To assess Parkin-mediated ubiquitination of proteins associated with mitochondria by immunostaining we used an antibody (FK1) that specifically recognizes polyubiquitinated proteins but not free ubiquitin or monoubiquitinated proteins.¹⁶ In HeLa cells expressing EYFP-Parkin treated with vehicle (DMSO) very little ubiquitinated protein can be detected localized with mitochondria (Fig. S1). Following depolarization with CCCP for 1 hr, however, a clear polyubiquitination signal is seen in cells expressing EYFP-Parkin but not in adjacent untransfected cells, which lack endogenous Parkin (Fig. 2A and B). We assessed the kinetics of mitochondrial ubiquitination relative to Parkin recruitment by live cell imaging. In single cells, mCherry-Parkin and EGFP-Ubiquitin consistently accumulated on mitochondria within the same 10 minute window following mitochondrial depolarization (Fig. S2A and B).

Although the Parkinson disease-linked mutant of Parkin, R275W, does not exhibit decreased ubiquitination activity in vitro,¹⁷ it is defective in mediating mitophagy of uncoupled mitochondria.^7,9 Therefore, we tested whether the clumping and mitophagy deficiencies might result from the inability of the R275W mutant to polyubiquitinate mitochondrial substrate(s) in vivo. Consistent with this hypothesis, we found that in cells with similar levels of wild-type and R275W Parkin on depolarized mitochondria, the R275W mutant exhibits a relative deficiency in mitochondrial polyubiquitination (Fig. 2A and B). This association also supports the hypothesis that the mitochondrial clumping and mitophagy induced by Parkin are a consequence of polyubiquitination of mitochondrial substrates.

The effect of polyubiquitination on a protein depends, in part, on how individual ubiquitin monomers are linked in the polyubiquitin chain. While polyubiquitinated proteins with linkages other than K63 (the most common of which is K48) are usually targeted for degradation by the proteasome, K63-linked chains are thought to serve a signaling function and have been implicated in DNA repair, activation of NF-KappaB signaling, sorting of endocytosed receptors to lysosomes, as well as the aggregation and autophagic degradation of misfolded proteins.^18,19 As previous reports suggest that Parkin is able to induce K48 and K63 polyubiquitination,²⁰ we used recently developed antibodies, Apu2 and Apu3, which are specific for K48 and K63 linkages, respectively,²¹ to assess which linkages are predominant following recruitment of Parkin to depolarized mitochondria. We found little K48 polyubiquitination signal could be detected co-localizing with depolarizing mitochondria in HeLa cells expressing ECFP-Parkin (Fig. 2C and D), even though a known K48 polyubiquitinated substrate, SOD1 G93A,²² was readily detected with the same immunostaining procedure (Fig. S3). In contrast, a clear K63-linked polyubiquitin signal colocalized with depolarized mitochondria in the presence of ECFP-Parkin but not in untransfected HeLa cells or Parkin expressing HeLa cells not treated with CCCP (Fig. 2E and F, data not shown). Significantly less K63-linked polyubiquitination was seen co-localizing with depolarized mitochondria in HeLa cells expressing Parkin R275W, consistent with the hypothesis that the R275W mutant is deficient in the polyubiquitination of mitochondrial substrate(s) in vivo (Fig. 2E and F). The relative signals observed for K48- and K63-linked polyubiquitination are unlikely to reflect their relative affinities for native polyubiquitin chains, as the antibody against K48, Apu2, has greater affinity for K48 diubiquitin than the antibody against K63, Apu3, has for K63 diubiquitin in vitro.²¹ Parkin has been shown to ubiquitinate an N-terminal tag following treatment with CCCP.^10,23 To assess if ubiquitination of Parkin's N-terminal EYFP tag can account for the K63 immunoreactivity following Parkin recruitment, we tested the ability to untagged Parkin, which is not robustly ubiquitinated under these conditions,¹⁰ to induce K63-linked mitochondrial polyubiquitination by immunocytochemistry. We found untagged Parkin induces K63-linked polyubiquitination to a similar extent as EYFP-Parkin, suggesting that a substrate other than Parkin is K63 polyubiquitinated following the recruitment of Parkin to depolarized mitochondria (Fig. S4A). Together these results suggest that Parkin induces K63-linked polyubiquitination of mitochondrial protein(s) following its recruitment to depolarized mitochondria and supports the hypothesis that at least some mitochondrial substrates polyubiquitinated by Parkin are regulated in a proteasome-independent manner. However, we cannot rule out that Parkin may also induce mono-ubiquitination or polyubiquitination of mitochondrial protein(s) with other linkages such as K27.

