Activated Mitofusin 2 Signals Mitochondrial Fusion, Interferes with Bax Activation, and Reduces Susceptibility to Radical Induced Depolarization* □ S

Mitochondrial fusion in higher eukaryotes requires at least two essential GTPases, Mitofusin 1 and Mitofusin 2 (Mfn2). We have created an activated mutant of Mfn2, which shows increased rates of nucleotide exchange and decreased rates of hydrolysis relative to wild type Mfn2. Mitochondrial fusion is stimulated dramatically within heterokaryons expressing this mutant, demonstrating that hydrolysis is not requisite for the fusion event, and supporting a role for Mfn2 as a signaling GTPase. Although steady-state mitochondrial fusion required the conserved intermembrane space tryptophan residue, this requirement was overcome within the context of the hydrolysis-deficient mutant. Furthermore, the punctate localization of Mfn2 is lost in the dominant active mutants, indicating that these sites are function-ally controlled by changes in the nucleotide state of Mfn2. Upon staurosporine-stimulated cell death, activated Bax is recruited to the Mfn2-containing puncta; however, Bax activation and cytochrome c release are inhibited in the presence of the dominant active mutants of Mfn2. The dominant active form of Mfn2 also protected the mitochondria against free radical-induced permeability transition. In contrast to stauro-sporine-induced outer membrane permeability transition, pore opening induced through the introduction of free radicals was dependent upon the conserved intermembrane space residue. This is the first evidence that Mfn2 is a signaling GTPase

The mitochondria sit at the crossroad of hundreds of chemical reactions that are essential for the life and death of a cell. The dynamic behavior of these organelles has only just begun to be examined, and the implications of steady-state fission, fusion, motility, and cristae remodeling events in the control of the mitochondrial activity are not yet known. Studies in different model organisms are addressing this question by investigating the molecular mechanisms that govern mitochondrial dynamics to gain insights into the physiological triggers and consequences of these events. Mitochondrial fusion in mammalian cells requires at least two essential outer membrane GTPases, Mitofusin 1 (Mfn1) 1 and Mitofusin 2 (Mfn2) (1)(2)(3)(4)(5)(6). These proteins span the outer membrane twice, and in addition to their amino-terminal GTPase domain, they have two conserved hydrophobic heptad repeats, HR1 and HR2, which are exposed to the cytosol (2,7). The second HR2 domain of Mfn1 has been shown to facilitate mitochondrial tethering, and the crystal structure of the purified HR2 domain demonstrates that it can form a 100 Å antiparallel structure that could bind in trans to bridge mitochondria (7). Another important region of Mfns is their short, 2-3-amino acid intermembrane space loop, which contains a highly conserved tryptophan residue. In yeast, this region of the protein is required for mitochondrial fusion and has been shown to anchor Fzo1p to sites of membrane contact between the inner and outer membrane (8). Although Mfn1 and Mfn2 are 60% identical, recent evidence with both in vitro mitochondrial docking assays and in rescue experiments of Mfn1 knock-out cells has shown that Mfn1 appears to play a more direct role in mitochondrial docking (7,9) and that it functions in cooperation with the intermembrane space dynamin like GTPase Opa1 (autosomal dominant optical atrophy 1) (10). The role of Mfn2 in mitochondrial fusion has remained elusive, although it is clearly required for fusion and can be found in heterodimeric complexes with Mfn1. Recent studies have shown that the nucleotide binding and hydrolysis properties of the two Mfn proteins are distinct (9), consistent with the idea that the two GTPases regulate different steps along the fusion pathway (5). These steps may include the processes that drive mitochondrial motility, tethering, assembly of a fusion pore to facilitate lipid bilayer mixing and eventually leading to inner mitochondrial membrane fusion. Given the complexity of these molecular requirements, the Mfn proteins do not act alone, and studies in yeast have identified a number of additional proteins required for mitochondrial fusion, including the outer membrane protein, Ugo1p (11)(12)(13), the inner membrane serine protease Rbd1p (14 -16), an F-box-containing protein Mdm30p (17), along with other candidates like Mdm35p, Mdm34p, and Mdm39p (18).
One of the outstanding questions in the field of mitochondrial dynamics remains the physiological importance of mitochondrial fission and fusion under steady-state conditions. Knock-outs of either Mfn1 or Mfn2 are embryonic lethal (6), demonstrating an essential role of mitochondrial fusion for viability. In addition, mutations within the Mfn2 gene have been found in patients suffering from Charcot-Marie-Tooth neuropathy type 2A, and six of seven of these mutations were found within the conserved GTPase region (19,20). Interestingly, evidence that Mfn2 may exhibit intracellular signaling activity has come from one study that identified the rat Mfn2 (called hyperplasia suppressor gene HSG) as an important antiproliferative protein, which interferes with the Ras pathway and blocks signaling from growth factor receptors at the plasma membrane (21). Although the mechanism for this inhibition is unknown, these findings suggest that Mfn2 and/or the morphological state of the mitochondria is highly integrated into cellular signaling cascades.
Another example of how the mitochondrial morphology is integrated into cellular signaling events is the growing evidence for a role of mitochondrial dynamics in the progression of apoptosis. For example, two of the proteins required for mitochondrial fission, Fis1p and DRP1, are also essential for programmed cell death (22)(23)(24)(25). Fis1p knock-down by small interfering RNA blocks recruitment and activation of Bax at the surface of mitochondria after a death stimulus, indicating an essential role for this small integral membrane protein in apoptosis (25). Similarly, although loss of DRP1 does not dramatically interfere with Bax activation, cytochrome c release is partially inhibited, and mitochondrial fission is blocked in these cells (25,26). In addition, small interfering RNA knockdown of Opa1, a protein required for mitochondrial fusion, results in fragmented mitochondria, which are highly sensitized to the loss of electrochemical potential and cytochrome c release (25,27). In contrast, overexpression of the two mitofusin proteins together provides some protection against different apoptotic stimuli (28). These recent data highlight the dual roles of mitochondrial GTPases in the regulation of both mitochondrial dynamics and the mitochondrial contribution to programmed cell death.
