Autophagy is a non-selective bulk degradation process in which isolation membranes enclose a portion of cytoplasm to form double-membrane vesicles, called autophagosomes, and deliver their inner constituents to the lytic compartments. Recent studies have also shed light on another mode of autophagy that selectively degrades various targets. Yeast Atg8 and its mammalian homologue LC3 are ubiquitin-like modifiers that are localized on isolation membranes and play crucial roles in the formation of autophagosomes. These proteins are also involved in selective incorporation of specific cargo molecules into autophagosomes, in which Atg8 and LC3 interact with Atg19 and p62, receptor proteins for vacuolar enzymes and disease-related protein aggregates, respectively. Using X-ray crystallography and NMR, we herein report the structural basis for Atg8–Atg19 and LC3–p62 interactions. Remarkably, Atg8 and LC3 were shown to interact with Atg19 and p62, respectively, in a quite similar manner: they recognized the side-chains of Trp and Leu in a four-amino acid motif, WXXL, in Atg19 and p62 using hydrophobic pockets conserved among Atg8 homologues. Together with mutational analyses, our results show the fundamental mechanism that allows Atg8 homologues, in association with WXXL-containing proteins, to capture specific cargo molecules, thereby endowing isolation membranes and/or their assembly machineries with target selectivity.
Autophagy is an intracellular bulk degradation system conserved among eukaryotes from yeast to mammals. In autophagy, isolation membranes enclose a portion of cytoplasm to form double-membrane vesicles, called autophagosomes, and deliver their inner constituents to the lysosome/vacuole (Baba et al. 1994). Yeast Atg8 and its mammalian homologue LC3 are conjugated to phosphatidylethanolamine in a similar manner with ubiquitination (Ohsumi 2001) and localize to isolation membranes, thereby play crucial roles in the formation of autophagosomes (Nakatogawa et al. 2007). Although autophagy is in principle a non-selective degradation process, recent studies have also shed light on another mode of autophagy that selectively degrades aggregated proteins, surplus or damaged organelles and even invasive bacterial cells (Mizushima 2007).
In yeast Saccharomyces cerevisiae, aminopeptidase I (Ape1) is selectively and constitutively transported into the vacuole via an autophagy-like process called the cytoplasm-to-vacuole targeting (Cvt) pathway (Baba et al. 1997; Scott et al. 1997). Although it is a biosynthetic process, the Cvt pathway has been regarded as a well-studied model of selective autophagy. Immediately after translated in the cytoplasm, Ape1 self-assembles into a huge complex (Kim et al. 1997) and interacts with the receptor protein Atg19 (Scott et al. 2001). Atg19 further interacts with Atg8 (Shintani et al. 2002), thus Atg19 recruits membrane-forming machineries to the Ape1 complex. This pathway is similar to the p62-mediated autophagic degradation of protein aggregates, although Atg19 and p62 are unrelated to each other over their entire sequences.
Here, we report the structures of LC3–p62 peptide and Atg8–Atg19 peptide complexes by NMR and X-ray crystallography, respectively. Remarkably, LC3 and Atg8 were shown to interact with p62 and Atg19, respectively, in a quite similar manner, which could be a fundamental mechanism that endows autophagy with target selectivity.
