Mutations in the gene encoding Parkin are a major cause of recessive Parkinson's disease. Recent work has shown that Parkin translocates from the cytosol to depolarized mitochondria and induces their autophagic removal (mitophagy). However, the molecular mechanisms underlying Parkin-mediated mitophagy are poorly understood. Here, we investigated whether Parkin interacts with autophagy-regulating proteins. We purified Parkin and associated proteins from HEK293 cells using tandem affinity purification and identified the Parkin interactors using mass spectrometry. We identified the autophagy-promoting protein Ambra1 (activating molecule in Beclin1-regulated autophagy) as a Parkin interactor. Ambra1 activates autophagy in the CNS by stimulating the activity of the class III phosphatidylinositol 3-kinase (PI3K) complex that is essential for the formation of new phagophores. We found Ambra1, like Parkin, to be widely expressed in adult mouse brain, including midbrain dopaminergic neurons. Endogenous Parkin and Ambra1 coimmunoprecipitated from HEK293 cells, SH-SY5Y cells, and adult mouse brain. We found no evidence for ubiquitination of Ambra1 by Parkin. The interaction of endogenous Parkin and Ambra1 strongly increased during prolonged mitochondrial depolarization. Ambra1 was not required for Parkin translocation to depolarized mitochondria but was critically important for subsequent mitochondrial clearance. In particular, Ambra1 was recruited to perinuclear clusters of depolarized mitochondria and activated class III PI3K in their immediate vicinity. These data identify interaction of Parkin with Ambra1 as a key mechanism for induction of the final clearance step of Parkin-mediated mitophagy.
Parkinson's disease (PD) is an incurable movement disorder characterized by the relentless loss of substantia nigra dopaminergic neurons (Lees et al., 2009). Mutations in PARK2, the gene encoding the E3 ubiquitin ligase Parkin, are a major cause of autosomal recessive PD, probably through a loss-of-function mechanism (Kitada et al., 1998; Shimura et al., 2000). Parkin-deficient flies (Greene et al., 2003; Clark et al., 2006; Park et al., 2006) and mice (Palacino et al., 2004) and fibroblasts from PD patients with PARK2 mutations (Mortiboys et al., 2008) show evidence of mitochondrial dysfunction. This is intriguing, because Parkin is essentially a cytosolic protein (Narendra et al., 2008). However, recent work has shown that Parkin selectively and rapidly translocates from the cytosol to depolarized mitochondria and subsequently induces their autophagic removal (mitophagy) (Narendra et al., 2008, 2010a; Geisler et al., 2010; Matsuda et al., 2010; Vives-Bauza et al., 2010). Thus, Parkin may protect neurons by triggering the clearance of damaged mitochondria that would otherwise be a source of oxidative stress and apoptosis-inducing factors.
The biogenesis of autophagosomes is tightly regulated by a complex network of autophagy-regulating proteins (Behrends et al., 2010; Yang and Klionsky, 2010). The process begins with the formation of a crescent-shaped, double-membrane precursor structure (the phagophore), which then expands and closes to form an autophagosome, capable of fusion with a lysosome. How translocated Parkin activates the engulfment of mitochondria by autophagic membranes is not well understood. Recent work has shown that translocated Parkin ubiquitinates the mitochondrial outer membrane protein voltage-dependent anion channel 1 (VDAC1) (Geisler et al., 2010). Ubiquitinated VDAC1 may then connect with ambient autophagic membranes via recruitment of the adaptor protein p62/SQSTM1 (Geisler et al., 2010). p62 can bind simultaneously to ubiquitin and LC3, a protein associated with autophagosomes and expanding phagophores (Pankiv et al., 2007). However, other studies have found p62 to be dispensable for Parkin-mediated mitophagy (Narendra et al., 2010b; Okatsu et al., 2010). Whether Parkin also interacts directly with the autophagic machinery or only indirectly via VDAC1 ubiquitination and p62 recruitment is unknown. It is also unclear which factors trigger the initiation of new phagophore formation around depolarized mitochondria after Parkin translocation.
Here, we have used tandem affinity purification (TAP) and mass-spectrometric analysis to investigate whether Parkin interacts with components of the autophagic machinery. We have identified the autophagy-promoting protein Ambra1 (activating molecule in Beclin1-regulated autophagy) as a Parkin interactor. Ambra1 activates autophagy in the CNS by stimulating the activity of the class III phosphatidylinositol 3-kinase (PI3K) complex that is essential for the initiation of phagophore formation (Fimia et al., 2007). Functional deficiency of Ambra1 in mice leads to impaired autophagy, excessive apoptosis, accumulation of ubiquitinated proteins, and embryonic lethality (Fimia et al., 2007). We show that the interaction of Parkin with Ambra1 strongly increases during mitochondrial depolarization. Ambra1 is recruited to perinuclear clusters of depolarized mitochondria and activates class III PI3K in their immediate vicinity. This mechanism allows for localized induction of the final clearance step of Parkin-mediated mitophagy.
