Reactive oxygen species trigger Parkin/PINK1 pathway–dependent mitophagy by inducing mitochondrial recruitment of Parkin

Defective mitophagy linked to dysfunction in the proteins Parkin and PTEN-induced putative kinase 1 (PINK1) is implicated in the pathogenesis of Parkinson's disease. Although the mechanism by which Parkin mediates mitophagy in a PINK1-dependent manner is becoming clearer, the triggers for this mitophagy pathway remain elusive. Reactive oxygen species (ROS) have been suggested as such triggers, but this proposal remains controversial because ROS scavengers fail to retard mitophagy. Here we demonstrate that the role of ROS in mitophagy has been underappreciated as a result of the inefficiency of ROS scavengers to control ROS bursts after high-dose treatment with carbonyl cyanide m-chlorophenylhydrazone. Supporting this, combinatorial treatment with N-acetyl-l-cysteine and catalase substantially inhibited the ROS upsurge and PINK1-dependent Parkin translocation to mitochondria in response to carbonyl cyanide m-chlorophenylhydrazone treatment. In addition to the chemical mitophagy inducer, overexpression of voltage-dependent anion channel 1 (VDAC1) induced Parkin translocation to mitochondria, presumably by stimulating ROS generation. Similarly, combined N-acetyl-l-cysteine and catalase treatment also suppressed VDAC1-induced redistribution of Parkin. Alongside these observations, we also found that the elevated protein level of PINK1 was not necessary for Parkin translocation to mitochondria. Thus, our data suggest that ROS may act as a trigger for the induction of Parkin/PINK1-dependent mitophagy. In addition, our study casts doubt on the importance of protein quantity of PINK1 in the recruitment of Parkin to mitochondria.

Parkinson's disease (PD) 2 is the second most common neurodegenerative disorder after Alzheimer's disease. Accumulating evidence from studies of mutations in familial PD-related genes, including Parkin and PINK1, has revealed that dysfunctions in mitochondrial quality control contribute to the pathogenesis of PD (1). Recently, Parkin and PINK1 have been found to play a pivotal role in the final checkpoint of this quality control mechanism, i.e. the autophagic clearance of damaged mitochondria (mitophagy) (2)(3)(4)(5)(6). Upon mitochondrial depolarization, PINK1 selectively accumulates on the damaged mitochondria and undergoes autophosphorylation at Ser-228 and Ser-402 (7). Accumulated PINK1 activates Parkin by phosphorylating the Ser-65 residue of Parkin (8), releasing it from an autoinhibitory state (9). Meanwhile, PINK1 phosphorylates ubiquitin, which, in turn, further activates Parkin (10 -12). Subsequently, activated Parkin translocates to mitochondria, where it physically associates with mitochondrial substrates such as VDAC1 (13) and mitofusin 2 (MFN2) (14). When anchored to mitochondria, Parkin mediates global polyubiquitination of the outer mitochondrial membrane proteins, triggering convergence of the autophagic machinery with mitochondria to ensure proper execution of mitophagy (15)(16)(17).
Notwithstanding advances in the knowledge of this area, there are concerns regarding the existing model of Parkin/ PINK1-dependent mitophagy. In particular, our understanding of the triggers of such mitophagy makes it difficult to reconcile this model with the physiological scenario (18). For example, chemically induced mitochondrial depolarization has popularly been utilized to trigger Parkin/PINK1-dependent mitophagy. However, the commonly used mitophagy inducer, i.e. the protonophore CCCP, is usually applied at high concentrations, which disrupts mitochondrial potential to an extent beyond physiological relevance (19,20). Alternatively, the accumulation of misfolded proteins in the mitochondrial matrix was shown to promote mitophagy (21), but its role in PD or neurodegeneration is not unequivocal (22). Of note, ROS are promising candidates of the physiological triggers for mitophagy, as ROS naturally exist in cells and have been linked to the pathogenesis of PD because of their detrimental effects on lipids, proteins, and DNA (23). Moreover, mitochondrial DNA defects caused by ROS may result in the presence of faulty proteins in the electron transport chain, which boosts ROS generation and mitochondrial DNA impairment, creating a vicious cycle (24). It is therefore conceivable that the mitophagy machinery may sense the excessive mitochondrial damage caused by ROS and subsequently eliminate impaired mitochondria beyond repair. However, several independent groups reported the failure of ROS scavengers to suppress CCCP-induced Parkin recruitment in HeLa cells (2,25,26), challenging the establishment of ROS as a trigger for mitophagy.