Given the increasing evidence that Mfn1 plays a direct role in mitochondrial tethering (5,9), we have specifically investigated the function of Mfn2 in the process of mitochondrial fusion and examined further how the GTPase activity of Mfn2 might contribute to the regulation of programmed cell death.

EXPERIMENTAL PROCEDURES
Construct Preparation and Reagents-The cDNA encoding human Mfn2 (KIAA0214) was graciously provided by Kazusa DNA Research Institute, Japan. Mfn2 cDNA was PCR amplified using standard protocols, for insertion into pECFP-C1 (Clontech) and pcDNA3.1 (Invitrogen) with BamHI and HindIII restriction sites. Mfn2 RasG12V -CFP was prepared with QuikChange mutagenesis (Stratagene, La Jolla, CA) using pECFP-C1-Mfn2 as the template and a set of oligonucleotides designed to replace amino acids GRTSNGKS with GAVGVGKS. The restriction site NarI was introduced into the primers for screening purposes. The pcDNA3-GST, Mfn2-His 6 , and Mfn2 RasG12V -His 6 construct vectors for protein purification from transfected cell lysates were also prepared using subcloning techniques, and all sequences used in this work were confirmed. Mfn2 W631P -CFP was prepared with QuikChange mutagenesis using pECFP-C1-Mfn2 as the template and a set of oligonucleotides designed to replace the tryptophan amino acid at position 631 with a proline. The ApaI restriction site was introduced into the primers for screening purposes. Mfn2 RVWP -CFP was prepared by isolating the DNA fragment encoding amino acids 1-431 of Mfn2 RasG12V -CFP by digestion with HindIII and SalI and subcloning this fragment containing the RasG12V mutation into the Mfn2 W631P -CFP construct cut with the same enzymes, thereby replacing the wild type GTPase domain of Mfn2 W631P -CFP with the Mfn2 RasG12V mutation. The cDNA encoding DsRed2 was amplified from pDsRed2-C1 (Clontech) using primers designed for digestion with BamHI and XbaI and ligation into the pcDNA3-pOCT vector (29). Tom7-GFP was obtained from Mike Ryan, La Trobe University, Melbourne Australia. Mfn2 mouse polyclonal antiserum was generated against a mixture containing both recombinantly expressed Mfn2 (710 -757) -GST and a synthetic Mfn2 NH 2 -terminal peptide CNSIVTVKKNKRIIM-OH (Dalton Chemical Laboratories, Toronto, Canada), conjugated to 5 mg of maleimideactivated keyhole limpet hemocyanin. Antiserum was generated following a 56-day standard immunization protocol. Polyclonal antibodies against fluorescent proteins (anti-FP) used for immunoelectron microscopy were purchased from Clontech. Dihydroethidium (Molecular Probes, Eugene, OR) was utilized to determine steady-state radical levels in transfected cells. Monoclonal 7H8.2C12 anti-cytochrome c antibodies were obtained from BD Biosciences, and rabbit polyclonal anti-Bax antibodies were obtained from Upstate Biotechnology cell signaling solutions. Alexa Fluor 350 and 594 goat anti-mouse or rabbit secondary antibodies from Molecular Probes were used for staurosporine (STS) experiments. Z-VAD was obtained from Enzyme Systems Products (Aurora, Canada), and STS was obtained from Sigma.
Transfection and Imaging of COS-7 Cells-Transfection and imaging methods were exactly as described in Ref. 29. For electrochemical potential determination, cells were incubated with 50 nM MitoFluor Red 589 (Molecular Probes) at 37°C for 20 min. Whole cell images were acquired for untransfected and transfected cells by exciting at 589 nm with the CFP/YFP/DsRed triple pass filter (Chroma, Brattleboro, VT). The presence of transfected CFP-tagged protein was confirmed by exciting at 434 nm using the same filter set. Areas of interest were selected for each cell, and total fluorescence arbitrary units were summed for each cell. The total fluorescence intensity/cell was quantified as the sum of the values of each pixel within the area of interest minus the average background signal obtained/pixel. The number of cells within each fluorescence distribution range was scored and tabulated as a percentage.
For MitoFluor Red 589 (Molecular Probes) flickering experiments, 50 nM dye was added to the chamber medium and incubated with the cells at 37°C for 20 min. After equilibration of the dye, 400 images were collected by exciting at 589 nm with the CFP/YFP/DsRed triple pass filter (Chroma) for 1 s followed by a 2.5-s delay. The presence of transfected CFP-tagged protein was confirmed as indicated above. The time series were analyzed, and each frame where all mitochondrial dye was released from the whole cell was plotted.
