Syntaxin 17 regulates the localization and function of PGAM5 in mitochondrial division and mitophagy

Abstract

PGAM5, a mitochondrial protein phosphatase that is genetically and biochemically linked to PINK1, facilitates mitochondrial division by dephosphorylating the mitochondrial fission factor Drp1. At the onset of mitophagy, PGAM5 is cleaved by PARL, a rhomboid protease that degrades PINK1 in healthy cells, and the cleaved form facilitates the engulfment of damaged mitochondria by autophagosomes by dephosphorylating the mitophagy receptor FUNDC1. Here, we show that the function and localization of PGAM5 are regulated by syntaxin 17 (Stx17), a mitochondria‐associated membrane/mitochondria protein implicated in mitochondrial dynamics in fed cells and autophagy in starved cells. In healthy cells, loss of Stx17 causes PGAM5 aggregation within mitochondria and thereby failure of the dephosphorylation of Drp1, leading to mitochondrial elongation. In Parkin‐mediated mitophagy, Stx17 is prerequisite for PGAM5 to interact with FUNDC1. Our results reveal that the Stx17‐PGAM5 axis plays pivotal roles in mitochondrial division and PINK1/Parkin‐mediated mitophagy.

Introduction

Mitochondria are double membrane‐bound and highly dynamic organelles that fuse, divide, and move along the cytoskeleton to form a mitochondrial network in response to cellular energy status (Labbé et al, 2014; Mishra & Chan, 2016; Wai & Langer, 2016). Mitochondrial dynamics is mediated by GTPases: Drp1 promotes mitochondrial fission, and Opa1 and mitofusins (Mfns) catalyze the fusion of the inner and outer membranes, respectively (Labbé et al, 2014; Mishra & Chan, 2016; Wai & Langer, 2016).

Mitochondria in coordination with the endoplasmic reticulum (ER) play key roles in cellular metabolism and cell survival/death. The close proximity between mitochondria and a subdomain of the ER, termed the mitochondria‐associated membrane (MAM; Paillusson et al, 2016; Herrera‐Cruz & Simmen, 2017), allows efficient Ca²⁺ transfer from the ER to mitochondria. Uptake of Ca²⁺ into the matrix of mitochondria activates the tricarboxylic acid cycle, leading to the production of ATP, whereas prolonged excess Ca²⁺ uptake results in the release of apoptosis‐inducing factors from mitochondria (Herrera‐Cruz & Simmen, 2017; Marchi et al, 2017). Mitochondria generate ATP via the proton gradient across the mitochondrial inner membrane that is formed as a consequence of the passage of electrons through the electron transport chain. Electrons that have leaked from the electron transport chain produce reactive oxygen species (ROS; Murphy, 2009). While low levels of ROS function as signaling molecules (Holmström & Finkel, 2014), excess ROS production causes damage to cellular components, including the mitochondria themselves. Damaged parts of mitochondria are segregated from healthy parts by fission and cleared through a specific mode of autophagy, called mitophagy (Hamacher‐Brady & Brady, 2016).

PGAM5 is a member of the phosphoglycerate mutase family, but exhibits protein phosphatase activity toward Ser/Thr (Takeda et al, 2009) and His (Panda et al, 2016). Recent studies highlighted the importance of PGAM5 in a variety of cellular processes, including mitochondrial dynamics (Wilkins et al, 2014; Lavie et al, 2017; O'Mealey et al, 2017), mitophagy (Chen et al, 2014; Wu et al, 2014; Hawk et al, 2018), programmed cell death (Wang et al, 2012; Lin et al, 2013; Zhuang et al, 2013; Xu et al, 2015; He et al, 2017), the WNT/β‐catenin pathway (Rauschenberger et al, 2017; Bernkopf et al, 2018), immune responses (Kang et al, 2015; Moriwaki et al, 2016), and longevity (Borch Jensen et al, 2017). Several lines of evidence suggest the close relationship of PGAM5 with Parkinson's disease. PGAM5 deficiency suppresses the loss‐of‐function mutation of the mitochondrial protein kinase PINK1 in Drosophila (Imai et al, 2010) and causes a Parkinson's‐like movement disorder and resistance to metabolic stress in mammals (Lu et al, 2014, 2016; Sekine et al, 2016). Moreover, PINK1 and PGAM5 are alternative substrates for the mitochondrial rhomboid protease PARL: while PARL complexed with SLP2 (Wai et al, 2016) continuously cleaves the intramembrane domain of PINK1 in healthy cells (Jin et al, 2010), upon mitochondrial injury PARL cleaves PGAM5 instead of PINK1 (Sekine et al, 2012), leading to the expression of PINK1 on the mitochondrial surface (Jin et al, 2010). In Parkin‐expressing cells, PINK1 appearance on the mitochondrial surface facilitates the recruitment of cytosolic Parkin to dysfunctional mitochondria, followed by ubiquitination and engulfment of the mitochondria for mitophagy (Pickrell & Youle, 2015; Hattori et al, 2017).

Syntaxin 17 (Stx17) is a SNARE protein located in the MAM and mitochondria, and functions not only as a fusion protein but also as a scaffold at the ER–mitochondria interface (Arasaki et al, 2015; Tagaya & Arasaki, 2017). Our previous results suggest that Stx17 binds to microtubules through MAP1B‐LC1 phosphorylated at Thr217 (Arasaki et al, 2018) and promotes mitochondrial fission in fed cells by preventing Drp1 from binding to Rab32 (Arasaki et al, 2015), a mitochondrial GTPase serving as an anchor for protein kinase A that inactivates Drp1 through phosphorylation at Ser637 (Bui et al, 2010). Starvation causes the dephosphorylation of MAP1B‐LC1 at Thr217 and allows Stx17 to dissociate from Drp1 (Arasaki et al, 2018) and associate with Atg14L (Hamasaki et al, 2013), a subunit of the Vps34‐containing class III phosphatidylinositol (PI) 3‐kinase (Matsunaga et al, 2010). This binding induces the attachment of this kinase to the MAM, promoting the formation of phosphatidylinositol 3‐phosphate (PI3P), followed by the expansion and formation of autophagosomes (Axe et al, 2008). The function of Stx17 as a scaffold requires not the SNARE motif, but the C‐terminal hydrophobic domain (CHD) separated by Lys254 (Arasaki et al, 2015; Fig 1A). Stx17 and Atg14L function not only at the early stage of autophagy, but also at the late stage, that is, the fusion of autophagosomes with lysosomes (Itakura et al, 2012; Diao et al, 2015).

Figure 1. Stx17 binds to PGAM5.

1. Schematic representation of Stx17 and PGAM5.
2. 293T cells were transfected with a plasmid encoding FLAG‐Stx17 wild type (WT) or the indicated constructs. At 24 h after transfection, cell lysates were immunoprecipitated (IP) with anti‐FLAG M2 beads, and analyzed by IB using antibodies against PGAM5 and FLAG. Five percent of lysates was analyzed as input.
3. HeLa cells stably expressing FLAG‐Stx17 WT or the K254C mutant were fixed and subjected to PLA using antibodies against FLAG and PGAM5. Scale bar, 5 μm. Values are means ± SEM (n = 3). ***P < 0.001 as compared with WT (paired Student's t‐test).
4. MBP or the MBP‐Stx17 constructs attached to amylose resin were mixed with GST‐PGAM5, and the proteins bound to the resin were separated by SDS–PAGE and blotted onto PVDF membranes. The blots were detected by an anti‐GST antibody (upper panels) or stained with Coomassie Brilliant Blue R‐250 (lower panels). Ten percent of the proteins used for each experiment was analyzed as input. Asterisks and double asterisk represent possible MBP dimers and degradation products, respectively.
5. 293T cells were cotransfected with plasmids encoding FLAG‐Stx17 WT and the indicated PGAM5‐GFP constructs and analyzed as described in (B).
6. 293T cells were cotransfected with plasmids encoding FLAG‐tagged Stx17 or Stx18 and the C‐terminally GFP‐tagged transmembrane domain of PGAM5 (amino acids 1–35) constructs and analyzed as described in (B).

Source data are available online for this figure.

