Decision letter: Mfn2 ubiquitination by PINK1/parkin gates the p97-dependent release of ER from mitochondria to drive mitophagy

Article Figures and data Abstract Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract Despite their importance as signaling hubs, the function of mitochondria-ER contact sites in mitochondrial quality control pathways remains unexplored. Here we describe a mechanism by which Mfn2, a mitochondria-ER tether, gates the autophagic turnover of mitochondria by PINK1 and parkin. Mitochondria-ER appositions are destroyed during mitophagy, and reducing mitochondria-ER contacts increases the rate of mitochondrial degradation. Mechanistically, parkin/PINK1 catalyze a rapid burst of Mfn2 phosphoubiquitination to trigger p97-dependent disassembly of Mfn2 complexes from the outer mitochondrial membrane, dissociating mitochondria from the ER. We additionally demonstrate that a major portion of the facilitatory effect of p97 on mitophagy is epistatic to Mfn2 and promotes the availability of other parkin substrates such as VDAC1. Finally, we reconstitute the action of these factors on Mfn2 and VDAC1 ubiquitination in a cell-free assay. We show that mitochondria-ER tethering suppresses mitophagy and describe a parkin-/PINK1-dependent mechanism that regulates the destruction of mitochondria-ER contact sites. https://doi.org/10.7554/eLife.32866.001 Introduction Loss of PRKN or PINK1 results in an early-onset form of hereditary Parkinson’s disease (PD), a neurological disorder that is linked to mitochondrial dysfunction (Kitada et al., 1998; Ryan et al., 2015; Valente et al., 2004). Accordingly, parkin and PINK1 promote mitochondrial health through several mitochondrial quality control mechanisms; the turnover of outer mitochondrial membrane (OMM) proteins by the proteasome, the generation of mitochondrial-derived vesicles, and whole-organellar degradation by mitophagy, a form of selective autophagy (Sugiura et al., 2014; Yamano et al., 2016). During mitophagy, PINK1, a mitochondrial kinase, builds up on the surface of damaged mitochondria where it activates parkin directly via phosphorylation and allosterically through the generation of phosphoubiquitin (pUb) (Kane et al., 2014; Kazlauskaite et al., 2014; Kondapalli et al., 2012; Koyano et al., 2014; Shiba-Fukushima et al., 2012). Parkin, an E3 ubiquitin (Ub) ligase, mediates the ubiquitination of resident OMM proteins, recruiting Ub-binding autophagic machinery through a feed-forward mechanism to ultimately degrade the organelle via the lysosome (Heo et al., 2015; Lazarou et al., 2015; Ordureau et al., 2015; Ordureau et al., 2014). Contact sites between mitochondria and the endoplasmic reticulum (ER) act as crucial signaling hubs in the context of non-selective, starvation-induced autophagy, where they serve as the site of autophagosome formation (Hamasaki et al., 2013; Kishi-Itakura et al., 2014). Indeed, autophagosome biogenesis is impaired in cells with defective mitochondria-ER tethering (Hamasaki et al., 2013), as lipid transfer between organelles may be important for their formation (Hailey et al., 2010; Klecker et al., 2014). As steady-state mitophagy in yeast requires mitochondria-ER contacts (Böckler and Westermann, 2014), it has been assumed that parkin-dependent mitophagy follows a similar mechanism (Yoshii and Mizushima, 2015). However, this model directly conflicts with the observation that mitofusin-2 (Mfn2) – a mitochondria-ER tether required for starvation-induced autophagosome formation in mammals (de Brito and Scorrano, 2008; Hamasaki et al., 2013; Naon et al., 2016) – is ubiquitinated by parkin and rapidly turned over by the proteasome (Tanaka et al., 2010). Thus, how mitophagy is regulated by contacts between mitochondria and the ER (if at all), and the location from which the mitophagic membrane originates, remain open questions in the field. Results Parkin and PINK1 destroy mitochondria-ER contact during mitophagy We hypothesized that PINK1 and parkin may regulate contact between both organelles during mitophagy, based on studies demonstrating high levels of parkin ubiquitination activity on Mfn2 in both cells and in organello ubiquitination assays (Tanaka et al., 2010; Tang et al., 2017). To first determine whether parkin destroys the OMM-ER