Structural basis of LC3–p62 interaction
To determine the minimum region of human p62 required for interaction with LC3, we analyzed their interaction by NMR spectroscopy, which showed that residues 335–345 of p62 are mainly involved in the interaction with LC3 (Fig. S1 in the Supporting Information). We then determined the solution structure of the complex between LC3 and a peptide corresponding to residues 332–347 of p62 by NMR spectroscopy (Fig. 1A, Supporting Information Table S1, Supporting Information Fig. S2). The overall structure of LC3 is similar to the crystal structure of free LC3 (Sugawara et al. 2004), being composed of an N-terminal domain containing two α-helices (α1, α2) and a C-terminal ubiquitin-like domain. Among the 16 residues (332–347) of the p62 peptide, the C-terminal 13 residues (335–347) adopt a converged structure and interact with LC3. The WTHL moiety of p62 peptide adopts an extended β conformation and forms an intermolecular parallel β-sheet with the β2 of LC3. The side-chain of Trp338 of WTHL is bound deeply into the hydrophobic pocket formed at the interface between the N-terminal domain (α2) and the ubiquitin-like domain, whereas the side-chain of Leu341 of WTHL is bound to the hydrophobic pocket on the ubiquitin-like domain. The former pocket is comprised of the side-chains of Asp19, Ile23, Pro32, Ile34, Lys51, Leu53 and Phe108, whereas the latter pocket is comprised of the side-chains of Phe52, Val54, Pro55, Leu63, Ile66 and Ile 67 (Fig. 1B). All of these residues are well conserved among Atg8 homologues. In vitro pull-down assay showed that Trp338 and Leu341 of p62 are essential for the interaction with LC3 (Fig. 1C). Trp338 is strictly conserved, and Leu 341 is type-conserved among p62 homologues (Pankiv et al. 2007). Complex formation buries approximately 1500 Å2 of the exposed surface, of which approximately 820 Å2 is attributed to the WTHL moiety of the peptide. p62 has three consecutive aspartate residues (DDD) N-terminal to the WTHL sequence, which were shown to be responsible for the interaction with LC3 (Pankiv et al. 2007), although the contribution of each aspartate residue was limited (Fig. 1C). Recently, crystal structure of LC3 complexed with a peptide corresponding to residues 334–344 of murine p62 was reported, which also showed similar interactions between LC3 and the WTHL moiety (Ichimura et al. 2008).
Structural basis of Atg8–Atg19 interaction
It was known that the C-terminal six residues of Atg19 are important for the interaction with Atg8 (Shintani et al. 2002). To determine the minimum region of Atg19 required for the interaction with Atg8, we studied their interaction by NMR spectroscopy, which showed that the C-terminal 8 residues are mainly involved in the interaction (Supporting Information Fig. S3A). Furthermore, in vitro pull-down assay using Atg19 mutants showed that both Trp412 and Leu415 residues are required for the interaction with Atg8 (Fig. 2C). In order to further elucidate the interaction between Atg8 and Atg19, we determined the crystal structure of Atg8 complexed with a tetra peptide, WEEL (Fig. 2A, Supporting Information Table S2, Supporting Information Fig. S4). The structure of Atg8 in the complex is similar to that of LC3 (Sugawara et al. 2004). Surprisingly, the interaction between Atg8 and the WEEL peptide is quite similar to that between LC3 and the WTHL moiety of p62 (Figs 1B and 2B). WEEL adopts an extended β conformation and forms an intermolecular parallel β-sheet with β2 of Atg8. The side-chains of both Trp412 and Leu415 of WEEL are bound deeply into two hydrophobic pockets on Atg8 in a manner quite similar to that of the WTHL–LC3 interaction. In addition to these hydrophobic interactions, the side-chains of the Glu413 and Glu414 of WEEL form ionic interactions with those of Arg67 and Arg28 of Atg8, respectively (Fig. 2B). In vitro pull-down and yeast two-hybrid assays using Atg8 mutants showed that both Arg28 and Arg67 of Atg8 are indeed involved in Atg19 binding (Supporting Information Figs S5 and S6). The DDD sequence N-terminal to the WTHL sequence of p62, which is essential for the interaction with LC3, is not conserved in Atg19. Furthermore, two arginine residues (Arg10 and Arg11) of LC3 that interact with the DDD sequence of p62 (Ichimura et al. 2008) are not conserved in Atg8. Complex formation buries approximately 800 Å2 of the exposed surface, which is compatible with the buried surface of LC3 by WTHL. It should be noted here that the crystal structure of the complex is consistent with the results of the chemical shift perturbation study by NMR (Supporting Information Fig. S3B), supporting the hypothesis that the crystal structure of the complex is also maintained in solution.
WXXL sequence of Atg19 is crucial for the Cvt pathway in vivo
Structural studies and in vitro assays showed that the WXXL sequence in both p62 and Atg19 is crucial for the interaction with LC3 and Atg8, respectively. Both Trp and Leu residues of the WXXL sequence in p62 were shown to be crucial for the degradation of p62 by autophagy (Ichimura et al. 2008). Furthermore, p62-deficient mouse cells expressing p62 with mutations at these residues were shown to accumulate ubiquitin-positive inclusion bodies (Ichimura et al. 2008). We then examined whether the WEEL sequence in Atg19 was essential to the Cvt pathway in vivo (Fig. 3). When the wild type and mutant forms of Atg19 bearing an alanine substitution in the WEEL sequence were expressed from single-copy plasmids in yeast cells lacking endogenous ATG19, they accumulated at a similar level (Fig. 3A, lower panel). However, the vacuolar transport of Ape1 was found to be defective in the Atg19 mutants (Fig. 3A, upper panel). The extent of their defects was in good agreement with their affinity to Atg8 observed in vitro (Fig. 2C). In addition, we also showed that the simultaneous replacement of Trp412 and Leu415 resulted in more severe defects (Fig. 3A, upper panel). These results suggest that the interaction between Atg19 and Atg8 through the WEEL sequence is required for the selective transport of Ape1.
