Parkin promotes proteasomal degradation of misregulated BAX

Mitochondrial apoptosis is a key form of programmed cell death (Hotchkiss et al., 2009). The activities of the BCL-2 proteins BAX and BAK commit mammalian cells to apoptosis by inducing outer mitochondrial membrane (OMM) permeabilization. The release of cytochrome c (cyt c) and other proteins from the mitochondrial intermembrane space into the cytosol initiates the caspase cascade, efficiently dismantling the cell (Martinou and Youle, 2011; Tait and Green, 2010). Although cells can survive limited mitochondrial permeabilization (Ichim et al., 2015), activation of BAX and BAK is usually the first irreversible step in mitochondrial apoptosis signaling (Youle and Strasser, 2008). Consequently, BAX and BAK are regulated by a complex network of protein interactions both within and outside the BCL-2 family. BAX and BAK contain three BCL-2 homology domains (BH1–BH3) and their primary antagonists are pro-survival BCL-2 proteins such as BCL-2 itself, BCL-x_L (BCL2L1) and MCL-1, which have four BH domains (BH1–BH4). All these proteins are structurally similar and contain a C-terminal transmembrane domain (TMD) mediating association with mitochondria to different extents.

BAX and BAK specifically target the voltage-dependent anion channel 2 (VDAC2) on the OMM, forming large protein complexes (Cheng et al., 2003; Lauterwasser et al., 2016; Lazarou et al., 2010; Ma et al., 2014). Pro-survival BCL-2 proteins antagonize OMM localization and activation of BAX by constant retrotranslocation of mitochondrial BAX into the cytosol (Edlich et al., 2011). Permanent translocation and retrotranslocation establish an equilibrium between cytosolic and mitochondrial BAX that regulates BAX and thus BAX-dependent OMM permeabilization (Edlich, 2015; Edlich et al., 2011; Schellenberg et al., 2013; Todt et al., 2013, 2015). BH3-only proteins (containing only one BH motif) can trigger mitochondrial BAX accumulation by inhibition of pro-survival BCL-2 proteins and are also thought to induce active BAX conformations by direct interaction (Edlich et al., 2011; Letai et al., 2002; Llambi et al., 2011; Willis et al., 2005, 2007). The functionally redundant BAK is controlled by the same process, but different shuttling rates dependent on the C-terminal TMDs result in primarily cytosolic BAX and predominantly mitochondrial BAK (Edlich, 2015; Todt et al., 2015). Exposure of the TMD plays an important role in BAX activation (Annis et al., 2005; George et al., 2007; Nechushtan et al., 1999; Suzuki et al., 2003). However, the structural analysis of BAX lacking the TMD (helix α9) after induction of conformational changes by various means also suggests the separation of the ‘latch’ (helix α6, α7 and α8) from the ‘core’ domain (helices α1–α5) during BAX activation (Czabotar et al., 2013).

Inhibition of BAX has also been suggested to occur independently of BCL-2 proteins. Cytosolic interactions between BAX and Ku70 (also known as XRCC6), 14-3-3 isoforms (Samuel et al., 2001; Tsuruta et al., 2004), Pin-1 (Shen et al., 2009) and other proteins have been implicated in BAX regulation. Recently, the E3 ligase parkin has been suggested to ubiquitylate the BAX BH3 domain, thereby targeting cytosolic BAX to proteasomal degradation prior to translocation to the mitochondria (Charan et al., 2014; Johnson et al., 2012). However, the comprehensive analysis of cytosolic BAX and its potential interactions revealed that cytosolic BAX interactions are not apparent (Hsu and Youle, 1998; Vogel et al., 2012). Nevertheless, the importance of BAX activity in particular for the commitment to apoptosis induced by a variety of apoptotic stresses (Wang and Youle, 2011) suggests additional layers of BAX-specific regulation.

