A Dimeric PINK1-containing Complex on Depolarized Mitochondria Stimulates Parkin Recruitment*

Background: PINK1 functions on depolarized mitochondria. However, details regarding its mechanism remain limited. Results: We reveal the formation of a high molecular weight complex composed of two phosphorylated PINK1 molecules that stimulates Parkin recruitment. Conclusion: The dimeric PINK1-containing complex is important for mitochondrial quality control. Significance: The PINK1 molecular process is reminiscent of receptor kinase dimerization in signal transduction. Parkinsonism typified by sporadic Parkinson disease is a prevalent neurodegenerative disease. Mutations in PINK1 (PTEN-induced putative kinase 1), a mitochondrial Ser/Thr protein kinase, or PARKIN, a ubiquitin-protein ligase, cause familial parkinsonism. The accumulation and autophosphorylation of PINK1 on damaged mitochondria results in the recruitment of Parkin, which ultimately triggers quarantine and/or degradation of the damaged mitochondria by the proteasome and autophagy. However, the molecular mechanism of PINK1 in dissipation of the mitochondrial membrane potential (ΔΨm) has not been fully elucidated. Here we show by fluorescence-based techniques that the PINK1 complex formed following a decrease in ΔΨm is composed of two PINK1 molecules and is correlated with intermolecular phosphorylation of PINK1. Disruption of complex formation by the PINK1 S402A mutation weakened Parkin recruitment onto depolarized mitochondria. The most disease-relevant mutations of PINK1 inhibit the complex formation. Taken together, these results suggest that formation of the complex containing dyadic PINK1 is an important step for Parkin recruitment onto damaged mitochondria.

Parkinsonism typified by sporadic Parkinson disease is a prevalent neurodegenerative disease. Mutations in PINK1 (PTENinduced putative kinase 1), a mitochondrial Ser/Thr protein kinase, or PARKIN, a ubiquitin-protein ligase, cause familial parkinsonism. The accumulation and autophosphorylation of PINK1 on damaged mitochondria results in the recruitment of Parkin, which ultimately triggers quarantine and/or degradation of the damaged mitochondria by the proteasome and autophagy. However, the molecular mechanism of PINK1 in dissipation of the mitochondrial membrane potential (⌬⌿m) has not been fully elucidated. Here we show by fluorescence-based techniques that the PINK1 complex formed following a decrease in ⌬⌿m is composed of two PINK1 molecules and is correlated with intermolecular phosphorylation of PINK1. Disruption of complex formation by the PINK1 S402A mutation weakened Parkin recruitment onto depolarized mitochondria. The most disease-relevant mutations of PINK1 inhibit the complex formation. Taken together, these results suggest that formation of the complex containing dyadic PINK1 is an important step for Parkin recruitment onto damaged mitochondria.
Parkinsonism is a pervasive neurodegenerative disease caused by a loss of dopaminergic neurons in the substantia nigra. The predominant form of parkinsonism, clinically defined as Parkinson disease (PD), 3 is sporadic. Less prevalent is the heritable familial early onset parkinsonism. Although the symptoms of hereditary parkinsonism are not the same as sporadic PD, they share a number of clinical features (1). Thus, the functional study of genes/proteins involved in familial parkinsonism can also expand our understanding of the pathogenic mechanisms underlying PD.
The mitochondrial Ser/Thr kinase PINK1 (PTEN-induced putative kinase 1) was identified as a causal gene for autosomal recessive early onset parkinsonism (2) and has been studied genetically using PINK1-deficient model organisms. For example, Drosophila melanogaster lacking pink1 is characterized by flight muscle degeneration, short life span, and male sterility (3)(4)(5). Furthermore, mitochondria in the pink1 mutants have morphological and functional defects. Factors that regulate mitochondrial morphology have been shown to interact genetically with pink1 (5-7), and the pink1-deficient phenotype in Drosophila can be rescued by a component of the mitochondrial electron transport chain complex or a mitochondrial electron carrier (8,9). Respiratory chain defects have been reported in Pink1 Ϫ/Ϫ mouse embryonic fibroblasts (10), and some morphological and functional defects of mitochondria have been observed in the striatum of Pink1 null mice; Pink1 knock-out mice, however, do not exhibit any obvious morphological changes or loss of dopaminergic neurons in the substantia nigra (11,12). These findings indicate that PINK1 contributes to mitochondrial integrity.
Reductions in the activity of the electron transport chain complex I and mitochondrial DNA mutations have been observed in sporadic PD (13,14). Further, toxic parkinsonism is caused by various inhibitors of the mitochondrial respiratory chain, such as rotenone and MPTP(1-methyl-4-phenyl-1,2,3,6tetrahydropyridine). Thus, a correlation between parkinsonism and mitochondrial impairment has been well established.
