Mitofusin 2 Regulates STIM1 Migration from the Ca2+ Store to the Plasma Membrane in Cells with Depolarized Mitochondria*

Store-operated Ca2+ channels in the plasma membrane (PM) are activated by the depletion of Ca2+ from the endoplasmic reticulum (ER) and constitute a widespread and highly conserved Ca2+ influx pathway. After store emptying, the ER Ca2+ sensor STIM1 forms multimers, which then migrate to ER-PM junctions where they activate the Ca2+ release-activated Ca2+ channel Orai1. Movement of an intracellular protein to such specialized sites where it gates an ion channel is without precedence, but the fundamental question of how STIM1 migrates remains unresolved. Here, we show that trafficking of STIM1 to ER-PM junctions and subsequent Ca2+ release-activated Ca2+ channel activity is impaired following mitochondrial depolarization. We identify the dynamin-related mitochondrial protein mitofusin 2, mutations of which causes the inherited neurodegenerative disease Charcot-Marie-Tooth IIa in humans, as an important component of this mechanism. Our results reveal a molecular mechanism whereby a mitochondrial fusion protein regulates protein trafficking across the endoplasmic reticulum and reveals a homeostatic mechanism whereby mitochondrial depolarization can inhibit store-operated Ca2+ entry, thereby reducing cellular Ca2+ overload.

In eukaryotic cells, a variety of different agonists, including hormones, neurotransmitters, and growth factors, elicit cellular responses through a rise in cytoplasmic Ca 2ϩ concentration (1). Cytoplasmic Ca 2ϩ can be increased following Ca 2ϩ release from intracellular stores, by Ca 2ϩ entry across the plasma membrane via Ca 2ϩ -permeable ion channels, or by both processes. In many cell types, the emptying of intracellular Ca 2ϩ stores opens store-operated Ca 2ϩ channels in the plasma membrane (2)(3)(4). The best characterized store-operated Ca 2ϩ channel is the CRAC 2 channel, which constitutes the major Ca 2ϩ entry pathway in immune cells (5,6). Ca 2ϩ entry through CRAC channels activates a range of temporally distinct responses, including exocytosis, enzyme activation, and gene transcription (2,7).
Recently, the molecular basis of store-operated Ca 2ϩ entry has been teased apart. Targeted RNAi screens first identified STIM1, a single transmembrane-spanning domain protein found predominantly in the endoplasmic reticulum (ER), as being essential for CRAC channel activation (8,9). STIM1 has a Ca 2ϩ -binding EF-hand that faces the lumen of the ER, and sitedirected mutagenesis has revealed that this likely senses the Ca 2ϩ content of the store (9 -11). Upon store depletion, Ca 2ϩ dissociates from STIM1, and this is thought to promote multimerization through the N-terminal sterile ␣-motif (12), a step that is central to CRAC channel activation (13). STIM1 multimers then migrate to punctate structures Ͻ25 nm from the plasma membrane that correspond to ER-PM junctions (14). At such sites, STIM1 activates Orai1, a four-transmembrane domain plasma membrane protein that is also required for CRAC channel activity (10,15,16). Mutagenesis studies have established that Orai1 is at least part of the CRAC channel pore (17)(18)(19). A cytoplasmic domain of STIM1 binds to both the N and C termini of Orai1 (20 -23), leading to CRAC channel activation.
Although STIM1 trafficking toward the plasma membrane is a critical early step in CRAC channel activation, an important but unresolved question is as follows. How is the migration process controlled? One idea is that STIM1 diffuses randomly in the ER but becomes trapped at the ER-PM junctions upon store depletion (24). Alternatively, it has been suggested that STIM1 reaches the ER-PM junctions by active transport along microtubules (25), and STIM1 binds directly to the microtubule plus-end tracking protein EB1 (26). Interestingly, STIM1 has been found to co-localize with ␣-tubulin, and microtubule depolymerization reduces STIM1 puncta formation, supporting a role for the microtubule cytoskeleton (25). However, STIM1 can form puncta even when intracellular ATP levels have been depleted, suggesting its movement is a passive process (27).
