PINK1 stabilized by mitochondrial depolarization recruits Parkin to damaged mitochondria and activates latent Parkin for mitophagy

Parkinson's disease (PD) is a very common movement disorder characterized by dopaminergic neuronal loss. The majority of PD cases are sporadic; however, the discovery of genes linked to rare familial forms of this disease has provided important insight into the molecular mechanisms of disease pathogenesis (Moore et al., 2005; Hardy et al., 2006). In 2000, we and others found that dysfunction of an E3 ubiquitin ligase (Imai et al., 2000; Shimura et al., 2000; Zhang et al., 2000) termed Parkin causes autosomal recessive juvenile Parkinsonism (Kitada et al., 1998). Since then, a multitude of papers have been published, but the mechanism by which dysfunction of Parkin causes autosomal recessive juvenile Parkinsonism has largely remained obscure, and claims of pathogenicity remain controversial (Lim, 2007; Matsuda and Tanaka, 2010). In addition, PINK1 (PTEN-induced putative kinase 1) was identified in 2004 as the gene responsible for another form of early-onset PD (Valente et al., 2004). PINK1 functions in mitochondrial maintenance, suggesting that mitochondrial integrity is another key factor in disease pathogenesis (Dodson and Guo, 2007; Schapira, 2008). Intriguingly, genetic studies using Drosophila melanogaster revealed that PINK1 and Parkin function in the same pathway, with PINK1 functioning upstream of Parkin (Clark et al., 2006; Park et al., 2006; Yang et al., 2006). Little is known about how PINK1 regulates Parkin, and our knowledge, especially in mammals, of their relationship is limited. In this study, we describe the mechanism underlying the functional interplay between ubiquitylation catalyzed by Parkin and mitochondrial quality control regulated by PINK1.

We initially sought to study the subcellular localization and E3 activity of Parkin using HeLa cells, which reportedly lack a functional Parkin gene (Denison et al., 2003). In support of that study, we found that endogenous Parkin was barely detectable in HeLa cells even when PRK8, the best-characterized specific anti-Parkin antibody (Pawlyk et al., 2003), was used (Fig. S1 A). Consequently, HA-Parkin was exogenously introduced into HeLa cells. Under steady-state conditions, HA-Parkin was diffusely localized throughout the cytosol and did not overlap with mitochondria, whereas Parkin was rapidly recruited to the mitochondria when HeLa cells were treated with the mitochondrial uncoupler CCCP (carbonyl cyanide m-chlorophenylhydrazone; Fig. 1 A), as reported by Narendra et al. (2008). Next we tried to confirm the redistribution of Parkin from the cytoplasm to the mitochondria using a biochemical approach. In fractionation experiments, detection of Parkin in the mitochondria-rich fraction was faint, probably because Parkin was weakly associated with the mitochondria and thus unstable during fractionation. Inclusion of the cross-linker DSP (dithiobis[succinimidyl propionate]) significantly strengthened the signal and further confirmed redistribution of exogenous (Fig. 1 B, left) and endogenous (Fig. 1 B, right) Parkin from the cytoplasm to a mitochondria-enriched fraction. (Note that endogenous Parkin in SH-SY5Y cells is detectable as a doublet, which is consistent with a previous study [Pawlyk et al., 2003].) To more convincingly demonstrate that Parkin is selectively recruited to depolarized mitochondria, we used MitoTracker red CMXRos, which accumulates in mitochondria with an intact membrane potential. Incomplete treatment with CCCP can generate cells in which healthy and damaged mitochondria coexist. Under these conditions, signals of Parkin and MitoTracker were mutually exclusive, and Parkin selectively localized on mitochondria with lower MitoTracker red staining (Fig. 1 D), indicating that Parkin was selectively targeted to mitochondria whose membrane potential had been lost.

Subsequently, we performed immunofluorescence staining using an antiubiquitin antibody. Under normal conditions, the ubiquitin signal was spread throughout the cell. In contrast, when cells were treated with CCCP, the ubiquitin signal was concentrated in the mitochondria (Fig. 2, A and B). Mitochondrial ubiquitylation was only observed in Parkin-expressing cells (Fig. 2 A and Fig. S1 B) and disappeared when Parkin mutants deficient in E3 activity (T415N and G430D) were introduced (Fig. 2 A). Triple staining using mitochondria-targeting GFP (Mt-GFP), anti-Parkin, and antiubiquitin antibodies further confirmed the colocalization of Parkin, ubiquitin, and mitochondria after CCCP treatment (Fig. 2 C). Staining with single antibodies or Mt-GFP alone indicated that the aforementioned merged data were not derived from channel cross talk (Fig. S1, C and D). These results demonstrate that Parkin ubiquitylates mitochondria in response to a reduction in mitochondrial membrane potential.

