A pseudo-receiver domain in Atg32 is required for mitophagy

ABSTRACT

Mitochondria are targeted for degradation by mitophagy, a selective form of autophagy. In Saccharomyces cerevisiae, mitophagy is dependent on the autophagy receptor, Atg32, an outer mitochondrial membrane protein. Once activated, Atg32 recruits the autophagy machinery to mitochondria, facilitating mitochondrial capture in phagophores, the precursors to autophagosomes. However, the mechanism of Atg32 activation remains poorly understood. To investigate this crucial step in mitophagy regulation, we examined the structure of Atg32. We have identified a structured domain in Atg32 that is essential for the initiation of mitophagy, as it is required for the proteolysis of the C-terminal domain of Atg32 and the subsequent recruitment of Atg11. The solution structure of this domain was determined by NMR spectroscopy, revealing that Atg32 contains a previously undescribed pseudo-receiver (PsR) domain. Our data suggests that the PsR domain of Atg32 regulates Atg32 activation and the initiation of mitophagy.

Abbreviations:AIM: Atg8-interacting motif; GFP: green fluorescent protein; LIR: LC3-interacting region; NMR: nuclear magnetic resonance; NOESY: nuclear Overhauser effect spectroscopy; PDB: protein data bank; PsR: pseudo-receiver; RMSD: root-mean-square deviation

Introduction

Damaged or superfluous mitochondria must be degraded to maintain cellular homeostasis and prevent the accumulation of harmful reactive oxygen species [1]. As such, defects in mitochondrial degradation have been implicated in Parkinson's disease, cancer, and metabolic disorders [2–4]. Mitochondrial clearance is mediated by macroautophagy (hereafter, autophagy), a cellular process in which cytosolic material is captured in double membrane vesicles, termed autophagosomes [5]. Completed autophagosomes fuse with the vacuole in yeast or lysosomes in mammals, leading to the degradation of the autophagic cargo. Autophagic cargo can be captured using either a non-selective or a selective mechanism. In selective autophagy, cargos including mitochondria, aggregated proteins, and pathogens, are marked for degradation by autophagy receptors [6]. These receptors recruit the autophagy machinery directly to the cargo to initiate selective autophagy.

In mammals, the selective autophagy of mitochondria, termed mitophagy, utilizes at least 7 different autophagy receptors [7]. Of these, 3 receptors bind directly to ubiquitin moieties on mitochondrial surface proteins, while the remainder are ubiquitin-independent [8]. The ubiquitin-dependent and ubiquitin-independent mitophagy receptors are activated independently in response to different stresses or signals. Significant progress has been made in understanding the molecular mechanisms of ubiquitin-dependent mitophagy. In particular, the determination of several crystal structures for the ubiquitin ligase Parkin, from rat and human, have provided insight into the molecular mechanism of ubiquitination during ubiquitin-dependent mitophagy [9–12]. However, while progress has been made in our understanding of ubiquitin-dependent mitophagy, insight into the molecular mechanisms of ubiquitin-independent mitophagy has been hindered by a lack of structural information for the ubiquitin-independent mitophagy receptors.

In contrast to mammalian systems, Atg32 is the only mitophagy receptor identified in the yeast Saccharomyces cerevisiae [13,14]. Atg32 is a 529 amino acid single-pass outer mitochondrial membrane protein that mediates ubiquitin-independent mitophagy in response to nitrogen deprivation [13]. Recent studies have begun to elucidate the mechanisms of Atg32, including the requirement of both proteolysis and phosphorylation for the induction of mitophagy [15,16]. During starvation, the mitochondrial protease Yme1 mediates cleavage of the C terminus of Atg32, which resides in the intermembrane space [15]. Phosphorylation of Atg32 on serine 114 facilitates recruitment of the selective autophagy scaffolding protein Atg11 to its binding region on Atg32 (residues 51–150) [16]. Finally, Atg32 contains an Atg8-interacting motif (AIM, residues 86–89), which anchors the mitochondria to Atg8, a ubiquitin-like protein conjugated to the phagophore membrane [17].

