Extracellular Regulated Kinase Phosphorylates Mitofusin 1 to Control Mitochondrial Morphology and Apoptosis

1Mitochondria sense and amplify apoptotic signals, releasing cytochrome c and other cofactors required for the activation of caspases and downstream cell death. Cytochrome c release requires mitochondrial outer membrane permeabilization (MOMP), is tightly regulated by proteins of the Bcl-2 family, and is associated with mitochondrial morphology changes. Mitochondrial morphology is modulated by large dynamin-like GTPases: the cytosolic dynamin-related protein 1 (DRP1), which translocates to mitochondria, binding to its receptor on the outer membrane (OMM) mitochondrial fission factor (MFF), constricting and fragmenting mitochondria. Pro-fusion proteins display pleiotropic functions: the inner mitochondrial membrane (IMM) protein Optic Atrophy 1 (OPA1) not only promotes fusion, but it also regulates cristae shape and remodeling to control cytochrome c release and mitochondrial function. Of the two mammalian OMM mitofusins (MFNs), MFN1 mediates organelle fusion together with OPA1, whereas MFN2 tethers mitochondria either in trans or to the endoplasmic reticulum (ER).

2The role of mitochondrial morphology in cell death is not completely understood. Hyperfusion represents a recognized strategy to allow survival during nutrient deprivation and cellular stress. Conversely, it has been questioned whether the fragmentation observed during apoptosis is causal for cytochrome c release (see, for example, for a review). Several models have been proposed to explain why mitochondrial fragmentation facilitates apoptosis. Fragmentation has been associated with mitochondrial hemifusion driven by the pro-fission molecule DRP1 that stimulates BAX accumulation on mitochondria or with the remodeling of mitochondrial cristae that is required for maximal cytochrome c release. The observed fragmentation is the result of DRP1-mediated fission and of impaired mitochondrial fusion. Interestingly, mitochondrial fusion and the apoptotic machinery are molecularly linked, since the Bcl-2 family proapoptotic members BAX and BAK complex with MFNs to facilitate MFN2 oligomerization and activity in healthy cells. Conversely, during apoptosis, BAK dissociates from MFN2 and associates with MFN1, albeit the functional role and the molecular mechanism of this switch is not understood.

3To understand how mitochondrial fusion changes during apoptosis, we tested whether key mitochondrial fusion molecules can be post-translationally modified. We show that MFN1 is phosphorylated at an atypical target site for the pleiotropic mitogen-activated protein (MAP) kinase cascade. This MFN1 phosphorylation modulates mitochondrial morphology, affecting its ability to form oligomers, and apoptosis, influencing BAK oligomerization in response to apoptotic stimuli.

4Whether and how phosphorylation regulates mitochondrial fusion is unclear: we therefore set out to identify if MFNs are phosphorylated. Phosphorylated and non-phosphorylated Flag-tagged MFN1 or MFN2 expressed in untreated mouse embryonic fibroblasts (MEFs) were separated by affinity chromatography on a column that specifically binds phosphorylated residues. Immunoblotting indicated that both Flag-MFN1 and MFN2 avidly bound to the column (Figure 1A). To identify the phosphorylated site(s) of these proteins, large-scale anti-Flag immuno-purification from Mfn1−/− and Mfn2−/− cells expressing Flag-MFN1 and Flag-MFN2 was successfully carried out (Figure S1A), and the bands corresponding to MFNs were excised from the gel and analyzed by mass spectroscopy. While 83% of MFN1 sequence was covered, coverage for MFN2 was only around 65% (Figure S1C), possibly explaining why we could not detect any MFN2 phosphorylation. Conversely, T562 and possibly T564 that lie in the HR1 region of MFN1 were found to be phosphorylated (Figure S1B). The sequence of this region, ASTPTAP, resembles a canonical ERK site, Pro-Xaa-Ser/Thr-Pro (X = neutral/basic amino acid), suggesting that this could be an atypical ERK site. Indeed, an in vitro phosphorylation assay using active recombinant ERK on immunoprecipitated FLAG-MFN1 and 32P-labeled ATP indicated that ERK can directly phosphorylate MFN1 (Figure 1B) like its canonical target myelin basic protein (MBP) (Figure 1C). Moreover, that MFN1 can be an ERK target was further substantiated by the finding that ERK co-immunoprecipitated with FLAG-MFN1 but not FLAG-MFN2 (Figure 1D).