To determine steady-state radical loads, cells were incubated with 5 M dihydroethidium at 37°C for 20 min. Dihydroethidium oxidation by superoxide to ethidium was visualized by excitation at 547 nm. Whole cell imaging and fluorescence quantification for untransfected and transfected cells were performed as indicated above. Mant GMP-PNP Binding Assay-COS-7 cells were transfected with pOCT-CFP, Mfn2-CFP, Mfn2 RasG12V -CFP, and Rab5-CFP fusion constructs. The day after transfection, the cells were harvested by trypsin treatment, washed with PBS and with nucleotide binding buffer (220 mM mannitol, 68 mM sucrose, 200 mM NaCl, 2 mM MgCl 2 , 0.5 mM EGTA, 2.5 mM KH 2 PO 4 , 10 mM Hepes, pH 7.4, 1 mg/ml bovine serum albumin) containing protease inhibitors. Cells were then broken in a cell cracker, and the whole lysate was centrifuged at 10,000 rpm to concentrate heavy membrane fractions. Pellets were resuspended in binding buffer, and 50-l aliquots were mixed with Mant GMP-PNP (1 M final concentration, Molecular Probes) and incubated at 37°C for different times. After incubation the aliquots were scanned for emission fluorescence of Mant nucleotides in a QuantaMaster 6000SE (Photon Technology International, London, Canada) (excitation at 360 nm). The peak emission at 448 nm was recorded, and the background fluorescence of the nucleotide alone was subtracted. Each value was also normalized for total protein concentration in the sample (determined using the DC protein assay (Bio-Rad)) and for the level of recombinant protein expression by measuring the CFP signal obtained in the fluorometer upon excitation at 434 nm, emission at 477 nm.
For the purification of GST, Mfn2-GST, and Mfn2 RasG12V -GST from transfected cell lysates, 10-cm dishes of COS-7 cells were transfected using Lipofectamine 2000 (Invitrogen), and after 12 h they were treated with trypsin, washed, and lysed with TNE buffer (25 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5 mM, EDTA, 1 mM dithiothreitol, 60 g/ml chymotrypsin, 1 M leupeptin, 25 g/ml antipain, 2 g/ml aprotinin, 40 M 4-aminidophenylmethane-sulphonyl fluoride, 1 mM pepstatin A), containing 1% Triton X-100 for 2 h at 4°C. Lysates were adjusted to 40% sucrose and centrifuged at 70,000 rpm for 1 h at 4°C. Supernatants were diluted to 10% sucrose with TNE ϩ Triton X-100 buffer and were incubated overnight with glutathione-Sepharose beads at 4°C. Beads were washed, and GST fusion proteins were eluted with 50 mM reduced glutathione in TNE buffer ϩ 10% sucrose. Eluted aliquots (50-l elutions) were incubated in duplicate with 0.2 M Mant GMP-PNP for 15 min. at 37°C and assayed for fluorescence (excitation, 360 nm; emission, 448 nm). Normalization was done by assaying the enzymatic activity of GST within each aliquot, with 1-chloro-2,4-dinitrobenzene (Sigma) and glutathione as substrates, with formation of a product with absorbance at 340 nm (Amersham Biosciences), being the product formation rate proportional to the amount of GST in the aliquots. Known concentrations of bacterially purified GST were used as a standard in this assay. Bacterially expressed Rab5-GST was purified as described previously (30).
GTP Hydrolysis Assay-COS-7 cells were transfected with Mfn2-His 6 , Mfn2 RasG12V -His 6 , or LacZ-His 6 , and the following day cells were washed with PBS, harvested by scraping, and centrifuged. Cell pellets were resuspended in a small volume of TNX solubilization buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 2% Triton X-100, 1 mM ␤-mercaptoethanol, 2 mM MgCl 2 , protease inhibitor mixture) and incubated for 1 h at 4 C. Lysates were then cleared at 70,000 ϫ g in a TLA 110 rotor for 30 min at 4°C, and supernatants were incubated with nickel-nitrilotriacetic acid-agarose beads for 1.5 h at 4°C. Beads were then centrifuged and washed twice with TNX buffer containing 200 M ATP (to remove chaperones and other nonspecific proteins of the lysate) followed by two more washes with TNX buffer containing 20 mM imidazole. Proteins were eluted with series of 100 and 200 mM imidazole in TNX buffer, and concentrations were determined by the Bio-Rad method. 10 g of the different eluted proteins were preloaded with 1 l of [␥-32 P]GTP (2,000 Ci/mmol) in exchange buffer (50 mM Tris, pH 7.4, 100 mM NaCl, 1% Triton X-100, 2 mM MgCl 2 , 1 mM dithiothreitol) by incubating for 30 s at 37°C in a 140-l volume. The reaction was then diluted 10 times to trap the nucleotide in place and begin [␥-32 P]GTP hydrolysis. From this mix, aliquots were taken, in triplicate, at different time points (0, 1, 2, 5, and 10 min), applied to nitrocellulose filters in a vacuum manifold, and flushed immediately with 1 ml of PBS (to remove unbound nucleotide and hydrolyzed 32 P). Filters were then placed in vials and assayed for 32 P scintillation counting.
Electron Microscopy-COS-7 cells were transfected with the appropriate cDNA in 10-cm dishes with Lipofectamine 2000 for 16 h. Following this, fluorescence was first examined using the light microscope to ensure that transfection was at least 70% prior to trypsin treatment and washing of the cells in PBS. The washed cells were then pelleted and fixed in 1.6% glutaraldehyde in 0.1 M sodium cacodylate buffer prior to osmium tetroxide and uranyl acetate staining, spur resin embedding, and final thin sectioning of the blocks, and the grids were stained with lead citrate. For immunoelectron microscopy, the same transfection procedure was followed. Cells were treated with trypsin, washed in PBS, fixed in 1.6% glutaraldehyde, and centrifuged at 3,000 ϫ g for 15 min. Cell pellets were embedded in LR white (Marivac, PQ, Canada), thin sections were cut with a Leica Ultracut E ultramicrotome, immunolabeled with polyclonal anti-FP antibodies and 10 nm gold-labeled secondary antibodies (Jackson Laboratories, Bar Harbor, ME). Cells were then counterstained with lead citrate and uranyl acetate. Digital images were taken using a JEOL 1230 TEM at 60 kV adapted with a 2,000 by 2,000 pixel bottom mount CCD digital camera (Hamamatsu, Japan) and AMT software.