Despite these advances in our understanding of Atg32, a lack of structural information has hindered our understanding of the molecular mechanism of Atg32 activation. In addition, structure prediction on Atg32 has failed to identify any conserved structural domains, leaving the overall structure of Atg32 a mystery. To begin to shed light on the structure of Atg32, we recently performed construct screening and limited proteolysis to determine if any structured domains were present in Atg32 [18]. Through this approach, we identified a structured domain within the cytosolic region of Atg32 comprising residues 200–341 (Figure 1(a)). In this study, we have demonstrated that this domain is essential for the induction of mitophagy. We have also determined the solution structure of this domain using nuclear magnetic resonance (NMR) spectroscopy. Our data provides the first insight into the overall architecture of Atg32 and sheds light on the molecular mechanisms regulating Atg32-mediated mitophagy.

Figure 1.

Atg32[200–341] is required for mitophagy following nitrogen starvation. (a) A schematic representation of S. cerevisiae Atg32 with the AIM and transmembrane helix (TM) labeled. Residues are numbered. (b) Fluorescence microscopy of atg32Δ yeast expressing GFP-Atg32 or GFP-Atg32Δ200–341 and stained with FM 4–64 to label the vacuole. Cells were grown in SMG or SD-N prior to imaging. Scale bar: 5 µm. (c) Quantification of the microscopy images in (b). SMG is shown as black bars and SD-N is shown as gray bars. Error bars represent the standard deviation of 3 or 4 independent measurements. (d) Fluorescence microscopy of yeast expressing GFP-Atg32 or GFP-Atg32Δ200–341 and stained with MitoTracker. Scale bar: 1 µm. (e) Western blot of GFP-Atg32 and GFP-Atg32Δ200–341 in SMG. Pgk1 was used as a loading control. (f) XX1 cells were transformed with empty vector, Atg32, or Atg32Δ200–341. Phosphatase activity for cells grown in SML is shown as black bars and SD-N is shown as gray bars. Results were normalized to phosphatase activity in cells expressing Atg32 that were starved. Error bars represent the standard deviation of 3 independent measurements. (g) TKYM130 cells were transformed with empty vector, Atg32, or Atg32Δ200–341. Cells were grown in SML or SD-N and OM45-GFP was detected by western blot. Significance was determined by 2-way ANOVA where p < 0.001 was considered to be significant (***). ns, not significant.

Results

Atg32[200–341] is required for nitrogen starvation-induced mitophagy

Both the Atg11 and Atg8 binding sites on Atg32 are required for nitrogen starvation-induced mitophagy [16,17]. However, it is unclear whether the recently identified structured domain, Atg32[200–341], is required for mitophagy. To address this question, we generated a construct lacking this structured domain (Atg32Δ200–341). This construct retains all other previously-identified regions of Atg32, including the proteolysis site, the transmembrane domain, and the Atg11 and Atg8 binding regions. We expressed GFP-tagged full-length Atg32 or GFP-tagged Atg32Δ200–341 in atg32Δ S. cerevisiae, using a galactose-inducible promoter, as previously described for Atg32 [13]. In nutrient-rich conditions, neither GFP-Atg32 nor GFP-Atg32Δ200–341 were observed in the vacuole, indicating that mitophagy is not active. Upon nitrogen starvation, GFP-Atg32 was efficiently targeted to the vacuole, as is expected during mitophagy (Figure 1(b) and (c)). The targeting of GFP-Atg32 to the vacuole was abolished in yeast lacking the essential autophagy protein Atg1 or the selective autophagy scaffold Atg11, demonstrating that the targeting of GFP-Atg32 to the vacuole is dependent on mitophagy (Figure S1). GFP-Atg32Δ200–341 was not observed in the vacuole during nitrogen starvation, suggesting that Atg32[200–341] is required for nitrogen starvation-induced mitophagy (Figure 1(b) and (c)). One possible explanation for this result is that Atg32Δ200–341 is not properly targeted to mitochondria or that it may be less stable than full-length Atg32. However, despite the loss of a large portion of the cytosolic domain, Atg32Δ200–341 is still localized to mitochondria and expressed at comparable levels to full-length Atg32, even at low levels of induction (Figure 1(d) and (e), S2).

