PINK1-dependent recruitment of Parkin to mitochondria in mitophagy

Protonophores Induce Parkin Relocalization to Mitochondria

Mounting evidence indicates that Parkin modulates mitochondrial dynamics and autophagy (4, 5, 7, 9). A prerequisite for Parkin’s actions on mitochondria may be its translocation from the cytosol to mitochondria, as shown after a dissipation of ΔΨ_m with the protonophore carbonyl cyanide m-chlorophenylhydrazone (CCCP) in clonal cell lines (9). In the present study, we show that cytosolic Parkin translocates to mitochondria in transiently transfected embryonic kidney HEK293T cells, expressing YFP-tagged Parkin (Parkin-YFP), upon exposure to 10 μM of either CCCP or its analog, carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP) (Fig. S1). As in the study by Narendra et al. (9), we found that >150 out of 250 (>60%) of our analyzed transfected cells, rather than showing a normal diffuse Parkin-YFP fluorescence, exhibited 1 large or several smaller discrete Parkin-YFP spots after only 1-h exposure to these protonophores (Fig. S1A). Under these experimental conditions, >85% of these spots colocalized with the mitochondrial protein TOM20, and >65% of the mitochondria colocalized with Parkin after protonophore exposure, vs. ∼10% after vehicle exposure (Fig. S1B). Identical results were obtained with GFP-tagged Parkin (Parkin-GFP) or c-Myc–tagged Parkin (Parkin-myc) and with Parkin-YFP–transfected human neuroblastoma SH-SY5Y and cervical carcinoma HeLa cells. In contrast to HeLa cells transfected with WT Parkin (Parkin^WT)-myc, cells transfected with mutated Parkin^T415N- or Parkin^G430D-myc, which are 2 PD pathogenic mutations (2), retained a normal diffuse cytosolic fluorescence whether cells were incubated with a protonophore or vehicle (Fig. 1B).

Dissipation of ΔΨ_m Triggers Parkin Relocalization

To examine further the effect of protonophores on Parkin cytosolic/mitochondrial partition, we prepared subcellular fractions from both non–neuronal-like HEK293T and neuronal-like SH-SY5Y cells, as before (10). These experiments demonstrated that endogenous Parkin was enriched in the mitochondrial fractions after only 1-h exposure to 10 μM CCCP (Fig. S1C). Furthermore, we found that Parkin contained in mitochondrial extracts from cells treated with 10 μM CCCP was accessible to digestion with proteinase-K (Fig. 1A). This result indicates that, upon translocation, Parkin associates with the outer surface of the mitochondria. Concentrations of CCCP or FCCP in excess of 1 μM, as used by Narendra et al. (9) and by us here, can affect cellular functions other than ΔΨ_m (11, 12). Nonetheless, as much as 30–50% of WT Parkin-YFP–transfected cells did show Parkin translocation to mitochondria triggered by mitochondrial depolarization, whether it was caused by lower concentrations of CCCP (10 nM–1 μM), by coincubation with 1 μM of the complex III inhibitor, antimycin A, plus 1 μM of the F₁F₀ ATPase inhibitor, oligomycin, or by a complete loss of mitochondrial respiratory function in Rho⁰ cells (Fig. 1 C and D). In each of these 3 conditions, Parkin-YFP translocation to MitoTracker Deep Red-labeled mitochondria was observed in cells that consistently had the lowest ΔΨ_m as assessed using tetramethyl rhodamine methyl ester (TMRM) fluorescence (Fig. 1E). Thus, these results provide further support to the notion that a marked mitochondrial depolarization triggers Parkin recruitment to the mitochondria. A loss of ΔΨ_m, as modeled here, is often regarded as a correlate to mitochondrial damage. If that correlation is valid, future studies will have to elucidate the actual nature of the mitochondrial damage that triggers Parkin relocalization.