Materials and Methods
The following primary antibodies were used for immunofluorescence (IF) or Western blot (WB): mouse anti-Parkin (IF, 1:500; WB, 1:4000; clone PRK8; Millipore), rabbit anti-Ambra1 (WB, 1:5000; FLJ20294; Strategic Diagnostics) (Di Bartolomeo et al., 2010a), rabbit anti-Ambra1 (IF, 1:250; Covalab) (Di Bartolomeo et al., 2010a), mouse anti-FLAG (IF, 1:1000; WB, 1:5000; clone M2; Sigma), rabbit anti-FLAG (WB, 1:1000; Sigma), mouse anti-β-actin (WB, 1:5000; clone AC-15; Sigma), mouse anti-presenilin-1 (WB, 1:1000; clone PS1-loop; Millipore), mouse anti-calretinin (WB, 1:1000; Swant), mouse anti-MAP2 (IF, 1:500; clone SMI 52; Covance), mouse anti-tyrosine hydroxylase (IF, 1:200; Millipore), mouse anti-GFAP (IF, 1:500; clone G-A-5; Sigma), mouse anti-2′,3′-cyclic nucleotide-3′-phosphodiesterase (CNPase) (IF, 1:200; Millipore), mouse anti-cytochrome c (IF, 1:200; clone 7H8.2C12; BD Biosciences), sheep anti-cytochrome c (IF, 1:200; c5723; Sigma), mouse anti-KDEL (Lys-Asp-Glu-Leu) (IF, 1:100; clone 10C3; Stressgen), mouse anti-GM130 (IF, 1:200; BD Biosciences), mouse anti-LC3 (IF, 1:50; MBL International), rat anti-LAMP-1 (IF, 1:200; 1D4B; Santa Cruz Biotechnology), rabbit anti-Tom20 (IF, 1:500; FL-145; Santa Cruz Biotechnology), mouse anti-HA (WB, 1:1000; Covance), goat anti-c-myc (IF, 1:200; Bethyl), rabbit anti-DNA polymerase γ (IF, 4 μg/ml; Ab2969; Abcam), mouse anti-GAPDH (WB, 1:5000; 6C5; HyTest), mouse anti-cytochrome c oxidase subunit 4 (COX IV) (WB, 1:1000; 20E8; Invitrogen), and rabbit anti-Pex14p (IF, 1:100; gift from Dr. M. Fransen, K.U. Leuven, Leuven, Belgium). Peroxidase-linked secondary antibodies for WB were from GE Healthcare. Secondary antibodies for IF were donkey anti-mouse IgG Alexa Fluor-488 and -647, anti-rabbit and anti-sheep IgG Alexa Fluor-555, anti-rat IgG Alexa Fluor-488, and anti-goat IgG Alexa Fluor-647 (Invitrogen).
cDNAs and siRNAs.
The constructs encoding untagged wild-type Parkin and HA-tagged wild-type and mutant Parkin were described previously (Winklhofer et al., 2003; Hampe et al., 2006; Van Humbeeck et al., 2008). A pcDNA3.1 vector containing the His6–FLAG cDNA for N-terminal tagging was a gift from Dr. M. Fukata (National Institute for Physiological Sciences, Aichi, Japan). To generate Parkin with an N-terminal His6–FLAG tag, PCR from untagged Parkin cDNA was performed using forward primer 5′-CCGTCGACATAGTGTTTGTCAGGTTCAACTCCAGCC-3′ and reverse primer 5′-GGTCTAGACTACACGTCGAACCAGTCGTCC-3′. The PCR product was digested with SalI/XbaI and cloned into XhoI/XbaI-digested His6–FLAG vector, resulting in full-length N-terminally His6–FLAG-tagged Parkin. The construct was verified by sequencing and WB. A pCMV6–XL4 vector containing human Ambra1 cDNA with a C-terminal Myc—FLAG tag and a pCMV6 Entry vector containing untagged human Ambra1 cDNA were purchased from Origene and verified by sequencing and WB. Ambra1 deletion constructs encoding amino acids 1–532 [N-terminal fragment (NF)], 533–751 [middle fragment (MF)], and 751–1208 [C-terminal fragment (CF)] with C-terminal Myc–FLAG tags were generated from full-length Ambra1 by PCR amplification and cloned into the SgfI/MluI restriction sites of the pCMV6 Entry vector in-frame with the Myc–FLAG tag provided by the plasmid. Resulting expression constructs were verified by sequencing and WB. PCR amplification of these deletion constructs was performed using the following oligonucleotides: NF forward, GCCGCGATCGCCATGAAGGTTGTCCC; NF reverse, CGTACGCGTCTCAGTCCGCTCGAGAGGTGG; MF forward, GCCGCGATCGCCATGGGCCAAACGCCCAGCTCCAGC; MF reverse, CGTACGCGTCCCTGCCAGGACCCTGTTAAC; CF forward, GCCGCGATCGCATGGCAGTGATCGGTGATGGACAG; and CF reverse, CGTACGCGTCCTGTTCCGTGG. FLAG-tagged Parkin deletion constructs encoding amino acids 1–76 [ubiquitin-like domain (UBL)], 77–237 (Linker), and 217–465 [RING1–IBR–RING2 domain (RIR)] were a gift from Dr. J. Feng (State University of New York at Buffalo, Buffalo, NY) (Yang et al., 2005). An HA-tagged ubiquitin plasmid construct was a gift from Dr. K. Winklhofer (Max Planck Institute for Biochemistry, Martinsried, Germany) (Henn et al., 2007). A FLAG–IκB kinase γ (IKKγ) plasmid construct was a gift from Dr. C. Scheidereit (Max Delbrück Center for Molecular Medicine, Berlin, Germany) (Tegethoff et al., 2003). A p40(phox)PX–EGFP plasmid construct was a gift from Dr. J. Jung (University of Southern California, Los Angeles, CA) (Liang et al., 2006). The target sequences of Ambra1 siRNA 1, Ambra1 siRNA 2, and control siRNA were 5′-AGAACTGCAAGATCTACAA-3′, 5′-GGCCTATGGTACTAACAAA-3′, and 5′-AATTCTCCGAACGTGTCACGT-3′, respectively (Qiagen) (Ahn et al., 2004; Fimia et al., 2007).
Cell cultures and transfection.