To explore the multiple functions of Stx17, we searched for its binding proteins by means of immunoprecipitation and identified PGAM5. In this study, we show that PGAM5‐mediated Drp1 dephosphorylation is regulated by Stx17 in healthy cells and that the PGAM5‐Stx17 interaction is essential for the function of the mitophagy receptor FUNDC1 (Wei et al, 2015), which is dephosphorylated by PGAM5 at the onset of mitophagy (Chen et al, 2014; Wu et al, 2014).

Results

Stx17 binds to PGAM5

Among the proteins co‐immunoprecipitated with FLAG‐Stx17, we focused on PGAM5, a protein phosphatase that was reported to catalyze Drp1 dephosphorylation under certain conditions such as cell death stimuli (Wang et al, 2012; Lin et al, 2013; Xu et al, 2015). We first determined which regions of Stx17 are responsible for the interaction with PGAM5. Deletion of the C‐terminal cytoplasmic tail had no effect on the binding to PGAM5 (Fig 1B, lane 7), and the CHD plus the C‐terminal cytoplasmic tail retained the ability to bind to PGAM5 (lane 8), suggesting that the CHD is responsible for the association with PGAM5. Notably, replacement of Lys254, which divides the CHD into two segments, markedly reduced the binding to PGAM5 (Fig 1B, lane 6). The requirement of Lys254 of Stx17 for PGAM5 association was corroborated by in situ proximity ligation assay (PLA) (Fig 1C). Using recombinant proteins, the direct binding between Stx17 and PGAM5 was verified (Fig 1D, lane 6). The results also confirmed that the CHD of Stx17 is required for binding to PGAM5 (lanes 7 and 8). Our previous study demonstrated that Stx17 also interacts with Drp1 (Arasaki et al, 2015). To define the Drp1‐binding site on Stx17, we used a GTPase‐deficient mutant of Drp1, i.e., Drp1 K38A, in which Lys38 was replaced by Ala (Smirnova et al, 1998). This mutant was found to co‐immunoprecipitate with Stx17 (Arasaki et al, 2015). Immunoprecipitation and pull‐down experiments demonstrated that although the CHD and the following C‐terminal cytoplasmic tail of Stx17 can bind to Drp1 (Fig EV1A, lane 8), as in the case for binding to PGAM5, deletion of the C‐terminal tail abolished the binding ability (Fig EV1A, lane 7 and B, lane 7). Thus, Drp1 and PGAM5 likely bind to overlapping but partially different sites on Stx17. PGAM5 has an N‐terminal transmembrane domain (amino acids 7–29), followed by a KEAP1 domain and PGAM‐like domain (Sadatomi et al, 2013; Fig 1A). Removal of the N‐terminal transmembrane domain abolished the binding to Stx17 (Fig 1E, lane 5), whereas the transmembrane domain itself could bind to Stx17 (lane 6). The transmembrane domain of PGAM5 was found not to bind to Stx18 (Fig 1F), a syntaxin localized in the ER (Hatsuzawa et al, 2000). The requirement of the transmembrane domain of PGAM5 for binding to Stx17 was confirmed by pull‐down experiments using recombinant proteins (Fig EV1C).

Figure EV1. Stx17 binds to Drp1 and PGAM5.

1. 293T cells were cotransfected with plasmids encoding GFP‐Drp1 K38A and FLAG‐Stx17 wild type (WT) or the indicated FLAG‐Stx17 constructs. At 24 h after transfection, cell lysates were immunoprecipitated (IP) with anti‐FLAG M2 beads and analyzed by IB using antibodies against GFP and FLAG. Five percent of lysates was analyzed as input.
2. MBP or the MBP‐Stx17 constructs attached to amylose resin were mixed with His₆‐Drp1 K38A, and the proteins bound to the resin were separated by SDS–PAGE and blotted onto PVDF membranes. The blots were detected by an anti‐penta‐His tag antibody (upper panels) or stained with Coomassie Brilliant Blue R‐250 (lower panels). Ten percent of the proteins used for each experiment was analyzed as input. Asterisks and double asterisk represent possible MBP dimers and degradation products, respectively.
3. MBP‐Stx17 WT attached to amylose resin was mixed with the indicated GST‐PGAM5 constructs, and the proteins bound to the resin were separated by SDS–PAGE and blotted onto PVDF membranes. The blots were detected by an anti‐GST antibody (upper panels) or stained with Coomassie Brilliant Blue R‐250 (lower panels). Ten percent of the proteins used for each experiment was analyzed as input. PGAM5 ΔTMD (amino acids 30–289) and ΔTMD# (amino acids 25–289, corresponding to the PARL‐cleaved form) (Sekine et al, 2012).

Source data are available online for this figure.

PGAM5 localizes to the ER–mitochondria interface

Although immunoreactivity for both Stx17 (Arasaki et al, 2015) and PGAM5 (Fig 2A, upper row) was predominantly detected on mitochondria, PLA combined with the expression of the ER and mitochondria fluorescence markers revealed that Stx17 interacts with PGAM5 principally on and in the vicinity of the ER (Fig 2B). This finding prompted us to examine whether PGAM5 localizes to the ER–mitochondria interface in addition to mitochondria. Subcellular fractionation revealed this to be the case. PGAM5 was recovered not only in the mitochondrial fraction (Fig 2C, lane 5), but also in the MAM fraction (lane 4). The localization of PGAM5‐GFP at the ER–mitochondria interface was confirmed by electron microscopy using an ascorbate peroxidase 2 (APEX2)‐GFP‐binding peptide (Fig 2D).

Figure 2. PGAM5 is localized at the ER–mitochondria interface.

1. HeLa cells were treated with DMSO (Vehicle) or 0.03 mg/ml digitonin (+Digitonin), fixed, and then double‐immunostained for PGAM5 and Tom20. Scale bar, 5 μm. The bar graph on the right shows the Manders’ coefficients for the colocalization of PGAM5 and Tom20. Values are means ± SEM (n = 3). ***P < 0.001 as compared with Vehicle (paired Student's t‐test).
2. HeLa cells stably expressing FLAG‐Stx17 wild type (WT) were transfected with a plasmid encoding Su9‐GFP (mitochondria) or Sec61β‐GFP (ER). At 24 h after transfection, the cells were subjected to PLA using antibodies against FLAG and PGAM5. Scale bar, 5 μm. The bar graph on the right shows the Manders’ coefficients for the colocalization of PLA dots and Su9‐GFP or Sec61β‐GFP. Values are means ± SEM (n = 3). ***P < 0.001 (paired Student's t‐test).
3. HeLa cells were treated with DMSO (Vehicle) or 20 μM CCCP (+CCCP) for 2 h, lysed, and subjected to Percoll‐based fractionation. Equal amounts of proteins were analyzed by IB using the indicated antibodies. PNS, postnuclear supernatant; MS, microsomes; Mt, mitochondria. The amounts of proteins recovered on fractionation were as follows for vehicle and CCCP treatment, respectively: PNS (6.6 mg and 5.5 mg), cytosol (4.8 mg and 4.9 mg), MS (2.0 mg and 2.2 mg), MAM (0.55 mg and 0.46 mg), and Mt (0.30 mg and 0.28 mg).
4. Electron microscopic analysis of HeLa cells expressing PGAM5‐GFP and APEX2‐GFP‐binding peptide. Samples were prepared as described in Materials and Methods. Arrows indicate the position of 3,3′‐diaminobenzidine reaction at the ER–mitochondria interface. Scale bar, 500 nm.
5. HeLa cells stably expressing FLAG‐Stx17 WT were mock‐transfected or transfected with siRNA for Mfn1, Mfn2, or PACS‐2. At 72 h after transfection, the cells were subjected to PLA using antibodies against FLAG and PGAM5. Scale bar, 5 μm. Values are means ± SEM (n = 3). ***P < 0.001 as compared with Mock (paired Student's t‐test).
6. HeLa cells were mock‐transfected or transfected with siRNA for Mfn1, Stx17, Mfn2, or PACS‐2. At 72 h after transfection, the cells were subjected to PLA using antibodies against Drp1 and PGAM5. Scale bar, 5 μm. Values are means ± SEM (n = 3). ***P < 0.001 as compared with Mock (paired Student's t‐test).