p62, which can directly bind polyubiquitin chains and LC3-II, a protein integral to autophagosomes, has been implicated as an adaptor for the aggregation and autophagy of misfolded ubiquitinated proteins.^12,24,25 More recently, it has also been implicated in the delivery of ubiquitinated intracellular bacteria,¹⁴ experimentally ubiquitinated peroxisomes,¹³ and Parkin-labeled mitochondria to lysosomes by autophagy.⁷

Given its role as a ubiquitinbinding adaptor protein, we assessed whether p62 is recruited to mitochondria following Parkin translocation. While endogenous p62 forms aggregates in HeLa cells, consistent with previous reports, p62 did not co-localize with mitochondria in untransfected HeLa cells, which have little or no endogenous Parkin, under depolarizing or basal conditions (Fig. 3A and B). However, in HeLa cells expressing EYFP-Parkin and treated with CCCP for 1 hr (but not the vehicle DMSO), endogenous p62 is readily recruited to mitochondria (Fig. 3A, B and S4B), as recently reported.⁷ In addition, live-cell imaging experiments demonstrate that, similar to EGFP-ubiquitin, translocation of EGFP-p62 to mitochondria occurs within the same 10-minute window as mCherry-Parkin translocation following CCCP treatment (Fig. S5). In contrast, Parkin R275W, which is deficient in the promotion of in vivo polyubiquitination, did not induce p62 recruitment to depolarized mitochondria. This is consistent with the hypothesis that ubiquitination of mitochondrial substrate(s) is required for p62 recruitment (Fig. 3A and B).

Parkin contains an N-terminal ubiquitin-like (UBL) domain. Thus, it is possible that Parkin recruits p62 through a direct interaction between the UBL domain of Parkin and the ubiquitin-associated (UBA) domain of p62. To determine whether Parkin's UBL is necessary for p62 recruitment, we assessed p62 recruitment to mitochondria in HeLa cells expressing EYFP-Parkin ΔUBL. Although Parkin ΔUBL is not recruited to mitochondria as robustly as wild-type Parkin, p62 co-localized with mitochondrial Parkin ΔUBL, in cells in which Parkin ΔUBL translocation was observed (Fig. S6). This is consistent with the hypothesis that p62 is recruited by substrate(s) on mitochondria that becomes ubiquitinated following recruitment of Parkin rather than by Parkin directly. That Parkin R275W fails to induce p62 localization to mitochondria, despite having an intact UBL domain, also suggests that p62 is recruited following ubiquitination of mitochondrial substrates and not by Parkin's UBL domain.

While endogenous p62 was recruited to some depolarized mitochondria in all HeLa cells expressing Parkin, p62 appeared to be most abundant on clumps of mitochondria coated with Parkin (Fig. 3C). To further investigate the localization and effect of p62 on mitochondria, we observed EGFP-p62 in single cells by live-cell imaging. Consistent with previous reports, we found the threadlike mitochondria begin to fragment 10–15 minutes following depolarization with CCCP, likely due to cleavage and inactivation of the fusion protein OPA1 with continued activity of the fission protein Drp1.²⁶ In the periphery of the cell, around 25–30 minutes following the addition of CCCP, EGFP-p62 appears to coat the fragmented mitochondria (Fig. 4A, top right parts). When the individual, p62-coated mitochondria come into proximity they appear to stick together, forming aggregates of round, fragmented mitochondria, resembling bunches of grapes (Fig. 4A, top right parts). Likewise, near the nucleus p62 appears to connect mitochondria as they are compressed together into a perinuclear aggregate (Fig. 4A and bottom). Occasionally, groups of mitochondria within these aggregates appear stretched apart as one group is pulled toward the nucleus—presumably by microtubule motors, as such events were not observed in the presence of nocodazole (data not shown). Thin filaments of p62 appear stretched between the aggregates under tension, suggesting a force resisting the separation (Suppl. Movie 1). This is consistent with the notion that in order to separate fragmented mitochondria that appear tethered together by polymerized p62, binding between polymerized p62 molecules (and possibly other molecules binding p62, such as NBR1) must be overcome.^12,27