PEG Fusion Assay-The whole cell fusion assay was performed as described previously (4). Given that the broad excitation spectra of DsRed overlaps with the GFP excitation wavelength, special precautions were taken to ensure that the images of fused mitochondria did not include any bleed-through between the channels. For this, live cells were imaged on the Olympus IX70 microscope with a 100 ϫ objective U Plan Apochromat, NA 1.35-0.50 objective, exited at 488 nm (GFP) and 560 nm (DsRed) with the Polychrome IV monochrometer (TillPhotonics, Grä felfing, Germany) through a fluorescein isothiocyanate/Cy3/Cy5 triple pass filter (Chroma). The emitted light was filtered through an additional dual view beam splitter (Optical Insights, Santa Fe, NM) equipped with two filters HQ520/20 and D600/40 to separate the GFP and DsRed emission signals, and images were captured as described above. The transfected CFP-tagged protein was imaged using the CFP/ YFP dual pass filter (TillPhotonics) along with the beam splitter equipped with D465/30 and HQ535/30 filters (data not shown). The acquired images were saved as .tiff images and overlaid in Adobe Photoshop for image assembly.

RESULTS
Creation of a Hydrolysis-deficient Mutant of Mfn2-To determine whether nucleotide hydrolysis of Mfn2 was essential for mitochondrial fusion, we constructed a RasG12V mutant. Our purpose in designing the mutant was to maintain nucleotide binding and exchange properties of Mfn2 and yet reduce the intrinsic rate of nucleotide hydrolysis. Therefore we replaced the residues within the P-loop with those from the activated RasG12V (31-33) by replacing amino acids GRTSNGKS with GAVGVGKS (the consensus for GTP binding is GxxxxGKS (34)). We first compared the nucleotide exchange properties of this mutation (Mfn2 RasG12V ) with wild type Mfn2 and controls by using crude extracts from cells transfected with CFP fusion proteins. COS-7 cells were transfected with matrix-targeted CFP (negative control), wild type Mfn2-CFP, Mfn2 RasG12V -CFP, or Rab5-CFP as a positive control. Rab5 is the best characterized of the Rab GTPases and is known to regulate early endosome fusion (35). Regulatory GTPases of the Ras family are all characterized by their low intrinsic rates of hydrolysis and high nucleotide affinities, which results in stable nucleotide states that require accessory proteins for their activity. Given the evidence that Mfn2 exhibits low rates of nucleotide hydrolysis and high affinity (9), it is relevant to use Rab5 as a positive control in these assays. After transfection with CFPtagged constructs, the cells were harvested, broken, and the mitochondria-enriched heavy membrane fractions were incubated with the environment-sensitive nonhydrolyzable GTP analog Mant GMP-PNP (36,37). Increased Mant nucleotide fluorescence is a direct measurement of nucleotide exchange. As expected, untransfected cells, or cells transfected with the matrix CFP control plasmid, bound a constant level of Mant GMP-PNP because of the endogenous GTPases present in the extracts. Interestingly, overexpression of Mfn2-CFP did not significantly alter the basal amount of nucleotide binding in the heavy membrane fraction (Fig. 1A). Overexpression of Rab5 provided a ϳ4-fold increase in Mant GMP-PNP binding within the first 10 min after incubation at 37°C (Fig. 1A). This signal is specific because addition of unlabeled GTP competes for the binding. Notably, although transfection of Mfn2-CFP had no effect on basal nucleotide binding in this assay, transfection of Mfn2 RasG12V -CFP demonstrated a significant increase in nucleotide binding (Fig. 1A). We next isolated Mfn2 RasG12V -GST or Mfn2-GST from transfected mammalian cells using glutathione-Sepharose beads. We employed a quantitative GST enzyme assay with 1-chloro-2,4-dinitrobenzene as substrate and determined the total amount of GST fusion proteins purified to normalize/mol of purified protein (Fig. 1B). GST-tagged protein was purified from transfected cell lysates, and, as in total cell lysates, the nucleotide binding experiments demonstrate a ϳ2fold increase in activity of Mfn2 RasG12V -GST relative to the wild type protein (Fig. 1B). In addition, the stimulated exchange activity of the Mfn2 RasG12V mutant reached levels similar to those obtained with molar equivalents of recombinant Rab5-GST (Fig. 1B). Although we could detect no significant differences in nucleotide binding to Mfn2-CFP within whole cell extracts (Fig. 1A), the GST-purified Mfn2 did demonstrate nucleotide binding/exchange relative to the GST control (Fig.  1B). These biochemical data indicate that purified Mfn2 has extremely low rates of nucleotide exchange, consistent with high affinity binding. These data are consistent with published data that have also shown much stronger binding of Mfn2 to nucleotide relative to Mfn1 (9). Our data now show that mutations within the P-loop of Mfn2 increase nucleotide exchange relative to the wild type protein, thereby bringing the rates closer to the intrinsic rates observed for the Rab GTPases.
We next examined the intrinsic rates of [␥-32 P]GTP hydrolysis using the established filter-based assay quantifying the release of the labeled tertiary phosphate from the GTP-bound protein purified from transfected COS-7 cells (Fig. 1D) (38).