5To discriminate between the two identified sites of phosphorylation, we first analyzed them in silico. T562 is conserved throughout mammalian MFN1s (Figure S1E). Of note, this residue is conserved also in MFN2, but the phosphoproteomic data excluded its phosphorylation, (data not shown) and MFN2 did not bind to ERK (Figure 1D). We next generated single and double mutants mimicking phosphorylated (MFN1T562D, MFN1T564D, and MFN1T562D,T564D) or dephosphorylated MFN1 (MFN1T562A, MFN1T564A, and MFN1T562A,T564A). Upon phosphocolumn affinity purification of the expressed mutants, only MFN1T562A did not bind to the substrate, whereas all the T564 mutants showed comparable binding (Figure 1E). In the same in vitro phosphorylation assay used in Figure 1B, recombinant ERK was unable to phosphorylate the MFN1T562A mutant (Figure 1F), further indicating that T562 could be the residue phosphorylated by ERK. To unequivocally prove that ERK phosphorylate MFN1T562, we produced a specific antibody directed to pT562 of MFN1. This antibody immunodecorated wild-type (wt), but not the T562A MFN1 mutant re-expressed in Mfn1−/− MEFs (Figure 1G, arrowhead), lending strong support to the fact that MFN1 was phosphorylated by ERK at T562. Moreover, treatment of cells with epidermal growth factor (EGF) that activates MEK and ERK, as well as genetic ERK activation (Figure S3) by a constitutively active MEK (MEKCA), increased T562 phosphorylation levels; pharmacological or genetic MEK inhibition using the selective U0126 inhibitor or by expressing a dominant negative MEK mutant (MEKDN); and reduced T562 phosphorylation; finally, in Mfn1−/− cells re-expressing the MFN1T562A mutant, phosphorylation was undetectable even in the presence of EGF (Figure 1G). Taken together, these data indicate that ERK phosphorylates MFN1 at T562.

6We next explored the significance of ERK-mediated MFN1 phosphorylation in mitochondrial morphology. Expression of MEKDN caused mitochondrial elongation, whereas activation of ERK by MEKCA resulted in fragmentation (Figures 2A and 2B). Furthermore, in cells lacking Mek1 or Mek2 mitochondria were significantly elongated (Figures S2A and S2B). As the MEK mutants did not modify mitochondrial morphology in Mfn1−/− MEFs, we concluded that MEK requires MFN1 to modulate mitochondrial shape (Figures 2A and 2B), and we set out to investigate whether it depended on MFN1T562. As expected, MEKDN triggered mitochondrial elongation while MEKCA fragmentation in Mfn1−/− MEFs re-expressing wt MFN1. However, MEKDN and MEKCA were ineffective in Mfn1−/− MEFs reconstituted with the non-phosporylable MFN1T562A and MFN1T562D mutants, where mitochondria were constitutively elongated or fragmented (Figures 2C and 2D). These experiments indicate that ERK modulates mitochondrial morphology by impinging on MFN1T562. Since MFN1 is crucial in trans for mitochondrial docking, the initial step of mitochondrial fusion, we set out to understand the mechanism by which MFN1T562 phosphorylation impinges on mitochondrial morphology by turning to an established assay of mitochondrial tethering. We therefore mixed mtGFP- or mtRFP-labeled mitochondria isolated from cells stably expressing wt and/or mutated MFN1, imaged them by confocal microscopy, and scored the images for the appearance of yellow mitochondria indicative of tethering of individually labeled organelles. GTP-stimulated tethering of MFN1-MFN1 mitochondria was superimposable to that recorded with the MFN1-MFN1T562A or MFN1-MFN1T562D pairs. Conversely, tethering of MFN1T562D-MFN1T562D mitochondria was significantly reduced, whereas that of MFN1T562A-MFN1T562A mitochondria was increased (Figures 2E and 2F). During the docking reactions primed by GTP, MFN1 assembles oligomers at lower M.W. formed in cis as well as in a “docking” higher M.W. (∼450 kDa) complex. Upon sucrose gradient differential centrifugation of digitonin solubilized membranes from homotipically mixed, GTP primed Mfn1−/− mitochondria reconstituted with MFN1, MFN1T562A or MFN1T562D, MFN1T562A, and MFN1 were retrieved mostly in the docking, high-molecular weight (MW) oligomer. Conversely, MFN1T562D was mostly recovered in lower MW oligomers (fractions 18 and 19), where wt MFN1 and MFN1T562A are not found (Figure 2G). Thus, MFN1T562 regulates MFN1 oligomerization and hence mitochondrial tethering and fusion.