A solid body of evidence has demonstrated that mitochondria control CRAC channel activity (28 -33). Although some of the effects of mitochondria arise from their ability to buffer cytoplasmic Ca 2ϩ and thus reduce Ca 2ϩ -dependent inactivation of CRAC channels, growing evidence suggests that they might have an additional role in regulating CRAC channels that is unrelated to their ability to take up Ca 2ϩ , produce ATP, and generate reactive oxygen species (30,34). Mitochondria can be positioned adjacent to ER through interactions between proteins on the two organelles, with a major role for the mitochondrion-shaping protein mitofusin 2 (Mfn2) (35)(36)(37). Mouse embryonic fibroblasts lacking Mfn2 display loosened ER-mitochondria tethering and reduced rate of mitochondrial Ca 2ϩ uptake following InsP 3mediated Ca 2ϩ release from the ER (36). Whether this physical uncoupling impacts upon spatially more distal events is unclear.
We show here that mitochondrial depolarization suppresses STIM1 puncta formation and subsequent Orai1-dependent CRAC currents, and these inhibitory effects can be partially overcome by overexpression of either STIM1 or a STIM1 mutant that occupies ER-PM junctions in nonstimulated cells with intact stores. In cells lacking Mfn2, STIM1 puncta formation and CRAC channel activity were independent of mitochondrial status, and analysis of Mfn2 mutants revealed a major role for mitochondrially targeted Mfn2. Our results identify a new role for mitochondria in cell biology, i.e. these organelles help regulate the movement of an ER-resident multimeric protein complex to the plasma membrane. Furthermore, our findings reveal Mfn2 as an important component in the mechanism whereby mitochondrial depolarization inhibits CRAC channel activity.
I CRAC Recordings-Patch clamp experiments were conducted in the tight seal whole-cell configuration at room temperature (20 -25°C) as described previously (30,38). Sylgard-coated, fire-polished pipettes had resistances of 4.2-5.5 megohms when filled with standard internal solution that contained (in mM): Cs ϩ glutamate 145, NaCl 8, MgCl 2 1, Mg-ATP 2, EGTA 10, HEPES 10, pH 7.2, with CsOH. In some experiments, 30 M InsP 3 was added to this solution. A correction of ϩ10 mV was applied for the subsequent liquid junction potential that arose from this glutamate-based internal solution. The composition of the extracellular solution was (in mM) as follows: NaCl 145, KCl 2.8, CaCl 2 10, MgCl 2 2, CsCl 10, D-glucose 10, HEPES 10, pH 7.4, with NaOH. I CRAC was measured by applying voltage ramps (Ϫ100 to ϩ100 mV in 50 ms) at 0.5 Hz from a holding potential of 0 mV. Currents were filtered using an 8-pole Bessel filter at 2.5 kHz and digitized at 100 s. Currents were normalized by dividing the amplitudes (measured from the voltage ramps at Ϫ80 mV) by the cell capacitance. Capacitative currents were compensated before each ramp by using the automatic compensation of the EPC9 Ϫ2 amplifier. For I CRAC , leak currents were subtracted by averaging 2-3 ramp currents obtained just before I CRAC had started to develop and then subtracted this from all subsequent currents. Transfected cells were identified by expression of either eYFP or enhanced GFP.