To determine whether Atg32[200–341] is required for mitophagy when expressed at endogenous levels, we monitored mitophagy using the mitoPho8Δ60 assay [19]. Pho8 is a vacuolar phosphatase that contains a vacuole targeting sequence and an autoinhibitory region. In the vacuole, the autoinhibitory region of Pho8 is cleaved, generating an active phosphatase. In the mitoPho8Δ60 protein, the vacuolar targeting sequence has been removed and a mitochondrial targeting sequence has been added. As a result, the activity of the mitoPho8Δ60 protein depends on mitochondrial delivery to the vacuole by mitophagy. We performed the mitoPho8Δ60 assay in atg32Δ S. cerevisiae expressing either Atg32 or Atg32Δ200–341 under the control of its endogenous promoter. After 6 h of nitrogen starvation, mitoPho8Δ60 activity was significantly increased in cells expressing Atg32, but not in cells expressing Atg32Δ200–341, demonstrating that Atg32[200–341] is required for mitophagy when expressed at endogenous levels (Figure 1(f)). To confirm this result, we performed the OM45-GFP processing assay [20]. In this assay, the outer mitochondrial membrane protein OM45 is expressed with a C-terminal GFP tag. Mitophagy delivers OM45-GFP to the vacuole, where OM45 is rapidly degraded while GFP is more resistant to proteolysis. As such, the appearance of free GFP provides a means to monitor mitophagy. We again monitored mitophagy in atg32Δ S. cerevisiae expressing OM45-GFP and either Atg32 or Atg32Δ200–341 under the control of its endogenous promoter. After 6 h of nitrogen deprivation, free GFP was detectable in cells expressing Atg32, but not in cells expressing Atg32Δ200–341 (Figure 1(g)).

To determine whether there are any smaller motifs in Atg32[200–341] which confer mitophagy competence, we created a set of 4 smaller deletions and tested their role in mitophagy using the mitoPho8Δ60 assay (Figure S3). None of the smaller deletions showed any increase in mitoPho8Δ60 activity after 6 h of nitrogen starvation, suggesting that Atg32[200–341] functions as a single structured domain.

Atg32[200–341] is required for the initiation of mitophagy

During nitrogen starvation, Atg32 is proteolyzed at its C terminus and subsequently recruits Atg11 to initiate mitophagy [15]. To determine if Atg32[200–341] is required for the initiation of mitophagy, we monitored the proteolysis of Atg32Δ200–341. This assay was performed in atg1Δ S. cerevisiae to facilitate the accumulation of proteolyzed Atg32 [15]. Atg32 proteolysis was observed as early as 30 min after nitrogen starvation, whereas proteolysis of Atg32Δ200–341 was not observed (Figure 2(a)). To determine if Atg32Δ200–341 was still able to recruit Atg11 even in the absence of proteolysis at its C terminus, we monitored the interaction between Atg32 and Atg11 by co-immunoprecipitation. We found that Atg32Δ200–341 had significantly reduced Atg11 binding compared to Atg32 (Figure 2(b)). This data suggests that Atg32[200–341] is required for the induction of mitophagy.

Figure 2.

Atg32[200–341] is required for the initiation of mitophagy. (a) Western blot of GFP-Atg32 and GFP-Atg32Δ200–341 in atg1Δ cells grown in SMG media or in SD-N for 30 or 60 min. The arrowhead indicates the fragment of Atg32 that is produced from the proteolysis of the C-terminal domain. (b) KMB010 cells were transformed with combinations of HA-Atg11, FLAG-Atg32, or FLAG-Atg32Δ200–341 as indicated. Immunoprecipitation of FLAG-Atg32 or FLAG-Atg32Δ200–341 was performed using anti-FLAG resin. Western blots were performed against HA to monitor Atg11 or FLAG to monitor Atg32.

The solution structure of Atg32[200–341]

To gain insight into the molecular mechanism of Atg32[200–341], we determined the solution structure of this domain using NMR spectroscopy (Figure 3(a)). Structure calculation utilized 2422 nuclear Overhauser effect spectroscopy (NOESY)-derived distance constraints, generated using automated NOESY peak picking and assignment in ATNOS/CANDID, and 216 dihedral angle constraints, derived from TALOS-N (Table 1) [21–23]. The first 16 residues of Atg32[200–341] appeared disordered in the structure. Their inherent flexibility was in good agreement with the secondary structure probability calculated using TALOS-N and was further confirmed by a lack of cross peaks in the NOESY spectra for these residues. Excluding residues 200–215, we used 19.2 NOESY-derived distance constraints per residue during structure calculation. For residues 216–336, the bundle of the 20 lowest energy structures converged to a backbone root-mean-square deviation (RMSD) of 0.68 Å, contained standard geometry and a low number of violations, confirming the overall quality of the structure (Table 1 and Figure S4A). The structure of Atg32[200–341] contains 5 β-strands and 6 α-helices with the β-strands forming a single sheet that is surrounded by the 6 α-helices (Figure 3(a)). The β-strands and α-helices are in a near alternating β-sheet α-helix pattern except that 2 α-helices are located between β-strand 1 and 2 (Figure S4B and C).