7Since both the MEK/ERK pathway and changes in mitochondrial morphology have been implicated in the modulation of apoptosis, we wondered whether the MFN1 phosphorylation identified here could link the two pathways in cell death. When we measured apoptosis in cells lacking Mek1 or Mek2, we found that cell death was significantly reduced (Figure S3A). Indeed, MFN1T562 phosphorylation was increased in cells treated with the mitochondria-dependent intrinsic death stimulus etoposide (Figure S3B), and expression of MEKDN that, as expected, inhibited ERK phosphorylation (Figure S3C) delayed apoptosis not only by etoposide but also by the other intrinsic stimuli staurosporine and H2O2 in wt and Mfn2−/−, but not in Mfn1−/− MEFs (Figure 3A). Conversely, increased apoptosis was measured in Wt and Mfn2−/−, but not Mfn1−/− cells expressing MEKCA (Figure S3D). The finding that cytochrome c release by H2O2 was blunted by MEKDN in wt and Mfn2−/− cells, but not in Mfn1−/− MEFs (Figures 3B and 3C) further substantiated that ERK requires MFN1 to modulate apoptosis at the mitochondrial level. Cytochrome c release requires OMM permeabilization, effected by the multidomain proapoptotics Bcl-2 family members BAX and BAK. BAK oligomerization in response to recombinant caspases-8-cleaved BID (cBID) was increased in mitochondria from cells where ERK was activated by EGF treatment (Figures 3D–3F). Conversely in Mek1−/− mitochondria cBID-induced BAK oligomerization was reduced (Figure S3E), confirming a link between the MEK/ERK pathway and BAK activation, whereas MEKDN expression did not change H2O2-induced BAX activation in all the cell lines tested (Figure S4). We therefore decided to further explore how MFN1T562 modulated BAK activation. We first verified that in Mfn1−/− mitochondria reconstituted with phosphomimetic MFN1T562D cBID-induced BAK oligomerization was increased (Figures 4A and 4B). To directly verify if the increased BAK activation was operated by ERK, we turned to Mfn1−/− cells reconstituted with MFN1 or with the non-phosphorylable mutant MFN1T562A that per se does not affect BAK oligomerization (Figure 4B). While EGF treatment increased BAK oligomerization and cytochrome c release in Mfn1−/−::MFN1 cells, it was ineffective in Mfn1−/−::MFN1T562A cells (Figures 4C–4E). Along the same line, Mfn1−/−::MFN1T562A MEFs were less, while Mfn1−/−::MFN1T562D were more, susceptible to staurosporine. Both Mfn1−/−::MFN1T562A and Mfn1−/−::MFN1T562D MEFs were insensitive to MEK mutants expression (Figure 4F). Finally, in a pulldown experiment MFN1T562D bound more BAK than wt and MFN1T562A (Figure 4G), suggesting that T562 phosphorylation regulates MFN1-BAK binding and activation of the latter. In conclusion, ERK-phosphorylated MFN1 favors cytochrome c release by increasing BAK oligomerization.