The transport of inner mitochondrial membrane or mitochondrial matrix proteins generally utilizes ⌬⌿m as the driving force (32). In this transport process, the proteins are cleaved by various proteases. Numerous studies have shown that under steady-state conditions, PINK1 is cleaved by MPP (mitochondrial processing peptidase), PARL (presenilin-associated rhomboid-like protein), ClpXP, and AFG3L2 (33)(34)(35)(36)(37) and then recognized by the cytoplasmic N-end rule degradation pathway, which defines the rapid PINK1 turnover (38). The suppression of PINK1 transport into the inner mitochondrial membrane by a decrease in ⌬⌿m triggers an accumulation of PINK1 on OMM (18). Phosphorylation of Ser-228 and Ser-402 activates the accumulated PINK1, triggering the recruitment of Parkin to the depolarized mitochondria (17). Thus, PINK1 is not only quantitatively but also qualitatively regulated by mitochondrial conditions. However, the dynamics of PINK1 on depolarized mitochondria have not been fully elucidated.
In this study, we established a multicolor detection method for resolving with high sensitivity the dynamic interactions of PINK1. Using this method, we demonstrated that PINK1 forms a high molecular weight complex containing a PINK1 dimer. Complex formation is correlated with intermolecular phosphorylation. Defects in PINK1 complex formation significantly reduced Parkin translocation onto depolarized mitochondria, suggesting that PINK1 complex formation might be a pathophysiologically significant process for Parkin-mediated mitochondrial quality control.
Cell Fractionation and Proteinase K-resistant Assay-HeLa cells transfected with expression plasmids were treated with CCCP for 1 h at 37°C, suspended in fractionation buffer (0.25 M sucrose, 20 mM HEPES-NaOH (pH 8.1), and protease and phosphatase inhibitor mixture (Roche Applied Science)) and disrupted by passaging (30 times) through a 25-gauge needle with a 1-ml syringe. Debris was removed by centrifugation at 1,000 ϫ g for 7 min. The resulting supernatant was centrifuged at 10,000 ϫ g for 10 min at 4°C to separate the mitochondriarich fractions from the cytosolic fraction. The mitochondriarich fractions were incubated with 50 g/l proteinase K (Wako) on ice for 30 min, and the reaction was terminated by the addition of 1 mM PMSF prior to electrophoresis.
Phos-tag Assay-To detect phosphorylated proteins via PAGE, 7.5% polyacrylamide gels containing 50 M Phos-tag acrylamide (Wako) and 100 M MnCl 2 were used as reported previously (17). After electrophoresis, the gels were initially incubated with gentle shaking in transfer buffer containing 0.01% SDS and 1 mM EDTA for 10 min and then incubated in transfer buffer containing 0.01% SDS without EDTA for 10 min according to the manufacturer's protocol. Proteins were blotted onto PVDF membranes and analyzed by IB.
Blue Native PAGE (BN-PAGE), Clear Native PAGE (CN-PAGE), and NAMOS Assay-BN-PAGE was performed following the manufacturer's protocol, whereas CN-PAGE based on Wittig and Schagger's method (39) was optimized to facilitate observation of the PINK1 complex. The samples for BN-and CN-PAGE analysis were harvested using a Native TM PAGE Sample Prep Kit (Invitrogen). Cells were extracted with 1ϫ Native TM PAGE buffer for 15 min at 4°C, pipetted up and down 10 times, and centrifuged at 20,000 ϫ g for 30 min at 4°C. BN-PAGE and CN-PAGE were performed using Native TM PAGE Running Buffer (Invitrogen) containing either 0.002% G-250 (Invitrogen) or 0.05% sodium cholate (Wako) and 0.01% n-heptyl-␤-D-thioglucoside (DOJINDO), respectively. The fluorescence signal in clear native polyacrylamide gels was detected with a Typhoon 9400 fluorescence scanner (GE Healthcare). For BN-PAGE, gels were shaken in denaturation buffer (10 mM Tris-HCl pH 6.8, 1% SDS, and 0.006% 2-mercaptoethanol) for 15 min at 60°C after electrophoresis and then transferred to PVDF membranes for IB analysis. For the NAMOS assay, cell lysates were prepared as before, and antibodies were added to the lysate. The samples were incubated at room temperature for 1 h. Samples were subjected to CN-PAGE and fluorescence-detected as before.
Two-dimensional BN/SDS-PAGE-Two-dimensional gel electrophoresis was performed using BN-PAGE in the first dimension. The resulting gels were initially incubated with shaking in a reducing solution (SDS sample buffer with 50 mM DTT) before being transferred to an alkylating solution (SDS sample buffer with 50 mM N,N-dimethylacrylamide) at 60°C for 15 min. Alkylation was terminated by incubating the gels in a quenching solution (SDS sample buffer with 5 mM DTT and 20% methanol) according to the manufacturer's protocol. The equilibrated gel strip was then immediately subjected to second dimension electrophoresis using Phos-tag SDS-PAGE.
siRNA Assay-siRNA was designed to a portion of the PINK1 mRNA 3Ј-UTR comprising 3Ј-GGGUCAGCACGUUCAGU-UAdTdT-5Ј and 5Ј-UAACUGAACGUGCUGACCCAdT-3Ј. HeLa cells stably expressing GFP-or HA-Parkin were pretreated with PINK1 siRNA and then transfected with CMV d1 promoter-driven PINK1 WT, S228A, or S402A mutants. After 42 h, the cells were treated with CCCP (10 M) for the indicated times (1.5 or 16 h), and the number of cells with Parkin-positive mitochondria or Tom20 was determined in Ͼ100 cells. Because the siRNA targeted the 3Ј-UTR of PINK1, it had no effect on expression of the exogenous PINK1 mutants.