Figure 3.

Atg32[200–341] is a pseudo-receiver domain. (a) The solution structure of Atg32[200–341] shown as a cartoon representation in cyan. (b) The crystal structure of a response regulator domain from C. psychrerythraea (PDB ID: 3EQZ) shown as a cartoon representation in gray. (c) Overlay of Atg32[200–341] (cyan) and the receiver domain from C. psychrerythraea (gray). (d) Sequence alignment of Atg32[200–341] and 3 receiver domains (PDB IDs: 3EQZ, 3HDV, and 2QXY). Residue numbers are indicated. Secondary structure for Atg32[200–341] is shown in cyan above the alignment and secondary structure for the receiver domains is shown in gray below. Key conserved residues required for receiver domain function are highlighted in yellow. (e) Cartoon representation of Atg32[200–341] with residues that were mutated shown as sticks and labeled. (f) XX1 cells were transformed with Atg32, Atg32Δ200–341, or Atg32 mutants. Phosphatase activity for cells grown in SML is shown as black bars and SD-N is shown as gray bars. Results were normalized to phosphatase activity in cells expressing Atg32 that were starved. Phosphatase activity from cells expressing Atg32 mutants was compared to phosphatase activity from cells expressing Atg32 in SD-N. Error bars represent the standard deviation of 3 or 4 independent measurements. Significance was determined by 2-way ANOVA where p < 0.001 (***) and p < 0.1 (*) were considered to be significant. ns, not significant.

Table 1.

Structural statistics of Atg32[200–341].

	Atg32[200–341]
NMR distance and dihedral constraints
Distance constraints
Total NOE	2422
Intraresidue	543
Interresidue	1879
Sequential (|i-j| = 1)	680
Medium range (1<|i-j|≤ 4)	558
Long range (|i-j|≥ 5)	641
Dihedral angle constraints
ϕ (TALOS)	108
ψ (TALOS)	108
Structure statistics
Violations (mean ± SD)
Distance constraints (Å)	0.035 ± 0.004
Maximum distance constraint violation (Å)	0.632 ± 0.393
Dihedral angle constraints (°)	0.115 ± 0.100
Maximum dihedral angle constraint violation (°)	1.109 ± 0.970
Deviations from idealized geometry (mean ± SD)
Bond lengths (Å)	0.014 ± 0.000
Bond angles (°)	1.68 ± 0.047
Impropers (°)	1.71 ± 0.082
Average pairwise root mean square deviation (Å)
Heavy (216–336)	1.18 ± 0.110
Backbone (216–336)	0.68 ± 0.086

Sequence-based computational predictions using HHpred failed to predict the presence of any conserved domains within Atg32 [24]. To determine whether Atg32[200–341] is part of a structural family, we used the DALI server, which compares an input structure to structures deposited in the Protein Data Bank (PDB) [25]. To our surprise, Atg32[200–341] aligns well, with an RMSD of 2.7 Å over 110 amino acids, to receiver domains found in response regulator proteins (Figure 3(b) and (c)). Response regulator proteins typically function as the second part of bacterial two-component signaling [26]. In these systems, a stress signal activates a histidine kinase, which phosphorylates a conserved aspartic acid residue at the end of β-strand 3 in the receiver domain [26]. The phosphorylated receiver domain activates a neighboring domain in the response regulator protein, often a DNA binding domain that regulates transcription [27]. Receiver domains also coordinate a single metal ion, typically Mg²⁺ or Mn²⁺, which is used for autodephosphorylation of the receiver domain leading to inactivation of the response regulator protein [26]. Whereas two-component signaling is abundant in bacteria, only a few examples have been described in S. cerevisiae and none have been identified in more complex eukaryotes [28].

Despite their structural homology, Atg32[200–341] contains low sequence identity (~10%) to receiver domains. There are also some key structural differences that suggest a distinct function for Atg32[200–341]. Receiver domains contain 5 β-strands and 5 α-helices in a perfect alternating β-strand α-helix pattern. In contrast, Atg32 contains 2 short α-helices in the location of the first α-helix in canonical receiver domains (Figure 3(d)). β-strands 1 and 5 are also significantly shorter than those found in receiver domains. Most importantly, Atg32 lacks the conserved residues that are essential for receiver domain function. These residues include the aspartic acid, which is phosphorylated by a histidine kinase, and the aspartic acids, which coordinate metal binding (Figure 3(d)). These differences reveal that Atg32[200–341] is a member of the pseudo-receiver (PsR) domain family, which are defined by their lack of the conserved aspartic acid that is a target of phosphorylation by histidine kinases [26]. The function of PsR domains is not well understood, but they are thought to function as protein-protein interaction domains.