8To better understand the physiological relevance of MEK/ERK modulation of mitochondrial morphology and permeabilization, we turned our attention to oxidative stress, during which MEK/ERK are activated. Indeed, H2O2 induced rapid ERK phosphorylation in MEFs (data not shown) before the observed mitochondrial fragmentation, which was counteracted by the expression of a non-phosphorylable MFN1T562A mutant (Figures S5A and S5B). Comforted by these results, we turned our attention to oxygen glucose deprivation (OGD), an established in vitro model of ischemic damage where the MEK/ERK pathway is activated and contributes to primary cortical neurons damage. First we addressed if the MEK/ERK pathway and MFN1T562 participated in the regulation of mitochondrial morphology in neurons. Mitochondria appeared as short rods in the soma and in the neurites of primary cortical neurons. Enforced MFN1 expression triggered mitochondrial elongation, leading to an intricate network of long organelles in both soma and neurites. EGF treatment resulted in mitochondrial fragmentation in both soma and neurite. Conversely, in neurons expressing the non-phosphorylable MFN1T562A mutant, upon EGF treatment the mitochondria remained elongated and evenly distributed (Figures S5C and S5D), suggesting that also in neurons ERK activation impinges on mitochondria via MFN1T562 phosphorylation. We next verified whether this pathway was operational during OGD. Mitochondria appeared elongated and retained cytochrome c in healthy β-tubulin III-positive neurons from primary cortical cultures (top panels in Figures 5A and 5B). During OGD, the observed massive mitochondrial fragmentation was accompanied by cytochrome c release and followed by apoptosis, as indicated by the neuronal TUNEL positivity. MFN1 expression inhibited the mitochondrial phenotypes and cell death (bottom panels in Figures 5A and 5B and quantification in Figures 5C–5E). Of note, while EGF treatment abolished the MFN1 protective effect, it did not affect neurons expressing the non-phosphorylable MFN1T562A mutant (Figure 5), suggesting that active ERK impinges on MFN1T562 also during neuronal OGD. Our results indicate that MFN1 phosphorylation participates in relevant models of neuronal death where the MEK/ERK pathways are turned on.

9Here we demonstrate that the outer mitochondrial membrane fusion protein MFN1 is phosphorylated by ERK to regulate mitochondrial morphology and apoptotic permeabilization. Our results indicate that when MFN1 is phosphorylated by ERK, mitochondria fragment and BAK oligomerization is favored, increasing susceptibility to apoptotic stimuli.

10Considering that cytosolic DRP1 needs signals to translocate, form oligomers, and constrict mitochondria, it is not surprising that post-translation modifications including phosphorylation at different sites and SUMOylation tightly regulate it. Less is known on how components of the mitochondrial fusion machinery respond to cellular cues. OPA1 that sits at the interface between mitochondrial fusion and apoptosis, regulating fusion as well as cristae shape and remodeling, is post-transcriptionally cleaved by different proteases that produce short versions required for its biological activities. Conversely, post-translational modifications in MFNs are hard to retrieve: given their membrane localization, they are difficult to immunopurify and therefore to analyze by mass spectrometry. Despite these difficulties, MFNs have been found to be ubiquitinated, even if the exact site of ubiquitination has not yet been identified. Interestingly, Mfn is a substrate of the E3 ubiquitin ligase Parkin, mutated in Parkinson’s disease (PD), and Parkin mutations associated with genetic forms of PD compromise the ability to ubiquitinate Mfn, implying Mfns in the function of dopaminergic neurons and in pathogenesis of PD. More recently, MFN2 was also found to be phosphorylated: Pink1 phosphorylates MFN2 to convert it into a mitochondrial receptor for Parkin and stress-induced phosphorylation by JNK causes MFN2 degradation through the ubiquitin-proteasome pathway. However, whether the core mitochondrial fusion protein MFN1 is also post-translationally modified was unclear. Large-scale phosphorylation analysis of whole organs in mouse identified S590 as a site of MFN1 phosphorylation. However, phosphorylation of this site was restricted to kidney and lung, where the same site in MFN2 was also found to be phosphorylated, suggesting that S590 could be a tissue-specific regulator of mitochondrial OMM fusion. Conversely, our results indicate that MFN1 (and not MFN2) is specifically phosphorylated by the MEK/ERK pathway, suggesting that mitochondrial morphology is a previously unidentified target of this pleiotropic regulatory signaling cascade. Indeed, the OMM localization of MFNs appear ideal to relay cellular cues to mitochondria: the MFNs two heptad repeat (HR) and GTPase domains are exposed to the cytosolic signaling cascades, making mitochondrial fusion an attractive mechanism to regulate organelle shape, localization, and physiology in response to intra- or extra-cellular stimuli. Mechanistically, antiparallel interactions of MFN1 HR2 domains in trans constitute the basis for mitochondrial tethering. The phosphorylation site identified here is located; however, in the HR1 region and its mutation affects the oligomerization and docking ability of MFN1, raising the possibility, not investigated by Koshiba and colleagues, that also HR1 is crucial for MFN1-dependent mitochondrial tethering. During cellular stress, docking-competent MFN1 oligomers are reduced, thereby impairing mitochondrial tethering and fusion. Of note, the same residue is conserved in MFN2 (where the sequence is closer to a canonical ERK phosphorylation site), but there it was not retrieved to be phosphorylated. If the lack of phosphorylation is determined by different topology or subcellular localization of Mfn2 is unknown, but worth testing, especially in the light of the functional difference between the two Mfns.