Residues of Atg8 crucial for the Cvt pathway but not for autophagy
We also examined the significance of the WEEL-binding pockets on Atg8 in vivo. For this purpose, it was necessary to obtain mutations that abolished the interaction with Atg19 without affecting the Atg8 functions required for vesicle formation. Most mutations around the WEEL-binding pockets that reduced Atg19 binding ability also impaired the functions associated with vesicle formation (Nakatogawa et al. 2007) (Supporting Information Fig. S6 and data not shown). However, alanine replacement of Pro52 and Arg67, which are also located around the WEEL-binding pockets (Fig. 2B), significantly decreased the interaction with Atg19 and the efficiency of Ape1 transport (Fig. 3B, upper panel and Supporting Information Fig. S6) without affecting non-selective autophagy (Fig. 3C). The amounts of these mutant proteins and their lipidation were similar to those of the wild-type protein (Fig. 3B, lower panel). These results clearly show that Atg8 has a specific function in the recognition of the Ape1 complex by interacting with the WEEL sequence of Atg19, and that this function of Atg8 is separable from that involved in vesicle formation.
In the Cvt pathway, Atg proteins involved in vesicle formation are assembled to the Ape1 complex (Nair & Klionsky 2005). We can examine this assembly of the Atg proteins by observing the perivacuolar dot formation of their GFP-fused variants using fluorescence microscopy (Suzuki et al. 2001) (Fig. 3D). This fluorescently-labeled dot is called the preautophagosomal structure (PAS), in which vesicle formation and the incorporation of the Ape1 complex into the vesicle are considered to be initiated (Suzuki et al. 2001). The PAS is formed irrespective of nutrient conditions; however, its composition seems to vary depending on nutrient conditions (Kawamata et al. 2008). We showed that GFP-fused Atg8 bearing the Arg67Ala mutation failed to localize to the PAS in cells growing in nutrient rich media, whereas the same mutant could form the dots under nitrogen starvation conditions (Fig. 3D). This is consistent with the observation obtained using a C-terminal deletion mutant of Atg19 (Shintani et al. 2002), and suggests that under nutrient-rich conditions, the interaction between Atg8 and the WEEL sequence of Atg19 is necessary for the recruitment of Atg8 to the Ape1 complex.
Our present study showed that the Trp and Leu residues of the WTHL sequence in p62 and the WEEL sequence in Atg19 bind to the conserved hydrophobic pockets on LC3 and Atg8 in a similar manner. Intriguingly, these binding pockets are located at the opposite surface to the hydrophobic patch that is conserved among ubiquitin-like modifiers as a binding site for various target molecules (Sugawara et al. 2004). For the construction of the binding pockets, the N-terminal helical domain unique to Atg8 homologues plays a pivotal role. Furthermore, residues constituting these binding pockets are highly conserved among Atg8 homologues but not among other ubiquitin-like proteins (Sugawara et al. 2004). All these data suggest that the WXXL sequence is a common motif that is specifically recognized by Atg8 and its homologues. Another ubiquitin-like modifier SUMO also binds short peptides called SUMO interacting/binding motifs by forming an intermolecular β-sheet using the second β-strand (Kerscher 2007); however, SUMO cannot bind the WXXL motif because it lacks N-terminal helices that are essential for the interaction with a Trp residue in the WXXL motif.
In both complexes, the WXXL motif adopts an extended β-conformation, which is considered to be essential for its interaction with Atg8/LC3. Therefore, in order to be recognized by Atg8/LC3, the WXXL motif needs to be presented on the exposed region with a flexible conformation. In addition, owing to the basic nature around the WXXL-binding pockets, XX part and/or adjacent regions of the WXXL motif need to contain acidic residues. In fact, acidic residues are abundant at the region adjacent to the WXXL motif both in Atg19 and p62. Therefore, although the sequence WXXL frequently appears in various proteins, the number of proteins satisfying these specific requirements is considered to be relatively restricted.