Parkin Translocation Is PINK1-Dependent

Cereghetti et al. (13) have reported that mitochondrial depolarization also can stimulate the translocation of the fission protein Drp1 from the cytosol to the mitochondria through a calcineurin-dependent mechanism. Despite the apparent similarity between the Drp1 and Parkin observations, in our hands, the calcineurin inhibitor, cyclosporine A, failed to prevent CCCP-induced Parkin translocation (Fig. 1D), thus suggesting that Drp1 and Parkin translocation to mitochondria is governed by distinct molecular underpinnings. Given the reported Parkin/Pink1 genetic interaction observed in Drosophila (4, 5, 7) and our revised PINK1 topology (10), we then asked whether PINK1 plays any role in the mitochondrial recruitment of Parkin. To address this question, we used a PINK1 siRNA construct and HeLa cells, because we have previously shown that this reagent reduces PINK1 mRNA by >80% in these specific cells (10). When PINK1 was silenced in Parkin^WT-YFP–transfected HeLa cells, the CCCP-induced collapse of ΔΨ_m was no longer associated with a relocalization of cytosolic Parkin to the mitochondria (Fig. 2 A–C). A similar observation was made in primary cortical neurons from Pink1-knockout mice (Fig. 2D). The proto-oncogene DJ-1, when mutated, also causes a familial form of PD (2) and was suggested to interact with Parkin and PINK1 to form a mitochondrial multiprotein complex (14). However, unlike PINK1 silencing, DJ-1 knockdown by >75% in HeLa cells (Fig. 2B) had no effect on CCCP-mediated Parkin relocalization (Fig. 2C). Thus, whatever the functional nature of DJ-1 interaction with Parkin and PINK1 may be, our data exclude the possibility that DJ-1 is required for the PINK1-dependent translocation of Parkin to mitochondria.

PINK1 Causes Parkin Relocation in Cells with Normal ΔΨm

Next, we sought to determine the effect of increased PINK1 expression on ΔΨ_m and Parkin subcellular distribution. To investigate this effect, we took advantage of stable rat fetal mesencephalic N27 cell lines developed by M.C. and K.T., which have an ecdysone-inducible mammalian expression system to regulate the expression of either human WT PINK1 (PINK1^WT), 2 PD-linked PINK1 mutants (the truncating nonsense mutation PINK1^W437× and the missense mutation PINK1^L347P) (2), or, as control, an empty vector. These 4 cell lines, which share phenotypic similarities with dopaminergic neurons, were transfected transiently with Parkin-YFP as above and, 6 h later, were exposed to ponasterone A to induce the expression of human PINK1 (Fig. 3). In these cells, ΔΨ_m was comparable to that of Parkin-YFP–transfected/PINK1-noninduced and Parkin-YFP–transfected/empty vector-induced cells (Fig. S2 A and B). However, despite having a normal ΔΨ_m—as evidenced by TMRM fluorescence in live-cell imaging (Fig. S2A)—cytosolic Parkin-YFP relocalized to the mitochondria at a time point corresponding to marked PINK1^WT induction (Fig. 3 A–C). Although the PINK1^W437× and PINK1^L347P cells had expression levels comparable to that of the PINK1^WT cells (Fig. 3F), no Parkin-YFP relocalization to the mitochondria was noted (Fig. 3 D and E). The effects of WT but not of mutated PINK1 on Parkin translocation were confirmed by Pink1/Parkin-YFP cotransfection in HeLa cells (Fig. S2C). We also transfected these cells with the artificial kinase dead mutant PINK1^K219M, which we showed to be overexpressed to comparable levels as PINK1^WT (10). Here, the proportion of cells overexpressing PINK1^K219M with Parkin-YFP relocalization (4.9 ± 3.0%, n = 100) was lower than that of cells overexpressing PINK1^WT (97.0 ± 1.4%, n = 100; Student’s t test: t(198) = 27.7, P < 0.001). These results suggest that WT PINK1, but neither pathogenic nor functionally dead PINK1 mutants, is instrumental in the relocalization of cytosolic Parkin and operates downstream of mitochondrial depolarization.

Because both a loss of ΔΨ_m and an increase in PINK1 expression promote Parkin translocation, we wondered if mitochondrial depolarization could enhance PINK1 expression. However, because CCCP triggers Parkin translocation within 1 h, we reasoned that any effect that a loss of ΔΨ_m might have on PINK1 must be posttranslational in nature. Consistent with this view, we found, as before (10), that untreated HeLa cells were the only cells of the varied cell types used in this work in which endogenous mitochondrial PINK1 was detectable, albeit barely (Fig. 4), but endogenous PINK1 was seen clearly after 1-h exposure to 10 μM CCCP (Fig. 4). Remarkably, the mitochondrial contents of both full-length 63-kDa and cleaved 52-kDa PINK1 species increased after dissipation of ΔΨ_m, suggesting that mitochondrial depolarization may enhance PINK1 stability. Although the latter hypothesis may have to be tested formally in future studies, it has been reported that Parkin may indeed stabilize PINK1 (15).