To further test the hypothesis that p62 may be necessary for the observed aggregation and autophagy of mitochondria following recruitment of Parkin to mitochondria, we assessed changes in mitochondrial aggregation and number in p62^+/+ and p62^−/− MEFs by confocal microscopy. In p62^+/+ MEFs expressing Parkin treated with CCCP for 24 hrs, mitochondria often appeared clumped in aggregates (Fig. 4B and C), similar to depolarized mitochondria in HeLa cells expressing Parkin (Fig. 1A). In contrast, mitochondrial aggregates were rarely observed in p62^−/− MEFs (Fig. 4B and C). To quantitatively assess of the relative aggregation of mitochondria in p62^+/+ and p62^−/− MEFs, we measured the compaction index of mitochondria within each cell type. Consistent with our other observations, mitochondria within p62^+/+ MEFs had a higher compaction index than mitochondria within p62^−/− MEFs (Fig. 4D). Together these findings demonstrate that endogenous p62 can mediate the aggregation of depolarized mitochondria expressing the E3 ubiquitin ligase, Parkin.

Given the role of p62 in mediating the selective autophagy of misfolded proteins,¹² peroxisomes,¹³ and intracellular bacteria,¹⁴ we also assessed whether p62 is necessary for Parkin-induced mitophagy. Surprisingly, p62 appears to be dispensable for Parkin-induced mitophagy, with approximately equal numbers of p62^+/+ MEFs and p62^−/− MEFs lacking mitochondria following treatment with CCCP for 24 hrs (Fig. 4C). To assess whether p62 may be required for Parkin-induced mitophagy in other cell types or following acute depletion, we transfected HeLa cells with siRNA directed against p62. A substantial reduction in p62 protein levels was observed following transfection of HeLa cells with p62 siRNA but not control siRNA (Fig. 5A). Consistent with p62 being dispensable for Parkin-induced mitophagy, HeLa cells acutely depleted of p62 and transiently expressing EYFP-Parkin eliminated depolarized mitochondria to a similar extent as HeLa cells transfected with control siRNA (Fig. 5B and C). In particular, individual p62 siRNA transfected cells that lacked both the abundant p62 puncta apparent in control cells (indicating successful p62 knockdown at the level of individual cells) and Tom20 immunostaining (indicating robust mitophagy) were observed, demonstrating at the single cell level that acute loss of p62 in HeLa cells does not prevent Parkin-induced mitophagy. Taken together these findings demonstrate that p62 is not necessary for Parkin-induced mitophagy.

p62 is believed to mediate the aggregation of misfolded proteins by binding the ubiquitinated protein with its UBA domain and then binding to other p62 molecules through its N-terminal PB1 domain (Fig. 6A).^12,24,25 To assess whether a similar mechanism is responsible for Parkin-induced aggregation of depolarized mitochondria, we tested the ability of wild-type and mutants of p62 to rescue the mitochondrial aggregation defect in p62^−/− MEFs. We found that whereas wild-type p62 and p62 lacking the LIR domain, responsible for the binding of p62 to LC3, were able to rescue mitochondrial aggregation in p62^−/− MEFs, p62 with a mutation in the PB1 domain, D69A, failed to restore mitochondrial aggregation (Fig. 6A and B). These findings strongly suggest that p62 mediates mitochondrial aggregation by polymerizing through its PB1 domain.