11While DRP1 post-translational modifications are well established players in apoptosis, necrosis, and autophagy, our knowledge of how mitochondrial fusion is modulated during cell death is scarcer. Fusion inhibition is an established consequence of apoptosis induction, and it occurs around the time of BAX activation. However, how fusion is blocked is not fully understood; OMM permeabilization with the release of the soluble fraction of OPA1 and the proteolytic cleavage of the long OPA1 forms as a consequence of mitochondrial dysfunction could explain mitochondrial fusion inhibition only during late apoptosis. Proapoptotic BCL-2 family members appear to participate in the earlier inhibition of fusion, occurring before or around the activation of BAX. While at rest, BAK and BAX facilitate MFN2 assembly, promoting its activity; during apoptosis, BAK dissociates from MFN2 and associates with MFN1. Our results suggest that this increased BAK-MFN1 association could depend on the phosphorylation status of the latter. In fact, the Mfn1T562D mutant that mimics MFN1 phosphorylation by ERK binds more avidly to BAK, favoring its oligomerization, and causes mitochondrial fragmentation, perhaps by “stealing” MFN1 from the fusion reaction. Similarly EGF, which activates ERK, increases BAK oligomerization in a Mfn1T562-dependent manner, placing MFN1 at the crossroad between extracellular signals and activation of the mitochondrial pathway of apoptosis.

12During apoptosis, blockage of DRP1-dependent fragmentation is protective. In addition to DRP1, our data indicate a role for the ERK-mediated inhibition of fusion in mitochondrial fragmentation. During apoptosis induced by the intrinsic stimuli tested here, ERK is phosphorylated (not shown), as previously reported in the case of oxidative stress, but its role during apoptosis remains unclear. Our model indicates that the activation of MEK/ERK inhibits the pro-fusion function of MFN1 and triggers its association with BAK, facilitating its oligomerization and therefore cytochrome c release and cell death. Our data therefore place MFN1 in the core apoptotic pathway of OMM permeabilization. Whether phosphorylated MFN1 facilitates the dissociation of BAK from BCL-2, or it shifts the equilibrium of the BH3-only BAK reaction toward their interaction, remains to be clarified.

13The effect of phosphorylated MFN1 extended beyond cellular stress. Since H2O2-induced mitochondrial fragmentation requires MFN1 phosphorylation, we investigated oxygen-glucose deprivation in primary cortical neurons, a condition mimicking disease-associated tissue damage and pathological ROS production. Our findings that MFN1 phosphorylation is involved in the cellular and neuronal response to oxidative stress and OGD extend the role of this modification to a relevant pathological condition. Finally, our results provide a conceptual framework for the analysis of mitochondrial fusion in the many kinds of cancer where the MEK/ERK pathway is constitutively upregulated.

14FLAG-MFN1 and FLAG-MFN2 plasmids were a gift from Dr. Yuka Eura (Kyushu University). PBabe-MEKDN and PBabe -MEKCA were a gift from Prof. Chris Marshall (Institute of Cancer Research, London, UK). Amphotrophic viruses were generated by co-transfecting the HEK293-packaging cell lines with the packaging vector pIK and the required pMIG constructs as previously described. Viral supernatants were retrieved and used to transduce MEFs in the presence of 4 μg/ml of hexadimethrine bromide (Sigma). Following an overnight transduction, cells were selected on 5 μg/ml puromycin or 400 μg/ml hygromycin. Details on the primers used to generate FLAG-MFN1 mutants by site directed mutagenesis can be found in the Supplemental Information.