To clarify the role of ROS in Parkin/PINK1-dependent mitophagy, Parkin translocation to mitochondria was utilized in this study as an indicator for the initiation of mitophagy. ROS status was closely monitored in CCCP-treated or VDAC1overexpressing cells in the presence or absence of ROS scavenger(s). We show that ROS, induced by CCCP treatment or VDAC1 overexpression, play an essential role in the recruitment of Parkin to mitochondria. In addition, we reveal that an elevated protein level of PINK1 is not necessary for translocation of Parkin to mitochondria. Thus, our study contributes to the understanding of the role of ROS and PINK1 in mitophagy.

Effect of ROS scavengers on ROS-induced Parkin translocation
The results from previous studies using CCCP treatment and ROS scavengers indicated a dispensable role of ROS in Parkin/ PINK1-dependent mitophagy (2,25,26). Here we revisited the current model of mitophagy induced by CCCP treatment, notwithstanding that the protonophore CCCP may produce other effects that complicate mitophagy (27,28). First we examined the mitochondrial potential and ROS levels in HeLa cells in response to CCCP stress using the fluorescent dyes JC-1 and dihydroethidium (DHE), respectively. As expected, treatment with 10 M CCCP led to mitochondrial depolarization in a time-dependent manner. CCCP treatment significantly disrupted mitochondrial potential 30 min after it was added to the cells and reached its maximal effects at 1 h (Fig. 1, A-C). In addition, the effects of CCCP treatment on mitochondrial potential were dose-dependent. Treatment of cells with a low concentration (2.5 M) of CCCP enhanced mitochondrial potential, whereas CCCP treatment at concentrations higher than 5 M depolarized mitochondria. Mitochondrial depolarization became more severe with increasing concentrations of CCCP beyond 5 M (Fig. 1D). Similarly, treatment of cells with 10 M CCCP led to an increase in ROS level with time ( Fig. 1, E-G). Low-dose CCCP treatment (2.5 M) mildly decreased ROS production, whereas CCCP treatment at concentrations higher than 5 M boosted ROS generation, and this effect was amplified with increasing concentrations of CCCP (Fig. 1H). Given our findings that CCCP treatment led to a ROS burst in the cells, we then asked whether ROS scavengers could quench CCCP-stimulated ROS efficiently and affect CCCP-induced Parkin recruitment to mitochondria. Consistent with previous studies, neither NAC nor catalase treatment alone suppressed Parkin translocation to mitochondria (supplemental Fig. S1, A  and B). This has been interpreted previously as evidence to support that ROS play an insignificant role in CCCP-induced Parkin recruitment (2,25). However, the efficacy of these scavengers to control ROS production induced by CCCP treatment was unknown. Therefore, we examined the ROS status of HeLa cells exposed to both CCCP and ROS scavengers. First, we examined whether catalase added to cells did ablate the intra-cellular ROS spike using flow cytometry. Indeed, catalase treatment substantially reduced 2Ј,7Ј-dichlorofluorescein intensity in the cells in response to Earle's balanced salt solution treatment, which does not contain pro-oxidant, indicating that catalase treatment can attenuate endogenous ROS production (supplemental Fig. S2). A total concentration of 1000 units/ml of catalase quenched ROS generation induced by 10 M CCCP at the first 30 min, but this effect gradually diminished beyond 30 min (Fig. 1I). In contrast, 2 mM NAC treatment suppressed ROS production induced by 10 M CCCP from 1 h after the drug was added but failed to do so at 30 min (Fig. 1J). These data indicate that failure of the ROS scavengers to inhibit CCCPinduced Parkin translocation to mitochondria may be due to their inability to extinguish CCCP-induced ROS generation at appropriate time points.