The structure of active BAX has been subject to extensive research. An intramolecular disulfide tether between helix α5 and α6 (V121C/I136C) was shown to interfere with BAX activity in organellar cyt c release assays, suggesting that separation of helix α5 from helix α6 occurs during BAX activation (Czabotar et al., 2013). We studied the importance of conformational changes separating helix α5 and α6 by introducing this disulfide tether between BAX V121C and I136C (BAX 5-6) expressed in HCT116 BAX–BAK double knockout (DKO) cells (Fig. 1A). This approach parallels our previous analysis of BAX 1-2/L-6 in the context of living cells, leading to the discovery of BAX retrotranslocation from the mitochondria into the cytosol (Edlich et al., 2011). Similar to BAX 1-2/L-6, the engineered disulfide bond between helix α5 and α6 is largely oxidized and shielded from the reducing environment of the cytosol, as shown by altered SDS-PAGE migration and the lack of Mal–PEG labeling for the vast majority of BAX 5-6 also in denaturing conditions (Fig. 1B and Fig. S1A).

Fractionation of cells expressing BAX 5-6 showed a shift of tethered BAX from the mitochondria to the cytosol compared with that in cells expressing BAX ΔSH (which localizes like wild-type BAX; Edlich et al., 2011) at similar expression levels (Fig. 1C,D, Fig. S1B and Fig. S2A). The voltage-dependent anion channel 2 (VDAC2) mediates mitochondrial BAX association and is essential for BAX retrotranslocation (Lauterwasser et al., 2016; Ma et al., 2014). Therefore, we tested the requirement of conformational changes separating helix α5 and α6 for VDAC2 interactions with BAX. BN-PAGE analysis shows a substantial reduction of BAX-containing VDAC2 complexes when the tether between BAX helices α5 and α6 was formed (Fig. 1E and Fig. S2B). A BAX variant lacking helix α6 also did not form complexes with VDAC2. We, therefore, hypothesized that the stabilized BAX 5-6 structure prevents interactions between BAX helix α6 and VDAC2. Therefore, alanine substitutions of the bulky hydrophobic residues W139 and L141 on both sides of helix α6, corresponding to residues F157 and V159 in the similarly regulated BAK (Fig. 1F), were tested and confirm the importance of both sides of helix α6 by impairing the formation of VDAC2-containing BAX complexes (Fig. 1E). W139A also abolishes contact between BAX and the OMM (Fig. 1G). Fluorescence loss in photobleaching (FLIP) experiments to analyze BAX 5-6 retrotranslocation from the mitochondria into the cytosol also revealed diminished mitochondrial BAX 5-6 association, rendering the measurement of BAX 5-6 retrotranslocation in comparison to the wild type impossible (Fig. 1H). The loss of cytosolic BAX fluorescence during FLIP emphasizes the significant reduction of a mitochondrial BAX pool by the α5–α6 tether (Fig. 1C,D,H). Taken together, the separation of helix α5 from α6 and thus exposure of helix α6 is important for efficient BAX association with the mitochondria in healthy cells.

Impaired mitochondrial association of BAX 5-6 could cause reduced BAX activity, as observed with mitochondrial BAX 5-6 S184L isolated from MEF cells (Czabotar et al., 2013). Surprisingly, BAX 5-6 commits HCT116 BAX/BAK DKO cells to apoptosis at least as potent as wild-type BAX (Fig. 2A). Profound BAX 5-6 activity despite reduced OMM-association of this variant could result from impaired inhibition in the absence of BAX retrotranslocation complexes. However, we cannot rule out that a minor BAX 5-6 pool lacking the α5–α6 tether (Fig. 1B) due to reduction of the involved disulfides has triggered the measured activity. Increased BAX 5-6 activity compared with the wild type is also observed by the analysis of caspase-dependent cleavage of poly (ADP-ribose) polymerase (PARP), the presence of active BAX detected by the monoclonal antibody 6A7 and the release of cyt c following apoptosis stimulation (Fig. 2B-D). Conformational changes such as the separation of BAX helices α5 and α6 are dispensable for BAX activation, which is in line with the possibility of helices α5 and α6 inserted in concert into the OMM in active BAX (Annis et al., 2005). Therefore, separation of the ‘core’ domain and ‘latch’ is important for association of BAX with mitochondria and inhibition in non-apoptotic cells, but is apparently not required for BAX activation.