HEK293 cells and HeLa cells were grown in DMEM (Invitrogen) with 25 mm glucose, supplemented with 10% fetal bovine serum (Greiner), l-glutamax (4 mm), penicillin (100 U/ml), streptomycin (100 μg/ml), and (only for HEK293 cells) non-essential amino acids (1%). SH-SY5Y cells were grown in DMEM with l-glutamax (4 mm) and 25 mm glucose, supplemented with 15% fetal bovine serum, gentamycin (200 μg/ml), and 1% non-essential amino acids. Midbrain cultures were prepared from E14 mouse embryos or postnatal day 1 mouse pups of either sex based on a hippocampal culture protocol (Kaech and Banker, 2006). Procedures were approved by the local ethical committee. Midbrain cultures were grown in Neurobasal medium (Invitrogen) supplemented with B-27 (Invitrogen), glutamine (2 mm; Invitrogen), and GDNF (10 ng/ml; Alomone Labs) and immunostained after 6 d in vitro. HEK293, HeLa, and SH-SY5Y cells were transiently transfected with 80 ng/cm2 cDNA, unless otherwise indicated. TransIT-293 Reagent (Mirus Bio) was used for cDNA transfections in HEK293 cells and TransIT-HeLaMONSTER kit (Mirus Bio) for cDNA transfections in HeLa and SH-SY5Y cells. For cotransfections, we used 60 ng/cm2 of each cDNA. HEK293, HeLa, and SH-SY5Y cells were transfected with 50 nm siRNA using the TransIT-TKO Reagent (Mirus Bio). Transfection efficiency for siRNAs in these cell lines was 75–90%, as determined in control experiments with siRNAs with a 3′ fluorescent tag (Qiagen). For the combined transfection of cDNA and siRNA in HeLa cells, TransIT-HeLaMONSTER kit and TransIT-TKO Reagent were used according to the instructions of the manufacturer. Carbonyl cyanide m-chlorophenylhydrazone (CCCP) (Calbiochem), valinomycin (Sigma), or DMSO were added 1 d after transfection.
TAP and mass spectrometry.
Twenty-four hours after transfection, HEK293 cells were washed with ice-cold PBS, collected with a cell scraper, and resuspended in buffer A (350 mm NaCl, 2.7 mm KCl, 8.1 mm Na2PO4, 1.8 mm KH2PO4, and 1% Triton X-100, pH 7.4) containing protease inhibitor cocktail (Roche). After solubilization for 30 min at 4°C, the sample was cleared by centrifugation for 15 min at 20,000 × g. The supernatant was then incubated with pre-equilibrated anti-Flag M2 affinity gel (Sigma) for 1 h at 4°C. The beads were washed with buffer A, and bound proteins were eluted with buffer B (50 mm Tris-HCl and 150 mm NaCl, pH 7.4) containing FLAG peptides (0.1 μg/μl; Sigma). The eluate was supplemented with imidazole (10 mm) and incubated with nickel beads (PDC Sepharose CL-4B matrix; Affiland) at 4° for 1 h. The beads were washed with buffer A supplemented with 10 mm imidazole. Bound proteins were eluted two consecutive times with buffer A supplemented with 250 mm imidazole. To concentrate the eluate, 4 vol of methanol was added, followed by overnight storage at −20°C. After centrifugation at 20,000 × g for 10 min at 4°C and removal of the supernatant, the pellet was air dried for 1 h, resuspended in 5× Laemmli's buffer, and heated at 95°C for 10 min. After SDS-PAGE and staining with Coomassie blue (Pierce), stained bands and the corresponding regions in the control lane were excised, in-gel digested with trypsin, and analyzed by liquid chromatography tandem mass spectrometry using an LTQ Orbitrap XL mass spectrometer (Thermo Fisher Scientific) as described (Ghesquière et al., 2009). The Mascot Distiller software (version 2.2.1; Matrix Science) was used to generate Mascot generic files (Ghesquière et al., 2009) to identify peptides and proteins using the Mascot database search algorithm in the Swiss-Prot database (release 56) with restrictions set to human proteins. Acetylation of protein N termini, deamidation of asparagine and glutamine, pyroglutamate formation of N-terminal glutamine, methionine oxidation to methionine sulfoxide, and S-propionamide modification on cysteine were set as variable modifications. Trypsin was set as enzyme allowing for one missed cleavage. The mass tolerance on the precursor ion was set to 10 ppm and on fragment ions to 0.5 Da. Only peptides that were ranked first and scored higher than the corresponding Mascot threshold score for identity at 99% confidence were withheld. Two complete runs of TAP-MS were performed. Only proteins that were not present in the negative control lane were considered positively identified.
Coimmunoprecipitation and Western blot.
For coimmunoprecipitation from HEK293 cells, cells were washed with ice-cold PBS, removed with a scraper, and resuspended in lysis buffer (20 mm Tris-HCl and 1% Triton X-100, pH 7.4). After solubilization for 30 min at 4°C, insoluble material was removed by centrifugation at 20,000 × g for 5 min. Protein concentrations were determined using the Bio-Rad Protein assay. The sample was precleared by incubation with protein A or G Sepharose (GE Healthcare) for 30 min at 4°C. The beads were removed by centrifugation and the samples were incubated for 1 h at 4° with mouse anti-Parkin (Millipore), rabbit anti-Ambra1 (Strategic Diagnostics), or control IgG (purified mouse or rabbit IgG; Millipore). The antigen–antibody complexes were captured by addition of protein A or G Sepharose for 1 h at 4°C and washed three times with buffer A. Proteins were released from the beads by heating at 95°C for 10 min in 5× Laemli's buffer, followed by SDS-PAGE, blotting onto PVDF membranes, and incubation with blocking solution, primary and secondary antibodies. Immunoreaction was visualized with ECL (Pierce) or ECL Plus (GE Healthcare) and exposure to Amersham Hyperfilm ECL (GE Healthcare). The density of scanned signals was measured with UN-SCAN-IT gel 6.1 (Silk Scientific). For coimmunoprecipitation from brain, brain from 2- to 4-month-old mice was homogenized with a glass pestle in 10 vol of 20 mm Tris-HCl, pH 7.4, followed by trituration through a 200 μl pipette tip. Homogenates were centrifuged at 200 × g for 5 min, followed by centrifugation of the supernatant for 5 min at 350 × g. After an additional spin of the supernatant at 20,000 × g for 2 min, Triton X-100 was added to the supernatant to a final concentration of 1%. After incubation for 30 min at 4°C, the sample was centrifuged at 20,000 × g for 5 min. The protein concentration of the supernatant was determined, followed by preclearing and coimmunoprecipitation as described above.