Parkin Binds to PINK1 Without Modifying Each Other

The data presented in the previous sections raise the possibility that, once recruited to mitochondria, Parkin is physically apposed to PINK1, and this apposition may have important functional consequences. To ascertain this physical proximity, we used fluorescence lifetime imaging microscopy (FLIM) in living HEK 293T cells as reported previously (16). The fluorescence lifetime of CFP tagged at the C-terminus of PINK1 in transfected cells was 2.61 ± 0.04 ns (mean ± SEM; n = 7), but when cells were cotransfected with Parkin-YFP, the lifetime was reduced to 2.03 ± 0.03 ns (n = 4; Student's t test: t(9) = 8.57, P < 0.001; Fig. S3). However, when cells were cotransfected with DJ-1-YFP instead of Parkin-YFP, the lifetime of PINK1-CFP was unchanged, 2.61 ± 0.02 ns (n = 5; Student's t test: t(10) = 0.79, P = 0.449) (Fig. S3A). These results indicate that a positive energy transfer occurred specifically between PINK1-CFP and Parkin-YFP, supporting the close proximity of these 2 proteins.

We further assessed the physical proximity of Parkin and PINK1 by coimmunoprecipitation using human neuroblastoma SH-SY5Y cells [because these cells have relatively high levels of endogenous parkin (10)] stably transfected with a cDNA plasmid expressing full-length PINK1 tagged at the C-terminus with Flag (PINK1-Flag). On incubation of these cell extracts with a rabbit polyclonal anti-Parkin antibody (Abcam), endogenous Parkin immunoprecipitated, and PINK1-Flag did, also (Fig. S3B). Because of the lack of anti-PINK1 antibodies that reliably immunoprecipitate endogenous PINK1, it cannot be determined at present whether the immunoprecipitation of endogenous PINK1 can pull down Parkin.

These findings raise the possibility that Parkin may be a substrate for PINK1 or that PINK1 may be a client for Parkin. However, in our hands, we found no evidence of Parkin phosphorylation by PINK1 on [γ-³²P]ATP autoradiography or by use of phosphoserine- and phosphothreonine-specific antibodies (Fig. S4 A and B). We also found no electrophoretic indication that the phosphorylation status of Parkin extracted from mouse brain tissues was altered by the lack of Pink1 (Fig. S4C). To test the effect of Parkin on PINK1, Myc-PINK1, FLAG-Parkin, and HA-ubiquitin were coexpressed in SH-SY5Y cells (Fig. S5). This experiment showed that PINK1 was not covalently modified by HA-ubiquitin and that Parkin expression did not decrease the basal levels of PINK1. When these coexpressing cells were treated for 24 h with 5 μM of the proteasomal inhibitor MG132, there was no evidence that Parkin promoted the accumulation of PINK1-ubiquitin conjugates (Fig. S5B). Thus, we found that Parkin failed to ubiquitinate PINK1, to decrease its steady-state level, or to promote its proteasomal degradation. Collectively, these findings suggest that Parkin and PINK1 may collaborate on some aspects of mitochondrial dynamics but probably not via posttranslational modification of each other.