Source data are available online for this figure.

The MAM is rich in cholesterol and sphingolipids, thus resembling lipid rafts (Herrera‐Cruz & Simmen, 2017). Because the cholesterol‐rich structure is sensitive to low concentrations of digitonin (Oliferenko et al, 1999), we treated cells with 0.03 mg/ml digitonin, a concentration only effective to solubilize cholesterol‐rich membranes such as the plasma membrane, but not intracellular membranes. Similar to the case of Stx17 (Arasaki et al, 2015), this treatment caused partial dissociation of PGAM5 from mitochondria (Fig 2A, lower low). Albeit less effective, treatment of cells with cholesterol depletion reagents, methyl‐β‐cyclodextrin (MβCD; Ohtani et al, 1989) and nystatin (Rothberg et al, 1992), also caused an increase in diffuse staining for PGAM5 (Fig EV2A).

Figure EV2. PGAM5 dephosphorylates and activates Drp1 in healthy cells.

1. HeLa cells were incubated with DMSO (Vehicle), 5 mM MβCD for 1 h (MβCD), or 10 μg/ml nystatin (Nystatin) for 20 min and then double‐immunostained for PGAM5 (Alexa Fluor 488) and Tom20 (Alexa Fluor 594).
2. HeLa cells with mock treatment (Mock) or depleted of PGAM5 (PGAM5 KD) were fixed and then double‐immunostained for PGAM5 and Tom20.
3. HeLa cells with mock treatment or depleted of PGAM5 were fixed after treatment with 20 μM CCCP for 2 h and then double‐immunostained for PGAM5 and Tom20.
4. HeLa cells with mock treatment or depleted of PGAM5 or Stx17 were lysed and analyzed IB using the indicated antibodies.
5. HeLa cells were transfected with a plasmid encoding C‐terminally FLAG‐tagged PGAM5 or the H105A mutant. At 24 h after transfection, the cells were double‐immunostained for FLAG and Tom20.

Data information: Scale bars, 5 μm. Values are means ± SEM (n = 3). **P < 0.01 and as compared with Mock; ***P < 0.001 as compared with PGAM5 (WT) by paired Student's t‐test.Source data are available online for this figure.

To determine whether ER–mitochondria contact is required for the association of FLAG‐Stx17 with PGAM5, we knocked down Mfn2. Although the role of Mfn2 in MAM–mitochondria tethering is under debate (Naon et al, 2016), it is, at least, widely accepted that Mfn2 depletion abolishes the MAM function (Hailey et al, 2010; Hamasaki et al, 2013; Arasaki et al, 2015). PLA revealed that the FLAG‐Stx17‐PGAM5 proximity was diminished upon depletion of Mfn2, but not the non‐tethering protein Mfn1 (Fig 2E). Knockdown of PACS‐2, a multifunctional sorting protein required for maintaining MAM integrity (Simmen et al, 2005), also reduced the PLA signal for the FLAG‐Stx17‐PGAM5 proximity (Fig 2E).

PGAM5 promotes mitochondrial fission through interaction with Stx17

Previous studies demonstrated that PGAM5 dephosphorylates Drp1 in the context of cell death (Wang et al, 2012; Lin et al, 2013; Xu et al, 2015) and that its overexpression causes mitochondrial fragmentation (Wilkins et al, 2014). As calcineurin and PP2A were reported to be also responsible for the dephosphorylation of Drp1 (Cereghetti et al, 2008; Merrill et al, 2013), we first sought to confirm that PGAM5 regulates mitochondrial division in healthy cells. Knockdown of PGAM5 caused mitochondrial elongation (Fig EV2B, lower row), and the defect in mitochondrial fission was confirmed by incubation of PGAM5‐depleted cells with the protonophore carbonyl cyanide m‐chlorophenylhydrazone (CCCP; Fig EV2C), which is known to induce robust mitochondrial fission in a Drp1‐dependent manner (Ishihara et al, 2006). Immunoblotting (IB) demonstrated that PGAM5 depletion (Fig EV2D, lane 2), as well as Stx17 depletion (lane 4), increased the phosphorylation level of Drp1 at Ser637. Contrary to PGAM5 depletion, its overexpression caused mitochondrial fission (Fig EV2E, upper row), as reported previously (Wilkins et al, 2014), whereas the overexpression of a phosphatase‐dead H105A mutant (Takeda et al, 2009), in which His105 was replaced by Ala, caused mitochondrial elongation (lower row). The proximity between endogenous PGAM5 and Drp1 was detected on PLA, and this proximity was also abrogated upon Stx17 depletion as well as depletion of Mfn2 or PACS‐2 (Fig 2F). These results suggest that PGAM5 promotes mitochondrial fission by dephosphorylating Drp1 at Ser637 in a Stx17, MAM‐dependent manner.

To understand why Stx17 depletion abrogates the link between PGAM5 and Drp1, we examined the localization of PGAM5 in Stx17‐depleted cells. Strikingly, Stx17 depletion caused aggregation of PGAM5 within mitochondria, and some Tom20‐positive tubules were found to be devoid of PGAM5 (Fig 3A, lower row), suggesting that Stx17 regulates PGAM5 localization.

Figure 3. Localization of PGAM5 is regulated by Stx17.

1. HeLa cells with mock treatment (Mock) or depleted of Stx17 (Stx17 KD) were fixed and double‐immunostained for PGAM5 and Tom20. Scale bar, 5 μm.
2. 293T cells with mock treatment or depleted of Stx17 were incubated with ethanol (Vehicle) or 20 μM CCCP (+CCCP) for 2 h, lysed, and then analyzed by IB using the indicated antibodies.
3. 293T cells transiently expressing FLAG‐Stx17 wild type (WT) or the K254C mutant were incubated with ethanol (Vehicle) or 20 μM CCCP (+CCCP) for 2 h, lysed, immunoprecipitated (IP) with anti‐FLAG M2 beads, and then analyzed by IB using the indicated antibodies.
4. HeLa cells stably expressing FLAG‐Stx17 WT were incubated with ethanol (Vehicle) or 20 μM CCCP (+CCCP) for 2 h, and subjected to PLA using antibodies against FLAG and PGAM5. Scale bar, 5 μm. Values are means ± SEM (n = 3). ***P < 0.001 as compared with Vehicle (paired Student's t‐test).

Source data are available online for this figure.

PGAM5 was often observed as two bands on IB: the upper and lower bands correspond to the full‐length and PARL‐cleaved forms, respectively (Fig 3B, lane 1; Sekine et al, 2012; Wai et al, 2016). Loss of mitochondrial membrane potential halts PINK1 cleavage and instead enhances PGAM5 cleavage by PARL (Sekine et al, 2012). When Stx17 was knocked down in 293T cells, the amount of the upper band decreased concomitant with an increase in the amount of the lower band (Fig 3B, lane 2 vs. lane 1), implying a link between PGAM5 and Stx17. As reported previously (Sekine et al, 2012), CCCP treatment caused almost complete cleavage of PGAM5 in both mock‐ and Stx17‐depleted cells (Fig 3B, lanes 3 and 4). Taken together, the results suggest that Stx17 is a critical factor for the regulation of PGAM5 localization and function.

When PGAM5 was cleaved as a consequence of CCCP treatment, the resultant short form was found not to bind to FLAG‐Stx17, as demonstrated by immunoprecipitation (Fig 3C, lane 7) and PLA (Fig 3D). This is consistent with the finding that a PARL‐cleaved form of recombinant PGAM5 did not bind to Stx17 (Fig EV1C, lane 6). A part of the cleaved PGAM5 was redistributed to the microsomes/ER (Fig 2C, lane 8).