The demonstrated activity of BAX 5-6 suggests that HCT116 cells have the means to inhibit this BAX variant in order to survive. We reasoned that the requirement of BAX inhibition would be elevated if BAX is targeted to the OMM. BAX binding to the OMM is mediated by helix α6 and TMD interactions with VDAC2 (Lauterwasser et al., 2016). Since the hydrophobicity of the TMD determines the BAX shuttling rate (Todt et al., 2015), we introduced an S184V substitution to increase the mitochondrial pool of BAX 5-6 (Nechushtan et al., 1999; Fig. 3A). Blue native (BN)-PAGE analysis showed a vastly increased level of VDAC2-containing BAX complexes mediated by the S184V substitution (Fig. 3B). These results suggest that BAX TMD functions are not impaired by the BAX α5–α6 tether. Cells expressing BAX 5-6 S184V thus contain a pool of VDAC2 complexes as observed with wild-type BAX. These complexes of mitochondrial BAX 5-6 (and BAX 5-6 S184V) were similar in size and likely composition to the wild type and provide no rationale for increased apoptotic activity of BAX 5-6. In parallel to BAX 5-6, the apoptotic activity of BAX 5-6 S184V was significantly increased compared with the wild type in response to apoptotic stimuli (Fig. S3). Pro-apoptotic activity of BAX 5-6 S184V can also be detected in the absence of apoptotic stimuli, mirroring the properties of BAX S184V (Nechushtan et al., 1999). Confocal analysis revealed cytosolic and mitochondrial localization of BAX 5-6 S184V (Fig. 3C). Surprisingly, ectopic expression of BAX 5-6 S184V resulted in significantly lower protein levels than expected. However, in the presence of the proteasome inhibitor MG132 or the ubiquitin-activating enzyme E1 inhibitor PYR41, the levels of BAX 5-6 S184V were substantially increased (Fig. 3D,E), suggesting proteasomal degradation of misregulated BAX. By contrast, wild-type BAX and BAX 5-6 levels were affected to a much lesser extent in the presence of MG132 or PYR41 (Fig. 3F,G and Fig. S4). Therefore, proteasomal degradation represents an additional layer of BAX regulation that is dependent on mitochondrial BAX localization. Analysis of BAX 5-6 S184V retrotranslocation by FLIP measurements revealed decreased shuttling rates compared with BAX S184V (Fig. 3H,I). The S184V substitution results in a marked reduction of BAX retrotranslocation (Edlich et al., 2011; Todt et al., 2015) and this analysis was performed with ectopically expressed BCL-x_L to prevent BAX activation from obscuring the results. The seemingly moderate reduction of BAX retrotranslocation by the α5–α6 tether could have a strong impact on the regulation of mitochondrial apoptosis, as BAX activation depends on the dwell time of BAX molecules on the OMM (Todt et al., 2015). BAX 5-6 S184V showed prominent ubiquitylation (Fig. 3J) whereas ubiquitylated wild-type BAX was only detectable in the presence of a ubiquitin ΔK variant. Ubiquitin ΔK is randomly incorporated into poly-ubiquitin chains in the presence of endogenous ubiquitin and stabilizes ubiquitylated protein species by less-efficient proteasomal degradation of shorter ubiquitin chains. Therefore, ubiquitylation occurs only with a mitochondrial minor pool of wild-type BAX, but this pool increases when BAX retrotranslocation is impaired. Endogenous BAX levels increased in a time-dependent manner following inhibition of the proteasome or ubiquitylation (Fig. S4C). Proteasomal BAX degradation could represent a quality control mechanism for BAX retrotranslocation, removing slowly shuttling BAX molecules from the OMM and thus preventing OMM permeabilization (Fig. 3K).