ROS scavengers suppress CCCP-induced Parkin recruitment
Based on our findings that the ROS scavengers did not inhibit CCCP-induced ROS production completely, we reconsidered the role of ROS in the recruitment of Parkin. 2 mM NAC treatment failed to impede 10 M CCCP-induced Parkin translocation to mitochondria (supplemental Fig. S1). However, as demonstrated above, 2 mM NAC was also insufficient to quench 10 M CCCP-induced ROS at 30 min (Fig. 1J). Because 7.5 M CCCP triggered substantially less ROS production than 10 M CCCP (Fig. 1H), we examined Parkin recruitment induced by 7.5 M CCCP and the effect of 2 mM NAC under this condition. At this concentration, CCCP treatment induced Parkin translocation to mitochondria in ϳ90% of the cells (Fig. 2, A and B). However, in the presence of 2 mM NAC, co-localization of Parkin and mitochondria could only been seen in ϳ40% of cells, indicating significant attenuation of CCCP-induced Parkin translocation (Fig. 2B). Similarly, our earlier time course experiments demonstrated that suppression of 10 M CCCP-induced ROS using 2 mM NAC did not occur until after 30 min. Therefore, we checked ROS levels after 30-min incubation of cells with 7.5 M CCCP and 2 mM NAC. Our result showed that 2 mM NAC reduced 7.5 M CCCP induced-ROS production to a similar level as that in the NAC-treated cells (Fig. 2C). Moreover, depolarization of mitochondria was not significantly affected (supplemental Fig. S3A). In addition, treatment of cells with 10 mM NAC suppressed Parkin translocation to mitochondria induced by 10 M CCCP (supplemental Fig. S4). However, 10 mM NAC failed to inhibit 20 M CCCP-induced Parkin translocation to mitochondria (supplemental Fig. S4), which is consistent with a previous report (25). These data suggest a dose-dependent effect of NAC treatment on Parkin translocation to mitochondria induced by CCCP treatment. The earlier time course experiments also showed that catalase treatment efficiently suppressed ROS production during the first 30 min (Fig. 1I), but not beyond this time, whereas NAC treatment had an opposite effect (Fig. 1J). Therefore, we asked how the combination of NAC and catalase would affect Parkin recruitment and whether they could act in synergy to attenuate the burst of ROS production stimulated by CCCP treatment. Indeed, we found that combinatorial treatment with 2 mM NAC and 1000 units/ml catalase almost completely blocked 10 M CCCP-induced Parkin accumulation on mitochondria (Fig. 2, D and E).

ROS trigger Parkin/PINK1-dependent mitophagy
Furthermore, the same combinatorial treatment quenched ROS production stimulated by treatment with 10 M CCCP (Fig. 2F) and inhibited removal of mitochondria after the cells were exposed to CCCP treatment for 24 h (supplemental Fig.  S5). However, we also noted that this combinatorial treatment affected mitochondrial potential in the presence or absence of CCCP treatment (supplemental Fig. S3B). In addition, to determine whether CCCP treatment induces Parkin translocation to mitochondria by altering the bioenergetics of the cells, an ATP assay was performed 2 h or 24 h after cells were exposed to CCCP treatment at a series of concentrations. Treatment of cells with CCCP moderately decreased ATP generation at 2 h, and this effect became more substantial at 24 h in a dosedependent manner (supplemental Fig. S6), indicating that changes in the bioenergetics may not be involved in Parkin translocation to mitochondria shortly after CCCP was added to the cells. Taken together, our data suggest that ROS may be important for Parkin recruitment to mitochondria, as ROS scavengers are able to block CCCP-induced Parkin recruitment when they effectively suppress ROS production.