Purified Mfn2-His 6 hydrolyzed 7.33 ϫ 10 Ϫ3 mmol of GTP/mmol of protein/min, which is ϳ4 times faster than the rates obtained for molar equivalents of bacterial expressed Rab5 protein (1.55 ϫ 10 Ϫ3 mmol of GTP/mmol of protein/min) (Fig. 1C). Although part of this signal may be the result of complete loss/exchange of nucleotide rather than nucleotide hydrolysis, we consider this to be negligible because the rates of exchange for wild type Mfn2 are lower than Rab5 (Fig. 1, A and B). The rates of GTP hydrolysis we have determined for Rab5 are similar to previously calculated rates of 2 ϫ 10 Ϫ3 /min (38,39), further demonstrating the validity of the assay. Although the mutant Mfn2 exhibited increased rates of nucleotide exchange, Mfn2 RasG12V -His 6 demonstrated significantly reduced levels of hydrolysis (2.08 ϫ 10 Ϫ3 /min) relative to the wild type protein (7.33 ϫ 10 Ϫ3 /min) (Fig. 1C). Although we cannot exclude the contribution of trace levels of copurifying GTPases, GTP exchange factors or GTPase-activating proteins in the Mfn2 purification from cell extracts, our data clearly demonstrate that specific mutations in Mfn2 alter the nucleotide binding and hydrolysis properties of the purified protein.
Mfn2 Is Localized in Specific Subdomains along Mitochondrial Tubules in a Nucleotide-dependent Manner-We examined the cellular consequences of the GTPase mutant by transfecting COS-7 cells with wild type Mfn2-CFP, GTPase mutant Mfn2 RasG12V -CFP, or a truncation mutant completely lacking the GTPase domain Mfn2 (430 -757) -CFP. Transfection of COS-7 cells with the wild type Mfn2 resulted in increased interconnectivity among the mitochondrial reticulum (Fig. 2), which upon high levels of expression appears as a cluster (1, 2, 4). As reported previously (40), Mfn2-CFP was localized in puncta along the mitochondrial tubules. Transfection of the GTPase mutant Mfn2 RasG12V -CFP resulted in the clustering of mitochondria, which appeared to be fused even at the lowest levels of detectable expression (Fig. 2). Importantly, unlike Mfn2-CFP, Mfn2 RasG12V -CFP was distributed evenly along the surface of these clusters (Fig. 2). In contrast, transfection of the Mfn2 (430 -757) -CFP mutant resulted in mitochondria visibly fragmented into small, spherical units that cluster together in a stable manner (Fig. 2 and data not shown), similar to those found in the Mfn1 GTPase truncation mutant (7). These phenotypes are specific to Mfn2 because transfection of another mitochondrial outer membrane protein, Tom7 (24), did not cause significant mitochondrial clustering (Fig. 2). In addition, these phenotypes were not affected by the presence of the CFP tag because transfection of untagged constructs gave similar results (data not shown). We considered that the assembly of the Mfn2 puncta may depend upon the conserved IMS residues. Mammalian Mfn2 contains only 2-3 amino acids separating the two transmembrane domains, one of which is a highly conserved tryptophan residue that we replaced with a proline residue (Mfn2 W631P ). This proline residue is naturally occurring within the IMS loop in Drosophila melanogaster Fzo, indicating that this substitution should not alter the topology of Mfn2. Mfn2 W631P -CFP remains sharply localized in distinct foci; however, the tubular morphology of the mitochondria is reduced, and the organelles eventually fragment (Fig. 1). Creation of a double mutant where both the Mfn2 RasG12V and Mfn2 W631P mutations are present (Mfn2 RVWP -CFP) also resulted in a loss of Mfn2 puncta, indicating that the punctate localization does not depend upon the conserved IMS residues, rather it is determined by the nucleotide state (Fig. 2). Video analysis showed that the Mfn2-CFP puncta remained highly immotile and were not observed at sites of mitochondrial fusion, nor did they significantly colocalize with the cytoskeleton (actin filaments or microtubules), and they did not accumulate at sites of endoplasmic reticulum/mitochondrial contact (data not shown).
To ensure that the clustered mitochondria within cells transfected with Mfn2 and mutants maintained their electrochemi-cal potential, we quantified the total fluorescence intensity within cells loaded with a ⌬⌿-dependent dye and scored the results in a distribution profile as shown in Table I. In untransfected cells, the fluorescence intensity was between 1 and 3 ϫ 10 6 arbitrary units in 90% of the cells examined. Transfection of Mfn2 W631P showed some loss in total dye uptake, with 61% of the cells examined exhibiting fluorescent arbitrary units of fewer than 1 ϫ 10 6 . Transfection of the other constructs did not show significant loss in potential, but showed a broader distribution profile than untransfected cells, with many transfected cells exhibiting fluorescence units higher than normal. For example, transfection of Mfn2, Mfn2 RasG12V , or Mfn2 RVWP showed 15-20% of cells having greater than 4 ϫ 10 6 fluorescent arbitrary units. These data demonstrate that electrochemical potential is maintained, with a slight reduction in cells transfected with Mfn2 W631P .
PEG-induced Cell Fusion Demonstrates That Mfn2 RasG12V Is a Dominant Active Mutant-To assess directly the fusion competence of mitochondria in cells expressing these proteins, we employed a well characterized assay that induces fusion between whole cells transfected with different matrix marker proteins (4 -6). Mitochondria from heterokaryons transfected with marker proteins alone fused completely within 8 -12 h (Fig. 3, top left panels) (4 -6). Mitochondria from donor cells expressing Mfn2-CFP also fused with acceptor mitochondria within a similar time course (Fig. 3, middle left, n ϭ 29); however, ϳ20% of the heterokaryons expressing Mfn2-CFP scored complete content mixing within 2 h. Strikingly, fusion between mitochondria expressing Mfn2 RasG12V -CFP with acceptor GFP organelles occurred within 30 min after the addition of PEG in Ͼ90% of the observed cell fusions (Fig. 3, bottom  left, n ϭ 20). Video analysis of the PEG fusion assay shows that the Mfn2 RasG12V -CFP cluster remains immobile, whereas the GFP-containing acceptor mitochondria are motile within the heterokaryon (see supplemental video). This efficient fusion between mitochondria was highly significant because the migration of mitochondria from one cell into another is a slow process in mammalian cells. In control experiments very few mitochondria within the heterokaryon had migrated across the cell boundary in the 1st h (Fig. 3). In contrast, the rapid and/or directed migration of the GFP-expressing mitochondria toward the Mfn2 RasG12V -CFP-expressing DsRed mitochondria allowed complete fusion of all mitochondria to occur within 30 min. This indicates that the presence of Mfn2 RasG12V within one population of mitochondria initiates a cascade of events that lead to highly efficient mitochondrial fusion. Therefore, we consider the Mfn2 RasG12V to be the first characterized dominant active, GTPase-deficient mutant that stimulates mitochondrial fusion.