The E3 ligase parkin is recruited to the OMM to induce mitophagy (Narendra et al., 2008). To investigate whether parkin could mediate mitochondrial ubiquitylation of BAX, we analyzed potential transient interactions between BAX and parkin using a biotin ligase (BirA) protein fusion approach (Huang and Jacobson, 2010). Here, we expressed a BirA–BAX fusion protein in cells to detect transient BAX-interacting proteins (Fig. 4A). Confirming the functionality of this approach, BCL-x_L was specifically biotinylated following BirA–BAX expression, in line with known transient interactions between BAX and BCL-x_L. In contrast, GAPDH, an abundant protein without specific interactions to BAX, remained unlabeled. Importantly, using this approach, biotinylation of parkin was also observed, demonstrating potential transient interactions between BAX and parkin (Fig. 4B). Interestingly, the pool of labeled parkin and therefore potentially the probability of BAX interaction increases with an increased pool of misregulated BAX on the mitochondria due to the α5–α6 tether. Ectopic parkin expression enhanced BAX 5-6/S184V ubiquitylation, suggesting that parkin ubiquitylates misregulated BAX on the mitochondria (Fig. 4C). VDAC ubiquitylation by parkin was not previously reported (Geisler et al., 2010), but this might depend on massive recruitment of parkin by the mitochondria. Strikingly, two hours longer expression of wild-type parkin eliminated BAX 5-6 S184V from human cells (Fig. 4D and Fig. S5A). BAX 5-6 S184V levels are restored in the presence of parkin when proteasomal degradation is inhibited by MG132 or ubiquitylation is inhibited by PYR41.

Extensive parkin-dependent ubiquitylation of mitochondrial proteins has been demonstrated to follow loss in mitochondrial membrane potential [induced by the mitochondrial uncoupler carbonyl cyanide m-chlorophenyl hydrazone (CCCP)] and subsequent activation of the kinase PINK1 (Narendra et al., 2010). PINK1 phosphorylates ubiquitin and parkin directly, causing mitochondrial parkin accumulation (Kane et al., 2014; Koyano et al., 2014). Since BAX ubiquitylation is dependent on mitochondrial BAX localization, the effect of mitochondrial parkin recruitment was tested. Induction of increased recruitment of mitochondrial parkin following CCCP treatment resulted in decreased BAX 5-6 S184V levels (Fig. 4E). Increased BAX degradation after the loss of the mitochondrial membrane potential suggests enhanced BAX ubiquitylation following massive recruitment of parkin in mitochondria. By contrast, extensive mitochondrial parkin recruitment is not required for steady-state BAX ubiquitylation, and parkin activation by phosphorylation is not influenced by the presence of BAX (Fig. S5B). This also excludes the requirement of parkin S56 phosphorylation for BAX ubiquitylation (Ordureau et al., 2014). These results show that parkin ubiquitylates BAX on the OMM and this activity does not require but is enhanced by mitochondrial parkin accumulation, suggesting enhanced BAX targeting by ectopic parkin expression follows increased mitochondrial parkin levels. BAX misregulation only slightly influences transient interactions with parkin, indicating that parkin-mediated BAX ubiquitylation is dependent on the conformation of BAX.

Parkin specifically targets BAX but not BAK for proteasomal degradation. The protein levels of the predominantly mitochondrial BAK are not affected by parkin in the presence of ubiquitin ΔK (Fig. 5A). BAX-specific proteasomal degradation and accelerated BAX retrotranslocation ensuring low mitochondrial BAX levels reveal the paramount importance of mitochondrial BAX inhibition for cell survival. Specific ubiquitylation of BAX by parkin could theoretically occur at all nine lysine residues present in BAX. Strikingly, lysine 128 is readily exposed in the globular BAX fold between helices α5 and α6 and the corresponding residue in BAK is glycine 146 (Fig. 5B). We reasoned that ubiquitylation of K128 could confer BAX-specific degradation and tested this hypothesis introducing the K128R substitution in BAX 5-6 S184V. Indeed, the K128R substitution reduces BAX 5-6 S184V polyubiquitylation substantially (Fig. 5C). However, the inhibition of BAX polyubiquitylation is incomplete, suggesting that other lysines in overexpressed BAX are also ubiquitylated. Importantly, the BAX variant containing K128R is completely protected against parkin-dependent proteasomal BAX degradation (Fig. 5D). Therefore, polyubiquitylation in position K128 is essential for parkin-dependent proteasomal degradation of BAX.