Ubiquitination was assayed as described (Henn et al., 2007). One day after transfection, proteins were extracted in denaturating lysis buffer (50 mm Tris-HCl, pH 7.4, 5 mm EDTA, 1% SDS, 15 U/ml DNase I, and protease inhibitor cocktail) and boiled for 5 min. Protein extracts were then diluted 1:10 with non-denaturating lysis buffer (50 mm Tris-HCl, pH 7.4, 300 mm NaCl, 5 mm EDTA, 1% Triton X-100, and protease inhibitor cocktail), followed by preclearing and immunoprecipitation as described above.
Cytosolic and mitochondria-enriched cell fractions were prepared as described (Gegg et al., 2010). In brief, SH-SY5Y cells were scraped off the plate, washed in PBS, and homogenized with a glass pestle in isolation medium (250 mm sucrose, 1 mm EDTA, and 10 mm Tris, pH 7.4). Nuclei were removed by centrifugation (1500 × g) for 2 min at 4°C. After centrifugation at 11,800 × g for 15 min at 4°C, the pellet (enriched in mitochondria) was separated from the supernatant (containing cytosol) and resuspended in isolation medium.
Immunocytochemistry, confocal microscopy, and image analysis.
Cells grown on poly-d-lysine (Sigma)-coated coverslips were washed with PBS, fixed in 4% paraformaldehyde in PBS for 30 min, washed with PBS, and blocked and permeabilized for 45 min in PBS containing 5% donkey serum and 0.1% Triton X-100. Cells were incubated for 1 h with primary antibodies in PBS containing 0.1% Triton X-100 and 2% donkey serum. After washing, secondary antibodies were applied for 1 h in PBS containing 0.1% Triton X-100 and 2% donkey serum. Nuclei were stained with TOTO-3 iodide (1 μm in the secondary antibody solution; Invitrogen) or with DAPI in the mounting medium. Cells were mounted onto microscope slides with Vectashield (Labconsult). Confocal images with 1 μm slice thickness were acquired with an Axioskop 2 microscope (LSM510 META; Carl Zeiss) at room temperature equipped with an AxioCam HR camera (Carl Zeiss), using a 63×, NA 1.4 oil Plan-Apochromat objective. Brightness and contrast were adjusted with NIH ImageJ software. In translocation, clustering, and mitophagy experiments, random images were captured and analyzed by an investigator blinded to the experimental condition. A minimum of 100 cells per condition were analyzed per experiment.
Significance of differences was analyzed with two-tailed Student's t test for comparison between two groups and with one-way ANOVA and post hoc Holm–Sidak test for comparison between more than two groups (SigmaStat 3.5; Systat). Values and error bars represent mean ± SEM.
Ambra1 is a non-substrate interactor of Parkin
TAP combines two high-affinity binding and two highly specific elution steps, resulting in the enrichment of the protein of interest and its interacting proteins (Fig. 1A) (Collins and Choudhary, 2008; Wakabayashi et al., 2009). A His6–FLAG TAP tag was linked to the N terminus of Parkin (Fig. 1B). Similar to untagged Parkin, TAP-tagged Parkin translocated to depolarized mitochondria and induced their removal, indicating that the tag did not interfere with Parkin-mediated mitophagy (data not shown). For purification, TAP-tagged Parkin was expressed in HEK293 cells in basal culture conditions, whereas HEK293 cells transfected with untagged Parkin served as negative control. Expression levels of TAP-tagged and untagged Parkin were similar (data not shown). Cell extracts were subjected to TAP in native conditions, followed by concentration and SDS-PAGE of the final TAP eluates. Figure 1C illustrates the strong enrichment of TAP-tagged Parkin. Approximately 10% of the TAP-tagged Parkin present in the initial extract was recovered in the final TAP eluate (data not shown). After Coomassie blue visualization of bound proteins, stained bands (Fig. 1D), along with the corresponding portions of the negative control lane, were excised, in-gel digested with trypsin, and analyzed by mass spectrometry. We confirmed a total of eight previously reported Parkin interactors (Table 1). Among these, histone deacetylase 6 (HDAC6) (Lee et al., 2010) and α- and β-tubulin (Lee et al., 2010; Okatsu et al., 2010; Vives-Bauza et al., 2010) have been implicated previously in Parkin-mediated mitophagy.
Interestingly, we also identified the 142 kDa autophagy-activating protein Ambra1 in the TAP eluate (Fig. 1D; Table 1). This was confirmed by SDS-PAGE and Western blot analysis of the TAP eluate with an Ambra1-specific antibody (Fig. 1E–G). In addition, the interaction of Parkin and Ambra1 was validated by reciprocal coimmunoprecipitation of the endogenous proteins from HEK293 cells and adult mouse brain (Fig. 1H–K).