The Mfn2 (430 -757) -CFP construct completely inhibited the fusion, confirming the requirement for the GTPase domain (Fig. 3, top right). In addition to a role in contact site formation, previous studies with yeast Fzo1p have demonstrated the requirement for the IMS loop of Fzo1p for mitochondrial fusion (8). Consistent with this, fusion was inhibited between mitochondria containing Mfn2 W631P -CFP and acceptor GFP-containing mitochondria. Surprisingly, the inhibition of mitochondrial fusion conferred by the Mfn2 W631P mutation (Fig. 3, middle right) was rescued by the Mfn2 RasG12V mutation within the double mutant (Fig. 3, bottom right), indicating that the conserved IMS domain is not directly required for the fusion event, but likely plays a more regulatory role in the activation of Mfn2.
Fused Mitochondria Are Interconnected by Novel Membrane Elements-Ultrastructural analysis revealed that the mitochondria do not fuse into a single organelle, but instead they are interconnected through unusual membranous networks FIG. 2. Mitochondrial morphology and distribution of Mfn2 mutants. COS-7 cells were cotransfected with pOCT-YFP (left panels) and the CFP-tagged constructs indicated (middle panels). As a control Tom7-GFP was cotransfected with pOCT-DsRed2 (top panels). Images were taken from living cells 16 h after transfection. In the overlay, the matrix marker is shown in red and FP fusion proteins in green. Note the punctate appearance of Mfn2-CFP and Mfn2 W631P -CFP (arrows) compared with the smooth distribution of Mfn2 RasG12V -CFP and Mfn2 RVWP -CFP. Scale bars are 1 m. (Fig. 4A). The fused clusters in Mfn2 RasG12V -transfected cells were qualitatively similar to the wild type (Fig. 4A), indicating that the end point of the fused mitochondrial reticulum is similar, even though the PEG fusion assay demonstrated that kinetics of the fusion event are different (Fig. 3). Analysis of Mfn2 W631P -transfected cells did not reveal any interconnecting membranes between the clustered organelles. Notably, some of the mitochondria in the Mfn2 W631P -transfected cells appeared to "unravel," with membranous material emanating beyond the clear boundaries of the outer membrane (Fig. 4A). This may be caused by the loss of contact site formation (8), leading to aberrant membrane architecture. Mitochondria in cells expressing Mfn2 RVWP -CFP contained stacks of parallel membranes and extensive membrane whorls that were interconnected throughout the cluster. Immunolabeling of the fused mitochondrial clusters indicates that the Mfn2 protein is found within the interconnecting membranes, demonstrating that these membranes are derived, at least in part, from the outer mitochondrial membrane (Fig. 4B). Regardless of this massive alteration in membrane architecture, the cristae and electron dense matrix compartments appeared normal, consistent with their ability to maintain electrochemical potential (Table I) and stimulate mitochondrial fusion (Fig. 3). Mitochondria within cells expressing Mfn2 (430 -757) -CFP were docked together within a cluster, consistent with the proposed tethering func-tion of the HR2 domain (7). However, there was no fused membrane material between the mitochondria (Fig. 4A), demonstrating that the fused membranes do not arise simply because of nonspecific mitochondrial clustering.
Activated Mfn2 Represses Bax Activation, Cytochrome c Release, and Permeability Transition-Given the increasing involvement of dynamic changes in mitochondrial morphology during the progression of apoptosis, we next examined the specific consequence of Mfn2 activation on the mitochondrial response to two different types of stimuli. Stimulation of programmed death by STS treatment resulted in the efficient recruitment and activation of Bax to the mitochondria, as revealed by an antibody that specifically recognizes the conformationally active form of Bax (41). As expected (42,43), the amount of Bax activation correlated with the amount of cytochrome c release (ϳ50% by 3 h, Fig. 5, A and B, n ϭ 379). As a negative control, we transfected cells with DRP1(K38E) (44 -46), which has been shown to block mitochondrial fission and inhibit cytochrome c release from mitochondria (26,47). As expected, only 3% of these cells showed cytochrome c release after 3 h of STS treatment, with ϳ10% of cells showing Bax activation (n ϭ 66). By 5 h of treatment, this level of Bax activation increased without a release of cytochrome c (data not shown), as has been reported previously within cells transfected with DRP1(K38A) (40). Transfection of Mfn2-CFP  showed a reduction of STS-induced Bax activation and cytochrome c release (Fig. 5A), with only ϳ35% of cells showing susceptibility to STS treatment (Fig. 5B, n ϭ 92). Interestingly, in the 35% of STS-sensitive cells, many of the Mfn2-CFPcontaining puncta colocalized with activated Bax (insets, Fig.  5A). Mfn2 RasG12V -CFP and Mfn2 RVWP -CFP both repressed Bax activation and cytochrome c release, where only ϳ20% of cells were susceptible to STS treatment (Fig. 5B, n ϭ 69 and 72, respectively). In contrast, the fusion-incompetent Mfn2 W631P -CFP did not provide protection against Bax activation or cytochrome c release (Fig. 5A, n ϭ 73). As with Mfn2-CFP, activated Bax often colocalized with Mfn2 W631P -CFP-containing puncta (see insets, Fig. 5A). Cells transfected with the dominant negative mutant Mfn2 (460 -757) -CFP were highly sensitive to STS treatment, with 80% of cells showing complete cytochrome c release by 3 h (Fig. 5A, n ϭ 95). Oddly, the amount of cytochrome c release in this condition (ϳ80%) did not mirror the amount of Bax activation (ϳ40%), which could be explained because cytochrome c was released in Mfn2 (430 -757) -CFP-transfected cells in a Bax-independent manner, without any death stimuli (Fig. 5A). As expected, transfection of the other Mfn2 constructs showed normal, mitochondrial cytochrome c staining in the absence of any death stimuli (data not shown). Taken together, the data show that activated Mfn2 is a repressor of Bax activation and outer membrane pore formation.