Recent findings suggest that experimental fusion of the ubiquitin mutant G76V to peroxisomal membrane proteins is minimally sufficient to induce their selective autophagy, provided ubiquitin faces the cytosol.¹³ As Parkin induces both the polyubiquitination of substrate(s) on mitochondria and mitochondrial degradation, we created an inducible heterodimerization system to recruit ubiquitin directly to the mitochondrial outer membrane.^9,28 This system allows us to assess whether mitochondrial anchored ubiquitin is minimally sufficient for mitophagy, p62 recruitment and/or mitochondrial clumping. The C-terminal mitochondrial anchoring tails of Bcl-X_L (213-233), Bax (EYFP20),²⁹ and Fis1 (92–152) were fused to the FRB domain, while UbG76V-GFP or GFP were fused to tandem FKBP domains. While treatment with the rapamycin analog AP21967 (rapalog) caused at least partial co-localization of UbG76V-EGFP-FKBP with mitochondria following co-transfection with each of the FKBP fused outer mitochondrial membrane anchors, increased expression of ubiquitin on mitochondria did not cause a detectable reduction in mitochondrial mass by confocal microscopy after 24 hrs (Fig. 7A). We investigated the UbG76V-EGFP-FKBP paired with the FRB-Fis1 construct further, as this combination resulted in the most stably expressed ubiquitin on the outer mitochondrial membrane. We found no difference in mitochondria mass in HeLa cells expressing UbG76V-EGFP-FKBP and FRB-Fis1 at 24 hrs or 96 hrs of treatment with the rapalog relative to that of HeLa cells expressing EGFP-FKBP and FRB-Fis1 (Fig. 7A and data not shown). Mitochondrial fission, which occurs following mitochondrial depolarization, is thought to precede and possibly promote mitophagy, during basal mitochondrial turnover.³⁰ To test whether mitochondrial fission or another consequence of depolarization is required in conjunction with stable expression of ubiquitin on the outer mitochondrial membrane for mitophagy, we depolarized HeLa cells (which lack endogenous Parkin) expressing UbG76V-EGFP-FKBP/FRB-Fis1 with CCCP for 24 or 96 hrs in conjunction with rapalog treatment (Fig. 7B and C and data not shown). No mitophagy was apparent in the UbG76V-EGFP-FKBP/FRB-Fis1 expressing cells comparable to untransfected cells or cells expressing EGFP-FKBP/FRB-Fis1 (data not shown).

Interestingly, UbG76V-EGFP-FKBP but not EGFP-FKBP was sufficient to recruit p62 to clumped mitochondria in a substantial proportion of cells co-expressing FRB-Fis1 and treated with rapalog in the presence of CCCP (Fig. 7B and C). These findings suggest that anchoring mono-ubiquitin on the outer mitochondrial membrane is sufficient for p62 recruitment to mitochondria and further suggest that p62 recruited to ubiquitinated mitochondria leads to mitochondrial clumping but not mitophagy. These data are also consistent with the hypothesis that ubiquitination of particular mitochondrial substrate(s) and/or modification of the substrate(s) with a particular polyubiquitin chain (e.g., K63-linked) is required for Parkin-induced mitophagy (but not for p62 recruitment and mitochondrial clumping).

VDAC1 was recently reported to be ubiquitinated by Parkin following Parkin translocation to depolarized mitochondria, suggesting that VDAC1 ubiquitination may mediate mitochondrial phenotypes following Parkin recruitment.⁷ To verify that VDAC1 is ubiquitinated under depolarizing conditions, we treated HeLa cells stably expressing EYFP-Parkin (HeLa^EYFP-Parkin) with CCCP for 0, 2 or 5 hrs. Using anti-porin 31HL (VDAC1) mAB6, higher molecular weight immunoreactive bands (near ~39 and 48 kDa) were observed following CCCP treatment in addition to a band near VDAC1's predicted molecular weight, 31 kDa (Fig. 8A). These findings are consistent with the appearance of mono- and di-ubiquitinated forms of VDAC1, respectively, following CCCP treatment in the presence of Parkin and confirm previous data obtained with this antibody. The immunoreactivity of the 39 and 48 kDa bands compared to the 31 kDa band suggests that a relatively small proportion of VDAC1 is modified.