The present study showed an unexpected, common mechanism by which Atg8 homologues recognize the WXXL motif in receptor proteins that are totally unrelated over their whole sequence as well as in their targets. Whereas Atg8 homologues, which localize on the isolation membrane, can potentially serve as a universal connector, the WXXL motif can work not only in receptor proteins but also in cargos themselves. Therefore, it is possible for the interaction between Atg8 homologues and the WXXL motif to work broadly in various types of selective autophagy.
Protein expression and purification
Plasmid construction, expression and purification of LC3 were performed as described previously (Sugawara et al. 2003). To construct Escherichia coli expression plasmids encoding Atg19 (395–415), p62 (321–341), p62 (332–347) and the C-terminal glycine-exposed form of Atg8, the appropriate genes were amplified by polymerase chain reaction and cloned into pGEX-6P-1 (GE Healthcare). The ATG8 gene was also cloned into pHT1 (Matsushita et al. 2006). Mutations leading to the specific amino acid substitutions were introduced by PCR-mediated site-directed mutagenesis. All the constructs were sequenced to confirm their identities and transformed into E. coli BL21 (DE3). Atg8 for crystallization was expressed in E. coli with a hexa-histidine tag at its N-terminus. After purification using a Ni-NTA column (QIAGEN), hexa-hisitidne tag was excised form the protein with TEV protease (GE Healthcare). Further purification was performed using HiTrap CM cation-exchange column (GE Healthcare). Atg8 for NMR spectroscopy was expressed in E. coli with a glutathione S-transferase (GST) tag at its N-terminus. After purification with a glutathione-Sepharose 4B column (GE Healthcare), GST was excised from the protein with PreScission protease (GE Healthcare). Further purification was performed using Resource S cation-exchange column (GE Healthcare). Peptides derived from Atg19 (395–415), p62 (321–341) and p62 (332–347) for NMR spectroscopy and pull-down assay were expressed in E. coli with a GST tag at their N-terminus. After purification with a glutathione-Sepharose 4B column, the GST-fused peptides were applied to HiTrap Desalting column (GE Healthcare) for pull-down assay. For NMR spectroscopy, GST was excised from the GST-fused peptides with PreScission protease followed by purification using Superdex75 gel-filtration chromatography (GE Healthcare). WEEL peptide for crystallization with Atg8 was purchased from Sigma. 15N-labeled and 15N-13C-double labeled proteins and peptides for NMR spectroscopy were prepared by culturing E. coli in M9 medium containing 15NH4Cl and 13C6-glucose as sole nitrogen and carbon sources, respectively.
Crystals of the Atg8–WEEL complex were obtained by the sitting drop vapor diffusion method at 293 K. A 0.5-µL of protein solution was mixed with an equal volume of reservoir solution consisting of 2.0 M ammonium sulfate and 0.1 M sodium citrate pH 4.2. Crystals belong to the orthorhombic space group P212121 with unit-cell dimensions a = 44.15, b = 104.45, c = 113.02 Å. The asymmetric unit contains four Atg8–WEEL complexes with an estimated solvent content of 45%. For data collection, crystals were soaked into the reservoir solution supplemented with 20% glycerol, flash-cooled and kept in a stream of nitrogen gas at 95 K during data collection. Diffraction data were collected on the ADSC Quantum 210 charge-coupled device detector using beamline NW12A, KEK, Japan, at a wavelength of 1.00 Å. Diffraction data were processed using the HKL2000 program suite (Otwinowski & Minor 1997). Molecular replacement was performed using the program molrep (Vagin & Teplyakov 1997) in the CCP4 software suite (Collaborative Computational Project, Number 4, 1994). The crystal structure of GABARAP was used as a search model (Knight et al. 2002) (PDB code 1GNU). Manual building and modification was performed with the molecular modeling program COOT (Emsley & Cowtan 2004), followed by iterative rounds of refinement using the cns program (Brunger et al. 1998). Atomic coordinates and structure factors (code 2ZPN) have been deposited in the Protein Data Bank, Research Collaboratory for Structural Bioinformatics, Rutgers University, New Brunswick, NJ <http://www.rcsb.org/>.