15Simian virus 40 (SV40)-transformed wild-type, Mfn1−/−, and Mfn2−/− MEFs were from D. Chan (California Institute of Technology) and cultured as described previously. Wild-type and Bax−/−Bak−/− MEFs were cultured as described previously. Transfection of MEFs with DNA was carried out using Transfectin (Biorad) according to the manufacturer’s instructions. When indicated, cells were transfected 24 hr after plating.

16Primary cultures of cortical neurons from P0 Wistar rats were prepared by modifying a previously described method. The brain tissue was minced and trypsinized (0.1% for 15 min at 37°C), triturated, plated on poly-ornithine-coated coverslips, and cultured in Neurobasal-A medium (Invitrogen) supplemented with B-27 (Invitrogen) and L-glutamine (2 mM). For immunocytochemistry experiments, 1 × 105 neurons were plated on 13-mm glass coverslips. Cultures were maintained at 37°C in a humidified atmosphere of 5% CO2 and 95% air, fed twice a week, and maintained for a minimum of 10 days before use. The procedures for OGD, mitochondrial morphology analysis, and TUNEL staining can be found in Supplemental Information.

17MEFs treated as indicated were stained with propidium iodide (PI) and Annexin-V-FITC (Bender MedSystem). Where indicated, cells were cotransfected with pEGFP and the indicated vector. After 24 hr cells were treated as described and stained with Annexin-V-PE (Bender MedSystem) according to the manufacturer’s protocol. Cell death was measured by flow cytometry (FACSCalibur) as the percentage of Annexin-V-positive events in the GFP-positive population and viability as the percentage of Annexin-V-negative, PI-negative cells for transfected and untransfected cells, respectively.

18For immunoprecipitation experiments, cells were lysed in CPBS buffer (6 mM CHAPS in PBS [pH 7.4]). Lysates were incubated with the indicated antibody (1:50, 14 hr, 4°C), and the protein-antibody complex was precipitated by centrifugation after incubation with protein G-coated magnetic beads (Dynal, 2 hr, 4°C). The immunoprecipitated material was washed twice in CPBS and resuspended in SDS/PAGE loading buffer (NuPAGE), boiled, and loaded on 4%–12% gels (NuPAGE). For phosphorylation studies, total cell lysates were loaded on a phosphoprotein binding column (QIAGEN) as previously described. Flow-through (unphosphorylated) and eluted (phosphorylated) proteins were collected and concentrated and 20 μg of proteins were separated by 4%–12% SDS-PAGE.

19Cell extracts (10 mg) from Mfn1−/− and Mfn2−/− MEFs transfected with Flag-MFN1 and Flag-MFN2, respectively, were immunoprecipitated using anti-Flag antibody. The immunoprecipitates were separated by 4%–12% Tris-MOPS SDS-PAGE (NuPage, Invitrogen) and stained with Simply Blue Safe Stain (Invitrogen). Bands corresponding to Flag-MFN1 and Flag-MFN2 were excised and reduced with TCEP, alkylated with iodoacetamide and digested with trypsin or AspN. Details on liquid chromatography-mass spectrometry analysis can be found in Supplemental Information.

20Mitochondria were isolated by standard differential centrifugation in isolation buffer (IB) as described in. For BAK oligomerization assay, mitochondria (0.5 mg/ml) were incubated in experimental buffer (EB: 125 mM KCl, 10 mM Tris-MOPS [pH 7.4], 1 mM Pi, 5 mM glutamate, 2.5 mM malate, and 10 μM EGTA-Tris [pH 7.4]), and treated as described at 25°C. At the indicated time, mitochondria were pelleted by centrifugation at 12,000 × g at 4°C for 3 min and resuspended in the same volume of EB. For crosslinking, mitochondria were incubated with 1 mM (30 min, 37°C) BMH (Pierce), and the reaction was quenched with 20 mM β-mercaptoethanol (15 min, 25°C).