Unexpectedly, while the 39 and 48 kDa bands were observed with mAb6, they were not observed with another commonly used VDAC1 antibody, mAB1, which was generated by the same antigen exposure as Ab6,³¹ and which exhibits fewer cross-reactive bands compared to mAb6 (Fig. S7). To assess whether the higher bands seen with mAb6 represent ubiquitinated VDAC1 or a related cross-reactive protein, such as VDAC2 or VDAC3, endogenous VDAC1 was immunoaffinity purified from HeLa^EYFP-Parkin cells treated with CCCP for 5 hrs with the mAb6 antibody. Four strong Coomassie blue labeled bands were observed on a SDS-PAGE gel, roughly corresponding to the molecular weight of unmodified VDAC1 (bands labeled 1 and 2) and mono- and diubiquitinated VDAC1 (bands labeled 4 and 5, respectively) (Fig. 8B). These bands were excised along with a control gel piece (band 3) for identification by LC-MS/MS. Tryptic peptides from VDAC1 were observed in all bands, but a greater number of peptides were identified by MS/MS from the strong Coomassie stained bands 1, 2, 4 and 5, consistent with these representing distinct VDAC1 species (Fig. 8C and Suppl. Table 1). The VDAC1 peptides observed from the control band likely represents a background of protein spread from the neighboring concentrated bands. Evidence for a background protein spread of the strong light chain immunoglobulin band around 25 kDa into bands 1–5 was also observed by MS/MS analysis (data not shown).

Three tryptic peptides unique to the VDAC2 isoform were identified from band 2 but not from the other bands (Fig. 8B and Suppl. Table 1). This demonstrates that although mAb6 is fairly specific for VDAC1, as reported previously,³² it also binds VDAC2. To assess whether VDAC2 is present in higher molecular weight forms, which may be ubiquitinated, we estimated the relative abundance of the three unique VDAC2 peptides and three homologous VDAC1 peptides in the five bands by examining the MS ion chromatograms for each peptide in each sample (Fig. 8D). While the three VDAC1 peptides were substantially more abundant in high molecular weight bands 4 and 5 than in control band 3, the three VDAC2 peptides were no more abundant in bands 4 and 5 than in control band 3. These findings suggest that VDAC2 is present primarily in band 2 and is not ubiquitinated following uncoupling in cells expressing Parkin. To provide independent evidence that VDAC2 is not ubiquitinated, we immunoblotted HeLa lysates treated with CCCP for 0, 2 or 5 hrs with a polyclonal antibody raised against a VDAC2 peptide. The membrane was then stripped of antibody and reprobed for VDAC1 using the anti-VDAC1 mAb6 antibody. While high molecular weight bands immunoreactive to anti-VDAC1 mAb6 appeared with 2 and 5 hrs CCCP treatment (but not 0 hrs of treatment), as was observed above, no corresponding bands reacted with the anti-VDAC2 antibody, indicating no change in VDAC2 mobility following CCCP treatment. Together these findings suggest that VDAC2 is not ubiquitinated following depolarization in cells overexpressing Parkin.

Although we did not identify site(s) of VDAC1 ubiquitination, two unique ubiquitin peptides, TITLEVEPSDTIENVK and TLSDYNKQK (hereafter, TITLE and TLS), were identified by MS/MS from bands 4 and 5 in this initial experiment (Fig. S8B and Suppl. Table 1). As identification of TLS, which fragmented poorly, did not have a p-value of lower than 0.05 in the first experiment in a MASCOT search, the peptide samples were concentrated and rerun to provide additional evidence of this peptide's identity. In the second technical replicate, the identification for the TLS peptide reached significance for band 4 and a third ubiquitin peptide, ESTLHLVLR (hereafter EST), was identified with significance from bands 4 and 5 (Suppl. Table 1). Prominent precursor to fragment ion transitions observed upon collision induced disassociation for these three peptides (894.47 → 1002.5, 541.28 → 867.4 and 534.31 → 637.4 for TITLE, TLS and EST, respectively) have been reported previously,³³ further supporting the correct identification of these three ubiquitin peptides in digests of bands 4 and 5 (Fig. S8A–E). No ubiquitin peptides were found from bands 1 or 2, while one, TLS, was found in control band 3 (most likely reflecting background of protein spread from adjacent band 4). The relative abundance of the ubiquitin peptides in each band was calculated from the MS ion chromatograms from the first technical replicate. Bands 4 and 5 exhibited substantially more abundant ubiquitin peptides than background (control band 3), while the abundance of ubiquitin peptides was similar in bands 1 and 2 and band 3 (Fig. 8E). Consistent with the mass shift observed on SDS-PAGE, these results suggest that the higher molecular weight bands (4 and 5) represent ubiquitinated forms of VDAC1, while the lower bands represent VDAC1 that is not ubiquitinated.