To determine which domains of Parkin and Ambra1 interact, we used various deletion constructs of Parkin (Fig. 1L) and Ambra1 (Fig. 1N). Coimmunoprecipitation experiments in HEK293 cells indicated that the interaction of Parkin and Ambra1 was mediated by the Linker domain of Parkin and the N-terminal region of Ambra1 (Fig. 1M,O).
We then transfected wild-type Parkin and the PD-linked Parkin mutants A82E, K161N, K211N, R256C, and R275W in HEK293 cells to compare the ability of the Parkin variants to coimmunoprecipitate with Ambra1 but found that these mutations did not disrupt Ambra1 binding (data not shown).
Ambra1 is expressed in both embryonic and adult CNS (Fimia et al., 2007). We compared the regional distributions of Ambra1 and Parkin in adult mouse brain using Western blotting. Like Parkin, Ambra1 was expressed throughout the brain, including midbrain and striatum (Fig. 2A). We then used a specific anti-Ambra1 antibody (Fig. 2B,C) for immunostaining of endogenous Ambra1 in embryonic and postnatal mesencephalic cultures. Ambra1 immunostaining was strong in neurons (Fig. 2D), including dopaminergic neurons (Fig. 2E,F), and was weak or undetectable in non-neuronal cells, such as astrocytes and oligodendrocytes (Fig. 2G,H). Similar to Parkin, Ambra1 was predominantly localized in the cytosol (Fig. 2I) and showed little colocalization with cellular organelles in basal conditions (Fig. 2J–N).
Because Parkin is an E3 ubiquitin ligase (Shimura et al., 2000), we asked whether Ambra1 might be a substrate for Parkin-mediated ubiquitination. We performed a ubiquitination assay (Fig. 3) (Henn et al., 2007) in HeLa cells, a cell line with little or no endogenous Parkin expression (Denison et al., 2003; Pawlyk et al., 2003). HeLa cells were transfected with various combinations of Parkin, Ambra1, and HA-tagged ubiquitin. Proteins were extracted in denaturing conditions to dissociate Ambra1 from its binding partners. The extracts were then diluted in non-denaturing buffer to allow immunoprecipitation of Ambra1 with anti-Ambra1 antibody. Finally, Western blot analysis with anti-HA was performed on the immunoprecipitates to detect ubiquitinated Ambra1 (Fig. 3A). As a positive control for Parkin-mediated ubiquitination, we transfected the Parkin substrate nuclear factor κB essential modulator (IKKγ) (Henn et al., 2007). As reported, cotransfection of Parkin clearly increased the amount of ubiquitinated IKKγ (Fig. 3B) (Henn et al., 2007). In contrast, ubiquitination of Ambra1 was not increased in the presence of Parkin (Fig. 3A). The detection of small amounts of ubiquitinated Ambra1 in HeLa cells lacking Parkin suggested that Ambra1 can be ubiquitinated by other E3 ligases than Parkin. We performed similar ubiquitination experiments in HEK293 cells but again did not find an increase in Ambra1 ubiquitination during cotransfection of Parkin (Fig. 3C). Together, these findings suggest that Ambra1 is not a target for Parkin-mediated ubiquitination but do not exclude the possibility that Parkin may ubiquitinate Ambra1 in other cell types or in different experimental conditions.
Mitochondrial depolarization promotes the interaction of Parkin with Ambra1
Parkin mediates mitophagy after prolonged dissipation of the mitochondrial membrane potential with protonophores, such as CCCP (Narendra et al., 2008, 2010a; Geisler et al., 2010; Matsuda et al., 2010; Vives-Bauza et al., 2010). We wondered whether mitochondrial depolarization might affect the interaction of Parkin with Ambra1. Indeed, exposure of untransfected HEK293 cells (Fig. 4A) or untransfected dopaminergic neural SH-SY5Y cells (Fig. 4B) to CCCP for 12 h strongly increased the binding of endogenous Parkin to Ambra1. In untransfected HEK293 cells, the amount of Parkin coimmunoprecipitated with anti-Ambra1 after 12 h treatment with CCCP (10 μm), normalized to the amount coimmunoprecipitated after 12 h treatment with DMSO, was 6.9 ± 1.8 (p < 0.05; n = 3). In untransfected SH-SY5Y cells, the amount of Parkin coimmunoprecipitated with anti-Ambra1 after 12 h of CCCP (25 μm) treatment, normalized to the DMSO condition, was 8.4 ± 1.4 (p < 0.01; n = 3). There was no clear increase in binding of the endogenous proteins after only 3 h exposure to CCCP (data not shown).
Next, we performed coimmunoprecipitation after cellular fractionation (Fig. 4C) in untransfected SH-SY5Y cells to determine whether the increased interaction of endogenous Parkin and Ambra1 during prolonged mitochondrial depolarization occurred in the cytosol or in the mitochondrial fraction. In basal conditions, binding of endogenous Parkin and Ambra1 was detected in both the cytosolic and mitochondria-enriched fractions (Fig. 4D). However, increased binding after prolonged exposure to CCCP was only observed in the mitochondrial fraction and not in the cytosolic fraction (Fig. 4D).
Parkin translocation to depolarized mitochondria is not Ambra1 dependent
The known role of Ambra1 as an activator of autophagy and the increase in Parkin–Ambra1 binding during mitochondrial depolarization prompted us to test for a possible role of Ambra1 in Parkin-mediated mitophagy. Parkin-mediated mitophagy occurs in two sequential steps: Parkin first translocates to depolarized mitochondria and subsequently activates their autophagic elimination (Narendra et al., 2008, 2010a; Geisler et al., 2010; Matsuda et al., 2010; Vives-Bauza et al., 2010). We first determined whether Ambra1 was involved in Parkin translocation. HeLa cells were transfected with Parkin and treated with 10 μm CCCP or DMSO for 3 h. As reported previously, CCCP induced a striking translocation of Parkin from the cytosol to mitochondria (Fig. 5A,B). Similarly, 3 h exposure to 1 μm valinomycin, a K+ ionophore that rapidly dissipates the mitochondrial membrane potential (Sharpe et al., 1995), redistributed Parkin from the cytosol to the mitochondria (Fig. 5A,B). Ambra1 did not translocate to depolarized mitochondria to the same extent as Parkin after 3 h of CCCP exposure, although there was some colocalization of Ambra1 with mitochondria (Fig. 5C).