Overexpression of VDAC1 causes Parkin recruitment
Because ROS scavenger may affect both mitochondrial depolarization and ROS production in response to CCCP treatment in our study, we sought to find a different approach to stimulate ROS production without altering mitochondrial potential. However, chemicals such as CCCP and paraquat usually depolarize mitochondria and trigger ROS generation concurrently (2), suggesting that these cellular phenomena may be linked. In addition, a photosensitizer, KillerRed, which stimulates ROS production upon photoactivation and triggers Parkin-mediated mitophagy, also disrupts mitochondrial potential (29). Hence, we decided to employ a genetic approach to separate these two events. VDAC1 is known to mediate the release of ROS from the mitochondrial intermembrane space to the cytoplasm (30), whereas VDAC1 overexpression increases the cellular levels of ROS (31). Moreover, VDAC1 has recently been identified as a mitochondrial substrate for Parkin after CCCP treatment (13). Our immunoprecipitation assay replicated a physical interaction between Parkin and VDAC1 (supplemental Fig. S7). We next examined the effect of VDAC1 overexpression on mitochondrial recruitment of Parkin. As shown in Fig.  3, A and B, Parkin was recruited to mitochondria in ϳ30% of the cells concurrently overexpressing VDAC1. Western blot analysis was performed to evaluate the protein level of the overexpressed VDAC1 (supplemental Fig. S8). To determine whether VDAC1-induced Parkin recruitment is PINK1-related, a PINK1 knockdown HeLa cell line was generated. Both full-length PINK1 and its processed form failed to accumulate in response to CCCP or MG132 treatment in PINK1 knockdown cells (supplemental Fig. S9). Next, we co-transfected VDAC1 and Parkin into stable PINK1 knockdown cells and found that Parkin recruitment to mitochondria occurred in less than 5% of the cells expressing VDAC1 and Parkin (Fig. 3, C and D). Notably, VDAC1 overexpression in native HeLa cells did not affect the level of PINK1 expression compared with that in vector-overexpressing control cells (Fig. 3E), suggesting that the elevated level of PINK1 may not be necessary for PINK1dependent Parkin recruitment to mitochondria. This result also suggests that mitochondrial potential is not disrupted by VDAC1 overexpression, as PINK1 would otherwise accumulate upon mitochondrial depolarization.

ROS contribute to Parkin recruitment induced by VDAC1 overexpression
To test whether ROS play a fundamental role in Parkin translocation induced by VDAC1 overexpression, we first examined whether the cellular level of ROS was changed by VDAC1 overexpression. VDAC1 plasmids with fluorescent tags at either terminus were constructed for subsequent analysis using flow cytometry. However, VDAC1 with either an N-terminal or C-terminal GFP tag was distributed evenly or with a punctumlike pattern in the cells (supplemental Fig. S10), suggesting that the GFP tag compromised proper localization of VDAC1 to the outer mitochondrial membrane. We therefore designed a blue fluorescent protein (BFP)-T2A-FLAG-VDAC1 plasmid construct that consists of a T2A sequence between the BFP and the FLAG tag. The T2A sequence is skipped by the ribosome during translation, leading to its "self-cleavage" and separation of the adjacent proteins (32). When expressed in the cells, the BFP tag is used as a reporter to indicate the presence of ectopic VDAC1, whereas the FLAG tag can be immunostained to show the distribution pattern of VDAC1. Immunostaining showed that the BFP signals were ubiquitously distributed in the cells, whereas the FLAG signals showed perfect colocalization with the mitochondrial marker Tom20 (supplemental Fig. S11A), indicating functional VDAC1 for the downstream live-cell assays. The protein level of overexpressed VDAC1 using BFP-T2A-FLAG-VDAC1 is shown compared with that in cells expressing vector (supplemental Fig. S11B). We then examined ROS production and mitochondrial potential in HeLa cells overexpressing BFP-T2A-FLAG-VDAC1 by flow cytometry. VDAC1 overexpression increased ROS production by approximately 30% (Fig. 4A) but had little effect on mitochondrial potential (Fig. 4B). Next, we asked whether the ROS scavengers were able to attenuate VDAC1-induced Parkin translocation to mitochondria. Similar to its inhibition of CCCP-induced Parkin translocation to mitochondria, combinatorial treatment with 2 mM NAC and 1000 units/ml catalase dramatically impeded VDAC1-induced Parkin accumulation to mitochondria, whereas treatment with NAC or catalase alone had little effect (Fig. 4, C and D). Consistently, combinatorial treatment, but not treatment with NAC or catalase alone, was sufficient