To confirm that translocation of Parkin to depolarized mitochondria is etiologically important, we selected nine pathogenic mutations and examined their subcellular localization (Fig. 1 C). In vitro experiments have previously shown that two of the mutations (T415N and G430D) in the RING2 domain abolish E3 activity of Parkin, whereas E3 activity is unaffected by the other mutations (Hampe et al., 2006; Matsuda et al., 2006). These mutants were serially introduced into HeLa cells, followed by CCCP treatment, and their subcellular localization was examined. Parkin with the D280N or G328E mutation in RING1 or the IBR (in between RING) domain, respectively, was recruited to the mitochondria in a manner similar to wild-type Parkin (Fig. 1, E and F). In contrast, the other pathogenic mutations altered to some degree the mitochondrial localization of Parkin; in particular, the K161N, K211N, and T240R mutations, which lie in or near the RING0 domain in the linker region (Hristova et al., 2009), severely compromised the mitochondrial localization of Parkin (Fig. 1, E and F). The aforementioned results suggest that mitochondrial localization of Parkin is pathologically significant and that the RING0 domain is important for the translocation of Parkin to the damaged mitochondria.

As shown in Fig. 2 C and Fig. S1 B, mitochondrial ubiquitylation was dependent on Parkin translocation to the mitochondria. Thus, we tried to determine whether the subcellular localization of Parkin modulates its E3 activity. To address this issue, we monitored the E3 activity of Parkin using an artificial substrate fused to Parkin. In vitro experiments have shown that Parkin can ubiquitylate an N-terminally fused lysine-rich protein as a pseudosubstrate (Matsuda et al., 2006). Similar ubiquitylation of pseudosubstrates even in the cytoplasm under normal conditions in vivo would be evidence that the E3 activity is constitutive; otherwise, E3 activity of Parkin is dependent on mitochondrial retrieval. A GFP tag is lysine rich and thus a good candidate for an in vivo pseudosubstrate, whereas an HA tag possesses no lysine residues and thus cannot function as a pseudosubstrate. GFP- and HA-Parkin expressed in HeLa cells treated with CCCP were both recruited to damaged mitochondria (Figs. 1 A and 2 D), but interestingly, a higher molecular mass population of only GFP-Parkin was observed (Fig. 2 F). Immunoprecipitation experiments demonstrated that GFP-Parkin was indeed ubiquitylated (Fig. 2 E). This was not based on autoubiquitylation of Parkin itself because mitochondria-associated HA-Parkin did not undergo ubiquitylation after CCCP treatment (Fig. 2 F and not depicted). Moreover, ubiquitylation of GFP-Parkin was absent in the T415N and G430D mutants, which lack E3 activities, suggesting that ubiquitylation of GFP-Parkin is not derived from other E3s (Fig. 2 G). The K161N and K211N mutants that impaired mitochondrial localization also inhibited ubiquitylation of GFP-Parkin (Fig. 2 G). Collectively, the aforementioned results indicate that Parkin ubiquitylates fused GFP only when it is retrieved to the mitochondria, suggesting that the latent E3 activity of Parkin is dependent on decreased mitochondrial membrane potential.

Recessive mutations in the human PINK1 gene are also the cause of autosomal recessive early-onset PD (Valente et al., 2004). We next examined whether the subcellular localization of PINK1 was affected by mitochondrial membrane potential. As reported previously (Valente et al., 2004; Beilina et al., 2005; Takatori et al., 2008), N-terminal Myc– or N-terminal Flag–tagged PINK1 clearly localized to the mitochondria, whereas C-terminal Flag– or C-terminal V5–tagged PINK1 mainly localized to the cytoplasm (Fig. 3 A and not depicted). Exogenous nontagged PINK1 also localized to the cytoplasm under steady-state conditions (Fig. 3 B), suggesting that mitochondrial localization of PINK1 is an artifact of the N-terminal epitope. More importantly, similar to Parkin, untagged PINK1 and C-terminal Flag– or C-terminal V5–tagged PINK1 localized to the mitochondria after CCCP treatment (Fig. 3, A and B; and not depicted). These results suggest that the subcellular localization of PINK1 is also regulated by the mitochondrial membrane potential.