Development of a Multicolor Detection Method to Resolve a High Molecular Weight PINK1 Complex in Response to a
Decrease in ⌬⌿m-Under ordinary cellular conditions, PINK1 on healthy mitochondria is immediately cleaved and degraded in the presence of a normal ⌬⌿m. On depolarized mitochondria, however, PINK1 eludes ⌬⌿m-dependent degradation,   and the accumulated PINK1 recruits cytosolic Parkin to the damaged mitochondria (15,16). To observe the dynamics of PINK1 on depolarized mitochondria, we performed BN-PAGE, a method for detecting supermolecular complexes of membrane proteins (40). In BN-PAGE, Coomassie G-250 functions as the negatively charged molecule that binds to the non-denatured proteins and drives their migration toward the anode during electrophoresis. When HeLa cells transiently expressing non-tagged PINK1 (ϳ60 kDa) were treated with CCCP to decrease ⌬⌿m, exogenous PINK1 formed a supermolecular complex (predicted size of ϳ850 kDa; referred to as 850-kDa complex hereafter) ( Fig. 1A) (18,41). We next used differential centrifugation to examine whether this complex is formed on mitochondria. HeLa cells transiently expressing non-tagged PINK1 were treated with CCCP, and the resulting extracts were centrifuged to obtain a mitochondria-rich fraction. BN-PAGE analysis and concomitant immunoblotting indicated that the supermolecular PINK1 complex was indeed recovered in the mitochondrial fraction (Fig. 1B). Moreover, we confirmed that endogenous PINK1 formed a similar 850-kDa complex in CCCP-treated cells as well as exogenous PINK1 (Fig. 1C). We next performed CN-PAGE (39,42), in which sodium cholate is used as the negatively charged molecule. Similar to BN-PAGE, proteins bound to sodium cholate migrate toward the anode without denaturation. Because sodium cholate is colorless and the native folding structure of proteins is maintained during electrophoresis, it is possible to use fluorescently tagged proteins (e.g. GFP and mCherry) for in gel detection in CN-PAGE (39). When HeLa cells expressing PINK1-GFP were treated with CCCP and then subjected to BN-and CN-PAGE, the fluorescently tagged protein co-localized with the supermolecular complex (Fig. 1D, middle lane). This band disappeared A, exogenous PINK1 forms a supermolecular complex following CCCP treatment. HeLa cells expressing non-tagged PINK1 were treated with CCCP (10 M, 1 h), and the resulting cell lysates were subjected to BN-PAGE and immunoblotted with an anti-PINK1 antibody. The blue arrowhead indicates the supermolecular PINK1 complex (note that this designation will be used hereafter unless otherwise specified). B, mitochondria-rich fractions isolated from cells treated as before were analyzed by BN-PAGE and immunoblotted. C, endogenous PINK1 also forms a supermolecular complex following CCCP treatment. HeLa cells were treated with CCCP (10 M, 12 h) and subjected to BN-PAGE and immunoblotted. D, the presence of the PINK1 complex following a decrease in ⌬⌿m was detected using an anti-PINK1 antibody or GFP fluorescence. HeLa cells expressing PINK1-GFP were treated with CCCP (10 M, 3 h) and subjected to BN-PAGE or CN-PAGE. PINK1 was detected using an anti-PINK1 antibody following blotting to a PVDF membrane (left and middle) or was detected directly in gel by GFP fluorescence using a fluorescence imaging scanner (right). E, dual color imaging of PINK1 using distinct florescent proteins. HeLa cells expressing PINK1-GFP and/or PINK1-mCherry were treated with CCCP (10 M, 3 h) and subjected to CN-PAGE, and fluorescence signals derived from GFP and mCherry were detected.
when ⌬⌿m was recovered with CCCP washout (Fig. 1D, right). CN-PAGE in gel GFP fluorescence offered higher resolution and sensitivity for detecting the PINK1 complex than immunoblotting using an anti-PINK1 antibody (Fig. 1D, compare the right panel with the middle panel). Similar to PINK1-GFP, PINK1-mCherry likewise formed the 850-kDa complex following a decrease in ⌬⌿m and was detected by the multicolor detection method (Fig. 1E). These results confirmed the utility of the multicolor fluorescence detection method and suggested that this approach would be advantageous for monitoring PINK1 dynamics.