Stx17 and PGAM5 participate in PINK1/Parkin/FUNDC1‐mediated mitophagy

Because PGAM5 is closely related to PINK1 and Parkinson's disease (Imai et al, 2010; Sekine et al, 2012; Lu et al, 2014; Hawk et al, 2018), we examined whether PGAM5 and Stx17 are involved in PINK1/Parkin‐dependent mitophagy. To this end, we treated GFP‐Parkin‐expressing HeLa cells with CCCP. CCCP treatment for 16 h decreased the amount of a mitochondrial inner membrane protein, Tim23, as well as that of the outer membrane, Tom20, without affecting the amount of an ER/MAM marker, calnexin (CNX; Fig 4A, lane 5). Depletion of Stx17 or PGAM5 prevented the loss of the mitochondrial proteins (Fig 4A, lanes 6 and 7, respectively). On the other hand, depletion of PGAM5 did not affect the progress of starvation‐induced autophagy, as demonstrated by the formation of LC3‐II in PGAM5‐depleted cells (Fig EV3A, lane 6).

Figure 4. Mitophagy is inhibited upon depletion of Stx17 or PGAM5.

1. HeLa cells stably expressing GFP‐Parkin with mock treatment (Control) or depleted (KD) of Stx17, PGAM5 or Drp1 were incubated with ethanol (Vehicle) or 10 μM CCCP (+CCCP) for 16 h and analyzed by IB using the indicated antibodies.
2. HeLa cells stably expressing GFP‐Parkin with mock treatment (Mock) or depleted of Stx17, PGAM5, or Drp1 were incubated with ethanol (Vehicle) or 20 μM CCCP (+CCCP) for 2 h and analyzed by immunofluorescence microscopy. Scale bar, 5 μm. The bar graph on the right shows the Manders’ coefficients for the colocalization of GFP‐Parkin and Tom20. Values are means ± SEM (n = 3). ***P < 0.001 as compared with Mock (paired Student's t‐test).
3. PINK1‐FLAG and GFP‐Parkin stably expressing HeLa cells with mock treatment or depleted of Stx17 or PGAM5 were incubated in the absence (ethanol) or presence of 20 μM CCCP for 2 h and analyzed by IB using the indicated antibodies. F, full‐length PINK1; C, cleaved PINK1.

Source data are available online for this figure.

Figure EV3. FLAG‐Stx17 wild type, but not the K254C mutant, can compensate for Stx17 depletion in mitophagy.

1. GFP‐Parkin stably expressing HeLa cells with mock treatment (Mock) or treated with siRNA (KD) for Stx17, PGAM5, or FUNDC1 for 72 h were untreated (Fed) or starved for 2 h (SV), lysed and then analyzed by IB using the indicated antibodies.
2. HeLa cells stably expressing FLAG‐Stx17 wild type (WT) or the K254C mutant were transfected with siRNA for Stx17. At 48 h after transfection, the cells were transfected with a plasmid encoding GFP‐Parkin and incubated for 24 h. They were then incubated with 10 μM CCCP (+CCCP) for 16 h, lysed, and analyzed by IB using the indicated antibodies.
3. HeLa cells stably expressing FLAG‐Stx17 WT or the K254C mutant were transfected with siRNA for Stx17. At 48 h after transfection, the cells were transfected with a plasmid encoding GFP‐Parkin and incubated for 24 h. They were then incubated with 20 μM CCCP for 2 h and then immunostained for Tom20. Scale bar, 5 μm.
4. GFP‐Parkin stably expressing HeLa cells incubated with siRNA for PGAM5 for 48 h were transfected with a plasmid encoding C‐terminally FLAG‐tagged PGAM5 or the H105A mutant without a mutation in the siRNA targeting sequence. After 24 h, the cells were incubated with 20 μM CCCP for 2 h and then immunostained for FLAG. PGAM5 proteins were found to be expressed likely due to the overexpression of their mRNAs. Scale bar, 5 μm.
5. GFP‐Parkin stably expressing HeLa cells, or ones depleted of PINK1 (PINK1 KD) or depleted of both PINK1 and Stx17 or PGAM5 (DKD) were incubated with 20 μM CCCP for 2 h and immunostained for Tom20. Scale bar, 5 μm.
6. GFP‐Parkin stably expressing HeLa cells or ones depleted of Stx17 or PGAM5 were incubated with 20 μM CCCP for the indicated times and immunostained for Tom20. Scale bar, 5 μm.

Source data are available online for this figure.

As previously reported (Jin et al, 2010), loss of mitochondrial potential induced the redistribution of GFP‐Parkin from the cytosol to mitochondria, leading to mitochondrial aggregation around the nucleus (Fig 4B, second row). In Stx17‐ and PGAM5‐depleted cells, on the other hand, Parkin appeared to translocate to membrane structures, but did not cover whole mitochondria, and mitochondrial aggregation was not obvious (Fig 4B, third and fourth rows, respectively). This is not due to the elongation of mitochondria caused by depletion of Stx17 or PGAM5 because knockdown of Drp1 also prevented mitochondrial clearance (Fig 4A, lane 8), but GFP‐Parkin fully covered elongated mitochondria (Fig 4B, bottom row). The phenotype of cells depleted of Stx17 was not observed for endogenous Stx17‐depleted, but FLAG‐Stx17 stably expressing cells (Fig EV3B, lane 3, and Fig EV3C, upper row), excluding the possible off‐target effect of the siRNA used. On the other hand, the phenotype of cells depleted of Stx17 was not compensated for by the expression of the K254C mutant (Fig EV3B, lane 4, and Fig EV3C, lower row), which did not interact with PGAM5 (Fig 1B and C) or Drp1 (Fig EV1A). Compensation was also observed upon PGAM5‐FLAG expression in PGAM5‐depleted cells (Fig EV3D, upper row). Of note is that the H105A mutant failed to compensate for the PGAM5 silencing effect (Fig EV3D, lower row), suggesting the requirement of the protein phosphatase activity of PGAM5.

A previous study demonstrated that PGAM5 knockout abolished the appearance of full‐length PINK1 on the mitochondrial surface (Lu et al, 2014). However, we found that depletion of PGAM5 or Stx17 did not affect the appearance of full‐length PINK1‐FLAG (Fig 4C, lanes 4 and 6, bottom panel). When PINK1 was silenced, no translocation of GFP‐Parkin to mitochondria was observed (Fig EV3E), and this phenotype was obviously different from those of Stx17‐ and PGAM5‐depleted cells (Fig 4B). The translocation of GFP‐Parkin to limited regions in or close to mitochondria observed in Stx17‐ and PGAM5‐depleted cells was seen at an early stage (30 min) of GFP‐Parkin translocation in control cells (Fig EV3F, upper panels). These results taken together suggest that Stx17 and PGAM5 are not involved in the initial recruitment of Parkin to membrane structures, but are required for later stage(s).

GFP‐Parkin translocates to cholesterol depletion‐sensitive structures upon CCCP treatment

As Stx17 and PGAM5 localization is sensitive to cholesterol depletion reagents (Arasaki et al, 2015; and Figs 2A and EV2A), we examined the effect of digitonin on the localization of GFP‐Parkin. When GFP‐Parkin‐expressing cells were treated with 0.02% Triton X‐100 before paraformaldehyde fixation, GFP‐Parkin leaked from the cells, whereas GFP‐Parkin targeted to mitochondria upon CCCP treatment was found not to be released from membranous structures (Fig EV4A). In contrast, digitonin treatment of CCCP‐treated cells caused the release of most GFP‐Parkin from the cells, and residual GFP‐Parkin was observed in punctate structures on and close to mitochondria (Fig EV4B, lower row), reminiscent of those observed in Stx17‐ and PGAM5‐depleted cells (Fig 4B). Mitochondrial aggregation also appeared to be disrupted upon digitonin treatment (Fig EV4B, lower row). Notably, GFP‐Parkin‐positive dots in Stx17‐depleted cells were resistant to digitonin treatment (Fig EV4C, lower row). When cells were treated with MβCD or nystatin, aggregation of mitochondria was moderately inhibited (Appendix Fig S1). These results suggest that upon CCCP treatment GFP‐Parkin translocates to cholesterol depletion‐sensitive structures, which are required for mitochondrial aggregation.

Figure EV4. GFP‐Parkin‐associated structures are sensitive to digitonin.