to quench ROS generation induced by VDAC1 overexpression (Fig.  4E). In addition, combinatorial treatment of NAC and catalase did not affect the protein expression levels of VDAC1 or Parkin (Fig.  4F), excluding the possibility that ROS scavengers blocked VDAC1-induced Parkin translocation by altering the levels of VDAC1 and/or Parkin. Moreover, combinatorial treatment of the ROS scavengers attenuated the overlap of LC-3 and mitochondria induced by VDAC1 overexpression (supplemental Fig. S12). Together, our data suggest that ROS may play an important role in VDAC1 overexpression-induced mitophagy. Starvation has been shown to stimulate ROS production and trigger general autophagy  (Hoechst 33258, blue). B, GFP-Parkin-overexpressing cells were scored for GFP-Parkin on mitochondria following treatment as described in A. Error bars represent standard deviation of three independent experiments; n Ն 100 cells/experiment. Statistical significance was calculated using Student's t test. *, p Ͻ 0.05. C, HeLa cells were treated with or without 2 mM NAC, 7.5 M CCCP, or in combination, as indicated, for 30 min. Cells were harvested and stained with DHE, followed by flow cytometric analysis. Error bars represent standard deviation of three independent experiments; n Ն 10,000 cells/experiment. Statistical significance was calculated using Student's t test. *, p Ͻ 0.05; NS, not significant (p Ͼ 0.05). D, HeLa cells were transfected with GFP-Parkin (green) and treated with 10 M CCCP for 2 h in the presence or absence of 2 mM NAC and 1000 units/ml catalase. Fixed cells were stained for Tom20 (mitochondria, red) and nuclei (Hoechst 33258, blue). E, cells were scored for GFP-Parkin on mitochondria following treatment as described in D. Statistical analysis was performed as in B. *, p Ͻ 0.01. F, HeLa cells were treated with or without 10 M CCCP, 2 mM NAC, and 1000 units/ml catalase, as indicated, for 30 min. The cells were harvested and stained with DHE, followed by flow cytometric analysis. Statistical analysis was performed as in C. *, p Ͻ 0.05; NS, p Ͼ 0.05. Scale bars ϭ 10 m. (33) and mitophagy (34). However, our results showed that Earle's balanced salt solution treatment failed to induce significant redistribution of Parkin or LC-3 to mitochondria (supplemental Fig.  S13), indicating that, under our experimental conditions, starvation may not be able to induce mitophagy. Furthermore, in addition to excessive ROS production, other factor(s) may be necessary for the initiation of mitophagy.

PINK1 protein level is not a determinant for Parkin recruitment
Previous studies have shown that Parkin is exclusively recruited to PINK1-accumulated mitochondria. Moreover, recruitment of Parkin to impaired mitochondria is blocked by depletion of PINK1 (3-6, 35). Thus, it has been proposed that the increase in protein level of PINK1 induced by CCCP treatment is responsible for the induction of Parkin recruitment (36 -38). However, our above result showed that PINK1 accumulation was not involved in Parkin translocation induced by VDAC1 overexpression. Thus, we next sought to determine whether the elevated protein level of PINK1 is required for mitochondrial translocation of Parkin using the stable PINK1 knockdown HeLa cell line. Consistent with previous studies, translocation of GFP-Parkin to mitochondria was almost completely abolished in PINK1 knockdown HeLa cells 2 h after incubation with 10 M CCCP (Fig. 5A). Interestingly, after 8 h of CCCP treatment, recruitment of Parkin to mitochondria was restored to a similar level as that in control cells (Fig. 5B). Similar Parkin dynamics in response to CCCP treatment were observed in another PINK1 knockdown stable HeLa cell line (supplemental Fig. S14). In addition, CCCP treatment induced a mild accumulation of PINK1 in PINK1-depleted HeLa cells after being added for 2 h, but no further accumulation of PINK1 occurred from 2 h to 8 h (Fig. 5C). Importantly, PINK1 protein levels in PINK1 knockdown cells after CCCP treatment were significantly lower than in control cells in the absence of CCCP treatment (Fig. 5C), in which Parkin was not recruited to mitochondria. Thus, our data suggest that Parkin translocation to mitochondria may occur in the absence of the elevated protein level of PINK1. In addition, PINK1 knockdown had little effect on ROS generation in cells exposed to 10 M CCCP for the various times (2-8 h) (Fig. 5D). We next assessed whether loss of PINK1 accumulation affected CCCP-induced mitophagy. We transfected GFP-Parkin into control or PINK1 knockdown cells and treated the cells with 10 M CCCP for 24 h. Unexpectedly, knockdown of PINK1 had little effect on completion of mitophagy (supplemental Fig. S15).