The 850-kDa High Molecular Weight Complex Contains Two Molecules of PINK1-We next performed a NAMOS assay (43) to determine the component constituency and stoichiometry of the supermolecular complex. In this assay, an antibody against one component causes the complex to shift to a higher molecular weight position in native PAGE. HeLa cell extracts containing PINK1-GFP were subjected to CN-PAGE with either an anti-PINK1 or anti-GFP antibody, followed by detection of PINK1-GFP fluorescence ( Fig. 2A). In this experiment, we used multiple antibodies, including a polyclonal anti-PINK1 antibody, BC100-494 (immunoreactive to amino acids 175-250 of PINK1); polyclonal anti-GFP antibodies A6455 and ab6556 (which recognize full-length GFP); and the monoclonal anti-GFP antibody 3E6. When the anti-PINK1 antibody was incubated with the 850-kDa complex, a shift in the PINK1 complex was clearly observed (Fig. 2, A and B, lane 2). When the monoclonal antibody 3E6 was utilized, PINK1-GFP resolved as a single band (probably because the antibody recognizes a unique GFP site), whereas PINK1-GFP resolved as multiple bands with the polyclonal antibodies (e.g. BC100-494, A6455, and ab6556) ( Fig. 2A). We next determined if components of the TOM machinery were also included in the PINK1 complex (18). In this assay, antibodies against Tom20, Tom22, and Tom40 caused a shift in the PINK1 complex in CN-PAGE (Fig. 2B, lanes [3][4][5], suggesting that all three TOM components are included in the PINK1 complex. In contrast, an anti-Tom70 antibody had no effect on the position of PINK1 fluorescence (lane 6). Although we cannot exclude the possibility that Tom70 interacts with the PINK1 complex in a digitonin-sensitive manner or a Tom70 antibody recognition site is masked, this result suggests that the complex is devoid of Tom70 (see "Discussion").
We next used PINK1 proteins fused to a distinct tag (V5, GFP, or mCherry) to obtain more detailed stoichiometric information regarding the number of PINK1 molecules contained in the complex. PINK1-GFP and PINK1-mCherry were co-expressed in HeLa cells, treated with CCCP, and subjected to the NAMOS assay using anti-GFP and anti-mCherry monoclonal antibodies. If the 850-kDa complex is composed of only PINK1-GFP, then the addition of an anti-mCherry antibody should have no effect on the position of the PINK1-GFP fluorescent signal in CN-PAGE. A shift in the PINK1-GFP signal by the anti-mCherry antibody, however, would indicate that the 850-kDa complex contains both PINK1-mCherry and PINK1-GFP. When PINK1-GFP alone was expressed in cells, the GFP fluorescence signal did not change following the addition of the anti-mCherry antibody (Fig. 2C, lane 7). In contrast, when co-expressed with PINK1-mCherry, the PINK1-GFP-derived fluorescence signal was noticeably shifted upward following the addition of the anti-mCherry antibody (Fig. 2C,  lane 9).
The above findings revealed that the 850-kDa complex contains at least two PINK1 molecules. We next sought to determine the stoichiometry. PINK1-GFP was co-expressed with both PINK1-V5 and PINK1-mCherry, subjected to a NAMOS assay using anti-V5 and anti-mCherry monoclonal antibodies, and detected by GFP-derived fluorescence in CN-PAGE. If the 850-kDa complex contains two PINK1 molecules, only a onestep shift in GFP-fluorescence (equivalent to one antibody) would be observed under the aforementioned conditions regardless of the type and combination of antibodies used, whereas, if the 850-kDa complex contains more than two PINK1 molecules, a two-step shift in GFP fluorescence would be observed with the anti-V5 and anti-mCherry antibodies. HeLa cells transfected with PINK1-GFP alone, PINK1-GFP and PINK1-V5, PINK1-GFP and PINK1-mCherry, or all three constructs were treated with CCCP and then subjected to the NAMOS assay using anti-V5 and anti-mCherry antibodies. No two-step shift in GFP-fluorescence was observed (Fig. 2D), strongly suggesting that the 850-kDa complex contains only two PINK1 molecules. A two-step shift was only observed following the inclusion of an anti-GFP antibody along with the anti-V5 and anti-mCherry antibodies, thus confirming that the supermolecular complex is limited to two PINK1 molecules (Fig. 2D, far right panel).
The High Molecular Weight PINK1 Complex Is Composed of Phosphorylated PINK1-PINK1 rapidly accumulates on depolarized mitochondria and undergoes autophosphorylation following a decrease in ⌬⌿m (17,44). Hence, we next examined the phosphorylation state of PINK1 with polyacrylamide gels conjugated to a 1,3-bis(bis(pyridin-2-ylmethyl)amino)propan-2-olato diMn(II) complex (referred to here as Phos-tag (45)). Phos-tag specifically retards the migration of phosphorylated proteins; thus, phosphorylated PINK1 resolved as a slower migrating band compared with non-phosphorylated PINK1 (Figs. 3, A and B). To determine if PINK1 in the high molecular weight complex is phosphorylated, we performed twodimensional electrophoresis in which the proteins were first resolved using BN-PAGE and then subsequently Phos-tag SDS-PAGE. Following CCCP treatment, almost all of the PINK1 in the complex underwent a molecular shift in the Phos-tag SDS-PAGE, revealing that PINK1 in the high molecular weight complex is the phosphorylated form (Fig.  3, C and D). These results suggest that PINK1 complex formation is related to PINK1 phosphorylation.