Parkin/PINK1 Promotes Mitochondrial Clustering

Because we found that increased PINK1^WT expression suffices to recruit Parkin to the mitochondria, we assessed the effects of PINK1 and Parkin on mitochondrial distribution by overexpressing either or both proteins in SH-SY5Y cells to take advantage of their neuronal-like nature and their highly interconnected tubular mitochondrial network. Once transfected, these cells were immunolabeled to detect EndoG and TOM20 as validated previously (10), and with MitoTracker 633. As expected, these different mitochondrial markers colocalized (Fig. S6). In untransfected cells and cells transfected with empty vectors, mitochondria appeared to be primarily tubular and organized in an interconnected network throughout the cell body, as expected (Fig. S6). Neither PINK1^WT nor Parkin^WT overexpression alone caused overt alteration of the mitochondrial network (Fig. 5A). In contrast, when SH-SY5Y cells were cotransfected with PINK1^WT and Parkin^WT, the normal mitochondrial network became altered. By 24–48 h after transfection, ∼90% of the cells (n = 250) exhibited Parkin-positive fragmented mitochondria, primarily in the vicinity of the nucleus, and/or large, perinuclear clusters of MitoTracker-positive mitochondria (Fig. 5A and Fig. S6). Even at 48 h after transfection, ∼10% of the cotransfected cells still had a normal tubular mitochondrial network (Fig. S6). Of note, in our pilot studies, we found that these changes in the mitochondrial network were similar to those observed in Parkin-YFP–transfected cells exposed to 10 μM CCCP for ∼2 h. By co-overexpressing Parkin^WT and PD-linked mutated PINK1 (A217D, G309D, L347P) or kinase dead mutant PINK1^K219M —all of which have markedly reduced kinase activities (17)—these mitochondrial changes were attenuated (Fig. 5A and Fig. S7). A similar observation was made with co-overexpression of functionally defective Parkin (produced by deletion of the RING2 domain) and PINK1^WT (Fig. S7). As confirmed by Western blots, in all the different combinations of coexpression, levels of mutated Parkin or PINK1 were at least comparable to those of their WT counterparts (Fig. 5B). These findings support the notion that the disruption of the mitochondrial network and the formation of mitochondrial aggregates and perinuclear clusters depend on both Parkin and PINK1 activity. We also found that the formation of perinuclear mitochondrial aggregates and clusters appeared to be specifically caused by Parkin and PINK1, because co-overexpression of Parkin/DJ-1 and PINK1/DJ-1 at comparable levels did not cause these mitochondrial structures (Fig. S7). Incidentally, we saw identical mitochondrial perinuclear phenotypes with PINK1/Parkin co-overexpression in other cell lines, such as human neuroblastoma M17 and HEK 293T cells.

The Mitochondrial Clustering Is Microtubule-Dependent

The results reported in the previous sections support the view that Parkin and PINK1 may act in concert to modulate mitochondrial location, a complex function that typically relies on the microtubule motors (18). Consistent with this notion, in PINK1-stable SH-SY5Y cells transfected with Parkin, we found that >90% of the perinuclear mitochondrial clusters dispersed after only 1-h incubation with 1 μM nocodazole, a microtubule depolymerizing agent (Fig. S8A). Furthermore, in cells with perinuclear mitochondrial clusters, there was >80% colocalization between at least a part of these large perinuclear clusters and γ-tubulin, suggesting that they gather in the vicinity of the centrosome, an organelle that serves as the main microtubule-organizing center (Fig. S8B). However, we saw no obvious effect of nocodazole on the smaller perinuclear mitochondrial aggregates nor a definite colocalization between γ-tubulin and these mitochondrial aggregates. Further studies may be needed to elucidate whether these mitochondrial aggregates represent a distinct arrangement or a preceding stage [i.e., thanks to the microtubule motor, mitochondrial aggregates coalesce into larger perinuclear structures preferentially localized in the vicinity of the microtubule-organizing center, as supported by our time-lapse live imaging analyses (Video S1)]. Nonetheless, microtubules serve as railways for the transport of organelle cargos other than mitochondria (18). Remarkably, PINK1/Parkin co-overexpression seemed to modulate mitochondrial location specifically, because other microtubule organelle cargos, such as endoplasmic reticulum, never showed any change in their cellular organization (Fig. S9A). Thus, the interaction between PINK1 and Parkin may operate collaboratively on specific molecules of the mitochondrial trafficking machinery. In keeping with this idea is the demonstration that PINK1 can interact with the mitochondrial protein Miro and the adaptor protein Milton (19), which connects kinesin heavy chain to Miro on mitochondria. However, our view of Parkin/PINK1 collaboration in mitochondrial trafficking in mammalian cells does not agree with the idea that, in Drosophila, Parkin operates downstream of Pink1 and that Parkin overexpression makes Pink1 dispensable (4, 5, 7). At this point, we cannot exclude the possibility that this apparent molecular divergence may result from an incomplete conservation of the Parkin/PINK1 pathway between invertebrate and vertebrate organisms. It also should be taken into account that, here, we investigated the role of Parkin/PINK1 interaction on mitochondrial distribution and disposition, whereas in all the Drosophila studies the authors ascertained mitochondrial morphology and fission/fusion, very different aspects of mitochondrial dynamics that are not necessarily governed by an identical molecular machinery.