To investigate the effects of chronic, low-dosage CCCP treatment on PINK1 expression and mitochondrial recruitment of Parkin, HeLa cells were treated with 5 M CCCP for 8 h (Fig. 5,  E and F). Recruitment of Parkin to mitochondria was modest at 8 h of incubation with 5 M CCCP (Fig. 5F). Interestingly, the levels of PINK1 gradually increased with time upon 5 M CCCP treatment, reaching a similar level after 8 h as that induced by 10 M CCCP treatment for 2 h (Fig. 5G). To investigate whether 5 M CCCP treatment led to an accumulation of PINK1 on mitochondria, PINK1-GFP-transfected HeLa cells were incubated with 5 M CCCP for 8 h or 10 M CCCP for 2 h and analyzed by immunofluorescence. PINK1-GFP was distributed throughout the cells under control treatment. Following acute CCCP treatment, PINK1-GFP accumulated on mitochondria, consistent with previous reports. Interestingly, chronic exposure to 5 M CCCP for 8 h also induced an accumulation of PINK1 on mitochondria (supplemental Fig. S16). Therefore, our data suggest that other factor(s) might influence PINK1-dependent Parkin translocation to mitochondria because similar accumulation of PINK1 on mitochondria had different effects on Parkin translocation. Notably, treatment with 5 M CCCP within 8 h induced less ROS production compared with 2-h treatment of 10 M CCCP (Fig. 5H), suggesting that the variable cellular levels of ROS might be associated with distinct Parkin recruitment.

Discussion
The mechanism by which Parkin is recruited to mitochondria and triggers subsequent autophagic clearance of mitochondria has been elusive, partially because of the multiple effects on cellular functions of CCCP treatment, which has been widely utilized to study mitophagy. CCCP treatment may induce ROS production, mitochondrial depolarization, and collapse of cellular bioenergetics. The role of ROS in Parkin/ PINK1-dependent mitophagy is under debate, largely because ROS scavengers have an inconclusive effect on Parkin translocation. Here we address the controversy by showing that the inability of ROS scavengers to inhibit Parkin accumulation onto mitochondria may, in fact, be a result of the inefficiency of ROS scavengers to quench CCCP-induced ROS production. By decreasing CCCP concentration or combining ROS scavengers, we successfully blocked mitochondrial translocation of Parkin. Notably, treatment with ROS scavengers also altered mitochondrial depolarization in response to CCCP treatment. Given that the collapse of mitochondrial potential is considered to be a major driving force for Parkin translocation to mitochondria, the essential role of ROS in Parkin translocation would be more convincing if mitochondrial depolarization is not involved in this event. Indeed, VDAC1 overexpressioninduced Parkin translocation is independent of mitochondrial depolarization, and translocation under this condition is inhibited by ROS scavengers. Here we also showed that 2 h of CCCP treatment failed to substantially decrease cellular ATP generation, suggesting that cellular bioenergetics are less likely to be involved in CCCP-induced Parkin translocation to mitochondria in HeLa cells. Previous studies have also demonstrated that disruption of bioenergetics in cells even suppresses Parkin translocation to mitochondria (39,40). In short, ROS outburst may have an essential role in Parkin recruitment. Of note, ROS alone may not be sufficient to trigger such an event, as starvation, which is a classical trigger of autophagy, may stimulate marked ROS production but fail to induce Parkin translocation to mitochondria.
In addition, the accumulation of PINK1 may only accelerate Parkin recruitment to mitochondria rather than play an indispensable role in the process. PINK knockdown resulted in an initial suppression of Parkin translocation to mitochondria in response to acute treatment with high concentrations of CCCP, as reported previously (5,41). However, this suppression was alleviated with time despite the absence of PINK1 accumulation, indicating that, to a certain extent, increased PINK1 expression is not necessary for Parkin translocation to mitochondria, whereas decreased PINK1 expression only delays the redistribution of Parkin to mitochondria. In contrast, chronic exposure to low-dose CCCP induced just modest Parkin translocation and ROS generation despite a considerable increase in PINK1 protein level, indicating that ROS may have a significant impact on PINK1-dependent Parkin translocation to mitochondria. Moreover, VDAC1 overexpression induced Parkin translocation to mitochondria without altering mitochondrial potential. The protein level of PINK1 was not changed by VDAC1 overexpression, supporting our hypothesis that the quantity of PINK1 is not a determining factor in Parkin recruitment to mitochondria. However, the presence of PINK1 is still necessary for Parkin recruitment, as VDAC1-induced Parkin translocation to mitochondria was significantly reduced by depletion of PINK1. Moreover, Parkin translocation to mitochondria is completely abolished in PINK1 knock-out mouse embryonic fibroblasts even with long-term exposure of CCCP (4).