Disease-relevant Mutations of PINK1 Inhibit Complex Formation-We next sought to determine if various pathogenic mutations or a kinase-dead (KD) mutation consisting of the triple missense mutation K219A/D362A/D384A (46) inhibit formation of the PINK1 complex in cells. Constructs harboring the PINK1 KD mutation or one of 10 pathogenic mutations, including C92F, A168P, E240K, H271Q, G309D, L347P, G386A, G409V, E417G, and a pathogenic mutation that results in the insertion of glutamine at amino acid position 534 (referred to hereafter as 534insQ), were generated. These PINK1 mutants, with the exception of C92F, partially or completely impaired phosphorylation and inhibited Parkin recruitment onto depolarized mitochondria without altering mitochondrial localization (16,17,47). HeLa cells expressing PINK1-GFP harboring the aforementioned mutations were subjected to CN-PAGE following CCCP treatment. Formation of the 850-kDa PINK1 complex was completely or severely inhibited by the KD mutation as well as the A168P, E240K, H271Q, L347P, G386A, E417G, and 534insQ mutations (Fig.  4A). In contrast, the C92F, G309D, and G409V PINK1 mutants formed the 850-kDa complex equivalent to wild type (WT) PINK1 (Fig. 4A). Immunoblotting by conventional SDS-PAGE with an anti-PINK1 antibody confirmed that the mutant PINK1 proteins were almost equally expressed (Fig. 4B). The NAMOS assay revealed that the C92F, G309D, and G409V PINK1 mutants interact with components of the TOM complex as well as PINK1 WT (Fig. 4C). Taken together, the mutation data demonstrated that various PINK1 mutants were unable to form the 850-kDa complex in cells, suggesting an etiological importance of PINK1 complex formation. were incubated with anti-GFP and/or anti-mCherry antibodies and subjected to CN-PAGE, and GFP or mCherry fluorescence was detected as before. The green asterisks indicate a PINK1-GFP shift caused by interaction between the anti-mCherry antibody and PINK1-mCherry. The red asterisks indicate a PINK1-mCherry shift caused by the anti-GFP antibody. D, HeLa cells expressing PINK1-GFP, PINK1-mCherry, and/or PINK1-V5 treated with CCCP (10 M, 3 h). These cell extracts were incubated with anti-mCherry, anti-V5, and/or anti-GFP antibodies, and GFP fluorescence was detected as before.

PINK1 Complex Formation Is Correlated with Intermolecular Phosphorylation of PINK1 following the Reduction of ⌬⌿m-
We next used CCCP exposure and subsequent washout to determine which regulation process is correlated with PINK1 complex formation. To date, several aspects of PINK1 regulation following a change in ⌬⌿m have been reported e.g. PINK1 accumulates on OMM and undergoes ⌬⌿m dissipation-dependent phosphorylation but is reimported and degraded when ⌬⌿m is recovered following CCCP washout (Fig. 5A) (reviewed in Ref. 48). Hence, we examined the relationship between these processes and formation of the 850-kDa complex. First, HeLa cells expressing PINK1 WT and the pathogenic mutants were treated with CCCP, homogenized, and subjected to an in vitro proteinase K-resistant assay. In this assay, OMM proteins are easily degraded, whereas matrix proteins are more resistant. We used Tom20 (a single spanning OMM protein), Tom40 (a ␤-barrel protein in OMM), Tim23 (an inner mitochondrial membrane protein facing the intermembrane space), and HSP60 (a matrix protein) as control proteins (Fig. 5B, depicted on the right). We observed that the accumulation of some PINK1 mutants on mitochondria was attenuated; however, there was no correlation between protease resistance and complex formation (Fig. 5B). Complete degradation of PINK1 by proteinase K was similar to that of Tom20, suggesting that PINK1 WT and the pathogenic mutants localize to OMM (41,49).