Distinct surfaces on Atg32[200–341] are required for mitophagy

To identify residues in Atg32[200–341], hereafter Atg32[PsR], which may be important for mitophagy, we examined the conservation of solvent-exposed amino acids using the ConSurf Server (Figure S5A) [29,30]. The majority of the Atg32[PsR] surface has low to average conservation, with only a few highly-conserved residues. To gain additional insight into solvent-exposed residues, we examined the electrostatic surface of the Atg32[PsR]. The surface of the Atg32[PsR] is enriched in basic residues that form patches on the surface of the protein (Figure S5B). One of these basic patches is located on the surface formed by β-strand 5 and α-helices 5 and 6. The corresponding surface on receiver domains frequently serves as a protein interaction site, suggesting that this surface in the Atg32[PsR] may also play a role in protein binding [31].

To determine which surfaces on the Atg32[PsR] might be important for mitophagy, we mutated a series of solvent-exposed residues to alanine. We chose 3 basic patches on distinct surfaces of the Atg32[PsR] that contain at least one lysine or arginine with a high conservation score and mutated the pairs of residues to 2 alanines (Figure 3(e)). As a control, we disrupted the overall fold of the Atg32[PsR] by mutating a residue buried in the hydrophobic core of the protein (F328A). As expected, Atg32[PsR]^F328A expressed well in E. coli but was insoluble, suggesting that the Atg32[PsR] is no longer folded correctly (Figure S6A). Like Atg32Δ200–341, Atg32^F328A localized correctly to mitochondria in yeast cells (Figure S6B). We performed the mitoPho8Δ60 assay as described above, using cells transformed with Atg32, Atg32Δ200–341, or various Atg32 mutants (Figure 3(f)). Mitophagy was almost entirely lost in cells expressing the F328A mutant, and this was confirmed by monitoring GFP-Atg32^F328A localization after 1 h of nitrogen starvation (Figure S6C). In contrast, the surface mutations showed only a partial loss of mitophagy capacity. Atg32^N273A,R274A had a 53% reduction in mitophagy while Atg32^N320A,K321A had a 31% reduction in mitophagy. Combining the N273A,R274A and N320A,K321A mutations did not lead to a further reduction in mitophagy, suggesting that these sites may have redundant roles in mitophagy regulation. The R305A,N306A mutation displayed no significant reduction in mitophagy.

Discussion

We have identified a previously unrecognized structured domain in Atg32 and found that it is required for the induction of mitophagy. This domain most likely acts upstream of previously described Atg32 activation events, since it is required for Atg32 proteolysis and Atg11 recruitment. Our data also shows that the Atg8 and Atg11 binding regions located at the N terminus of Atg32 are not sufficient to induce mitophagy in the absence of a functional Atg32[PsR]. Thus, the Atg11 and Atg8 binding sights are necessary but not sufficient to induce mitophagy [13,17].

We have determined the structure of the Atg32[PsR] and identified it as a member of the PsR domain family, providing the first insight into the overall structure of Atg32. This surprising structural homology suggests that the Atg32[PsR] is a protein-protein interaction domain. In agreement with this possibility, putative protein interaction surfaces on Atg32[PsR] are important for mitophagy. The upstream role of this domain in mitophagy and the structural homology to PsR domains suggest that the Atg32[PsR] is a regulatory domain, which may control the activation of Atg32 and thereby regulate the induction of mitophagy. There are 2 potential mechanisms that would enable the Atg32[PsR] to perform this regulatory role. First, the Atg32[PsR] might undergo a conformational change to activate neighboring domains, as receiver domains commonly do [26]. However, post-translational modifications have not been identified within the Atg32[PsR], leaving the nature and mechanism of any conformational change unknown. As the neighboring domains of the Atg32[PsR] are predicted to be unstructured it is also unclear how this signal would be transmitted to the remainder of Atg32. Second, the Atg32[PsR] could recruit additional factors to mitochondria, which may regulate Atg32 activation or be required for mitophagy. Indeed, PsR domains are thought to function as protein-protein interaction domains. We note that these 2 possibilities are not mutually exclusive, as binding to other proteins may induce a conformational change in the Atg32[PsR]. Identifying binding partners of the Atg32[PsR] will certainly shed more light on this domain’s role in Atg32 activation and mitophagy and may reveal how different cellular pathways communicate stress to trigger mitophagy.