• A, B
  GFP‐Parkin stably expressing HeLa cells were incubated with ethanol (Vehicle) or 20 μM CCCP (+CCCP) for 2 h, treated with 0.02% Triton X‐100 (A) or 0.03 mg/ml digitonin (B), and then immunostained for Tom20. Scale bars, 5 μm.
• C
  GFP‐Parkin stably expressing HeLa cells depleted of Stx17 were incubated with 20 μM CCCP (+CCCP) for 2 h, untreated or treated with 0.03 mg/ml digitonin, and then immunostained for Tom20. Scale bar, 5 μm.

Autophagy progress is required for the expansion of GFP‐Parkin‐positive structures and mitochondrial aggregation

At the onset of mitophagy, autophagosome formation is enhanced by a mechanism not fully understood (Yoshii & Mizushima, 2015). As Stx17 participates in autophagosome formation (Hamasaki et al, 2013; Arasaki et al, 2015, 2017), we wondered whether the phenotype of Stx17‐depleted cells would also be observed in cells defective in autophagosome formation. To examine this possibility, we knocked down Atg14L, a subunit of the PI3‐kinase responsible for the formation of PI3P‐positive omegasomes (Matsunaga et al, 2010), and Atg9, a transmembrane protein responsible for the supply of membrane components to isolation membrane/autophagosome expansion (Young et al, 2006). In both cases, GFP‐Parkin was found to translocate into only a small fraction of non‐aggregated mitochondria (Appendix Fig S2A, second and third rows). Similar results were obtained when cells were treated with CCCP following starvation in Earle's balanced salt solution (EBSS) (Appendix Fig S2A, bottom row) or with a PI3‐kinase inhibitor, wortmannin (Appendix Fig S2B). These results suggest that autophagosome expansion during CCCP treatment is required for GFP‐Parkin accumulation on mitochondria and mitochondrial aggregation.

Autophagosome formation does not occur close to ubiquitinated mitochondria in PGAM5‐depleted cells

To determine which step of the PINK1/Parkin‐dependent mitophagic pathway is blocked by PGAM5 depletion, we examined the localization of early mitophagic marker proteins. In mock‐treated cells, puncta positive for FLAG‐tagged DFCP1, a PI3P‐binding protein and a marker for omegasomes (Axe et al, 2008), were found to be on and/or close to GFP‐Parkin‐localized mitochondria (Fig 5A, top and second rows). In PGAM5‐depleted cells, on the other hand, FLAG‐DFCP1‐positive puncta were not colocalized with GFP‐Parkin‐positive structures (Fig 5A, third and fourth rows). Consistent with the involvement of Stx17 in the autophagosome formation step (Hamasaki et al, 2013; Arasaki et al, 2015, 2017), no discrete FLAG‐DFCP1‐positive puncta were observed in Stx17‐depleted cells (Fig 5A, fifth and bottom rows). Similar results were obtained for LC3 distribution, although LC3 dots in PGAM5‐depleted cells were smaller than those in mock‐treated cells (Figs 5B and EV5A). No LC3 dots were observed in Stx17‐depleted cells. On the other hand, ubiquitination and association of the ubiquitin‐binding autophagy receptor p62 with mitochondria occurred in cells depleted of PGAM5 or Stx17 as well as mock‐treated cells (Fig EV5B and C). It should be noted that PGAM5 per se is not involved in autophagy because LC3‐II, a hallmark for autophagy progression, was detected in PGAM5‐depleted cells (Fig 5C, lane 6) as well as in mock‐treated cells (lane 4). These data raise the possibility that the step impaired by the loss of PGAM5 is the link between autophagosomes and ubiquitinated mitochondria.

Figure 5. PGAM5 depletion abrogates the proximity of GFP‐Parkin‐associated mitochondria to omegasomes.

1. GFP‐Parkin stably expressing HeLa cells were mock‐transfected (Mock) or transfected with siRNA (KD) for Stx17 or PGAM5. At 48 h after siRNA transfection, the cells were transfected with a plasmid encoding FLAG‐DFCP1, incubated for 24 h, treated with 20 μM CCCP for 2 h, and then immunostained for FLAG. Scale bars, 5 μm. The bar graph below shows the Manders’ coefficients for the colocalization of GFP‐Parkin and FLAG‐DFCP1. Values are means ± SEM (n = 3). **P < 0.01 as compared with Mock (paired Student's t‐test).
2. GFP‐Parkin stably expressing HeLa cells with mock treatment or depleted of Stx17 or PGAM5 were incubated with 20 μM CCCP for 2 h and immunostained for LC3. Scale bar, 5 μm. The bar graph below shows the Manders’ coefficients for the colocalization of GFP‐Parkin and LC3. Values are means ± SEM (n = 3). **P < 0.01 as compared with Mock (paired Student's t‐test).
3. GFP‐Parkin stably expressing HeLa cells with mock treatment or depleted of Stx17 or PGAM5 were incubated in the absence (Vehicle) or presence of 20 μM CCCP (+CCCP) for 2 h and analyzed by IB using the indicated antibodies.

Source data are available online for this figure.

Figure EV5. Ubiquitination occurs at the sites to which GFP‐Parkin has translocated even in the absence of Stx17 or PGAM5.

• A–C
  GFP‐Parkin stably expressing cells with mock treatment (Mock) or depleted of PGAM5 (PGAM5 KD) or Stx17 (Stx17 KD) were incubated with 20 μM CCCP for 2 h and immunostained for LC3 and ubiquitin (A), ubiquitin (B), or p62 (C). Scale bars, 5 μm.

Stx17 is required for the association of PGAM5 with the mitophagy receptor FUNDC1

FUNDC1 has been implicated in mitophagy under hypoxia, as well as carbonyl cyanide p‐trifluoromethoxyphenylhydrazone (FCCP) treatment (Liu et al, 2012). FUNDC1 is a mitochondria‐associated protein that facilitates the engulfment of damaged mitochondria by autophagosomes by linking the two organelles through its LC3‐interacting (LIR) motif at the N‐terminus: in healthy cells FUNDC1 at Ser13 is phosphorylated, preventing the interaction between the LIR of FUNDC1 and LC3 (Liu et al, 2012), whereas upon mitophagic stimuli, it is dephosphorylated and activated by PGAM5 (Chen et al, 2014; Wu et al, 2014). More recently, FUNDC1 was reported to localize to the MAM in association with CNX in healthy cells, and in response to mitophagy stress, it dissociates from CNX and associates with Drp1 to promote division of mitochondria required for effective engulfment by autophagosomes (Wu et al, 2016).

Given that Stx17 regulates PGAM5 localization and function in healthy cells, and that CCCP‐induced PGAM5 cleavage causes the release of Stx17 from PGAM5 (Fig 3C) and its partial redistribution to the microsomes/ER (Fig 2C), we were interested in whether the Stx17‐PGAM5 interaction is required for PGAM5 binding to FUNDC1 during mitophagy. First, we examined the translocation of GFP‐Parkin in FUNDC1‐depleted cells. Depletion of FUNDC1 not only inhibited CCCP‐induced mitochondrial clearance (Fig 6A, lane 4), but also prevented the whole coverage of mitochondria and mitochondrial aggregation by GFP‐Parkin (Fig 6B, lower row), as observed in cells depleted of PGAM5 or Stx17 (Fig 4B). We next examined the significance of Stx17 in the PGAM5‐FUNDC1 axis. When CCCP was added to GFP‐Parkin stably expressing HeLa cells, PGAM5 was found to be cleaved and to bind to FLAG‐FUNDC1 (Fig 6C, lower panel, lane 7, and Fig 6D), as reported previously (Chen et al, 2014). Strikingly, Stx17 depletion prevented the association of PGAM5 with FLAG‐FUNDC1 (Fig 6C, lower panel, lane 8, and Fig 6D).

Figure 6. Stx17 is required for the link between PGAM5 and FUNDC1.