The combination of the intramolecular disulfide tether between helices α5 and α6 of BAX and the S184V substitution in the BAX TMD shows the importance of parkin-dependent BAX regulation, but also vastly increases the amount of misregulated BAX on the mitochondria. Therefore, the impact of parkin-mediated BAX degradation by the proteasome on mitochondrial apoptosis signaling was analyzed using ectopic wild-type BAX expression in HCT116 BAX–BAK DKO cells at near-endogenous protein levels (Todt et al., 2015). Parkin expression protects human cells from wild-type BAX activity in the absence or the presence of the apoptotic stimulus actinomycin D (Fig. 6A). A substantial increase in BAX activity with or without apoptotic stimulus is observed, when proteasomal degradation of BAX is blocked by MG132. Parkin-dependent BAX inhibition by proteasomal degradation is corroborated by a reduction in the release of cyt c (Fig. 6B). Stabilization of MCL-1 in the presence of MG132 should counteract BAX activity (Carroll et al., 2014), but profound BAX activation is observed after inhibition of the proteasomal degradation pathway. Ectopic parkin expression inhibits wild-type BAX activity in HCT116 BAX–BAK DKO cells with or without staurosporine treatment, whereas BAX K128R shows no reduction in activity in the presence of overexpressed parkin (Fig. 6C and Fig. S6). Analysis of the clonogenic cell survival of parkin-deficient HeLa cells and SH-SY5Y cells, containing high levels of endogenous parkin, in response to MG132 treatment showed MG132-concentration-dependent cell death in SH-SY5Y cells (Fig. 6D). In contrast, HeLa cells deficient in parkin-dependent BAX inhibition and thus requiring inhibition of BAX by other means, e.g. BAX retrotranslocation, show no toxic effect of MG132. YFP–parkin expression in HeLa cells significantly reduced the apoptotic response to a variety of chemotoxic stresses (Fig. 6E,F). Consequently, MG-132 treatment led to cell death in YFP–parkin-expressing HeLa cells (Fig. 6G). Parkin knockout in SH-SY5Y cells results in increased sensitivity to apoptotic stimuli (Fig. 6H). Therefore, parkin-dependent degradation of BAX provides additional protection against BAX activation when misregulated BAX accumulates on the OMM (Fig. 6I).

The pro-apoptotic BCL-2 protein BAX is regulated by constant translocation to the mitochondria and retrotranslocation back into the cytosol (Edlich, 2015; Edlich et al., 2011; Schellenberg et al., 2013; Todt et al., 2015). Therefore, pro-survival BCL-2 proteins inhibit BAX (and BAK) in cells by transient interactions during retrotranslocation from the OMM (Lauterwasser et al., 2016), whereas previously described heterodimeric complexes between BAX (and BAK) and pro-survival BCL-2 proteins were probably induced artificially because of the use of detergents (Hsu and Youle, 1997). BAX regulation requires major conformational changes, such as exposure of the C-terminal TMD and the BH3 domain from the globular fold (Edlich et al., 2011; Suzuki et al., 2000). Here, we show that conformational changes involving helices α5 and α6 are essential for BAX association with VDAC2 on the OMM, but also for BAX regulation by retrotranslocation into the cytosol. BAX variants suggest a central role of helix α6 in the formation of VDAC2-containing complexes on the OMM. These results support the importance of the separation of the ‘latch’ from the ‘core’ domain in BAX regulation (Czabotar et al., 2013), but also argue for a role of this conformational change in the mitochondrial association of BAX in healthy cells rather than it being solely a BAX-activation-specific event. Noteworthy, the mitochondrial BAX pool prior to apoptotic stress determines the cellular predisposition to apoptosis (Reichenbach et al., 2017; Todt et al., 2013). Therefore, the large impact of the BAX α5–α6 tether on mitochondrial BAX association is consistent with BAX activation models suggesting separation between helices α5 and α6 (Bleicken et al., 2014; Czabotar et al., 2013; Mandal et al., 2016). However, BAX 5-6 commits human cells to apoptosis and even increases apoptotic activity compared with the wild-type protein. These results suggest that BAX activation could alternatively circumvent major conformational changes between helices α5 and α6, as previously suggested (Annis et al., 2005). Therefore, BAX molecules that associate with the OMM but fail to undergo conformational changes required for BAX retrotranslocation impose the threat of OMM permeabilization upon the cell.