interface of depolarized mitochondria, we analyzed contacts between the two organelles by electron microscopy (EM) (Csordás et al., 2006). We quantified ER tubules within 100 nm of the OMM, as this distance is enough to capture tubules closely associated with the OMM (Figure 1A, left panel and inset). To induce PINK1-/parkin-mediated mitophagy, we treated U2OS cells stably-expressing GFP-parkin (U2OS:GFP-parkin) and control U2OS:GFP cells with CCCP for four hours, and observed by EM a decrease the total length of ER-OMM contact in both cell lines, although this decrease was greater in magnitude in cells expressing GFP-parkin (Figure 1A, quantified in 1B). However, when CCCP-induced, parkin-independent mitochondrial fragmentation was taken into account (Figure 1C), parkin had a specific effect on reducing the percentage of the OMM that remained in contact with the ER in depolarized cells (Figure 1D), as well as the percentage of total mitochondria that were still connected to the ER (Figure 1E). This effect was robust, as repeating our quantification using a variety of interorganellar tethering lengths – ER-OMM distances of 100, 50 and 25 nm (Figure 1—figure supplement 1A and B) – pointed us to the same conclusion; parkin disrupts mitochondria-ER contact upon activation of mitophagy. Indeed, this effect was indiscriminate in that it was not selective for one subset of ER-OMM distances (Figure 1—figure supplement 1C). Moreover, the subsets of remaining contacts observed after the ~75% reduction in CCCP-treated, GFP-parkin-expressing cells (Figure 1D and Figure 1—figure supplement 1C) were biased towards longer interorganellar distances (Figure 1—figure supplement 1D), consistent with parkin driving the OMM and ER apart. Given that the mitochondria observed in our EM analyses were still intact organelles and not yet engulfed by the isolation membrane (IM) of the autophagosome (Figure 1A, right panel), we concluded that parkin ablates contact between mitochondria and the ER as an early step during depolarization-induced mitophagy in cells. Figure 1 with 1 supplement see all Download asset Open asset Ultrastructural analysis of ER-mitochondria contact during mitophagy in U2OS cells and dopaminergic neurons. (A) Representative TEM images of mitochondria (‘M’) in contact with ER (pseudocoloured blue) in untreated and CCCP-treated U2OS:GFP-parkin cells. Scale bars, 500 nm. (B–E) Quantification of TEM from (A) in U2OS:GFP and GFP-parkin WT cells, left untreated (red bars) or treated with 20 μM CCCP for four hours (blue bars). Total apposition length (B), mitochondrial size (C), and the percent of OMM per mitochondrion (D) and mitochondria per field (E) in contact with the ER was quantified. Bars represent mean ± SEM, n = 82 to 152 mitochondria in 15 to 19 fields per condition. n.s., not significant; **, p<0.01; ***, p<0.001; ****, p<0.0001. (F) TEM image of an isolation membrane (‘IM’, broken green line) wrapping a mitochondrion (‘mito’). Blue arrowheads indicate the boundaries of OMM rupture, while red arrowheads indicate ER tubules in contact with the intact portion of the OMM. Scale bar, 500 nm. (G) Immunoblot analysis of whole-cell lysates from U2OS:GFP-parkin WT and C431S cells treated with 20 μM CCCP for four hours with or without 10 μM MG132. In the case of MG132 treatment, cells were first pre-incubated with 10 μM MG132 for 30 min prior to addition of CCCP. (H) Representative TEM images of mitochondria in contact with ER (pseudocoloured blue) in U2OS:GFP-parkin WT cells transfected with the indicated siRNA, and treated with 20 μM CCCP (‘+CCCP’) for four hours, in the presence or absence of 10 μM MG132 as in (G). Scale bar, 500 nm. (I,J) Quantification of TEM from (H) in cells treated with (blue bars) or without (red bars) 20 μM CCCP for four hours. The percent of OMM per mitochondrion (I) and mitochondria per field (J) in contact with the ER were quantified. Bars represent mean ± SEM, n = 101 to 203 mitochondria in 14 to 16 fields per condition. n.s., not significant; *, p<0.05; ***, p<0.001; ****, p<0.0001. (K) Immunoblot analysis of parkin levels in mouse brain cytosol from parkin+/+ and parkin-/- mice, along with whole-cell lysates from iDA neurons derived from iPSCs isolated from control (ctrl) individuals and a PRKN patient ("PRKN(del)"). (L) A representative wide-field image showing that iDA neurons express TH (green) and β-III tubulin (red) (Hoechst, blue). Scale bar, 20 microns. (M) Immunoblot analysis of whole-cell lysates from iDA neurons treated with 20 μM CCCP for one hour. The arrowhead indicates the unmodified Mfn2 band, while the red asterisk indicates ubiquitinated Mfn2. (N) Quantification of the percent of the OMM opposed to the ER in iDA neurons treated with 20 μM CCCP for one hour. Bars represent mean ± SEM, n = 80 to 131 mitochondria per condition. n.s., not significant; ****, p<0.0001. (O) Representative TEM images of mitochondria in contact with ER in iDA neurons. In the top row, total ER is pseudocoloured blue. In the second row, the red line denotes an area within 100 nm of the OMM. In the bottom row, ER tubules within the 100 nm area are pseudocoloured red. Scale bars, 200 nm. https://doi.org/10.7554/eLife.32866.002 Figure 1—source data 1 Numerical source data for Figure 1B–1D, E, I, J and N and Figure 1—figure supplement 1B to D. https://doi.org/10.7554/eLife.32866.004 Download elife-32866-fig1-data1-v2.xlsx We next took a closer look at how this process of contact site removal may occur (for the remainder of our study, we used the <100 nm interorganellar distance to quantify ER-OMM contacts). Parkin has been reported, through its ability to ubiquitinate OMM proteins and target them for proteasomal degradation, to eventually mediate the rupture of the OMM prior to or during engulfment by the autophagosome (Yoshii et al., 2011). Indeed, we observed rare (likely transient) mitochondrial structures where we believed OMM rupture to be occurring at the time of fixation (Figure 1F, the blue arrowheads indicate the limits of OMM rupture, where the organelle is being wrapped by the IM [indicated by the broken green line]). Concordantly, ER contacts with the still-intact OMM were observed (Figure 1F, red arrowheads), leading us to postulate that the removal of OMM-ER contacts may precede OMM rupture. To this end, we quantified ER-OMM contacts in CCCP-treated cells that were co-incubated with the proteasome inhibitor MG132, which stabilizes the unmodified band of OMM parkin substrates, including Mfn2, and prevents rupture of the OMM (Chan et al., 2011; Rakovic et al., 2011; Yoshii et al., 2011) (Figure 1G, GFP-parkinC431S, which cannot ligate Ub (Trempe et al., 2013), is used as a negative control). MG132 co-incubation rescued ER-OMM contact in U2OS:GFP-parkin cells treated with CCCP (Figure 1H,I and J). As expected, we also prevented OMM-ER disruption in cells depleted of PINK1 (Figure 1H,I and J). Finally, we replicated our U2OS cell data in induced pluriopotent stem cell (iPSC) -derived dopaminergic (iDA) neurons isolated from either control individuals or a patient carrying compound heterozygous deletions in the PRKN gene (PRKNdel; see Materials and methods). iDA neuronal cultures express endogenous parkin at a level comparable to that in the cytosolic fraction from mouse brain (Figure 1K), as well as the catecholinergic marker tyrosine hydroxylase (TH) (Figure 1L). Full-length parkin was undetectable in PRKNdel cells (Figure 1K), as expected given the genetic background of this line (Grünewald et al., 2010). Upon treatment of these neurons with CCCP for only one hour, we observed Mfn2 ubiquitination in both control lines but not in the parkin deletion line (Figure 1M). When we analyzed mitochondria-ER appositions in these cells, we again observed a CCCP-dependent decrease in the amount of <100 nm ER-OMM appositions in both control lines (Figure 1N and O). However, this decrease was absent in the parkin deletion line (Figure 1N and O), supporting our previous overexpression data in U2OS cells (Figure 1A to E ). Thus, PINK1 and parkin function to destroy contacts between the ER and mitochondria during mitophagy, likely through parkin-mediated OMM protein ubiquitination and turnover, as this process can be prevented by inhibiting proteasomal degradation. Moreover, this is a relevant biological process in human dopamine neurons, where it is regulated by endogenous parkin. Phosphoubiquitination of Mfn2 by the PINK1/parkin system disrupts its antagonistic effect on mitophagy Our EM data demonstrated that ER-mitochondria uncoupling occurs as an