BCL2L13 (BCL2 like 13) has recently been reported as the mammalian homolog of Atg32 as it was able to partially rescue mitophagy in atg32Δ S. cerevisiae [32]. BCL2L13 contains 4 BH domains at its N terminus followed by a predicted disordered linker that contains the mammalian version of the AIM, termed the LC3-interacting region (LIR), and then a single transmembrane helix. Our data demonstrates that Atg32 is similar to BCL2L13 in that it contains a structured domain fused to a disordered region containing the AIM. However, the structured domains of BCL2L13 and Atg32 belong to different families and appear to play different roles in mitophagy. Moreover, the BH domains of BCL2L13 are required for Drp1-independent mitochondrial fragmentation, and overexpression of BCL2L13 induces mitochondrial fragmentation and mitophagy [32]. In contrast, our data demonstrate that overexpression of Atg32 does not induce mitochondrial fragmentation. In addition, the Atg32[PsR] appears to be important for the activation of Atg32, further supporting the idea that these domains play different roles in mitophagy.

Intriguingly, the predicted architecture of Atg32Δ200–341 more closely resembles the mammalian mitophagy receptors BNIP3 (BCL2 interacting protein 3) and BNIP3L/NIX (BCL2 interacting protein 3 like), with an N-terminal disordered sequence that contains the LIR linked to a single transmembrane helix [33,34]. Since the Atg32[PsR] is necessary for the initiation of mitophagy in S. cerevisiae, it will be interesting to determine if an analogous domain might be required for BNIP3 and BNIP3L activation in mammalian cells. If so, this functionality may be provided in the form of a regulatory binding partner that interacts with BNIP3 and BNIP3L, but has yet to be identified.

Materials and methods

Yeast strains and plasmids

GFP-Atg32 and GFP-Atg32∆200–341 were cloned into pXP721, which was a gift from Nancy DaSilva (Addgene, 46,055), using SpeI and XhoI [35]. Atg32 and Atg32∆200–341 containing their endogenous promoter and terminator sequences were cloned into yCPLAC111 [36]. Site-directed mutagenesis was carried out using QuickChange (Agilent, 200,519) or Q5 site-directed mutagenesis (NEB, E0554S). Atg32 and Atg32∆200–341 containing an N-terminal FLAG tag were subcloned into pCu416CUP1 (ATCC, 87,729) for coimmunoprecipitation experiments [37]. All yeast strains used in this study are detailed in Table S1. The mitoPho8Δ60 S. cerevisiae strain (KWY90), OM45-GFP S. cerevisiae strain (TKYM130), and HA-Atg11 plasmid were generous gifts from Dan Klionsky [15,38].

Yeast microscopy

To determine whether mitochondria were delivered to the vacuole, atg1∆, atg11∆, or atg32∆ yeast cells (Invitrogen, 95,400) were transformed with pXP721 (Addgene, 46,055) containing GFP-Atg32, GFP-Atg32∆200–341, or GFP-Atg32^F328A. Cells were grown to log phase in SMD (0.67% [wt:vol] yeast nitrogen base with ammonium sulfate, 2% [wt:vol] glucose, supplemented with the appropriate amino acids). The day before imaging, cells were pelleted and resuspended in SMG (0.67% [wt:vol] yeast nitrogen base with ammonium sulfate, 2% [wt:vol] galactose, supplemented with the appropriate amino acids) to induce expression of Atg32. On the day of imaging, the cells were stained with 1 µg/OD₆₀₀ FM 4–64 (ThermoFisher, T13320) for 30 min, washed twice in media, then incubated in SMG for 1 h to allow the FM 4–64 to traffic to the vacuole. Next, cells were subjected to nitrogen deprivation in SD-N (0.17% [wt:vol] yeast nitrogen base without ammonium sulfate and amino acids, 2% [wt:vol] glucose) for 1 h to induce mitophagy. The cells were then plated on agarose pads (0.5% agarose in SMG or SD-N media) contained in 18 mm hanging drop slides (Fisher Scientific, 12-560A) and imaged immediately. To determine whether GFP-Atg32 variants were correctly localized, atg32∆ cells were transformed with pXP721 (Addgene, 46,055) containing GFP-Atg32, GFP-Atg32∆200–341, or GFP-Atg32^F328A and grown in SMD and SMG as above, then stained with 50 nM MitoTracker (Molecular Probes, M7512) for 30 min on the day of imaging. Cells were then washed and plated on agarose pads as above, and imaged immediately. All cells were imaged on a Zeiss 880 Airyscan with a 100× lens using ZEN-Black software. For each sample, 6–9 fields were acquired and >100 cells counted.