Proteasomal degradation of mitochondrial BAX represents an additional layer of cell protection from BAX activity. Experiments analyzing transient BAX interactions with the E3 ligase parkin suggest the importance of major conformational changes in BAX for parkin-dependent ubiquitylation. Parkin inhibits only a minor pool of endogenous BAX on the mitochondria, but in doing so provides considerable protection from BAX activity. Proteasomal degradation of BAX is induced by parkin-dependent ubiquitylation of BAX K128. Parkin has been suggested to ubiquitylate cytosolic BAX at K21 and/or K64 (Charan et al., 2014; Johnson et al., 2012). However, we do not find any support for these suggestions, as BAX ubiquitylation correlates with mitochondrial localization and BAX degradation is completely abolished by the K128R substitution. Parkin-dependent ubiquitylation regulates BAX but not the largely mitochondrial BAK, further emphasizing the importance of inhibiting BAX on the mitochondria. Parkin promotes mitophagy following selective recruitment to the OMM after loss of mitochondrial membrane potential (Narendra et al., 2008). Although mitochondrial parkin accumulation is not required for BAX ubiquitylation, loss of mitochondrial membrane potential induced by CCCP treatment increases BAX degradation. In addition, ectopic parkin expression enhances BAX ubiquitylation and degradation, depending on the ability of parkin to ubiquitylate mitochondrial proteins. Therefore, the size of the mitochondrial parkin pool determines ubiquitylation of misregulated BAX on the mitochondria.

Pro-survival BCL-2 proteins have been found to antagonize parkin translocation to the mitochondria (Hollville et al., 2014). This finding seems counterintuitive to the cell-protective function suggested for parkin in neuronal cells. Mutations in the parkin-encoding gene PRKN (also known as PARK2) occur in Parkinson's disease (PD), a condition associated with the death of dopaminergic neurons in the midbrain (Kitada et al., 1998). In addition, several studies have shown parkin-dependent cell protection from induced apoptosis (Ekholm-Reed et al., 2013; Sun et al., 2016; Wang et al., 2013). However, pro-survival BCL-2 protein-dependent inhibition of parkin translocation to the mitochondria does not contradict a central role of parkin in neuroprotection, considering the regulation of BAX. Efficient BAX retrotranslocation is ensured by high levels of pro-survival BCL-2 proteins, rendering additional inhibition of mitochondrial BAX less important (Fig. 6E). Mitochondrial parkin levels are increased when pro-survival BCL-2 proteins are inhibited and mitochondrial parkin can take over BAX inhibition on the OMM. This scenario suggests that parkin-dependent BAX ubiquitylation is a quality control mechanism balanced with BAX retrotranslocation. Pro-survival BCL-2 protein-dependent inhibition of apoptosis induced by mitochondrial depolarization further supports this connection (Carroll et al., 2014). Mitochondrial BAX inhibited by either BCL-2 protein-dependent retrotranslocation or parkin-dependent proteasomal degradation could exist in different conformations and may become active through different paths. It is possible that different activation pathways lead to different BAX complexes associated with BAX activity (Große et al., 2016; Nechushtan et al., 2001; Salvador-Gallego et al., 2016).

The bifunctional role of mitochondrial parkin in inhibiting BAX and promoting mitophagy could shape neuronal cell survival regulation. Increasing the cellular tolerance for stress could be beneficial for removal of post-mitotic neurons and thus their cell contacts could define their role. Mitochondrial parkin inhibits BAX-dependent apoptosis and induces mitophagy of impaired mitochondria, increasing stress resistance and providing the means of stress reduction at the same time. Therefore, dopaminergic neurons of patients bearing parkin mutations may undergo apoptosis that is dependent on misregulated BAX. Understanding the inhibition of mitochondrial BAX by parkin-dependent ubiquitylation and proteasomal degradation has the imminent potential to provide novel approaches for the prevention and therapy of Parkinson's disease.