early step in the mitophagy pathway, prior to autophagosomal engulfment of the organelle (Figure 1F), and we sought to understand the underlying mechanism of this phenomenon. Mfn2 is both a mitochondria-ER tether and parkin ubiquitination substrate (de Brito and Scorrano, 2008; Sarraf et al., 2013; Tanaka et al., 2010), and thus the modulation of interorganellar contact by PINK1/parkin may occur through their effect on Mfn2. We began by examining the ubiquitination (via the disappearance of the unmodified band) of various parkin substrates (Khan et al., 2016; Sarraf et al., 2013) during a CCCP time course in U2OS:GFP-parkin cells, using the A320R mutant – which fails to bind pUb and initiate mitophagy (Wauer et al., 2015a; Yamano et al., 2015) – as a negative control. Turnover of both Mfn1 and Mfn2 occurred early (almost complete disappearance by two hours) compared to other OMM proteins (Figure 2A). Upon higher exposure (Figure 2B) of these immunoblots (from Figure 2A), we observed a rapid ‘burst’ of Mfn2 ubiquitination that occurred between 30 and 60 min CCCP. When compared to TOM20, a protein that is not promptly ubiquitinated by parkin (Sarraf et al., 2013), the rapidity of this Ub burst on Mfn2 was emphasized as TOM20 ubiquitination occurs gradually over a period of hours, rather than rapidly over a period of minutes (Figure 2B). Thus, ubiquitination of the mitofusins is one of the very first steps after the induction of mitophagy. Figure 2 with 1 supplement see all Download asset Open asset Mfn2 is rapidly phosphoubiquitinated upon induction of mitophagy. (A) Immunoblot analysis of protein turnover in glucose-maintained U2OS:GFP-parkin WT and A320R cells treated with 20 μM CCCP for the indicated time. (B) Higher exposures of Mfn2 and TOM20 immunoblots from (A). Red asterisks indicate ubiquitinated forms of Mfn2 and TOM20. (C) Co-immunoprecipitation of parkin substrates with GFP-parkin WT or A320R in U2OS cells treated with 20 μM CCCP for the indicated time, using an anti-GFP antibody. Immunoprecipitates were separated, along with 4% input, by SDS-PAGE and immunoblotted for the indicated protein. The arrowhead indicates the unmodified form of the protein, while the red asterisks denote ubiquitinated forms. (D) Workflow for the on-bead deubiquitination of Mfn2. U2OS:GFP-parkin WT cells were treated for one hour with 20 μM CCCP, and GFP-parkin was immunoprecipitated as in (C). Immunoprecipitates were then treated with Usp2 deubiquitinase and the beads were re-isolated by centrifugation. (E) Immunoblot detection of Mfn2 after on-bead deubiquitination, as described in (D). Immunoprecipitates were either incubated at 37°C in the absence or presence of Usp2 catalytic domain for 30 min. Samples were then centrifuged to separate beads and supernatant (‘sup.’), which were denatured in sample buffer prior to separation by SDS-PAGE. Arrowheads indicate unmodified forms of Mfn2, while the red asterisks denote ubiquitinated forms. (F) Immunoprecipitation of Mfn2 for LC/MS analysis. Immunoprecipitates were separated, along with 4% input, by SDS-PAGE and immunoblotted for Ub. (G) Extracted ion chromatogram for the pS65 Ub peptide (TLSDYNIQKEpSTLHLVLR, a.a. 55–72) from Mfn2 immunoprecipitates from DMSO- (blue line) and CCCP- (red line) treated U2OS:GFP-parkin WT cells, immunoprecipitated as in (F). The red arrow indicates the peak corresponding to the peptide. (H) Immunoprecipitation of Mfn2 under denaturing conditions. Cells were lysed in buffer containing 1% SDS (see Materials and methods). Immunoprecipitates were separated, along with 4% input, by SDS-PAGE and immunoblotted for Ub and pS65 Ub. (I) Crystal structure of parkin complexed with pUb (PDB ID 5N2W, Kumar et al., 2017). The A320 residue at the pUb/parkin interface is highlighted in red, with parkin coloured blue and ubiquitin in green. (J) GST-R0RBR pulldown of pUb from U2OS:GFP-parkin WT cells. Pulldowns were performed with WT or A320R GST-R0RBR, with no GST-R0RBR (‘-’) as a further negative control. Pulldowns were separated, along with 10% input, by SDS-PAGE and immunoblotted for the indicated protein. The asterisk represents a cross-reaction between the pS65 antibody and the GST-R0RBR module. https://doi.org/10.7554/eLife.32866.005 Mechanistically, this Ub burst would require local