Mitophagy assays

XX1 cells were transformed with the appropriate vectors. Transformed cells were cultured in SMD, pelleted (1,400×g, 10 min) and resuspended in SML (0.17% [wt:vol] yeast nitrogen base without ammonium sulfate and amino acids, 2% [wt:vol] lactate) to an OD₆₀₀ of 0.05. Cells were grown in SML to an OD₆₀₀ of 1.0, pelleted and resuspended in SD-N for 6 h. Twenty OD₆₀₀ of cells were collected for each sample. Cells were resuspended in lysis buffer (20 mM PIPES [Sigma, P6757], pH 6.8, 0.5% [v:v] Triton X-100 [Alfa Aesar, A16046], 50 mM KCl, 100 mM potassium acetate, 10 mM MgSO₄, 10 μM ZnSO₄, 1 mM phenylmethyl sulfonyl fluoride [PMSF; Amresco, 0754–256]), and lysed by bead beating [Biospec Products mini bead beater-16, model # 607] for 30 sec repeated 5 times with 30 sec breaks between runs. Lysate (20 μl) was added to 80 μl of pre-warmed reaction buffer (250 mM Tris, pH 8.5, 0.4% [v:v] Triton X-100, 10 mM MgSO₄, 10 μM ZnSO₄) containing the phosphatase substrate p-nitrophenyl phosphate (pNPP; Sigma, S0942) and incubated in 37°C for 30 min. The reaction was stopped by adding 100 μl of 1 M NaOH. Phosphatase activity was determined by measuring absorbance at 405 nm. The experiment was performed in 3 or 4 independent repeats. The OM45-GFP processing assay was performed as previously described [20]. TKYM130 cells were transformed with the appropriate vectors. Cells were grown in SMD until an OD₆₀₀ of 1.0. Cells were pelleted and resuspended in SML at an OD₆₀₀ of 0.4 and grown until an OD₆₀₀ of 1.0. Cells were pelleted and resuspended in SD-N for 6 h. After starvation, cells were harvested and GFP was detected by western blot.

Immunoprecipitation

KMB010 cells were transformed with the appropriate vectors. Cells were grow to mid-log phase in SMD. Cells were then induced with 50 µM copper [Sigma, 209,198]. After 3 h of induction, cells were pelleted (1,400× g, 10 min) and resuspended in SD-N with 1mM PMSF for 30 min. Cells were harvested and resuspended in ice-cold lysis buffer (1× phosphate-buffered saline [Corning, 21-040-CV] containing 0.2 M sorbitol [Sigma, S7547], 1 mM MgCl₂, 0.1% Triton X-100, 1 mM PMSF, and 1 Roche Complete EDTA-free tablet [11,873,580,001]). Cells were lysed by bead beating and centrifuged at 16,100× g at 4°C. Supernatant was incubated with FLAG M2 magnetic beads (Sigma, M8823), rocking for 2 h at 4°C. After removal of the flow-through, beads were washed 5 times with ice-cold lysis buffer. 1× NuPAGE LDS sample buffer (Invitrogen, NP0007) was added to each tube. Tubes were then incubated at 37°C for 30 min to elute proteins bound to the magnetic beads. Supernatant was analyzed by western blot and probed with anti-FLAG (Sigma, A8592) or anti-HA (Abcam, ab173826) antibodies.

Atg32 proteolysis

atg1Δ cells were transformed with the appropriate vectors. Cells were grown in SMG to mid-log phase, pelleted, and resuspended in SD-N. Samples were taken prior to starvation and after 30 and 60 min in SD-N for western blot analysis using anti-GFP (Santa Cruz Biotechnology, sc-9996).