1. HeLa cells stably expressing GFP‐Parkin with mock treatment (Control) or depleted of FUNDC1 (FUNDC1 KD) were incubated for 16 h with ethanol (Vehicle) or 10 μM CCCP (+CCCP), and analyzed by IB using the indicated antibodies.
2. GFP‐Parkin stably expressing HeLa cells with mock treatment (Mock) or depleted of FUNDC1 (FUNDC1 KD) were incubated with 20 μM CCCP for 2 h and immunostained for Tom20. Scale bar, 5 μm.
3. 293T cells with mock treatment (Mock) or depleted of Stx17 (Stx17 KD) were incubated in the absence (Vehicle) or presence of 20 μM CCCP (+CCCP) for 2 h, immunoprecipitated with anti‐FLAG M2 beads and then analyzed by IB using the indicated antibodies. Three percent of lysates was analyzed as input. During Stx17 knockdown, cells were transfected with a plasmid encoding FLAG‐FUNDC1.
4. GFP‐Parkin stably expressing HeLa cells were treated as described in (C) and then subjected to PLA using antibodies against FLAG and PGAM5. During Stx17 knockdown, cells were transfected with a plasmid encoding FLAG‐FUNDC1. Scale bar, 5 μm. Values are means ± SEM (n = 3). *P < 0.05 and **P < 0.01 as compared with Mock (+CCCP) by paired Student's t‐test.

Source data are available online for this figure.

Stx17 and PGAM5 maintain mitochondrial integrity in Drosophila

We next examined the genetic association between Stx17 and PGAM5 in Drosophila. Stx17 has been well characterized as an autophagy regulator in mammals, which appears to be conserved in Drosophila (Takáts et al, 2013, 2014). Removal of a copy of Stx17 revealed a well‐preserved mitochondrial structure (Fig 7Ba), which was comparable to a normal control (Fig 7Aa). In contrast, the complete loss of Stx17 activity resulted in the disintegration of the crista architecture (Fig 7Ab), which was reversed by the re‐introduction of Stx17 (Fig 7Ac). Overexpression of PGAM5 partly compensated for the loss of Stx17 (Fig 7Ad). Although the loss of PGAM5 itself did not produce obvious mitochondrial degeneration (Fig 7Bb), the removal of a copy of Stx17 in this genetic background caused the crista disintegration (Fig 7Bc), suggesting that PGAM5 supports the Stx17 activity in mitochondria. However, PGAM5 overexpression failed to reverse the semi‐lethal phenotype of Stx17‐deficient flies (Appendix Fig S3), indicating that Stx17 has multiple functions in organelle maintenance other than in that of mitochondria unlike PGAM5. Although not conclusive, the partial rescue of mitochondrial morphology in Stx17‐deficient flies by PGAM5 overexpression suggests a possible genetic interaction between Stx17 and PGAM5 in flies.

Figure 7. PGAM5 modulates the mitochondrial phenotypes caused by Stx17 inhibition in Drosophila .

1. PGAM5 overexpression partially rescues the mitochondrial defects by Stx17 loss. Transmission electron microscope images of the indirect flight muscle in the indicated genotypes (a, Control; b, Stx17^−/−; c, Stx17^−/− and Stx17 overexpression; d, Stx17^−/− and PGAM5 overexpression) of 7‐day‐old adult flies are shown. The bar graph on the right shows frequency of healthy and abnormal mitochondria presented as percentages (mean ± SEM) using the scoring system: Class 0, normal; Class 1, fuzzy or dilated cristae; Class 2, fragmented cristae and loss of electron density. *P < 0.05, ^# P < 0.001 vs. the same class of control (Dunnett's test). n = 50–119 mitochondria from three or four independent samples. Genotypes used were as follows: +/y; Act5c‐GAL4/+ (a), +/y; Act5c‐GAL4/+; Stx17 ^LL06330 /Stx17 ^LL06330 (b), +/y; Act5c‐GAL4/UAS‐FLAG‐Stx17; Stx17 ^LL06330 /Stx17 ^LL06330 (c), +/y; Act5c‐GAL4/UAS‐PGAM5; Stx17 ^LL06330 /Stx17 ^LL06330 (d). Scale bars, 200 nm.
2. Simultaneous reduction in Stx17 and PGAM5 results in mitochondrial degeneration. a, Stx17^+/−; b, PGAM5^−/−; c, Stx17^+/−, PGAM5^−/−. The bar graph on the right represents the mitochondrial phenotypes classified as described in (A). *P < 0.05, **P < 0.01 vs. the same class of Stx17^+/− (Dunnett's test). n = 62–195 mitochondria from three independent samples. Scale bars = 1 μm. Genotypes used were as follows: +/y; Act5c‐GAL4/+; Stx17 ^LL06330 /+ (a), PGAM5 ¹ /y; Act5c‐GAL4/+ (b), PGAM5 ¹ /y; Act5c‐GAL4/UAS‐Stx17‐FLAG; Stx17 ^LL06330 /+ (c).

Source data are available online for this figure.

Discussion

Our previous study demonstrated that Stx17 defines the localization and activity of Drp1: Stx17 competes with the protein kinase A‐anchoring protein Rab32 for Drp1 binding, and thereby prevents protein kinase A‐mediated inactivating phosphorylation of Drp1 at Ser637 (Arasaki et al, 2015). In the present study, we showed that Stx17 also facilitates the dephosphorylation of Drp1 by regulating the localization of the mitochondrial protein phosphatase PGAM5. Depletion of Stx17 caused aggregation of PGAM5 within mitochondria and disrupted the proximity between Drp1 and PGAM5. The connection between Stx17 and PGAM5 was also demonstrated in Drosophila: overexpression of PGAM5 can partially rescue a mitochondrial morphology defect due to Stx17 ablation. We also demonstrated that Stx17 is important for the interaction between cleaved PGAM5 and the mitochondria‐associated mitophagy receptor FUNDC1 in PINK1/Parkin‐mediated mitophagy.

The exact localization of PGAM5 in mitochondria remains elusive. The first report demonstrated that PGAM5 mediates the dephosphorylation of the cytosolic protein kinase Ask1 (Takeda et al, 2009), suggesting that its active site faces the cytosol. However, the same authors later found that PGAM5 is resistant to protease treatment in isolated mitochondria and co‐fractionated with mitochondrial inner membrane markers, thus implying that PGAM5 is localized in the mitochondrial inner membrane (Sekine et al, 2012). Lu et al (2014) also reached the same conclusion based on the proteolytic insensitivity of PGAM5 in digitonin‐permeabilized cells. However, a number of cytosolic proteins, such as Drp1 (Wang et al, 2012; Lin et al, 2013; Wilkins et al, 2014; Xu et al, 2015; Kang et al, 2015; Lavie et al, 2017), NFAT (Kang et al, 2015), and FUNDC1 (Chen et al, 2014), have been reported to be dephosphorylated by PGAM5, strongly suggesting that its active site faces the cytosol. We showed that PGAM5 is localized not only in mitochondria, but also in the MAM with the PARL‐cleaved form in microsomes. Moreover, digitonin treatment not only permeabilized the plasma membrane, but also affected the localization of PGAM5 (Fig 2A), suggesting that the interpretation of the results of Lu et al (2014) needs reconsideration and that at least some fraction of PGAM5, as in the case of Stx17, is present on digitonin‐sensitive, possibly raft‐like structures. Therefore, previous results can be explained by the idea that PGAM5 is localized in mitochondrial outer–inner membrane contact sites and can shuttle between the two membranes depending on the mitochondrial inner membrane potential and cellular context. Because mitochondrial division sites are circumscribed and constricted by the ER/MAM (Friedman et al, 2011) and actin (Korobova et al, 2013), it is reasonable to assume that the mitochondria outer membrane at these sites is pushed into contact with the inner membrane. In these sites, PGAM5 may be predominantly localized in the outer membrane facing the cytosolic side and can interact and dephosphorylate cytosolic Drp1 with the help of Stx17. It is also possible that PGAM5 resides on the mitochondrial surface without embedding in the membrane. These possibilities and others should be addressed in future work.