We next sought to determine the subcellular localization of endogenous PINK1. Immunocytochemical experiments showed, as reported previously (Zhou et al., 2008), that the endogenous PINK1 signal was barely detectable in HeLa cells under steady-state conditions. However, a decrease in mitochondrial membrane potential resulted in a mitochondria-associated PINK1 signal (Fig. 3, C and D). We found that CCCP treatment promoted the gradual accumulation of endogenous PINK1 in immunoblots as well (Fig. 3 E) and the presence of endogenous PINK1 in a mitochondria-enriched fraction (Fig. 3 F). More importantly, when CCCP was washed out, the accumulated endogenous PINK1 rapidly disappeared (within 30 min) both in the presence and absence of cycloheximide (Fig. 3 G and not depicted). Moreover, the N-terminal 34 aa of PINK1 sufficiently recruited GFP to the mitochondria even in the absence of CCCP (Fig. 3 H). These results support the hypothesis in which PINK1 is constantly transported to the mitochondria but is rapidly degraded in a membrane potential–dependent manner. We speculate that PINK1 is stabilized by a decrease in mitochondrial membrane potential and, as a result, accumulates in depolarized mitochondria.

PINK1 normally exists as either a long (∼60 kD) or a short (∼50 kD) protein. Because the canonical mitochondria-targeting signal (matrix-targeting signal) is cleaved after import into the mitochondria, the long form has been designated as the precursor and the short form as the mature PINK1 (Beilina et al., 2005; Silvestri et al., 2005). The short (processed) form of PINK1 was clearly detected when untagged PINK1 was overexpressed (Fig. 3 E, sixth lane); however, this form of endogenous PINK1 was rarely detectable after CCCP treatment (Fig. 3 E, the first through the fifth lanes). Our subcellular localization study of endogenous PINK1 after CCCP treatment showed that the long form was recovered in the mitochondrial fraction (Fig. 3 F), suggesting that it is not the preimport precursor form. Moreover, by monitoring the degradation process of PINK1 after recovery of membrane potential, we realized that the short form of PINK1 transiently appeared soon after CCCP was washed out and then later disappeared (Fig. 3 G), suggesting that the processed form of PINK1 is an intermediate in membrane potential–dependent degradation. In conclusion, these results imply that PINK1 cleavage dose not reflect a canonical maturation process accompanying mitochondrial import as initially thought.

Because previous studies revealed that PINK1 functions upstream of Parkin (Clark et al., 2006; Park et al., 2006; Yang et al., 2006; Exner et al., 2007), we next examined the potential role of PINK1 in the mitochondrial recruitment of Parkin. To obtain clear-cut conclusions, we set up our experimental system using mouse embryonic fibroblasts (MEFs) derived from control (PINK1^+/+) or PINK1 knockout (PINK1^−/−) mouse (Gautier et al., 2008). Endogenous Parkin is undetectable in MEFs (Fig. S1 A); consequently, HA- or GFP-Parkin was introduced into these cells by retroviral transfection (Kitamura et al., 2003). In control MEFs (PINK1^+/+), Parkin was selectively recruited to the mitochondria after CCCP treatment (Fig. 4 A) and subsequently resulted in the disappearance of the mitochondria (Fig. 4, D and E). This mitochondrial clearance was considerably impeded by Atg7 (essential gene for autophagy) knockout (Fig. S2; Komatsu et al., 2005), suggesting that Parkin degrades mitochondria by selective autophagy as reported previously (Narendra et al., 2008). In sharp contrast, Parkin was not translocated to the mitochondria in PINK1 knockout (PINK1^−/−) MEFs after CCCP treatment (Fig. 4, A and B). Subsequent activation of Parkin and mitochondrial degradation were also completely impeded (Fig. 4, C–E). To exclude the possible role of retroviral integration of Parkin in the aforementioned phenotype, we checked whether reintroduction of PINK1 rescued this phenotype. Untagged or C-terminal Flag–tagged PINK1 complemented the mislocalization of Parkin in PINK1^−/− MEFs (Fig. 5, B and C; and not depicted), confirming that the aforementioned defects were caused by the loss of PINK1.

