the recognition mechanism of Ams1 by the ABD, structural determination of the Ams1-ABD complex by X-ray

our recent study, we identified the Ams1 binding domain (ABD) in Atg19 and Atg34 by limited proteolysis of full-length Atg19, an *in vitro* pull-down assay as well as sequence alignment (Watanabe et al., 2010). In *atg19*Δ cells expressing Atg19ΔABD, Ams1 transport to the vacuole was inhibited, suggesting that the Atg19 ABD is required for Ams1 transport to the vacuole through the Cvt pathway. In such cells, prApe1 transport to the vacuole is the normal process. These results indicate that the Atg19 ABD is specifically responsible for the

The Atg19 and Atg34 ABD structures were determined in solution using NMR spectroscopy (Figure 1A and B) (Watanabe et al., 2010). Both ABDs comprise eight β-strands (A-H), of which A, B, E, and H form an antiparallel β-sheet; the surface of this sheet faces a second antiparallel β-sheet comprising C, D, F, and G, thus forming a typical immunoglobulin-like β-sandwich fold. The Atg19 and Atg34 ABD structures are similar to each other with a root mean square difference of 2.1 Å for 102 residues (Z-score calculated by the Dalilite program (Holm and Park, 2000) is 12.8). There are relatively large structural differences between the Atg19 and Atg34 ABDs in the loops located at the bottom of the immunoglobulin fold (the loop connecting strands A and B (AB loop), CD, EF, and GH loops). In contrast, the loops located at the top of the immunoglobulin fold (the BC, DE, and FG loops) have a similar conformation. Furthermore, the residues comprising the top loops, especially those of the DE loop, are more strongly conserved between the Atg19 and Atg34 ABDs than those comprising the bottom loops (Figure 1). In the DE loop, His-310/296, Glu-311/297, Ile-314/300, and Lys-315/301 of Atg19/Atg34 are exposed. Among these exposed residues, His-310/296 and/or Glu-311/297 of the Atg19/Atg34 ABD are essential for Ams1 recognition. Further analysis showed that in *atg19*Δ*atg34*Δ cells expressing Atg19H310A (substitution of His-310 with alanine) but not Atg19E311A, transport of Ams1-GFP to the vacuole under autophagy-inducing conditions is inhibited. This indicates that the conserved His residue in the DE loop of the Atg19 ABD plays a critical role in Ams1 recognition and that Ams1 binding of the Atg19 ABD is essential for Ams1 transportation to the vacuole. Similar experiments using Atg34 mutants showed that His-296 of Atg34 ABD, which

transport of Ams1, but not prApe1, to the vacuole through the Cvt pathway.

corresponds to His-310 of Atg19 ABD, also plays a critical role in Ams1 recognition.

The ABDs in Atg19 and Atg34 have a β-sandwich fold that is observed in a variety of immunoglobulins and immunoglobulin-like domains responsible for recognizing various proteins. Because antibodies generally recognize antigens using the hypervariable loops from both the VH and VL regions, their manner of antigen binding should differ from that of monomeric ABDs with Ams1. Interestingly, however, the ABD-Ams1 interaction resembles that observed between camelid antibody fragments and their antigens, as camelid antibodies lack a light chain and function as a monomer where hypervariable loops of the VH are responsible for antigen binding (Muyldermans, 2001). It also mimics the interaction of monobodies (artificially designed proteins that use a fibronectin type III domain as a scaffold) and their targets, as monobodies interact with their targets using similar loops in a monomeric immunoglobulin fold. Camelid antibody fragments and monobodies interact with their target proteins using loops clustered at one side of their immunoglobulin fold; these loops are topologically equivalent to the BC, DE, and FG loops of the Atg19 and Atg34 ABDs, one of which was shown to be crucial for Ams1 recognition as mentioned above. Therefore, they might recognize Ams1 using these loops in a similar manner with camelid antibodies and monobodies. In order to further elucidate the recognition mechanism of Ams1 by the ABD, structural determination of the Ams1-ABD complex by X-ray crystallography is needed. We have already succeeded in overexpressing *S. cerevisiae* Ams1 in *Pichia pastoris* and purifying it on a large scale (Watanabe et al., 2009). Crystallization and structural determination of the Ams1-ABD complex are now in progress.

Fig. 1. Solution structures of Atg19 and Atg34 ABDs. (A), (B) Ribbon diagrams of the Atg19 ABD and Atg34 ABD structures, respectively. Strands are colored *light blue* and labeled. Loop residues conserved between Atg19 and Atg34 are colored *red*. *Left* and *right* are related by a 180° rotation along the *vertical axis*. (C) Sequence alignment between Atg19 and Atg34 ABDs. *Gaps* are introduced to maximize the similarity. Conserved or type-conserved residues are colored *red*. Secondary structure elements of the Atg19 and Atg34 ABDs are shown above and below the sequence, respectively.