All the NMR measurements were carried out at 25 °C. The LC3 sample complexed with the p62 peptide (residues 321–342 and 332–347) was prepared in 25 mm sodium phosphate buffer (pH 7.0), 100 mm NaCl (pH7.0) in the presence of 0.02 mm NaN3. The Atg8 sample complexed with the Atg19 peptide (395–415) was also prepared in the same buffer solution.
Two- and three-dimensional NMR experiments were performed on Varian UNITY inova spectrometers operating at 800 and 600 MHz. Spectra were processed using NMRPipe (Delaglio et al. 1995) and data analysis was performed using the Sparky program <http://www.cgl.ucsf.edu/home/sparky/>.
For structural determination of LC3 complexed with the p62 peptide (332–347), 13C–15N labeled LC3 with non-labeled p62 peptide at 1.5 molar equivalent and 13C–15N labeled p62 peptide with non-labeled LC3 at 1.5 molar equivalent were used. Interproton distance restraints for structural calculation were obtained from 13C-edited NOESY-HSQC and 15N-edited NOESY-HSQC spectra using a 100 ms mixing time. The spectra were analyzed using the Sparky program and the structure was calculated using the cyana 2.1 software package (Herrmann et al. 2002). The atomic coordinates (code 2K6Q) have been deposited in the Protein Data Bank.
GST pull-down assays
In the pull-down assay between Atg8 and Atg19, the purified GST-fused Atg19 (395–415) mutants were incubated with glutathione-Sepharose 4B beads for 10 min at 4 °C. After washing the beads with phosphate-buffered saline (PBS), Atg8 was added to the beads and they were further incubated for 10 min at 4 °C. After washing the beads three times with PBS, proteins were eluted with 10 mm glutathione in 50 mm Tris–HCl buffer (pH 8.0). In the pull-down assay between LC3 and p62, the purified GST-fused p62 (332–347) mutants and LC3 as well as glutathione-Sepharose 4B beads were simultaneously incubated for 10 min at 4 °C. After washing the beads three times with PBS, proteins were eluted with 10 mm glutathione in 50 mm Tris–HCl buffer (pH 8.0).
In vivo analyses
For immunoblotting analyses and fluorescence microscopy, the yeast strains KVY5 (SEY6210 (Darsow et al. 1997), Δatg8::HIS3) and GYS730 (SEY6210, Δatg19::kanMX; Suzuki K, pers comm) were used. Plasmids expressing Atg8 mutants and their GFP-fusion proteins were constructed using a QuickChange site-directed mutagenesis kit (Stratagene) with pRS316 carrying the wild-type ATG8 (Kirisako et al. 2000), GFP-ATG8 (Suzuki et al. 2001) genes as described previously (Nakatogawa et al. 2007). The Atg19 plasmids were similarly constructed using pRS316 encoding ATG19 (Suzuki K, pers comm). These plasmids were introduced into the above strains, and they were cultured in SD + CA + AT media (Nakatogawa et al. 2007) at 30 °C. For induction of autophagy, the cells were incubated in SD-N media (Nakatogawa et al. 2007) for 4.5 h. Immunoblotting was carried out as described previously (Nakatogawa et al. 2007). An alkaline phosphatase (ALP) assay was performed as reported previously using KVY54 (SEY6210, Δatg8::HIS3 pho8::Pho8Δ60) (Kirisako et al. 2000) carrying Atg8 plasmids. Pho8Δ60 is a mutant form of the vacuolar ALP that is expressed in the cytoplasm as an enzymatically inactive form. Whereas, it is delivered to the vacuole via autophagy under starvation conditions, where it is activated by a processing enzyme. Therefore, autophagic activity is assessed by measuring the ALP activity in cell lysates using a fluorogenic substrate. The experiments were repeated three times, and the average values are presented with error bars for the standard deviations.
The synchrotron radiation experiments were performed at the beamline NW12A in the KEK, Japan. This work was supported by Grants-in-Aid for Scientific Research on Priority Areas, the National Project on Protein Structural and Functional Analyses, and Targeted Proteins Research Program from the Ministry of Education, Culture, Sports, Science and Technology of Japan. This study was carried out under the NIBB Cooperative Research Program.