To examine whether pathogenic mutations of PINK1 affect its mitochondrial localization, we expressed PINK1 mutants harboring the missense mutations E240K and G309D, or a CAA nucleotide insertion behind C1602 (referred to hereafter as 1602-insert) in PINK1^−/− MEFs. Similar to wild-type PINK1, these PINK1 mutants colocalized with mitochondria after CCCP treatment (Fig. S3). Next, GFP-Parkin was introduced into these cells to examine whether pathogenic mutations of PINK1 affect the mitochondrial localization and activation of Parkin. The E240K and G309D mutants restored the mitochondrial localization and activation of Parkin as well as wild-type PINK1, whereas recruitment of Parkin to the mitochondria by the 1602-insert mutant was abolished (Fig. 5, B–D), suggesting that the pathology of this PINK1 mutation is mislocalization and consequent inactivation of Parkin.

Finally, we investigated the role of various PINK1 domains (Fig. 5 A) in the mitochondrial recruitment of Parkin. PINK1 is composed of an atypical N-terminal mitochondrial localization signal and transmembrane domain, a kinase domain in the middle, and a conserved C-terminal domain (Zhou et al., 2008). Deletion of the N-terminal 91 aa abolished the mitochondrial localization of PINK1 (Zhou et al., 2008). We also confirmed that the ΔN91 and ΔN155 mutants did not target to the mitochondria even after CCCP treatment (Fig. S3). We also generated a mutant containing the triple K219A, D362A, and D384A mutations that abolish kinase activity (Beilina et al., 2005) and a C-terminal domain deletion mutant associated with PINK1 dysfunction (Sim et al., 2006; Yang et al., 2006). The kinase-dead and ΔC72 mutants of PINK1 colocalized with damaged mitochondria similar to wild type (Fig. S3). When introduced into PINK1^−/− cells harboring GFP-Parkin, the mutants were unable to complement the mislocalization and inactivation of Parkin (Fig. 5, B–D), even though the mutant PINK1 proteins were expressed (Fig. 5 D and Fig. S3). These results indicate that the kinase activity and mitochondrial targeting of PINK1 are essential for the mitochondrial recruitment of Parkin.

In summary, we have shown that (a) PINK1 is a Parkin-recruitment factor that recruits Parkin from the cytoplasm to damaged mitochondria in a membrane potential–dependent manner for mitochondrial degradation, (b) endogenous PINK1 is constitutively degraded at the mitochondria, but its localization is specifically linked to a decrease in membrane potential, (c) under steady-state conditions, the E3 activity of Parkin is repressed in the cytoplasm but is liberated by PINK1-dependent mitochondrial localization, and (d) the aforementioned phenomena are presumably etiologically important in part because they were impeded for the most part by disease-linked mutations of PINK1 or Parkin. We believe that these results provide solid insight into the molecular mechanisms of PD pathogenesis not only for familial forms caused by Parkin and PINK1 mutations but also major sporadic forms of PD.

MEFs derived from embryonic day 12.5 embryos of PINK1 knockout mice (provided by J. Shen, Harvard Medical School, Boston, MA) were mechanically dispersed by repeated passage through a P1000 pipette tip and plated with MEF media containing DME, 10% FCS, β-mercaptoethanol (Sigma-Aldrich), 1× nonessential amino acids, and 1 mM l-glutamine. Various stable transformants of MEFs were established by infecting MEFs with recombinant retroviruses. HA-Parkin, GFP-Parkin, PINK1 (provided by Y. Nakamura, T. Iwatsubo, and S. Takatori, University of Tokyo, Bunkyo-ku, Tokyo, Japan), or various PINK1 mutants were cloned into a pMXs-puro vector. Retrovirus packaging cells, PLAT-E (provided by T. Kitamura, University of Tokyo; Kitamura et al., 2003), were transfected with the aforementioned vectors and were cultured at 37°C for 24 h. After changing the medium, cells were further incubated at 37°C for 24 h, and the viral supernatant was collected and used for infection. MEFs were plated on 35-mm dishes at 24 h before infection, and the medium was replaced with the aforementioned undiluted viral supernatant with 8 µg/ml polybrene (Sigma-Aldrich). 2 d later, transformants were selected by the medium containing 10 µg/ml puromycin.