It is possible that the energy barrier for the translocation of membrane proteins between the MAM and mitochondria may be low and that some proteins can shuttle between these organelles. Indeed, it was reported that certain proteins such as FKBP38 and Bcl‐2 can translocate from mitochondria to the ER during mitophagy (Saita et al, 2013). This idea may explain why the interaction between Stx17 and PGAM5 is detectable primarily at the ER (Fig 2B), and requires Mfn2‐mediated ER–mitochondria contacts and PACS‐2‐mediated MAM integrity (Fig 2E), although the two proteins exhibit a mitochondria‐like distribution (Arasaki et al, 2015; and Fig 2A). Moreover, this also may explain why the membrane‐anchoring regions of Stx17 and PGAM5 are important for their interaction (Figs 1 and EV1C). Given the requirement of both membrane‐anchoring regions for the binding, the most straightforward interpretation is that the two proteins should be on the same membrane. However, if this were the case, it would be difficult to explain why the MAM–mitochondria tethering and MAM integrity are required for the binding. It should be noted that translocation of cytochrome b5 from mitochondria to autophagosomes, perhaps through the MAM, occurs at the ER–mitochondria contact site (Hailey et al, 2010). For this transfer, its hairpinlike membrane‐anchoring structure is essential (Hailey et al, 2010), and Stx17 also possesses such a structure. Future studies should investigate the validity of this hypothesis and reveal the molecular mechanism underlying membrane protein transfer between the MAM and mitochondria.

Because FUNDC1 possesses transmembrane domains for the association with mitochondria, it is generally thought that Parkin is not necessary for FUNDC1 to function as a mitophagy receptor (Liu et al, 2012). Ubiquitination of mitochondria by Parkin is required for cytosolic mitophagy receptors to bind to mitochondria (Hamacher‐Brady & Brady, 2016). However, our results unequivocally showed that depletion of FUNDC1 inhibited Parkin‐mediated mitophagy (Fig 6A and B). Stx17 and PGAM5 are also required for PINK1/Parkin‐mediated mitophagy. Although a previous study revealed that PGAM5 is necessary for the stabilization of full‐length PINK1 (Lu et al, 2014), we did not observe that loss of PGAM5 affects the stabilization of full‐length PINK1 (Fig 4C). This may be due to different cell types and/or different experimental conditions. Consistent with the appearance of full‐length PINK1 on the mitochondrial surface, GFP‐Parkin was found to be translocated to membrane structures in Stx17‐ and PGAM5‐depleted cells, although the translocated area was limited in comparison with in control cells. This phenotype was obviously different from that of cells depleted of PINK1 (Fig EV3E), but rather similar to that observed in an early stage of GFP‐Parkin translocation (Fig EV3F, upper panels). Therefore, PGAM5 and Stx17 participate in mitophagy step(s) likely after the translocation of Parkin to membranes. It should be noted that GFP‐Parkin‐attached structures in Stx17‐depleted cells were insensitive to digitonin treatment (Fig EV4C), although most translocated GFP‐Parkin in control cells appeared to dissociate from membrane structures upon digitonin treatment (Fig EV4B). Moreover, cholesterol depletion by MβCD or nystatin affected mitochondrial aggregation (Appendix Fig S1). Thus, the initial GFP‐Parkin binding regions are digitonin‐insensitive structures, and cholesterol‐rich structures are necessary for GFP‐Parkin to cover whole mitochondria. As the MAM structure is rich in cholesterol and sphingolipids, thus resembling lipid rafts (Herrera‐Cruz & Simmen, 2017), and sensitive to digitonin (Oliferenko et al, 1999), it is tempting to speculate that the digitonin‐sensitive structures required for GFP‐Parkin expansion correspond to the MAM and ER–mitochondria interface.

Several studies reported that the ER–mitochondria interface is important for autophagosome formation in mitophagy. Yang and Yang (2013) demonstrated that contact regions between the ER and impaired mitochondria are initiation sites for local LC3 recruitment. Moreover, Gelmetti et al (2017) reported that PINK1 and a subunit of the PI3‐kinase complex, Beclin1, relocalize to MAM during mitophagy and promote ER–mitochondria tethering and autophagosome formation. In PGAM5‐depleted cells, FLAG‐DFCP1 was distant from GFP‐Parkin‐positive membrane structures, although the formation of LC3‐positive autophagosomes occurred (Figs 5A and B, and EV5A). The most straightforward interpretation of this is that loss of PGAM5 abrogates the link between ubiquitinated mitochondria and autophagosomes. FUNDC1 requires dephosphorylation at Ser13 by PARL‐cleaved PGAM5 for the interaction with autophagosome‐bound LC3 in mitophagy (Chen et al, 2014; Wu et al, 2014). We demonstrated that Stx17 is prerequisite for the interaction of cleaved PGAM5 and FUNDC1. Recently, Wu et al (2016) reported that in response to hypoxia FUNDC1 changes its binding partner from unknown protein(s) associating with CNX to Drp1 to drive mitochondrial division for promoting mitophagy. It is interesting to note that there is a reciprocal relationship between Stx17 and FUNDC1 such that Stx17 interacts with Drp1 under normal conditions, but is replaced by FUNDC1 during mitophagy (Fig 8). Stx17 assists FUNDC1 to function as a mitophagy receptor by regulating the localization and interaction of PGAM5 with FUNDC1. Stx17 also facilitates autophagosome formation by recruiting the PI3‐kinase complex to the MAM through interaction with Atg14L (Hamasaki et al, 2013; Arasaki et al, 2015). These complex reactions occur likely at the MAM that has unique membrane compositions and features.

Figure 8. Reciprocal relationship between Stx17 and FUNDC1 with respect to Drp1 binding.

In healthy cells, Stx17 facilitates normal division of mitochondria (Mt) by interacting with Drp1. Upon mitophagy stimulation, Stx17 dissociates from Drp1 and interacts with Atg14L for autophagosome formation. In the case of autophagy, mitochondria elongate due to inactivation of Drp1, whereas in mitophagy, FUNDC1 (FDC1) is dephosphorylated by cleaved PGAM5 (PGAM5*), which facilitates excessive division of mitochondria (Mt*) by interacting with Drp1. Stx17 supports this process by releasing PGAM5. Under normal division conditions, FUNDC1 associates with CNX through unknown protein(s).

Materials and Methods

Animals

All animal procedures and experiments were approved by the Animal Care Committee of Tokyo University of Pharmacy and Life Sciences and conducted according to the guidelines of the committee.

Chemicals and antibodies

Chemicals were obtained from the following sources: CCCP, digitonin, puromycin, hygromycin B, and MβCD were obtained from Wako Chemicals. Wortmannin and nystatin were obtained from Sigma‐Aldrich. MitoTracker Deep Red was purchased from Thermo Fisher Scientific. The following antibodies were obtained from Sigma‐Aldrich: monoclonal FLAG (No. F3165: 1/300 for IF and PLA) and polyclonal FLAG (No. F7425: 1/3,000 for IB). The following antibodies were obtained from BD Bioscience Pharmingen: Tom20 (No. 612278: 1/300 for IF, 1/500 for IB), Tim23 (No. 4339759: 1/3,000 for IB), CNX (No. 610523: 1/500 for IB), and Drp1 (No. 611112: 1/500 for IB). The following antibodies were obtained from MBL: monoclonal LC3 (No. M152‐3: 1/150 for IF), polyclonal LC3 (No. PM036: 1/3,000 for IB), Ub (clone FK2, No. D058‐3: 1/300), and p62 (No. PM045: 1/3,000 for IB). Antibodies against phospho‐Drp1 (Ser‐637), PGAM5, GFP, FUNDC1, GST, penta‐His, and maltose‐binding protein (MBP) were obtained from Cell Signaling: (No. 4867: 1/1,000 for IB), Abcam (No. ab126534: 1/200 for IF and 1/800 for IB), Thermo Fisher Scientific (No. A6455: 1/3,000 for IB), Atlas Antibodies (No. HPA038773: 1/1,000 for IB), Santa Cruz Biotechnology (No. sc‐459: 1/2,000 for IB), (No. 34660: 1/1,000), and New England Biolabs (No. E8030: 2,000 for IB), respectively. Alexa Fluor 350, 488, and 594 goat anti‐mouse and ‐rabbit antibodies (Nos. A‐21049, A‐11001, A‐11005, A‐11008, and A‐11012; 1:100 dilution, respectively) were obtained from Thermo Fisher Scientific. The preparation of an antibody against Stx17 was described previously (Arasaki et al, 2015).