Western blots

Yeast pellets were resuspended in ice-cold lysis buffer (1× phosphate-buffered saline containing 1 mM PMSF, and 1× NuPAGE LDS sample buffer). Glass beads (Sigma-Aldrich, G8772) were added, and the cells were lysed via bead beating. After centrifugation (16,100×g, 5 min), the supernatant was subjected to western blot analysis. Membranes were probed with anti-GFP primary (Santa Cruz Biotechnology, sc-9996) and goat anti-mouse IgG-HRP secondary (Sigma-Aldrich, A4416) antibodies. Anti-3-phosphoglyceric phosphokinase (Pgk1; Nordic MUbio, NE130/7S) was used as a loading control.

Protein purification

The Atg32[PsR] was expressed and purified as previously described [18]. Briefly, codon-optimized Atg32[200–341] was cloned into the ligation independent cloning vector (1B) which was a gift from Scott Gradia (Addgene, 29,653). The Atg32[PsR] was expressed in BL21 (DE3) star cells (Invitrogen, C601003) in Terrific Broth (Fisher, BP9728-2) to an optical density of approximately 3.0 at 600 nm. Cells were pelleted (3,000× g, 20 min) and resuspended in M9 minimal media comprising 3 g/L ¹⁵N ammonium chloride (Cambridge Isotope Laboratories, NLM-467-25) with or without 10 g/L ¹³C glucose (Cambridge Isotope Laboratories, CLM-1396-25) for uniform labeling. Expression was induced using 1 mM isopropylthio-β-D-galactoside (IPTG; Amresco, 97,061) and cultures were grown for 3 h at 37°C.

Cells were resuspended in 50 mM Tris, pH 8.0, 500 mM NaCl, 5 mM MgCl₂, 0.1% (v:v) Triton X-100 buffer containing protease inhibitors (Roche, 11,836,170,001) and were lysed using a French Press (Thermo Electron, FA-032). Cleared lysates were subjected to purification using TALON resin (Clontech, 635,504) which was preequilibrated with 50 mM Tris, pH 8.0, 500 mM NaCl. Recombinant TEV protease, made in the laboratory, was used to cleave off the hexahistidine tag. Cleaved protein was further purified using a 5 ml HiTrap SP column (GE Healthcare, 17-1152-01) preequilibrated in 50 mM Tris, pH 8.0, 100 mM NaCl. A final purification step was carried out using a HiLoad Superdex 75 PG column (GE Healthcare, 28–9893-33) equilibrated in 20 mM sodium phosphate, pH 6.5, 100 mM NaCl, 0.2 mM tris(2-carboxyethyl)phosphine (Amresco, K831-26).

NMR spectroscopy and structure determination

Chemical shift assignments were previously completed and deposited in the Biological Magnetic Resonance Data Bank (BMRB) under accession number 27,081. All NOESY experiments for structure calculation were performed at 308 K on a Bruker Avance 700 MHz magnet. A 2D ¹H-¹H NOESY was recorded on a 370 μM sample that had been dialyzed in buffer containing 100% D₂O (Cambridge Isotope Laboratories, OLM-4-50 #). A 3D ¹⁵N-resolved NOESY was recorded on a 620 μM ¹⁵N-labeled sample and a 3D ¹³C-resolved NOESY was recorded on 750 μM ¹⁵N,¹³C double-labeled sample. ¹H chemical shifts were externally referenced to 0 ppm methyl resonance of 2,2-dimethyl-2-silapentane-5-sulfonate (DSS), whereas ¹³C and ¹⁵N chemical shifts were indirectly referenced according to the IUPAC recommendations [39]. All NMR spectra were processed using Topspin 3.5 (Bruker). Processed spectra were analyzed using CARA (http://cara.nmr.ch/). UNIO was used for automated NOE peak picking and NOE assignment by ATNOS/CANDID [22,23]. UNIO was also used for initial structure calculation with CYANA v2.1 [40,41]. The 20 lowest energy structures were used for water refinement in CNS v1.3 with the RECOORD scripts [42]. The quality of the final ensemble was verified using NMR-Procheck and the RMSD of these structures was determined using MolMol [43,44]. The final structure was deposited in the PDB under ID 5WLP. All structure figures were generated using the PyMOL Molecular Graphics System, Version 1.8 (Schrödinger, LLC).

Funding Statement

This work was supported by the HHS | NIH | National Institute of General Medical Sciences (NIGMS) [GM113132].

Acknowledgments

We would like to thank Dan Klionsky for reagents. This work was supported by a Prouty Pilot Grant from the Friends of the Norris Cotton Cancer Center and a COBRE award from the National Institutes of Health (GM113132).

Disclosure statement

No potential conflicts of interest were disclosed.

Supplementary material

Supplemental data for this article can be accessed here.