Our immunofluorescence data showed that knockdown of PINK1 had little effect on the execution of mitophagy, suggesting that, when recruited to mitochondria, Parkin is able to pro-
We postulate that excessive ROS produced by dysfunctional mitochondria may activate PINK1, thereby triggering Parkin recruitment to mitochondria and succeeding mitophagy. Because excessive ROS are generated in damaged mitochondria, they may modify PINK1 exclusively on impaired mitochondria. This site-specific modification ensures the specificity of Parkin recruitment to unhealthy mitochondria. Another possible scenario is that ROS may cause mutations on the mitochondrial DNA in the affected mitochondria Error bars represent standard deviation of three independent experiments; n Ն 10,000 BFP-positive cells/experiment. Statistical significance was calculated using Student's t test. *, p Ͻ 0.05; NS, not significant (p Ͼ 0.05). C, HeLa cells were transfected with GFP-Parkin (green) and FLAG-VDAC1. Cells were simultaneously treated with 2 mM NAC or 1000 units/ml catalase separately or in combination. Medium was removed and replenished with fresh medium with NAC or catalase, separately or in combination as before, 4 h, 7 h, and 10 h post-transfection. After 12 h, cells were fixed and immunostained for Tom20 (mitochondria, red) and FLAG (purple). D, cells were scored for GFP-Parkin on mitochondria following treatment as described in C. Error bars represent standard deviation of three independent experiments; n Ն 100 cells/experiment. Statistical significance was calculated using Student's t test. *, p Ͻ 0.01; NS, p Ͼ 0.05. E, HeLa cells were transfected with BFP or BFP-T2A-FLAG-VDAC1 and simultaneously treated with 1000 units/ml catalase, 2 mM NAC, separately or in combination, as indicated. Medium was changed as in C. After 12 h, cells were harvested and stained with DHE, followed by flow cytometric analysis. DHE fluorescence was measured in BFP-positive cells. Statistical analysis was performed as in A. *, p Ͻ 0.01); NS, p Ͼ 0.05. F, HeLa cells were transfected with GFP-Parkin and FLAG-VDAC1. Cells were simultaneously treated with 2 mM NAC and 1000 units/ml catalase. Medium was changed as in C. After 12 h, cells were harvested, and protein extracts were assayed for levels of GFP-Parkin and FLAG-VDAC1 by immunoblotting. Scale bars ϭ 10 m. (44), triggering Parkin translocation to mitochondria and initiation of mitophagy thereafter (45).

ROS trigger Parkin/PINK1-dependent mitophagy
Taken together, our data indicate that ROS may act as a trigger for Parkin/PINK1-dependent mitophagy by inducing translocation of Parkin to mitochondria. The establishment of such a trigger for mitophagy has physiological significance, as disturbance of ROS occurs in a variety of physiological conditions. In contrast, the other experimental triggers of mitophagy are less physiologically relevant. Moreover, our findings may have important implications for therapeutic PD strategies based on  (Hoechst 33258, blue). B, cells were scored for GFP-Parkin on mitochondria following treatment as described in A. Error bars represent standard deviation of three independent experiments; n Ն 100 cells/experiment. Statistical significance was calculated using Student's t test. *, p Ͻ 0.01); NS, not significant (p Ͼ 0.05). C, stable PINK1 or control knockdown HeLa cells were treated with 10 M CCCP for the indicated durations. Lysates were harvested and subjected to Western blotting. Arrow, PINK1; asterisks, nonspecific bands. D, control or PINK1 knockdown HeLa cells were treated with 10 M CCCP for the indicated durations. Cells were stained with DHE and subject to flow cytometric analysis. E, HeLa cells were transfected with GFP-Parkin (green) and treated with 5 M or 10 M CCCP for the indicated durations. Fixed cells were stained for TRAP1 (mitochondria, red) and nuclei (Hoechst 33258, blue). F, cells were scored for GFP-Parkin on mitochondria following treatment as described in E. Statistical analysis was performed as in B. G, HeLa cells were treated with 5 M or 10 M CCCP for the indicated durations. Cell lysates were harvested and subjected to SDS-PAGE and immunoblotting for PINK1 or GAPDH. H, HeLa cells were treated with 5 M or 10 M CCCP for the indicated durations. Cells were stained with DHE and subject to flow cytometric analysis. Scale bars ϭ 10 m.