We then asked whether CCCP had any effect on the distribution of endogenous Parkin in neural cells. Immunocytochemical staining of endogenous Parkin in SH-SY5Y cells was weak but detectable (Fig. 5D). After exposure to CCCP (25 μm) for 3 h, 20.6 ± 1.9% of SH-SY5Y cells showed clusters of Parkin immunoreactivity that colocalized with mitochondria, compared with only 5.5 ± 2.1% of cells after DMSO treatment (p < 0.01; n = 3) (Fig. 5D).
Next, we used RNA interference to reduce Ambra1 expression to 20–30% of control levels (Fig. 5E,F). However, Ambra1 downregulation had no effect on CCCP-induced Parkin translocation in Parkin-transfected HeLa cells (Fig. 5G,H) or in SH-SY5Y cells (Fig. 5I,J), indicating that Ambra1 is not required for this process.
Ambra1 contributes to Parkin-mediated mitophagy
Next, we investigated whether Ambra1 was involved in the clearance of mitochondria after Parkin translocation. As reported previously, 24 h exposure of HeLa cells to 10 μm CCCP induced the removal of mitochondria in a Parkin-dependent manner (Fig. 6A–F) (Narendra et al., 2008, 2010a; Geisler et al., 2010; Matsuda et al., 2010; Vives-Bauza et al., 2010). CCCP-induced, Parkin-dependent mitochondrial clearance was observed using markers of various mitochondrial compartments: Tom20, an outer membrane protein (Fig. 6A,B); cytochrome c, an intermembrane space protein (Fig. 6C,D); and DNA polymerase γ, a matrix protein (Fig. 6E,F). As reported previously (Narendra et al., 2008), CCCP-induced mitochondrial clearance in Parkin-transfected HeLa cells was inhibited by the lysosomal inhibitor bafilomycin A1 (100 nm) and the PI3K inhibitor 3-methyladenine (10 mm), confirming its autophagic nature (Fig. 6G). CCCP-induced elimination of mitochondria was selective, because peroxisomes were not removed (Fig. 6H) (Narendra et al., 2008). Exposure to valinomycin (1 μm) for 24 h also triggered Parkin-dependent mitochondrial clearance, similar to CCCP (Fig. 6A,B).
We wondered whether mitochondrial depolarization also induced mitochondrial clearance in untransfected neural cells that endogenously express Parkin. Indeed, exposure to CCCP for 24 h clearly increased the percentage of SH-SY5Y cells devoid of mitochondrial staining (Fig. 6I). The percentage of SH-SY5Y cells without detectable Tom20 immunoreactivity was 1.3 ± 1.3% after 24 h DMSO treatment versus 8.4 ± 1.4% after 24 h treatment with 25 μm CCCP (p < 0.05; n = 3). Finally, we quantified mitochondrial mass in untransfected SH-SY5Y cells using Western blot for the integral inner mitochondrial membrane protein cytochrome c oxidase subunit 4 (COX IV) (Fig. 6J,K). Treatment with CCCP for 24 h substantially reduced the amount of COX IV (Fig. 6J,K) (Lee et al., 2010), and this decrease was inhibited by bafilomycin A1 (Fig. 6J,K).
Next, we used siRNA-mediated downregulation of Ambra1 to examine its role in Parkin-mediated mitophagy. Interestingly, Ambra1 knockdown significantly suppressed CCCP-induced mitochondrial clearance in both Parkin-transfected HeLa cells (Fig. 7A,B) and untransfected SH-SY5Y cells (Fig. 7C,D), indicating that this process was dependent on Ambra1.
We then asked whether CCCP-induced mitophagy might be enhanced by overexpression of Ambra1. Ambra1 overexpression in HeLa cells lacking Parkin did not result in mitochondrial clearance after 24 h exposure to CCCP (Fig. 7E,G). However, coexpression of Ambra1 and Parkin resulted in enhanced CCCP-induced mitophagy compared with overexpression of Parkin alone (Fig. 7F,G). Thus, Ambra1 overexpression promoted the elimination of depolarized mitochondria but only in the presence of Parkin.
Ambra1 locally activates class III PI3K around depolarized mitochondria
We then investigated the mechanism by which Ambra1 contributes to Parkin-mediated mitophagy. Ambra1 activates autophagy by stimulating the activity of the class III PI3K that is crucial for the initiation of phagophore formation (Fimia et al., 2007). As shown above (Fig. 4A,B), exposure to CCCP for 12 h strongly increased the interaction of Parkin with Ambra1. We therefore hypothesized that recruitment of Ambra1 might enable translocated Parkin to locally activate class III PI3K around depolarized mitochondria.