Plasmids

Plasmids for Su9‐GFP, the DFCP1 clone, and PGAM5‐FLAG were kindly supplied by Dr. Naotada Ishihara (Kurume University), Dr. Noboru Mizushima (University of Tokyo), and Dr. Shiori Sekine University of Tokyo) and Kosuke Takeda (Nagasaki University), respectively. The plasmid to express APEX2‐GFP‐binding peptide was obtained from Addgene.

The vectors pGEX4T‐3, pQE30, and pMalc‐2X were used for the production of GST‐, His₆‐ and MBP‐tagged constructs. Constructions of mutants were performed by inverse PCR.

Recombinant proteins and binding experiments

Proteins were expressed in Escherichia coli (BL21 codon plus RP strain) and then solubilized in buffer containing 25 mM HEPES‐KOH (pH 7.4), 500 mM NaCl, 1 mM MgCl₂, 1 mM dithiothreitol, and 1% Triton X‐100. The MBP‐, GST‐ and His₆‐tagged proteins were purified using amylose resin (New England Biolabs), glutathione Sepharose 4B (GE Healthcare), and Ni‐NTA agarose (Qiagen), respectively.

For binding experiments, recombinant MBP proteins (0.1 or 0.2 μM) were incubated with amylose resin (25 μl) in 200 μl of incubation buffer (25 mM HEPES‐KOH (pH 7.2), 100 mM NaCl, 1 mM MgCl₂, and 0.2% Triton X‐100) at 4°C for 60 min. The resin was washed two times with washing buffer (25 mM HEPES‐KOH (pH 7.2), 100 mM NaCl, and 1 mM MgCl₂), and then incubated with GST‐ or His₆‐tagged proteins (0.1 μM) at 4°C overnight with gentle rotation. The resin was washed two times with washing buffer, and SDS sample buffer was added. The samples were subjected to SDS–PAGE and analyzed.

Cell culture

293T cells were grown in DMEM supplemented with 50 IU/ml penicillin, 50 μg/ml streptomycin, and 10% fetal calf serum. Establishment and growth of HeLa cells stably expressing FLAG‐Stx17 wild type or FLAG‐Stx17 (K254C) were described previously (Arasaki et al, 2015). HeLa cells stably expressing GFP‐Parkin were grown in DMEM supplemented with 50 IU/ml penicillin, 50 μg/ml streptomycin, 10 μg/ml puromycin, and 10% fetal calf serum. To establish transfectants stably expressing PINK1‐FLAG, HeLa cells stably expressing GFP‐Parkin were transfected with pcDNA3.1/Hygro (+)‐PINK1‐FLAG and screened in the presence of 10 μg/ml puromycin and 250 μg/ml hygromycin B. For starvation of cells, the cells were rinsed with PBS twice and then incubated in EBSS.

Transfection

Transfection was carried out using Lipofectamine 2000 (Invitrogen).

RNA interference

The following siRNAs were used:

• PGAM5: 5′‐CCAUAGAGACCACCGAUAU‐3′

• Stx17 (440): 5′‐GGUAGUUCUCAGAGUUUGAUU‐3′

• Stx17 (3′ untranslated region) : 5′‐GGAAAUUAAUGAUGUAAGA‐3′

• PINK1: 5′‐GAAAUCGGACAACAUCCUUUU‐3′

• Mfn1: 5′‐CGAAACCAGAUGAACCUUU‐3′

• Mfn2: 5′‐AGAGGGCCUUCAAGCGCCA‐3′

• PACS‐2: 5′‐AACACGCCCGUGCCCAUGAAC ‐3′

• Drp1: 5′‐AAGCAGAAGAAUGGGGUAAAU‐3′

• FUNDC1: 5′‐GCAGCACCUGAAAUCAACA‐3′

• Atg14L: 5′‐UUUGCGUUCAGUUUCCUCACUGCGC‐3′

• Atg9: 5′‐GUACAUGAAUUGCUUCUUG‐3′

• siRNAs were purchased from Japan Bio Services (Asaka, Japan). HeLa cells were grown on 35‐mm or 10‐cm dishes, and siRNAs were transfected at a final concentration of 200 nM using Oligofectamine (Invitrogen) according to the manufacturer's protocol. In most cases, siRNA knockdown was performed for 72 h. When necessary, plasmid transfection was carried out at 48 h after siRNA transfection, and the cells were incubated for 24 h.

Immunoprecipitation

Unless otherwise stated, 293T cells expressing FLAG‐tagged proteins were lysed in lysis buffer (20 mM HEPES‐KOH (pH 7.2), 150 mM KCl, 2 mM EDTA, 1 mM dithiothreitol, 1 μg/ml leupeptin, 1 μM pepstatin A, 2 μg/ml aprotinin, and 1 mM phenylmethylsulfonyl fluoride) containing 1% Triton X‐100. After centrifugation, the supernatants were immunoprecipitated with anti‐FLAG M2 affinity beads (Sigma‐Aldrich). The bound proteins were eluted with SDS sample buffer and then analyzed by IB.

Immunofluorescence microscopy

For immunofluorescence microscopy, cells were fixed for 20 min with 4% paraformaldehyde at room temperature or with ice‐cold methanol and then observed under an Olympus Fluoview 300 or 1000 laser scanning microscope. Unless specifically stated, representative images of at least three independent experiments are shown in figures. To capture triple staining images, an Olympus BX53 microscope attached with a DP53 CCD camera was used.

Electron microscopy

HeLa cells were transfected with cDNAs for PGAM5‐GFP and APEX2‐GFP‐binding peptide (Ariotti et al, 2015) simultaneously. Two days after transfection, the cells were fixed with 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.4), for 1 h and treated with a reaction solution containing 1 mg/ml 3,3′‐diaminobenzidine and 0.03% hydrogen peroxide (Martell et al, 2012). They were further treated with 2% osmium tetroxide followed by 2% uranyl acetate, dehydrated, and embedded in epoxy resin. Ultrathin sections were observed by JEM1011 transmission electron microscope.

Subcellular fractionation

Subcellular fractionation was performed as described previously (Wieckowski et al, 2009). Experiments were repeated two or three times with similar results.

PLA

PLA was conducted using a PLA kit (Sigma‐Aldrich) according to the manufacturer's protocol. Thirty cells were analyzed in each assay. Determination of the number of PLA dots was performed using the ImageJ software (NIH). In each experiment, at least 25 cells were examined.

Digitonin treatment

Digitonin was dissolved in DMSO before use. Cells were incubated for 5 min at room temperature with 0.03 mg/ml digitonin in KHM buffer (25 mM HEPES (pH 7.2), 125 mM potassium acetate, 2.5 mM magnesium acetate, 1 mM dithiothreitol, and 1 mg/ml glucose).

Drosophila genetics

Fly culture and crosses were performed on standard fly food containing yeast, cornmeal, and molasses, and the flies were raised at 25°C. w ¹¹¹⁸ was used as a wild‐type genetic background. All fly stocks and GAL4 lines used in this study were obtained from the Bloomington Drosophila Stock Center and have been previously described: PGAM5 ¹ (Imai et al, 2010) and UAS‐Stx17‐FLAG (Takáts et al, 2014).

Statistics

The results were averaged, expressed as means with SEM, and, except for the data in Fig 7, analyzed by means of a paired Student's t‐test. Dunnett's test was used for the analysis of the data in Fig 7. The P‐values are indicated by asterisks in the figures with the following notations: *P < 0.05; **P < 0.01; ***P < 0.001.

Author contributions

This study was conceived and designed by MS, KA, and MT. MS, HK, KA, TA, and NH performed the experiments. YI, TI, KS‐F, and NH performed the Drosophila analysis. ND performed the LC‐MS/MS analysis. JC and TF performed the electron microscopic analysis of HeLa cells. HI and YW supported the experiments.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Source Data for Expanded View

Review Process File

Source Data for Figure 1

Source Data for Figure 2

Source Data for Figure 3

Source Data for Figure 4

Source Data for Figure 5

Source Data for Figure 6

Source Data for Figure 7

The EMBO Journal (2018) 37: e98899