Given that activated Mfn2 can provide protection against Bax activation and cytochrome c release triggered by external apoptotic stimuli, we next wanted to investigate whether Mfn2 could also protect the mitochondria from damage induced by internal metabolic stress. We therefore adapted an assay that artificially produces free radicals within the matrix of the mitochondria, thereby triggering permeability transition and permeability transition pore opening in the absence of Bax activation (48 -50). This approach allowed us to damage the mitochondria from the matrix side, simulating physiological systems where mitochondrial radical loads are increased because of excessive respiratory activity or other stresses. The ⌬⌿-sensitive dye tetramethylrhodamine ester and its derivatives, such as MitoFluor Red 589, become photoactivated upon exposure to light and subsequently produce free radicals within the matrix of the mitochondria. With the increasing accumulation of free radicals, the permeability transition pore opens, and protons become equilibrated across the inner membrane (49,50). Concomitant with the loss in ⌬⌿, the potential-sensitive dye is redistributed within the cell, a process observed using time lapse video microscopy. The mitochondria then regain ⌬⌿, which allows the reuptake of the potential-sensitive dye, and in time lapse they appear to "flicker" until they become terminally depolarized. 400 images were collected over 25 min by exciting MitoFluor Red 589 for 1,000 ms followed by a 2,500-ms delay. Fig. 6 shows that 84% of the untransfected cells incubated with 50 nM MitoFluor Red 589 reached terminal depolarization within the first frame category (frames 1-250, n ϭ 56). Mfn2-CFP-expressing cells revealed a moderate, but significant protection from free radical-induced damage, with a varied distribution of rates of dye loss across the four frame categories (p Ͻ 0.003 Fisher's exact test, n ϭ 19). Similar to the protection granted against STS treatment, 62% of the cells containing a fused mitochondrial reticulum induced by transfection of Mfn2 RasG12V -CFP were significantly protected against free radical-induced damage until the fourth frame category (350 -400 frames, p Ͻ 0.001 from Fisher's exact test, n ϭ 21). In contrast, the mitochondria within cells expressing Mfn2 W631P -CFP rapidly reached their damage threshold with 92% of cells losing their dye (n ϭ 13) within the first frame category, consistent with their susceptibility to STS. Unexpectedly, although the morphological phenotypes were similar between Mfn2 RVWP -CFP and Mfn2 RasG12V , and they both significantly protected mitochondria against STS treatment, the double mutant was not protected from the radical-induced damage, with 78% (n ϭ 18) of mitochondria terminally depolarized in the first frame category. These data suggest that both the activation of Mfn2 GTPase domain and the IMS tryptophan residue are essential for permeability transition. Fragmented mitochondria expressing Mfn2 (430 -757) -CFP but did not affect the loss of dye in these experiments because 82% (n ϭ 22) of the mitochondria were terminally depolarized within the first frame category. These effects are not caused by differences in either MitoFluor dye loading or absolute free radical production because fluorescent quantification of initial MitoFluor dye uptake and the radical dye dihydroethidium shows no differences between the transfected versus untransfected controls (Tables I and II). The specificity of this assay was further verified by creating a fused reticulum after transfection with the dominant interfering DRP1 mutant, DRP1(K38E). In contrast with the protection granted by DRP1(K38E) in terms of STS-induced cytochrome c release, the fused mitochondrial reticulum within cells expressing DRP1(K38E)-CFP were not protected from the free radical load (Fig. 6, n ϭ 26), demonstrating that a fused mitochondrial morphology alone does not inhibit permeability transition. DISCUSSION We have shown for the first time that Mfn2 exhibits properties of a signaling GTPase capable of regulating not only mitochondrial fusion, but that its nucleotide state also plays a critical role in regulating the mitochondrial response to apoptotic and free radical-induced damage. Through the creation of mutants, we have dissected the functional role of the GTPase activity and the intermembrane space domain within Mfn2. Biochemical evidence indicates that Mfn2 exhibits nucleotide properties similar to Rab5, with low intrinsic rates of hydrolysis and nucleotide exchange. These data are consistent with the recently published work showing that Mfn2 has a higher affinity to nucleotide and dramatically slower rates of hydrolysis relative to Mfn1 (9). Most importantly, we have characterized the nucleotide binding and hydrolysis properties of a mutant form of Mfn2, Mfn2 RasG12V , and showed that this mutant has slower hydrolysis and increased nucleotide exchange when compared directly with the wild type protein (Fig. 1). This mutant allowed us to examine the functional consequences of a GTP hydrolysis-deficient, dominant active form of Mfn2. Our mutants have demonstrated a number of important findings. First, the hydrolysis-deficient mutant of Mfn2 results in a dramatic stimulation of mitochondrial fusion, and the ultrastructural analysis of the fused mitochondrial clusters indi-cates a striking proliferation of interconnected membranes, shedding new light on the plasticity of the mitochondrial membranes. Furthermore, we have determined that the conserved tryptophan residue within the intermembrane space region of Mfn2 is not essential to form a fusion pore but is required to activate fusion within the context of the wild type GTPase. The mutational analysis also allowed us to determine that the punctate localization of Mfn2 (40) is regulated by the nucleotide state. The hydrolysis-deficient mutants of Mfn2 that stimulate mitochondrial fusion do not readily form puncta, whereas the mutants that inhibit mitochondrial fusion are found in these foci. Finally, the activated Mfn2 constructs significantly represses Bax activation, cytochrome c release, and free radical-induced permeability transition. Taken together, these data reveal a role for Mfn2 as a regulator of mitochondrial fusion and as a nucleotide-dependent modulator of the apoptotic response.