Parkin/PINK1 May Regulate Mitochondrial Trafficking

To examine the ultrastructure of perinuclear clustered mitochondria induced by Parkin/PINK1, we performed EM and observed a range of different types of mitochondrial perinuclear clusters that were not present in empty vector-transfected cells. In all cases, both the length and the width of perinuclear-clustered mitochondria in Parkin/PINK1-cotransfected cells were smaller than in mock-transfected control cells (Fig. S9B), suggesting that Parkin/PINK1 coexpression distorts the mitochondrial network, perhaps by promoting mitochondrial fragmentation. Furthermore, in some cases, clusters were made of nearly normal-appearing mitochondria, and both PINK1 and Parkin localized to the outside boundaries of each individual mitochondrion (Fig. 6 A and B). In other cases, multiple mitochondria were fused together (Fig. 6 A and C). Among clustered mitochondria, the gap between 2 mitochondria was ∼6 nm, similar to the gap of the mitochondria clusters induced by mitochondrial phospholipase-D (20). However, unlike mitochondrial phospholipase-D, Parkin/PINK1 overexpression was associated with mitochondrial outer-membrane fusion (Fig. 6A). We also identified perinuclear lysosomal vacuoles as well as autophagosomes, and some of these contained mitochondria (Fig. 6 D and F), suggesting a mitochondrial autophagic event. The autophagic nature of these vacuoles was confirmed by fluorescence for the autophagosome marker LC3-rFP and by immunofluorescence for the lysosome marker Lamp2 (Fig. 6 G and H). Notably, in untreated Parkin-GFP/LC3-rFP cotransfected cells, the LC3-rFP signal was detected throughout the cytoplasm (Fig. 6G). In contrast, in CCCP-treated cells, the LC3-rFP signal was localized mostly in the perinuclear region, where it colocalized with Parkin and the mitochondrial marker cytochrome c (Fig. 6G). Consistent with the preferential subcellular localization of lysosomes, Lamp2 immunofluorescence was detected primarily in the perinuclear area, which is the only subcellular region where we observed definite colocalization between Lamp2 and Parkin (Fig. 6H). Together with our results for the microtubule experiments, these data suggest that autophagosomes containing Parkin/PINK1-enriched mitochondria may form at some distance from the lysosomes and then are delivered by the microtubule motor to the perinuclear lysosomes for degradation. This scenario is reminiscent of that proposed for the clearance of aggresomes (21), in which proteinaceous inclusion bodies are thought to be targeted to the perinuclear area to be disposed of by autophagy. Although our study is in agreement with that of Narendra et al. (9), in that we also found that cytosplasmic Parkin can translocate to the mitochondria, we argue that the ensuing autophagy of mitochondria requires the trafficking of damaged mitochondria to the perinuclear area to be degraded. We thus propose that both PINK1 and Parkin are key elements of the trafficking machinery responsible for delivering defective mitochondria to the lysosome-rich perinuclear area, rather than being part of the actual autophagy systems. The interplay between PINK1 and Parkin in mitochondrial functioning also may modulate the trafficking of mitochondria in dendrites, perhaps accounting for the synaptic dysfunction that is observed in PINK1- or Parkin-knockout mice (22, 23).

Materials and Methods

All methods employed in this article are routinely used in our laboratories and are thus referenced (10, 16, 24, 25) and are described in SI Materials and Methods. For immunoblotting, the primary antibodies used were PINK1 (100-494; Novus), Parkin, GAPDH, Hsp60, and HA (Santa Cruz Biotechnology), TIM23, cytochrome c, and COX-I (Invitrogen). For immunostaining, primary antibodies were PINK1 (Novus), myc (9E10; Abcam), EndoG (ProSci), TOM20 (BD Biosciences), α- and γ-tubulin (Sigma-Aldrich), calreticulin (AbCam), and tyrosine hydroxylase (Chemicon-Millipore).

Supplementary Material