HCT116 cells and HCT116 BAX–BAK DKO cells with or without stable GFP–BAX expression were cultured in McCoy's 5A medium supplemented with 10% heat-inactivated fetal bovine serum and 10 mM HEPES in 5% CO₂ at 37°C. SH-SY5Y cells were cultured in a 1:1 mixture of DMEM and F12 medium supplemented with 10% fetal bovine serum and 10 mM HEPES in 5% CO₂ at 37°C. HeLa cells were cultured in DMEM supplemented with 10% heat-inactivated fetal bovine serum and 10 mM HEPES in 5% CO₂ at 37°C. Cells were transfected with Turbofect (Fermentas) or Lipofectamine LTX (Invitrogen), typically with 100 ng of BAX or BCL-x_L constructs in pEGFP vector according to the manufacturer's instructions.

BAX variants were transiently expressed in HCT116 BAX–BAK DKO cells. Cells were lysed and the cytosolic fractions were divided into two groups for non-denaturing and denaturing conditions, adding 0.5 mM DTT and 2 mM or 0.5 mM Mal–PEG with and without 8 M urea, respectively. Samples were incubated for 1 h on ice (protected from light), subjected to acetone precipitation and analyzed by SDS-PAGE and western blotting.

Parkin was expressed with and without BAX 5-6 S184 V in HCT116 BAX–BAK DKO cells for 4 h. Cells were treated with 10 µM CCCP for 3 h or left untreated. Membrane fractions were prepared from transfected cells. For the detection of protein phosphorylation, an acrylamide based Phos-tag gel (4.5%, Phos-tag and MnCl₂, without EDTA) was used according to the manufacturer's protocol (Wako Chemicals).

Cells were harvested and incubated in cell lysis buffer [10 mM HEPES, pH 7.4, 150 mM NaCl, 1% Triton X-100, protease inhibitor cocktail (Roche)] for 15 min on ice. Whole-cell extracts were obtained by centrifugation at 15,000 g for 10 min at 4°C. Samples were boiled in SDS sample buffer for 5 min at 95°C and subsequently subjected to SDS-PAGE and western blot analysis. Subcellular fractionations were performed as previously described (Todt et al., 2013).

HCT116 BAX–BAK DKO GFP–BAX-expressing cells were seeded on a chambered cover glass (Thermo Fisher Scientific) in McCoy's 5A medium, grown for 36 h and imaged using a Zeiss 510 META confocal LSM microscope equipped with argon (458/488/514 nm lines) and HeNe (543/633 nm) lasers.

Fluorescence loss in photobleaching (FLIP) experiments were performed as described previously (Edlich et al., 2011). In short, cells were imaged prior to bleaching then a single region (diameter of 1 μm) within the nucleus was repeatedly bleached with two iterations of a 488 nm laser line (100% output) using a Zeiss LSM510 META with a 63× PlanFluor lens. Two images were collected after each bleach pulse. After 15 cycles of bleaching and collecting 30 images, separate measurements on the mitochondria were taken to analyze loss of fluorescence. Unbleached cells neighboring analyzed cells served as controls for photobleaching during image acquisition of each measurement. Fluorescence intensities were normalized by setting the pre-bleach fluorescence to 100% signal.

For 6A7 staining and cyt c release, BAX variants and parkin constructs were transfected into HCT116 BAX–BAK DKO cells and then, cells were treated with either 20 µM Mg132 for 3 h or left as control; 1 µM ActD was used as apoptotic stimulus. After the treatment, cells were fixed with 4% paraformaldehyde and permeabilized with Triton X-100 (0.15%). Anti-BAX 6A7 antibody (Sigma, 1:500) or anti-cyt c antibody (6H2B4, BD Biosciences, 1:1000) were added for 1 h at room temperature (RT). After washing with PBS, cells were incubated with the secondary antibodies mouse Alexa Fluor 594 (Invitrogen, 1:750) and rabbit Alexa Fluor 647 (Invitrogen, 1:1000), for 1 h at RT. Cells were washed with PBS and GFP-positive cells were quantified.