To test this hypothesis, we used the p40(phox)PX–EGFP fusion protein. Because the PX domain of p40(phox) specifically binds to phosphatidylinositol 3-phosphate (PtdIns-3-P), the reaction product of class III PI3K, the p40(phox)PX–EGFP fusion protein can be used as a probe for measuring the subcellular levels and distribution of PtdIns-3-P (Kanai et al., 2001; Liang et al., 2006; Fimia et al., 2007). Hotspots of p40(phox)PX–EGFP disappeared in HeLa cells after treatment for 1 h with 75 nm of the specific PI3K inhibitor wortmannin (Fig. 8A), as reported previously (Kanai et al., 2001). We then cotransfected HeLa cells with p40(phox)PX–EGFP and Parkin. After 16 h exposure to 10 μm CCCP, mitochondrial staining had already disappeared in 16.6 ± 4.13% of Parkin-expressing cells (n = 6). In most of the Parkin-expressing cells that still contained mitochondria after 16 h of CCCP treatment, the mitochondria were concentrated in juxtanuclear clusters that have recently been called “mito-aggresomes” (Fig. 8B,D) (Geisler et al., 2010; Lee et al., 2010; Okatsu et al., 2010; Vives-Bauza et al., 2010). In contrast, mitochondria remained distributed as a network throughout the cytosol in cells lacking Parkin. Remarkably, the perinuclear mitochondrial clusters in Parkin-expressing cells were closely surrounded by and intermingled with p40(phox)PX–EGFP hotspots (Fig. 8B,D). A similar phenomenon was observed in SH-SY5Y cells transfected with p40(phox)PX–EGFP and treated with CCCP for 16 h (Fig. 8C). Parkin transfection or 16 h of CCCP treatment did not increase the total number of p40(phox)PX–EGFP hotspots per cell (data not shown). Interestingly, Parkin-transfected HeLa cells after 16 h of CCCP treatment showed a striking redistribution of Ambra1 to the juxtanuclear mito-aggresomes (Fig. 9A). In contrast, Ambra1 did not colocalize with mitochondria in Parkin-deficient HeLa cells after 16 h of CCCP treatment (Fig. 9A). Colocalization of Ambra1 with mitochondria in Parkin-expressing HeLa cells was clearly more prominent after 16 h (Fig. 9A) than after 3 h (Fig. 5C) of CCCP treatment. Endogenous Ambra1 in SH-SY5Y cells also redistributed to the mitochondria after 16 h of CCCP (Fig. 9B).
Finally, we tested whether recruited Ambra1 was responsible for the local activation of class III PI3K around the mitochondrial clusters. Downregulation of Ambra1 levels with RNAi did not interfere with Parkin-dependent mitochondrial clustering after 16 h of CCCP treatment but caused a significant reduction of PtdIns-3-P hotspots around mito-aggresomes (Fig. 10A,B). Together, these data showed that Parkin-dependent recruitment of Ambra1 led to local activation of class III PI3K around depolarized mitochondria.
Loss of function of the E3 ubiquitin ligase Parkin is one of the most common known causes of PD (Lees et al., 2009). Recent work has shown that Parkin surveys mitochondrial quality by promoting autophagic removal of depolarized mitochondria (Narendra et al., 2008, 2010a; Geisler et al., 2010; Lee et al., 2010; Matsuda et al., 2010; Okatsu et al., 2010; Vives-Bauza et al., 2010). Ambra1 was recently identified as an activator of the initial steps of phagophore formation in the nervous system (Fimia et al., 2007). Functional deficiency of Ambra1 leads to impaired autophagy, accumulation of ubiquitinated proteins, and excessive apoptosis (Fimia et al., 2007). Here, we demonstrate a Parkin–Ambra1 interaction, which strongly increases during mitochondrial depolarization and is crucial for subsequent mitophagy.
Parkin-mediated mitophagy involves a chronological sequence of events that are only beginning to be understood in molecular terms. The first step, translocation of Parkin from the cytosol to depolarized mitochondria, is dependent on PTEN (for phosphatase and tensin homolog)-induced putative kinase 1 (PINK1), a mitochondrial protein genetically linked to PD (Geisler et al., 2010; Matsuda et al., 2010; Narendra et al., 2010a; Vives-Bauza et al., 2010). Translocated Parkin then ubiquitinates outer mitochondrial membrane proteins, such as VDAC1 and mitofusins 1 and 2 (Ding et al., 2010; Gegg et al., 2010; Geisler et al., 2010; Lee et al., 2010; Okatsu et al., 2010; Ziviani et al., 2010). Ubiquitination of mitochondrial proteins induces recruitment of p62 (Ding et al., 2010; Geisler et al., 2010; Okatsu et al., 2010), which simultaneously binds to ubiquitin and LC3, a protein present on growing phagophores (Pankiv et al., 2007). Parkin-mediated mitochondrial ubiquitination also recruits HDAC6 (Lee et al., 2010), which interacts with both ubiquitin and microtubule-associated dynein motors (Kawaguchi et al., 2003). During this process, defective mitochondria are trafficked along microtubules into perinuclear clusters (mito-aggresomes) (Lee et al., 2010; Okatsu et al., 2010; Vives-Bauza et al., 2010). Finally, the clustered mitochondria are engulfed by autophagosomes.
We show that Ambra1 is not required for Parkin translocation or Parkin-mediated perinuclear clustering of depolarized mitochondria but is a key player in the final clearance of the mitochondria. More specifically, Ambra1 accumulates around juxtanuclear clusters of depolarized mitochondria and activates class III PI3K in their immediate vicinity. Production of PtdIns-3-P by class III PI3K is a powerful trigger for the formation of new phagophores (Simonsen and Tooze, 2009). This mechanism thus allows for local stimulation of new phagophore formation around depolarized mitochondria but spares the rest of the cell from autophagic degradation. Ambra1 recruitment to mito-aggresomes may therefore contribute to the spatially restricted, selective nature of mitophagy. Importantly, this activating role of Ambra1 in mitophagy is strictly dependent on Parkin: in the absence of Parkin, no CCCP-induced mitophagy occurs, even after overexpression of Ambra1.