We also examined the reimport and degradation ability of various PINK1 mutants in response to ⌬⌿m recovery following CCCP washout. Similar to WT PINK1, all of the PINK1 patho-genic mutants accumulated following CCCP treatment. Furthermore, when CCCP was washed out, all of the PINK1-FLAG pathogenic mutants examined underwent degradation with the cleaved form of PINK1 (⌬1) increasing (Fig. 5C) despite most of the mutants being unable to form the PINK1 complex (Fig. 4A). This result suggests that the correlation between PINK1 reimport activity and PINK1 association with the TOM complex is tenuous under our experimental conditions. Last, we examined the correlation between the supermolecular complex formation and intermolecular phosphorylation (17) of PINK1. PINK1-FLAG WT or the pathogenic mutants were co-expressed with WT PINK1-GFP, the cells were exposed to CCCP, and the phosphorylation state of PINK1-FLAG was determined. Although both phosphorylated and unphosphorylated forms of PINK1 resolve as a single band on   WT  G409V  G386A  L347P  G309D  H271Q  E240K  A168P  C92F  KD  534insQ  E417G  PINK1  - commercially available precast gels, such as NuPAGE 4 -12% BisTris gels (Figs. 4B and 5 (B and C)), PINK1 can be resolved as a doublet using hand-made 7.5% Tris-glycine gels with phosphorylated PINK1 migrating as a higher molecular weight band as shown in Fig. 6 (top) and Ref. 17. We also performed Phos-tag PAGE and immunoblotted using an anti-FLAG antibody to distinguish between phosphorylated PINK1-FLAG and degraded PINK1-GFP (Fig. 6, fourth panel). PINK1 WT and the C92F and G309D mutants underwent phosphorylation irrespective of PINK1-GFP co-expression (lanes 5, 9, and 17). The KD, A168P, E240K, H271Q, L347P, E417G, and 534insQ PINK1 mutants showed no detectable phosphorylation in the absence of PINK1-GFP (lanes 7, 11, 13, 15, 19, 25, and 27). The G386A and G409V PINK1 mutants exhibited an aberrant phosphorylation pattern in the absence of PINK1-GFP (lanes 21 and 23; compare the position of the WT phosphorylated band with that in lane  5). Interestingly, the G409V mutant, which can interact with the 850-kDa complex (Fig. 4), exhibited an aberrant phosphorylation pattern in the absence of WT PINK1 (lane 23) but demonstrated the expected phosphorylation-derived shift when PINK1(WT)-GFP was co-expressed (Fig. 6, fourth panel, lane  24, blue arrowhead). These results suggest that the two PINK1 molecules in the 850-kDa complex contribute to intermolecular phosphorylation. PINK1 Complex Formation Is Important for Efficient Parkin Recruitment on Mitochondria with Decreased ⌬⌿m-We previously identified Ser-228 and Ser-402 in PINK1 as the sites phosphorylated following a decrease in ⌬⌿m; a S228A/S402A PINK1 double mutant was not phosphorylated despite localizing on depolarized mitochondria (17). We consequently exam- . A, the proposed mechanisms described are illustrated. A reduction in ⌬⌿m results in the accumulation of PINK1 and cross-phosphorylation of the two PINK1 molecules. Reestablishment of ⌬⌿m following CCCP washout results in the reimport and degradation of PINK1. B, HeLa cells expressing PINK1-FLAG WT, KD, or various pathogenic mutants were treated with CCCP (10 M, 1 h) and fractionated. Mitochondria-rich fractions were incubated with 50 g/l proteinase K (PK) and subjected to SDS-PAGE using precast 4 -12% BisTris gels and immunoblotted using an anti-PINK1, anti-Tom20, anti-Tom40, anti-Tim23, or anti-HSP60 antibody. C, HeLa cells expressing PINK1-FLAG WT, KD, or pathogenic mutants were treated with CCCP (10 M, 3 h). The CCCP was subsequently washed out for 1 h, and SDS-PAGE was performed using precast 4 -12% BisTris gels and then immunoblotted using anti-PINK1 and anti-actin antibodies. DECEMBER 20, 2013 • VOLUME 288 • NUMBER 51

Dimeric PINK1-containing Complex on Mitochondria
ined whether mutations in Ser-228 and Ser-402 affect formation of the PINK1 complex. The PINK1 S228A mutation had no effect on formation of the PINK1 complex in CN-PAGE, whereas the S402A mutation strongly hindered formation (Fig.  7A); however, both sites affect the phosphorylation pattern of PINK1 (Fig. 7B) (17). Single S228A or S402A mutations are insufficient to block PINK1 phosphorylation, which is completely abolished by the S228A/S402A double mutation (17). Although the PINK1 S402A mutation inhibited formation of the 850-kDa complex (Fig. 7A), this result does not necessarily mean that phosphorylation at Ser-402 is essential for complex formation. Indeed, in cells expressing G409V alone, Ser-402dependent phosphorylation was not observed (Fig. 6, top, lane 23; a higher molecular weight band equivalent to Ser-402-phosphorylated PINK1 was not observed in non-Phos-tag PAGE), but formation of the complex was equivalent to WT PINK1 (Fig. 4A). These mutants should thus enable us to examine whether or not formation of the PINK1 complex is required for Parkin function on damaged mitochondria. We also examined the Parkin recruitment activity of the PINK1 S228A or S402A mutants following CCCP treatment. HeLa cells stably expressing GFP-Parkin were treated with PINK1 siRNA to deplete endogenous PINK1. Those cells were then transfected with plasmids containing PINK1 WT and the S228A or S402A mutants under the control of a modified cytomegalovirus (CMV) promoter (referred to as CMV d1), which contains a deletion that reduces the strength of the promoter and thus lowers overall expression. Because the siRNA was designed to a portion of the PINK1 3Ј-UTR, there should be no effect on expression of the exogenous PINK1, which lacks the corresponding sequence. Although excessive overproduction of PINK1 under the normal CMV promoter targets Parkin to mitochondria irrespective of a reduction in ⌬⌿m, and thus regardless of PINK1 complex formation (15, 50), co-expression of CMV d1 promoter-driven PINK1 and Parkin results in CCCP-dependent mitochondrial localization of Parkin (17). PINK1 reduction using siRNA inhibited Parkin recruitment onto depolarized mitochondria even following CCCP treatment (Fig. 7C, yellow arrowheads), whereas reintroduction of WT PINK1 rescued the Parkin recruitment defect (Fig. 7C, white asterisks). Expression of the CMV d1 promoter-driven PINK1 was too weak to be detected by immunocytochemistry using the anti-PINK1 antibody, and thus the transfected cells were marked by co-expression with DsRed. Reintroduction of the control vector did not complement the Parkin recruitment defect (Fig. 7C, red asterisks). Following CCCP treatment (10 M, 1.5 h), the number of cells with Parkin-positive mitochondria was determined in Ͼ100 transfected DsRed-positive cells. Interestingly, the PINK1 S228A mutant, which undergoes complex formation, complemented mislocalization of Parkin equivalent to WT PINK1, whereas the S402A mutant, which does not form the complex, exhibited significantly reduced Parkin recruitment to depolarized mitochondria (Fig. 7D).
We next examined the subsequent effect of these mutants on mitochondrial degradation. Endogenous PINK1 in HeLa cells stably expressing HA-Parkin was depleted by siRNA, and the S228A or S402A mutants PINK1 were introduced as before. To monitor OMM protein degradation, Tom20 was immunostained following prolonged exposure to CCCP (10 M, 16 h) as reported previously (16,21), and the number of Tom20 immunoreactive cells was determined in Ͼ100 transfected (GFP-positive) cells. The complex formation-competent PINK1 S228A mutant accelerated OMM protein degradation equivalent to WT PINK1, whereas the complex formation-deficient S402A mutant exhibited significantly reduced OMM protein degradation (Fig. 7, E and F). Taken together, these results suggest that formation of the PINK1 complex plays an important, albeit not essential, role in Parkin recruitment and mitochondrial degradation.

DISCUSSION
In this study, we revealed the stoichiometry and functional significance of a dyadic PINK1-containing complex that was formed following a decrease in ⌬⌿m. Although Lazarou et al. (18) previously reported formation of the PINK1 complex on depolarized mitochondria and suggested a role in the reimport process following recovery of ⌬⌿m, we show that the PINK1 complex also has an important function on damaged mitochondria. First, we confirmed that exogenous and endogenous PINK1 form a complex following CCCP treatment (Fig. 1, A-C) (18,41). Thereafter, we identified components of this PINK1 complex. Conventionally, immunoprecipitation was performed to determine the components of the protein complex. However, our initial immunoprecipitation attempts to identify components of the CCCP-treated PINK1 complex suffered from poor reproducibility. Exogenous PINK1 exists in two forms, a high molecular weight complex depending on reduction of ⌬⌿m and a low molecular species irrespective of mitochondrial state that is observed as a broad smear (Figs. 1A and 3 (C and D)). Separation and characterization of only the high molecular weight complex is difficult by immunoprecipitation and thus might account for the poor reproducibility. To address this problem, we developed an experimental system to detect the PINK1 complex using fluorescence in combination with CN-PAGE and NAMOS assays (Fig.  2B). This allowed us to determine the component constituency and stoichiometry of this multimeric complex. Several studies have proposed a direct interaction between PINK1 and Parkin (50 -52). We have also observed that overexpressed PINK1 results in Parkin recruitment to mitochondria even without CCCP treatment (15,17,50), which is easily explainable if PINK1 physically interacts with Parkin. We and others have reported that PINK1 phosphorylates Parkin (44,53,54). In addition, Lazarou et al. observed that ectopically overexpressed PINK1 translocates Parkin to the same organelle (18). These results seemingly suggest that PINK1 and Parkin interact directly. If so, does the 850-kDa PINK1 complex contain Parkin? In our initial experiments, we failed to detect Parkin in the complex following CCCP treatment, 4 which is consistent with Lazarou et al. (18). In addition, PINK1 reimport and degradation following ⌬⌿m recovery precedes the disengagement of Parkin from polarized mitochondria (18). These data are inconsistent with the hypothesis that Parkin localizes on mitochondria via stable interactions with PINK1. Because Parkin is not incorporated into the complex, we surmise that PINK1 does not stably interact with Parkin but rather accelerates the interaction between Parkin and the membrane. Further studies are needed to comprehensively understand this issue.