In our mass spectrometry analysis of endogenous VDAC1 no ubiquitin peptides were identified with a diglycine modification that would have suggested the topology of the polyubiquitin chain attached to VDAC1, if VDAC1 is polyubiquitinated. To assess whether ubiquitination of VDAC1 (or VDAC3) could account for the K63 polyubiquitin immunoreactivity associated with depolarized mitochondria following Parkin recruitment, we examined K63 polyubiquitin immunoreactivity in VDAC1/3^−/− MEFs (Fig. 9A). Following depolarization of mitochondria with CCCP for 3 hrs, we observed robust K63-linked polyubiquitination of mitochondria in cells expressing EYFP-Parkin but not in untransfected cells, demonstrating that a Parkin substrate other than VDAC1 is K63 polyubiquitinated in a Parkin-dependent manner. We next assessed whether VDAC1 or VDAC3 is necessary for the phenotypes observed following Parkin-dependent ubiquitination of mitochondria. Following treatment with CCCP for 24 hrs, a similar proportion of VDAC1/3^+/+ and VDAC1/3^−/− MEFs expressing EYFP-Parkin lacked mitochondria, demonstrating that VDAC1 and VDAC3 are dispensable for Parkin-induced mitophagy (Fig. 9B and C). In VDAC1/3^−/− MEFs that retained mitochondria, the mitochondria were often in perinuclear clusters and exhibited p62 immunoreactivity, demonstrating that VDAC1 and VDAC3 are also dispensable for Parkin-induced p62 recruitment and mitochondrial clustering (Fig. 9D).

The sequestration of ubiquitinated mitochondria by p62 appears to be analogous to its clustering of ubiquitinated proteins. By binding ubiquitin with its UBA domain and polymerization through its PB1 domain, p62 mediates the aggregation of misfolded proteins produced by a number of experimental challenges, including puromycin treatment, overexpression of insoluble mutant proteins and inhibition of autophagy, as has been demonstrated in cell culture as well as in Drosophila and mice.^12,37,38 In addition, p62 is found with ubiquitin in protein aggregates associated with disease, such as Mallory bodies in alcoholic cirrhosis and, notably, Lewy bodies in Parkinson disease.^27,39,40 By minimizing interaction between exposed hydrophobic regions of misfolded proteins and other cellular components, aggregation of misfolded protein may be protective in the context of neurodegenerative disorders—e.g., mutant Huntingtin and alpha-synuclein may be less harmful in larger aggregates compared with small oligomers.^41,42 However, in other contexts, such as in autophagy-deficient hepaptocytes, p62-mediated aggregation of misfolded proteins may be detrimental.³⁷

The pathophysiological significance of mitochondrial aggregation is less clear, but interestingly in MitoPark mice, a model of progressive parkinsonism, damaged mitochondria are sequestered in intracellular inclusions that bear some resemblance to Lewy bodies (although they lack alpha-synuclein, a key component of Lewy bodies in sporadic Parkinson disease).⁵ Parkinsonism in these mice results from conditional knockout of the mitochondrial DNA transcription factor Tfam from tyrosine hydroxylase expressing dopaminergic neurons, experimentally mimicking the electron transport chain dysfunction observed in sporadic Parkinson disease. Reducing the surface of area of impaired mitochondria within the cell by mitochondrial clustering may minimize their uptake of substrates and limit the spread of mitochondrial ROS to other cellular compartments. Additionally, clustering dysfunctional mitochondria would likely prevent their transport to parts of the cell with high-energy requirements, such as neuronal synapses, at the expense of more bioenergetically active mitochondria. Alternatively, as p62 clustering of ubiquitinated substrates has been shown to cause cell death in some cell types in the absence of its degradation by autophagy,³⁷ it is possible that p62 functions as a sensor of defective quality control mitophagy. In this model, p62 would trigger cell death in the face of a deficient removal of mitochondria that have been ubiquitinated by Parkin, in a manner analogous to the p62-dependent cell death triggered by Mallory body-like inclusions in the hepatocytes of Atg7-deficient mice. It will be interesting to learn whether crossing MitoPark mice with p62 null mice and Parkin null mice prevents mitochondrial aggregation and exacerbates or alleviates the MitoPark phenotype.