Apoptosis was induced using STS (Staurosporin, Enzo Life Technologies) or ActD (ActinomycinD, Sigma Aldrich). For caspase assays, cells were washed with ice-cold 1×PBS and resuspended in ice-cold cell lysis buffer (20 mM Tris-HCl, pH 7.4, 100 mM NaCl, 1 mM EDTA, with protease inhibitor cocktail and 0.5% Triton X-100). Whole-cell lysate was incubated with caspase-3/7 substrate (BD Pharmingen) for 60 min at 37°C and protein concentration was determined by a Bradford Assay (Roth). Substrate cleavage was measured for 50 cycles with 10 s delay (excitation at 380 nm, emission at 430–460 nm). Kinetics were determined and calculated to the amount of protein per sample.

To analyze native protein complexes, isolated mitochondria were solubilized in digitonin buffer (20 mM Tris-HCl, pH 7.4, 50 mM NaCl, 0.1 mM EDTA, 10% v/v glycerol, 2% w/v digitonin, 1 mM PMSF) and subsequently incubated at 4°C for 45 min in an end-over-end shaker. After insoluble material was removed by centrifugation (16,000 g, 10 min, 4°C) mitochondrial extracts were subjected to BN-PAGE (Schägger and von Jagow, 1991) using 4–13% continuous polyacrylamide gradient gels.

HCT116 BAX–BAK DKO cells transfected with HA-tagged ubiquitin, GFP-BAX and YFP-parkin as indicated, were harvested in ice-cold PBS and lysed in IP buffer [10 mM HEPES, pH 7.4, 150 mM NaCl, containing complete protease inhibitor mix (Roche) and 0.2% Triton X-100]. After clearing the lysate by centrifugation at 15,000 g for 10 min, input sample (1.25%) was separated. The remaining lysate was incubated with HA-antibody coupled agarose beads (Santa Cruz), end-over-end shaking at 4°C overnight. After incubation, the beads were washed five times with IP buffer and finally boiled in SDS loading buffer. Input and bead samples were resolved by 10% SDS-PAGE and analyzed by western blotting for the indicated proteins. Antibodies used were: GFP D5.1 (Cell Signaling), Myc 9B11 (Cell Signaling), AKT1 2H10 (Cell Signaling), VDAC/Porin ab5 (Calbiochem), VDAC N-18 (Santa Cruz), PARP (Cell Signaling), Actin C4 (Millipore), BAX E63 (Abcam), TOM20 FL-145 (Santa Cruz), HA Y11 (Santa Cruz), Bcl XL 54H6 (Cell Signaling), Parkin PRK8 (Santa Cruz), GAPDH (Sigma), BAK (Millipore), K48 Polyubiquitin (Cell Signaling) and K63 Polyubiquitin (Cell Signaling). pRK5-HA-ubiquitin-KO and pRK5-HA-ubiquitin-WT were obtained from AddGene and were deposited by Ted Dawson [plasmids #17603 and #17608 (Lim et al., 2005)].

HCT116 BAX–BAK DKO cells were transfected with pcDNA3–mycBioID–BAX plasmids, resulting in Myc-tagged BirA–BAX fusion expression. After cell harvest in ice-cold PBS, the cell pellet was resuspended and lysed in IP buffer [10 mM HEPES, pH 7.4, 150 mM NaCl, complete proteinase inhibitor cocktail (Roche) and 0.2% Triton X-100]. Cell lysate was centrifuged (120,000 g, 30 min, 4°C) and washed using a gel filtration column (GE Healthcare). Then, input sample (2.5%) was separated and the remaining lysate incubated with streptavidin agarose beads (Thermo, overnight at 4°C). Subsequently, beads were washed and boiled in SDS sample buffer. Input and pull-down samples were analyzed by SDS-PAGE and western blotting.

HeLa and SHSY-5Y cells were seeded to 70% confluency, treated for 6 h with 0, 5, 7.5, 10, 15 or 20 µM MG132 in DMSO. Cells were cultured for 10–12 days followed by fixation with 4% PFA in PBS and staining with Methylene Blue.

We thank R. Youle for providing YFP–parkin-expressing HeLa cells and X. Luo for sharing BAX ΔH6 with us.