21Sucrose-gradient separation of MFN1-containing complex was carried out following. Mitochondria isolated from Mfn1−/− MEFs reconstituted as indicated were incubated in EB (0.5 mg/ml) containing 1 mM GTP for 30 min at 25°C and solubilized with 1% digitonin for 30 min at 4°C. A discontinuous sucrose density gradient containing 5%–20% sucrose solutions was prepared by layering successive decreasing sucrose densities solutions (5%, 7.5%, 10%, 12.5%, 15%, 17.5%, and 20%) upon one another. The cleared mitochondrial lysate was layered on 7 ml of the sucrose gradient and centrifuged at 174,000 × g for 12 hr at 4°C (Beckman, 100Ti). The centrifuged solution was separated into 22 fractions, and which fractions MFN1 eluted in were checked. After confirming the elution of MFN1 in fractions 12–19, these fractions were immunoblotted using the indicated antibodies.

22For immunoblotting, proteins were separated by 7% Tris-acetate, 3%–8% Tris-acetate or 4%–12% Tris-MOPS SDS-PAGE (NuPage, Invitrogen); transferred onto PVDF membranes (Millipore); probed using the indicated primary antibodies; and isotype-matched secondary antibodies were conjugated to horseradish peroxidase (Amersham) and detected using ECL (Amersham). The following antibodies were employed: rabbit anti-ERK (1:1,000, Cell Signaling); rabbit anti-pERK (1:1,000, Cell Signaling), mouse anti-COX IV (1:1,000, Molecular Probes); mouse anti-MFN2 (1:1,000, Abnova); mouse anti-ACTIN (1:5,000, Chemicon) chicken anti-MFN1 (1:500, Abcam); rabbit anti-BAK (1:1,000, Millipore); mouse anti-FLAG (1:1,000, Sigma), p-MFN1 (custom made from 21st century biochemical) Densitometric quantification of western blots was performed using the Gel Pro Analyzer 4 or ImageJ softwares.

23For cytochrome c immunostaining, 2 × 104 MEFs of the indicated genotype grown on 13 mm round coverslips were transfected with as indicated and after 24 hr incubated as detailed in the text, fixed for 30 min at room temperature with 3.7% (V/V) ice-cold formaldehyde, permeabilized for 20 min with 0.01% (V/V) ice-cold Nonidet P-40, and incubated for 15 min with a 0.5% solution of BSA. Cells were then sequentially incubated for 60 min at 37°C with a mouse monoclonal anti-cytochrome c antibody (PharMingen, CA, clone 6h2.B4) and for 15 min at 37°C with a FITC-conjugated goat anti-mouse IgG.

24For BAX translocation, cells were transfected with mtRFP, fixed in 2% paraformaldehyde for 15 min at room temperature, and then washed five times with PBS. Samples were permeabilized in 0.5% CHAPS for 30 min, followed by incubation in blocking buffer (3% BSA in PBS) for 1 hr at room temperature. Primary antibody (anti-BAX antibody 6A7; 5 μg/ml in blocking buffer) was added and incubated overnight at 4°C, after which Alexa 488-conjugated goat anti-mouse secondary antibody (diluted 1:300 in blocking buffer) was added for 1 hr at room temperature.

25For confocal microscopy imaging of live cells, 1.8 × 105 cells seeded onto 24-mm round glass coverslips transfected and treated as indicated were incubated in Hank’s Balanced Salt Solution (HBSS) supplemented with 10 mM HEPES and coverslips were placed on the stage of a Zeiss LSM 510 inverted microscope. Cells expressing mtYFP and mtRFP were excited using the 488 nm or 543 nm line of the Argon laser using a 63c × c1.4 NA Plan Apochromat objective (Zeiss).

26For the mitochondrial docking assay, mitochondria from cells of the indicated genotype expressing mtRFP or mtYFP were isolated and resuspended in EB (0.5 mg/ml) at 25°C for 30 min with 1 mM GTP unless indicated. Mitochondria expressing mtRFP or mtYFP were mixed in 1:1 ratio, centrifuged, and incubated for 15 min. Mixed mitochondria were mounted on a 24-mm round coverslip, allowed to attach for 15 min, placed on the stage of a Leica SP5 confocal inverted microscope and excited using the 488 nm or 543 nm line of an Ar laser using a 63c × c1.4 NA Plan Apochromat objective (Leica).

27Data are presented as mean ± SEM of the indicated number of independent experiments. p values refer to two tailed Student’s t tests between the indicated samples.