p62 directly binds the autophagy proteins, LC3 and GABARAP, and is believed to serve as an autophagy receptor for ubiquitinated protein aggregates as well as peroxisomes and intracellular bacteria.^12–14 Thus, we were surprised to find mitophagy occurs equally well in cells null for p62 and wild-type cells, suggesting that p62 is not rate limiting for Parkin-induced mitophagy. We also assessed the dependence of Parkin-induced mitophagy on p62 in HeLa cells acutely depleted of p62 by siRNA. Similar to findings in p62 null MEFs, p62 depletion in HeLa cells did not inhibit p62-induced mitophagy. Notably, we observed p62 siRNA transfected HeLa cells that were absent immunostaining for both p62 and Tom20, demonstrating at the single cell level that mitophagy can still proceed in the face of acute p62 depletion. These findings suggest that the p62-independence of Parkin-induced mitophagy is likely not the result of a compensatory mechanism and may be characteristic of many mammalian cell types. These data are in contrast to findings in a recent report published while this manuscript was in preparation suggesting that knockdown of p62 in HeLa cells partially inhibits Parkin-induced mitophagy.⁷ It is not clear how to account for the discrepancy in these findings, as we appear to have achieved a similar level of p62 protein depletion, using the same p62 siRNAs, in the same cell line, and following a very similar protocol. Nonetheless, Parkin clearly induces mitophagy in single cells with no p62 protein expression, demonstrating that p62 is inessential for Parkin-induced mitophagy at least in some cell types. While this manuscript was in revision, Okatsu et al. reported that whereas p62 is required for mitochondrial clustering it is not essential in MEFs for mitophagy largely consistent with aspects of our study.⁴³

Consistent with the observation that p62 is not required for Parkin-induced mitophagy, knockout of Atg7 but apparently not p62 in hepatocytes leads to the accumulation of dysmorphic mitochondria, suggesting that p62 may not be required for quality control mitophagy in the liver.^37,44 In addition, while the phenotype of Drosophila lacking the p62 ortholog is similar to that of Parkin null Drosophila in some respects (both have male sterility), it lacks key features such as muscle degeneration.^38,45 This suggests that p62 is unlikely to be the sole downstream mediator of Parkin in Drosophila.

Although at present it is unclear how Parkin induces mitophagy, its ubiquitination activity and ubiqutination of mitochondrial targets appear essential. One possible explanation is the existence of redundant autophagy receptors (e.g., p62 and NBR1) for ubiquitinated mitochondria. Loss of one autophagy receptor, for instance, would not be sufficient to inhibit mitophagy, if a downstream step in the pathway were rate limiting, a possibility that may be addressed by the development of NBR1/p62 double knockout mice. Alternatively, Parkin may induce mitophagy through a different mechanism, such as modification of another autophagy receptor (e.g., Nix).⁴⁶

Parkin also appears to concentrate depolarized mitochondria within a perinuclear region, in a manner that is likely dependent on retrograde transport of mitochondria along microtubules. As retrograde transport of aggregated misfolded protein appears to enhance delivery to autophagosomes and lysosomes concentrated in this region, particularly in neurons,^34–36 it was recently proposed that Parkin's induction of mitophagy might be secondary to its effects on mitochondrial transport.⁸ This appears not to be the case. In HeLa cells, Parkin is able to eliminate depolarized mitochondria equally well in the presence and absence of a microtubule network. Additionally, Parkin appears to promote the encapsulation of mitochondria by autophagosomes in the periphery of the cell one hour following depolarization, before mitochondria have been concentrated in the perinuclear region.¹¹ It is, however, possible that retrograde mitochondrial transport induced by Parkin may enhance mitochondrial degradation in some cell types, such as neurons. Additionally, the movement of dysfunctional mitochondria, aggregated by p62, into a perinuclear aggresome may help isolate them from other cellular components.