To depolarize the mitochondria, HeLa and SH-SY5Y cells were treated with 10 µM CCCP, and MEFs were treated with 30 µM CCCP. For fractionation experiments, HeLa and SH-SY5Y cells were treated with CCCP for 1–5 h and subsequently treated with 1 mM DSP (Thermo Fisher Scientific) in PBS for 1 h on ice, inactivated by 10 mM glycine in PBS three times, and suspended in chappell-perry buffer (0.15 M KCl, 20 mM Hepes-NaOH, pH 8.1, 5 mM MgCl₂, and protease and phosphatase inhibitor [Roche]). Cells were disrupted by five passages through a 25-gauge needle (with 1-ml syringe), debris was removed by centrifugation at 1,000 g for 7 min, and the supernatant was subjected to 10,000 g for 10 min to separate the mitochondria-rich fraction from the cytosol-rich fraction. Immunoblotting and immunoprecipitation were performed by conventional methods. To detect the ubiquitylation of GFP-Parkin, the cell lysate of HeLa cells (10 µM CCCP for 1 h) or MEFs (30 µM CCCP for 3 h) was collected in the presence of 10 mM N-ethylmaleimide to protect ubiquitylated Parkin from deubiquitylation enzymes. To monitor the degradation of endogenous PINK1, HeLa cells were treated with 10 µM CCCP and 50 µg/ml cycloheximide as depicted in Fig. 3 G and were subjected to immunoblotting.

To depolarize the mitochondria, HeLa cells (provided by A. Tanaka and R. Youle, National Institutes of Health, Bethesda, MD) were treated with 10 µM CCCP for 1 h (exogenous Parkin and PINK1) or 5 h (endogenous PINK1), and MEFs were treated with 30 µM CCCP for 3–4 h (Figs. 4 A and 5 B; and Fig. S3 B) or 24 h (Fig. 4 D and Fig. S2 A). For immunofluorescence experiments, cells were fixed with 4% paraformaldehyde, permeabilized with 50 µg/ml digitonin, and stained with primary antibodies described in the next section and with the following secondary antibodies: mouse and/or rabbit Alexa Fluor 488, 568, and 647 (Invitrogen). N-terminal 34 aa of PINK1 were fused to GFP to stain mitochondria in the triple staining experiments. To monitor the mitochondrial membrane potential, MEFs were treated with 50 nM MitoTracker red CMXRos (Invitrogen) for 15 min, washed three times, and incubated for an additional 10 min before fixation, as reported previously (Narendra et al., 2008). Cells were imaged using a laser-scanning microscope (LSM510 META; Carl Zeiss, Inc.) with a Plan-Apochromat 63× NA 1.4 oil differential interference contrast objective lens. Image contrast and brightness were adjusted in Photoshop (Adobe).

Antibodies used in this study are as follows: antiactin (AC-40; Sigma-Aldrich), anti–cytochrome c (6H2.B4; BD), anti-Flag (M2; Sigma-Aldrich), anti-GFP (3E6 [Wako Chemicals USA, Inc.]; or A6455 [Invitrogen]), anti-HA (12CA5; Roche), anti-Hsp70 (MBL), anti–lactate dehydrogenase (Abcam), anti-Parkin (#2132 [Cell Signaling Technology] for immunocytochemistry; or PRK8 [Sigma-Aldrich] for immunoblotting), anti-PINK1 (Novus), anti-Tom20 (FL-145 and F-10; Santa Cruz Biotechnology, Inc.), antiubiquitin (P4D1 [Santa Cruz Biotechnology, Inc.]; or FK2 [MBL]), and anti-V5 (Invitrogen).

Fig. S1 shows various control experiments for immunoblotting and immunocytochemistry. Fig. S2 shows that Parkin promoted degradation of depolarized mitochondria via autophagy. Fig. S3 shows the subcellular localization of pathogenic and deletion mutants of PINK1 after CCCP treatment.

We thank Dr. T. Kitamura for providing the PLAT-E cells, Drs. A. Tanaka and R. Youle for NIH HeLa cells, Dr. J. Shen for PINK1 knockout MEFs, and Drs. Y. Nakamura, T. Iwatsubo, and S. Takatori for plasmids.

This work was supported by Grants-in-Aid for Young Scientist (B) (to N. Matsuda, K. Shiba, and S. Saiki) and by Specially Promoted Research (to K. Tanaka) from the Ministry of Education, Culture, Sports, Science and Technology of Japan and Heath and Labor Science Research grants (to N. Hattori and K. Tanaka).

Note added in review. While our manuscript was under review, Geisler et al. (2010), Narendra et al. (2010), and Vives-Bauza et al. (2010) independently published results that are consistent with those described herein.