activation of parkin by PINK1 in the vicinity of Mfn2, which could be achieved by PINK1-catalyzed phosphorylation of the resulting Ub chains – events that would dually serve to activate parkin and tether it in place (Okatsu et al., 2015). To test this, we first immunoprecipitated WT or A320R GFP-parkin from cells treated with CCCP over time. We observed robust coimmunoprecipitation of ubiquitinated Mfn1 and Mfn2 with GFP-parkinWT at one hour CCCP (corresponding to the Ub burst observed in Figure 2B), with no apparent binding at four hours (Figure 2C), likely due to turnover of the Mfns by the proteasome at this time (Figures 1G and 2B and [Tanaka et al., 2010]). When we analyzed other parkin substrates that are ubiquitinated less rapidly than the Mfns (Figure 2A), we observed binding to WT parkin only at four hours of CCCP treatment in the case of ubiquitinated Miro1, and binding of mono-ubiquitinated HK1 at one hour CCCP, which was further shifted at four hours, indicative of processivity of HK1 ubiquitination (Figure 2C). None of these ubiquitinated species coimmunoprecipitated with GFP-parkinA320R (Figure 2C). To confirm that GFP-parkin was indeed binding ubiquitinated Mfn2, we treated GFP-parkin immunoprecipitates from CCCP-treated cells with Usp2 deubiquitinase (see schematic in Figure 2D), which is active on both phosphorylated and unphosphorylated Ub chains (Wauer et al., 2015b), and observed the release of Mfn2 from the parkin-bound bead fraction into the supernatant after separation by centrifugation (Figure 2E). These results strongly suggested that, early on in the mitophagy pathway, parkin was binding ubiquitinated Mfn2, likely through interactions with pUb moieties. We next confirmed the phosphoubiquitination of Mfn2 during mitochondrial depolarization. When we immunoprecipitated Mfn2 from U2OS:GFP-parkinWT cells that were treated with CCCP for one hour, we detected Ub-modified species by immunoblot (Figure 2F). This was concomitant with a decrease in overall Mfn2 levels (Figure 2F), owing to its proteasomal turnover (Figure 1G). Liquid-chromatography coupled to mass spectrometry (LC/MS) confirmed that the Mfn2 immunoprecipitation contained pS65 Ub selectively in the CCCP-treated condition (Figure 2G), despite lower Mfn2 levels (Figure 2A and F and Figure 2—figure supplement 1). We then confirmed that both pS65 and unphosphorylated Ub were covalently attached to Mfn2 by its precipitation under denaturing conditions and detecting pS65 Ub and total Ub by immunoblot (Figure 2H). Finally, profiting from the nanomolar affinity of the parkin R0RBR module for pS65 Ub (Sauvé et al., 2015), we used GST-R0RBR to pull down phosphoubiquitinated species from CCCP-treated U2OS:GFP-parkinWT cell lysates. We again used the A320R mutant – which abolishes the parkin-pUb interaction (Figure 2I) (Wauer et al., 2015a; Yamano et al., 2015) – as a negative control. In a CCCP-dependent manner, pS65 Ub, Ub and (shifted) Mfn2 could be detected in GST-R0RBRWT pulldowns (Figure 2J). Strikingly, we did not observe any of these factors in pulldowns using GST-R0RBRA320R (Figure 2J). Mfn2 is therefore phosphoubiquitinated and, taken together with our previous data, a burst of phosphoubiquitination – parkin-mediated ubiquitination coupled to PINK1-catalyzed phosphorylation – occurs on Mfn2 at an early time point in the mitophagy pathway. Our observations so far demonstrated that mitochondria are separated from the ER during mitophagy, and that the OMM-ER tether Mfn2 is rapidly degraded at the onset of the pathway. We thus hypothesized that Mfn2 may antagonize mitophagy through its ability to tether mitochondria and the ER, necessitating its destruction. To test this, we silenced Mfn2 (siMfn2) in U2OS:GFP-parkinWT cells, as well as Mfn1 – which promotes mitochondrial fusion without any apparent role in interorganellar tethering (de Brito and Scorrano, 2008) – to control for resulting from fusion We confirmed Mfn1 and Mfn2 by immunoblot (Figure and observed mitochondrial fragmentation in both and cells (Figure and Figure supplement 1A and B) with an ER-OMM apposition to the condition (Figure supplement and as we the of parkin to depolarized mitochondria in these cells our a cell is to parkin the parkin the mitochondrial reticulum in its Moreover, we