Ca 2ϩ Imaging-Ca 2ϩ imaging experiments were carried out at room temperature using the IMAGO CCD camera-based system from TILL Photonics, as described previously (39). Cells were alternately excited at 356 and 380 nm (20-ms exposures), and images were acquired every 2 s. Images were analyzed offline using IGOR Pro for Windows. Cells were loaded with fura 2-AM (1 M) for 40 min at room temperature in the dark and then washed three times in standard external solution of composition (in mM) as follows: NaCl 145, KCl 2.8, CaCl 2 2, MgCl 2 2, D-glucose 10, HEPES 10, pH 7.4, with NaOH. Cells were left for 15 min to allow further de-esterification. Ca 2ϩ -free solution had the following composition (in mM): NaCl 145, KCl 2.8, MgCl 2 2, D-glucose 10, HEPES 10, EGTA 0.1, pH 7.4, with NaOH. The rate of Ba 2ϩ influx (on addition of 2 mM Ba 2ϩ ) was obtained by measuring the initial slope of the fluorescence rise following readmission of Ba 2ϩ to cells with depleted stores. Ca 2ϩ signals are plotted as R, which denotes the 356/380 nm ratio.
Mg 2ϩ Imaging-Cytoplasmic Mg 2ϩ was used to measure Mg-ATP (as explained in the text). Cells were loaded with magfura 2-AM (1 M) for 40 min at room temperature, and experiments were carried out as for fura 2.
TIRF Microscopy-RBL-1 cells were transfected with eYFP-STIM1 and plated onto Willco thin glass bottom dishes (Intracel, UK) 48 h in advance of experiments. Cells were bathed with the standard external solution that contained (in mM) the following: NaCl 145, KCl 2.8, CaCl 2 2, MgCl 2 2, D-glucose 10, HEPES 10, pH 7.4, with NaOH. Images from cells using TIRF microscopy were collected by a Hamamatsu ORCA-AG, Deep Cooled digital camera (model C4742-80-12AG), connected to an inverted TE2000 microscope with a through-the-lens (prismless) TIRF imaging attachment (Nikon). Samples were viewed through a CFI apo-TIRF 60ϫ oil immersion high resolution objective (1.49 N.A.) and excited by the 488-nm line of an argon laser (Spectra Physics 163-A120) via a FITC filter block (excitation 465-495 nm, dichroic 505 nm, barrier 515-555 nm). Image collection was controlled by IPLab software (BD Biosciences) at 2-s intervals for 8 min. Post-collection analysis consisted of each cell being selected by a region of interest, and the mean pixel intensity was calculated for each frame of the collection sequence.
Electron Microscopy-After specific treatments (as described in the text), eYFP-STIM1-transfected HEK cells were collected and fixed in 4% paraformaldehyde in 0.1 M phosphate buffer for 30 min at room temperature. The cells were blocked with 2% BSA and 1% goat serum in PBS for 1 h. Staining was performed by incubation with primary antibody (rabbit anti-GFP; 1:50 working dilution; Invitrogen) in PBS containing 0.2% BSA and 0.1% goat serum overnight at 4°C. Cells were then incubated with biotinylated secondary antibody (goat anti-rabbit IgG; 1:200 working dilution; Vector Laboratories) for 1 h at room temperature. The STIM1 protein was detected by using the Elite ABC peroxidase kit according to manufacturer's instructions (Vector Laboratories). Cell cultures of all treatments were processed simultaneously with the same solutions and incubation times. After post-fixation with 1% osmium tetroxide for 45 min, cells were stained with 2% aqueous uranyl acetate for 1 h. Cells were further processed as described previously (41). HEK293 cells were used for electron microscopy for two reasons. First, expression of YFP-STIM1 was much higher than in RBL-1 cells. Second, the HEK cells attached to the coverslips much better and hence less detached during the extensive washing/fixing procedures.
Distribution of mitochondria (supplemental Fig. 3) was measured by computing the distance of each mitochondrion from the plasma membrane in the X-Y direction, in sequences of 100 serial sections (each of 50 nm thickness) taken across each cell. Sections were scanned into a G5 Mac computer and superimposed, to avoid analyzing the same mitochondrion twice.