ROS trigger Parkin/PINK1-dependent mitophagy
ROS suppression. If cells use ROS bursts as a signal to initiate autophagy to eliminate dysfunctional mitochondria, then quenching ROS in the cells with functional Parkin or PINK1 may block the initiation of mitophagy, undermining cellular quality control. Thus, careful analysis and fine-tuning of ROS production should be considered when designing PD therapeutics.

Plasmid construction and transfection
The coding sequence of VDAC1 was PCR-amplified from a HEK-293T cell cDNA library and cloned into the pXJ40 vector (a gift from Dr. Low Boon Chuan, Department of Biological Sciences, National University of Singapore) with FLAG, BFP, and GFP tags. The VDAC1 gene was also inserted into the EGFP-N1 vector (Clontech) to fuse a GFP tag to the C terminus of VDAC1. For construction of the BFP-T2A-FLAG-VDAC1 plasmid, an oligo primer was designed to insert the T2A sequence upstream of the FLAG sequence from the pXJ40-FLAG-VDAC1 plasmid. The T2A-FLAG-VDAC1 sequence was then PCR-amplified and subcloned into the pXJ40-BFP vector. The pcDNA3-FLAG-Parkin plasmid has been described previously (46). GFP-Parkin was subcloned from a FLAG-Parkin plasmid and inserted into the pXJ-40-GFP vector. HA-Parkin (plasmid 38248) (47), GFP-LC3 (plasmid 21073) (48), and PINK1-GFP (plasmid 13316) (49) were obtained from Addgene. The PINK1 shRNAs (GCCAACAGGCTCACA-GAGAAGTGTT, adopted from Weihofen et al. (50), and GGAGCAGTCACTTACAGAA) and scrambled control shRNA (CCTAGACGCGATAGTATGGAC) were inserted into a pSUPER plasmid (Oligoengine) and transfected into cells using Lipofectamine 2000 TM (Life Technologies).

Cell culture and treatment
HeLa cells (ATCC) were maintained in DMEM with 10% fetal bovine serum and 1% penicillin/streptomycin at 37°C under 5% CO 2 conditions. To generate stable PINK1 knockdown cell lines, HeLa cells were transfected with control shRNA or PINK1 shRNAs and selected with puromycin (1.0 g ml Ϫ1 ) for 21 days. The stable HA-Parkin HeLa cell line was generated as described previously (51). MG132, CCCP, catalase, and NAC were purchased from Sigma. To stimulate Parkin translocation to mitochondria, cells were incubated with CCCP under the indicated conditions after GFP-Parkin was transfected into cells for 24 h. To induce Parkin translocation with overexpression of VDAC1, plasmids of FLAG-VDAC1 and GFP-Parkin were co-transfected into cells for 12 h prior to experimentation.

ROS level and mitochondrial potential detection
To detect ROS and mitochondrial potential, cells were harvested, washed in PBS, and stained with 5 M DHE or 2 M JC-1 dye at 37°C for 30 min in the dark. Fluorescence was measured immediately with LSRFortessa (BD Biosciences), and the data were analyzed using FACSDiva version 6.2 software (BD Biosciences). The fluorescence intensity of the red signal in DHEstained cells was used to indicate the levels of ROS. The ratio of red to green fluorescence intensity in JC-1-stained cells was used to determine mitochondrial membrane potential. To detect ROS and mitochondrial potential in BFP-T2A-FLAG-VDAC1-overexpressed cells, DHE or JC-1 signals were measured in BFP-positive cells.