took of of cells by cells in containing as a source than This generation through the electron and parkin-dependent mitophagy et al., 2015; et al., mitochondrial of and the of Ub, and on mitochondria are all in cells (Figure supplement we observed mitochondrial in not cells, under both conditions (Figure and A was within one hour of CCCP treatment in cells, and was in their owing to their in the control condition (Figure Strikingly, Mfn2 in cells to levels in glucose-maintained cells transfected with control (Figure Mfn1 and Mfn2 (Figure supplement did not further the of parkin cells (Figure supplement that this was and to a of mitochondrial Figure with see all Download asset Open asset Mfn2 mitophagy. (A) Immunoblot analysis of whole-cell lysates from cells in or transfected with control or Mfn1 or Mfn2 (B) in glucose-maintained cells transfected with the indicated siRNA, as by of TOM20 (red) (Hoechst, blue). Scale bar, 30 microns. (C) Representative images of GFP-parkin to mitochondria as a function of time in U2OS:GFP-parkin cells treated with 20 μM CCCP. Red asterisks indicate cells in which GFP-parkin has to Scale bar, 20 microns. (D) Quantification of parkin in cells from (C). Data represent mean ± SEM, n = cells per with cells per condition for (E) Parkin at one hour CCCP in cells from (C) as a Bars represent mean ± n.s., not significant; **, p<0.01; ***, (F) cells were transfected with the indicated and GFP-parkin WT or and were treated with 20 μM CCCP for four hours. in cells was using by at nm and nm The data are as of from at both The area cells mitophagy and the percentage of cells within this is indicated in the of (G) Quantification of the percent of cells mitophagy in cells from (F) treated with (red bars) or CCCP (blue bars) for four hours. Bars represent mean ± SEM, n = 2 n.s., not significant; *, p<0.05; **, p<0.01; ***, Figure data 1 Numerical source data for Figure E and Figure supplement Figures I, and Download We next this in could be induced by mitochondria-ER contacts via other than Mfn2. To test this, we silenced two other that been to promote mitochondria-ER and (Figure supplement et al., 2015; et al., Mfn2 we did not observe mitochondrial fragmentation in either or cells (Figure supplement When we parkin in these cells, we that, to Mfn2 of either or the of parkin to mitochondria (Figure supplement and the was in cells that were treated with CCCP for one hour, where parkin was to levels in and cells despite remaining cytosolic in cells transfected with control at this time point (Figure supplement and Thus, disruption of tethering increases the of parkin to depolarized We next directly the effect of Mfn2 on mitochondrial turnover using of a protein that its when by the lysosome et al., 2011). We transfected U2OS cells stably-expressing on either or with Mfn1 or Mfn2, by using the C431S mutant as a negative control. we treated these cells with CCCP for four hours and then the of per cell by (see Materials and as a of mitophagy et al., 2011; Tang et al., 2017). As expected, in the CCCP-treated a higher of control cells had an of compared with these cells were and this was replicated in cells (Figure and However, in cells, we observed a 2 in the of cells mitophagy (Figure and In we did not observe a at all in either control or cells but observed a level of mitophagy in cells similar to control cells in (Figure and These data demonstrate that, in cells, depolarization-induced mitophagy is in line with our parkin (Figure to and demonstrate that Mfn2 mitophagy at the level of To that we were from of our we replicated our data in Mfn2 U2OS cells that were using the system (see Materials and disruption was confirmed by in two and in which a was via a the corresponding to in the human Mfn2 gene (Figure supplement We these cells by along with a that the complete and but in which Mfn2 as a further negative Mfn1 levels remained similar all lines, and the of the mitochondrial remained (Figure supplement in the has been in isolated from et al., Accordingly, Mfn2 cells had mitochondrial that were but compared to WT U2OS cells (Figure supplement and our data in cells, Mfn2 cells on transfected with GFP-parkin (Figure supplement and and mitophagy (Figure supplement and Finally, we that parkin in Mfn2 cells (Figure supplement