The Parkin–Ambra1 interaction provides a mechanistic link between Parkin and the autophagy machinery. In a previously reported molecular connection between Parkin and autophagy, Parkin-catalyzed ubiquitination of VDAC1 led to p62-mediated recruitment of LC3 (Geisler et al., 2010). These two molecular models have different functional implications but are not mutually exclusive. Because LC3 only associates with phagophores after they have already started to form (Tooze and Yoshimori, 2010), the VDAC1/p62/LC3 connection can only recruit preexisting phagophores but cannot induce the formation of new phagophores. In contrast, Ambra1 stimulates new phagophore formation (Fimia et al., 2007). Thus, the two mechanisms could have complementary effects: Ambra1 induces the perimitochondrial formation of new phagophores, which, after incorporation of LC3, may be tethered to ubiquitinated mitochondria via p62. According to this scenario, Ambra1 should enhance the dynamics and efficiency of Parkin-mediated mitophagy. Interestingly, Ambra1 is only found in vertebrates and has no known orthologs in lower eukaryotes (Di Bartolomeo et al., 2010b). Ambra1 may have evolved to make removal of damaged mitochondria more efficient in the long-lived neurons of vertebrates.
We found that Parkin also interacted with Ambra1 in basal culture conditions (i.e., in the absence of mitochondrial uncouplers) and in adult mouse brain. Our cellular fractionation experiments suggest that this basal interaction occurs both in the cytosol and at the mitochondria. It is possible that Parkin and Ambra1 interact constitutively in the cytosol to cooperate in cytosolic autophagic processes, such as the removal of misfolded proteins. Indeed, genetic evidence has shown that continuous clearance of cytosolic proteins through basal autophagy is essential to prevent neuronal accumulation of ubiquitinated proteins and neurodegeneration (Hara et al., 2006; Komatsu et al., 2006). The basal interaction of Parkin and Ambra1 in the mitochondrial fraction could reflect the fact that a minority of mitochondria may become depolarized and dysfunctional even in basal, optimal environmental conditions (Narendra et al., 2008).
The biochemical basis for the strongly increased binding of Parkin and Ambra1 after mitochondrial depolarization remains to be elucidated. Mitochondrial depolarization may lead to posttranslational modifications of Parkin or Ambra1 that enhance their mutual affinity. Interestingly, Ambra1 was shown recently to be phosphorylated by ULK1 during induction of autophagy by starvation (Di Bartolomeo et al., 2010a). It will be important to determine whether Ambra1 phosphorylation similarly occurs during mitochondrial depolarization and contributes to the increased interaction with Parkin. An alternative mechanism for the increased binding could be that depolarized mitochondria recruit cofactors or binding partners of Parkin and Ambra1 that facilitate their mutual interaction.
The remarkable ability of Ambra1 to translocate to the perinuclear compartment during prolonged mitochondrial depolarization may be related to its recently demonstrated dynamic interaction with microtubule-associated dynein motors (Di Bartolomeo et al., 2010a). Interestingly, Ambra1 also translocates to the perinuclear region during induction of autophagy by starvation (Di Bartolomeo et al., 2010a). After starvation-induced perinuclear redistribution, Ambra1 was found to be localized in close proximity to omegasomes (Di Bartolomeo et al., 2010a). Omegasomes are PtdIns-3-P-enriched structures in dynamic equilibrium with the endoplasmic reticulum (ER), which provide a platform for accumulation of autophagy proteins and expansion of autophagosomal membranes (Axe et al., 2008). Intriguingly, recent work has shown that the outer mitochondrial membrane can also be a source of autophagosomal membranes and that mitofusin 2-dependent connections between ER and mitochondria are necessary for autophagosome formation during starvation (Hailey et al., 2010). The source of the autophagosomal membranes formed during mitophagy has not yet been determined.
The core components of the class III PI3K complex are Vps34 (the actual kinase in the complex), Beclin1, and p150 (Funderburk et al., 2010; Yang and Klionsky, 2010). Ambra1 activates class III PI3K by promoting the association of Beclin1 with Vps34 (Fimia et al., 2007). Interestingly, Beclin1 was recently reported to bind to PINK1 (Michiorri et al., 2010). Beclin1 only interacted with full-length PINK1 and not with its cleaved isoforms (Michiorri et al., 2010). Importantly, PINK1 is constitutively cleaved from healthy, polarized mitochondria by voltage-dependent proteolysis and only accumulates in its full-length form on depolarized mitochondria (Matsuda et al., 2010; Narendra et al., 2010a). Thus, the interaction of full-length PINK1 with Beclin1 could provide an additional mechanism for local induction of autophagy around depolarized mitochondria. Intriguingly, Beclin1 gene transfer ameliorated the neurodegenerative pathology in an α-synuclein transgenic mouse model of PD (Spencer et al., 2009). The class III PI3K complex and its regulators, including Ambra1, could be attractive therapeutic targets for PD.
W.V. is a Senior Clinical Investigator of the Research Foundation Flanders (FWO). This work was supported by FWO Grant G.0607.08 and Interuniversity Attraction Poles Grant IAP VI P6/43 (W.V.), by a Methusalem grant (Katholieke Universiteit Leuven and Flanders government) (B.D.S.), and by Concerted Research Actions Grant BOF07/GOA/012 of Ghent University and Interuniversity Attraction Poles Grant IAP VI P6/28 (K.G.). We are grateful to M. Fukata, K. Winklhofer, C. Scheidereit, J. Jung, and J. Feng for providing constructs and M. Fransen for providing the Pex14p antibody. We thank C. Reinhardt, V. Morais, A. Thathiah, and P. Verstreken for discussion.
- Correspondence should be addressed to Dr. Wim Vandenberghe, Department of Neurology, University Hospital Leuven, Herestraat 49, 3000 Leuven, Belgium.