Interestingly, Parkin was recently found to promote retrograde transport of mitochondria in Drosophila follicular tissue, in which loss of Parkin-mediated retrograde movement leads to aggregation of mitochondria. This suggests that in some systems Parkin likely modulates mitochondrial transport independently of its effects on mitophagy and p62-dependent mitochondrial clustering.⁴⁷

Using antibodies against endogenous K48- and K63-linked polyubiquitin chains, we found that Parkin induces abundant mitochondrial polyubiquitination with a preponderance of K63 linkages relative to K48 linkages. Recently, it was shown that ubiquitin anchored to peroxisomes is sufficient to induce p62 recruitment and pexaphagy.¹³ Similar to these findings, we observed that anchoring ubiquitin to mitochondria is sufficient to recruit p62 to mitochondria. However, the consequence of p62 recruitment to mitochondria appears to be different than that reported for peroxisomes. While mitochondria tagged with a ubiquitin fusion construct clustered similarly to those ubiquitinated by Parkin, ubiquitin anchored on mitochondria did not appear to induce mitophagy, even when the mitochondria were fragmented by depolarization. These findings suggest that a specific substrate or a polyubiquitin chain with a specific linkage may be required for the induction of mitophagy but not for p62 recruitment and mitochondrial clustering. These findings also support the hypothesis that p62 mediates mitochondrial clustering but not mitophagy. We cannot rule out, however, that mitophagy below our level of detection may have been induced by anchoring ubiquitin to the outer mitochondrial membrane.

Largely in agreement with our finding a predominance of K63-linked polyubiquitin chains following Parkin recruitment to depolarized mitochondria, a recent report, using over-expressed ubiquitin mutants, suggested that K27 and K63 are the predominate linkages in polyubiquitin on mitochondria induced by Parkin.⁷ The report identified VDAC1 as a likely K27 polyubiquitinated substrate of overexpressed Parkin and tagged Parkin as the primary K63 polyubiquitinated substrate. We confirm that VDAC1 but not VDAC2 is ubiquitinated following depolarization in the presence of Parkin and so is likely a substrate of overexpressed Parkin. Interestingly, a commonly used antibody against VDAC1's acetylated N-terminus failed to detect ubiquitinated VDAC1. The reason for the failure of this antibody to recognize ubiquitinated VDAC1 is unclear but may reflect direct or indirect epitope masking by the attachment of ubiquitin.

We also find that ubiquitination of VDAC1 cannot account for the K63-linked immunoreactivity observed following recruitment of Parkin to depolarized mitochondria, as this immunoreactivity was also observed in MEFs lacking VDAC1 and VDAC3. Similarly, ubiquitination of tagged Parkin is unlikely to account for the K63 polyubiquitin immunoreactivity, as untagged Parkin, which is not detectably ubiquitinated under depolarizing conditions, could also induce the K63-linked polyubiquitination of mitochondria. Thus, it appears that Parkin induces the K63-linked polyubiquitination of substrates other than VDAC1. Additionally, we observed that VDAC1 and VDAC3 are not required for Parkin-induced recruitment of p62, mitochondrial clustering or mitophagy, as these events occur in MEFs lacking VDAC1 and VDAC3. These findings are consistent with the hypothesis that ubiquitination of substrates other than VDAC1 by Parkin are required for p62 recruitment, mitochondrial clustering and mitophagy. Whether VDAC1 is ubiquitinated as a bystander, is functionally redundant with other Parkin substrates, or leads to another functional consequence is unclear. The relatively small pool of VDAC1 that appears to be ubiquitinated despite overexpression of Parkin and relatively long treatment with CCCP seems to support the view that VDAC1 may be ubiquitinated as a bystander or may play a minor and redundant role in the robust mitophagy induced by Parkin under these conditions. In any case, that Parkin appears to ubiquitinate VDAC1 but not VDAC2, which shares 74% sequence identity and 91% sequence similarity with VDAC1, suggests that Parkin does not ubiquitinate mitochondrial proteins indiscriminately and that there are likely contextual cues governing Parkin's choice of substrate. The identification of Parkin ubiquitination sites on VDAC1 and other Parkin substrates may better define what these contextual cues are.