The PEG fusion assay demonstrates that Mfn2 is a regulatory GTPase, which, when in the activated form, is capable of signaling to neighboring mitochondria to fuse in an accelerated manner. Because Mfn2 RasG12V -CFP is anchored within the mitochondrial outer membrane of the DsRed-containing mitochondria in Fig. 3, it follows that Mfn2 RasG12V initiated cytosolic events resulting in the efficient recruitment and fusion of GFP-containing mitochondria with the Mfn2 RasG12V -CFP/ DsRed reticulum. Clearly there are a number of molecular events that contribute to this dramatic increase in fusion, including increased motility events, as well as the activation of the fusion machinery. Our data therefore indicate that Mfn2 may play a dual role, both as a direct constituent of the tethering/fusion machinery (minimally through the coiled coil do-FIG. 6. Activation of Mfn2 protects against permeability transition. COS-7 cells, as indicated, were incubated with 50 nM MitoFluor Red 589, exposed to light (see "Experimental Procedures"), and scored for terminal depolarization. The results were converted to a percent value and plotted as a frequency distribution for each transfection condition, as indicated in the graph. mains of Mfn 430 -757 ) and as a GTPase capable of initiating secondary cellular events. These secondary cellular events likely include the cooperation and/or activation of Mfn1, which has been shown to function directly with Opa1 in driving mitochondrial tethering and fusion (7,9,10). Evidence supporting a primary role for Mfn2 as a signaling GTPase has come from previous work demonstrating that mouse embryonic fibroblast cells lacking Mfn2 show a loss of long range motility events (6), consistent with a role for Mfn2 in regulating mitochondrial movement. Our data are also consistent with the recently identified role for Mfn2 as a regulator of the Ras signaling pathway (21). Given that Ras signaling occurs at the plasma membrane, the ability of Mfn2 to invoke a signaling cascade would provide a mechanism for it to act upstream of events at a separate intracellular location.
What are the physiological triggers that might activate Mfn2? Because Mfn2 W631P -CFP results in fusion inhibition, we consider that this IMS residue is required for the stimulation of GTP nucleotide exchange in Mfn2. In this case, the GTPase domain within the Mfn2 W631P protein would remain in the inactive, GDP-bound state locked within the puncta where fusion could not be initiated, as observed in Figs. 2 and 3. Because the IMS tryptophan residue is so critical in providing protection against free radical-induced permeability transition (Fig. 6), we speculate that internal mitochondrial signals would communicate with Mfn2 through this residue to initiate mitochondrial fusion in response to local free radical production and other forms of metabolic damage. Mitochondrial fusion may then buffer and rescue local radical damage by sharing scavengers and other metabolites between healthy and damaged organelles. Under physiological conditions, this novel link between mitochondrial fusion and permeability transition pore inhibition suggests a robust system to protect against accumulated cellular damage that could potentially lead to inopportune cell death. Consistent with this, mouse embryonic fibroblast cells lacking Mfn2 showed a loss in membrane potential (6), and antisense experiments to reduce Mfn2 levels in cultured myotubes demonstrated a loss in glucose oxidation, membrane potential, and oxygen consumption (51).
In addition to providing protection against free radical-induced permeability transition, the data demonstrate that activated Mfn2 represses STS-induced Bax activation and cytochrome c release. This places a primary component of the mitochondrial fusion machinery as a regulator of the apoptotic response. Given that the activation of Mfn2 concomitantly blocks permeability transition and activates mitochondrial fusion, the data suggest that these two pathways are mutually exclusive. It has been shown previously that mitochondrial fusion events are inhibited during an apoptotic stimulus (52), consistent with our data indicating that Mfn2 must be in the inactive, GDP-bound form to allow for activation of Bax. Although the mechanism by which Mfn2 might regulate Bax activation is unknown, it is possible that the punctate Mfn2 sites represent preassembled channels competent for activated Bax recruitment and outer membrane permeability transition, and upon activation of Mfn2, critical components within these microdomains may disassemble. Alternatively, the activation of Mfn2 for fusion may directly inhibit the fission machinery so that the two events remain exclusive. Because fission precedes cytochrome c release and appears requisite for apoptosis (53,54), any interference of activated Mfn2 with components of the fission machinery would effectively block Bax activation and cytochrome c release. Obvious candidates would be Fis1 and DRP1 because they function together in promoting mitochondrial fission and eliciting an efficient apoptotic response (22,23,25,26,47). It will be important to characterize the Mfn2-interacting proteins, those both cytosolic and mitochondrial, to define better the molecular events that link the GTPase cycle of Mfn2 with mitochondrial fusion and the control of permeability transition.