We also examined the complex-forming ability of pathogenic PINK1 mutants that cause hereditary early onset parkinsonism. Although it has been reported, based on an in vitro mitochondrial import assay, that several pathogenic PINK1 mutants formed the complex (18), we found that most PINK1 mutants, with the exception of C92F, G309D, and G409V, completely or partially lost the complex-forming ability (Fig. 4A). The PINK1 G409V mutant exhibits typical complex formation ( Fig. 4) but failed to undergo complete phosphorylation (Fig. 6) and inhibited Parkin recruitment and activation on depolarized mitochondria (17,47). We thus surmise that the principle functional defect of this mutant is in the process of PINK1 autophosphorylation or substrate recognition. Various PINK1 mutants were unable to form the complex in cells, suggesting an etiological importance of PINK1 complex formation.
We have previously reported that PINK1 undergoes intermolecular phosphorylation (17). In this study, we noticed that the complex-forming ability of PINK1 correlated well with the intermolecular phosphorylation of PINK1 (Figs. 4 and 6). In addition, using a novel procedure, we determined that the 850-kDa complex contains two PINK1 molecules (Fig. 2). This result seems reasonable because two PINK1 molecules must localize in physical proximity to perform trans-phosphorylation. Interestingly, when we performed the NAMOS assay using an anti-GFP monoclonal antibody, 3E6, the second molecular weight shift was not observed (Fig. 2A). Nonetheless, the data obtained from the NAMOS assay using cells co-expressing PINK1-GFP and PINK1-mCherry clearly showed that two molecules of PINK1 are present in the 850-kDa complex (Fig. 2C). We surmise that PINK1 forms an asymmetric dimer such that one of the antibody recognition sites is masked by either the second PINK1 molecule or the TOM complex. Indeed, some cell surface receptor kinases dimerize asymmetrically following specific ligand binding prior to the subsequent autophosphorylation event (55,56). To our knowledge, this is the first report to demonstrate PINK1 dimerization in response to a decrease in ⌬⌿m. The formation of a complex containing two PINK1 molecules might play a pivotal role in intermolecular phosphorylation.
Although two groups have reported that PINK1 forms a supermolecular complex after dissipation of ⌬⌿m (18,41), the incorporation of TOM components in the PINK1 complex was controversial. In general, the TOM complex consists of distinct import receptors (Tom20 or Tom70) and a core complex (common import receptor Tom22, pore Tom40, and organizer Tom5, -6, and -7). The complex imports various mitochondrial proteins by distinct mechanisms. For example, proteins containing a presequence are imported in a Tom20-dependent manner, whereas carrier proteins and proteins containing multiple ␣-helical transmembrane segments are imported in a Tom70-dependent manner. Moreover, ␤-barrel proteins are integrated into the mitochondrial outer membrane by the cooperation of Tom20-and Tom70-dependent processes (32,57). Using immunoprecipitation in the presence of a crosslinker and with Tom20 and Tom22 under native conditions, Lazarou et al. reported that PINK1 interacts with Tom20, Tom22, Tom40, and Tom70 (18). Meanwhile, Becker et al. showed that an immunodetected Tom40 signal did not co-migrate with the PINK1 complex and thus concluded that the complex is unlikely to contain TOM components (41). Our data, which consist of two independent assays (the NAMOS assay and CN-PAGE), are consistent with the report by Lazarou et al. (18), although we did not observe the incorporation of Tom70 (Fig. 2). We surmise that the difference in experimental conditions influenced the interaction between Tom70 and PINK1. Although the physiological significance of TOM components in the PINK1 complex remains obscure, TOM components might assist in orientation of the PINK1 pair to facilitate the intermolecular phosphorylation.
Finally, to confirm the close relationship between PINK1 complex formation and Parkin function on depolarized mitochondria, we analyzed the complex-forming ability of PINK1 Ser-228 or Ser-402 mutants. We observed that the PINK1 S402A mutant, which is unable to form the 850-kDa complex, partly but significantly reduced mitochondrial localization of Parkin following CCCP treatment. In contrast, the S228A mutant, which is a component of the 850-kDa complex, recruited Parkin onto depolarized mitochondria equivalent to wild type PINK1 (Fig. 7). Complex formation and concomitant intermolecular phosphorylation of PINK1 is not imperative for Parkin translocation because the complex-deficient S402A mutant still partially recruited Parkin to the mitochondria (ϳ40%; Fig. 7D). Similarly, peroxisome-targeted PINK1 lets Parkin localize on the peroxisome, which is devoid of TOM components (18). Nevertheless, our finding indicates that the process is important for the efficient retrieval of Parkin to the mitochondria.
In conclusion, we propose that formation of a high molecular mass (850 kDa) supermolecular complex containing dimeric PINK1 contributes to intermolecular phosphorylation and the effective recruitment of Parkin onto depolarized mitochondria.