Confocal Microscopy-Cells were fixed in 4% paraformaldehyde in phosphate buffer for 30 min at room temperature, after stimulation with thapsigargin. All the washes used 0.01% phosphate-buffered saline (PBS (in mM): NaCl 137, KCl 2.7, Na 2 HPO 4 8, KH 2 PO 4 1). The cells were blocked with 2% bovine serum albumin (BSA) and 10% goat serum for 1 h. Mitofusin 2 was visualized using a monoclonal antibody (kindly provided by Dr. Richard Youle, National Institutes of Health; used at a dilution of 1:250). The secondary anti-rabbit IgG was a HandL chain-specific (goat) fluorescein conjugate (excitation at 495 nm, emission at 515 nm). This was used at 1:2000 in PBS for 2 h at room temperature. The cells were mounted in Vectashield mounting medium containing a propidium iodide counterstain for DNA (excitation 535 nm, emission 615 nm). Images were obtained using a Leica confocal microscope.
ER Distribution-HEK293 cells were transfected with cell light ER-red fluorescent protein, Bacman 2.0 (Invitrogen), and ER distribution was viewed using confocal microscopy of fixed cells.
Statistics-Data are presented as the mean Ϯ S.E. Statistical significance was determined using a Student's t test. Asterisk denotes p Ͻ 0.01.

STIM1 Migration and Orai1 Activity Is Regulated by
Mitochondria-We first confirmed, using patch clamp recordings, findings originally made in the HEK293 expression system (42,43) and Drosophila S2 cells (10) that co-expression of Orai1 and STIM1 in RBL cells increased the size of I CRAC . As shown in supplemental Fig. 1, A-C, dialysis of RBL-1 cells transfected with cDNA for both Orai1 and eYFP-STIM1 resulted in a larger I CRAC (ϳ3-5-fold) than was the case with corresponding controls transfected with eYFP or enhanced GFP plasmids alone. The current was identified as I CRAC on the basis of several characteristics, including its steep inward rectification and positive reversal potential (supplemental Fig. 1B), its absence when external Ca 2ϩ was removed (data not shown), inhibition by the CRAC channel blocker 2-aminoethyldiphenyl borate (50 M, data not shown), and ability to activate when stores were depleted passively (10 mM EGTA, supplemental Fig. 1C).
Transfection with eYFP-STIM1 alone increased Ca 2ϩ influx Ͻ1.5-fold compared with the control response (supplemental Fig. 2A). However, as observed previously (16,43), overexpression of Orai1 alone reduced Ca 2ϩ influx by ϳ35% (supplemental Fig. 2A). Hence recombinant co-expression of STIM1 and Orai1 increases the size of I CRAC . It is important to note that the increase in RBL cells is very modest when compared with common expression systems like HEK293 cells where a Ͼ500-fold increase in current can routinely be obtained (10,(42)(43)(44).
We designed experiments to see if Orai1 also contributed to native I CRAC and agonist-evoked Ca 2ϩ entry. Knockdown of Orai1 using an RNAi approach that we have described recently (40) reduced the amplitude of I CRAC by ϳ70% (supplemental Fig. 1D). We altered the expression levels of Orai1 before challenging intact cells, loaded with the Ca 2ϩ -sensitive fluorescent dye fura-2, with the endogenous P2Y receptor agonist ATP. Cells were stimulated with agonist in Ca 2ϩ -free solution, and then Ba 2ϩ was applied. Ba 2ϩ permeates CRAC channels but, unlike Ca 2ϩ , is not transported out of the cytoplasm by Ca 2ϩ ATPase pumps and therefore provides a good indication of CRAC channel activity (2,45,46). Exposure to Ba 2ϩ after challenge with ATP in Ca 2ϩ -free solution resulted in prominent Ba 2ϩ influx (supplemental Fig. 1E, labeled control). The rate and extent of Ba 2ϩ influx were significantly larger in cells co-expressing eYFP-STIM1 and Orai1 (supplemental Fig. 1E). Although STIM1 has been reported to control non-store-operated Ca 2ϩ entry pathways, this requires the protein to be in the plasma membrane (47). eYFP-STIM1 is not inserted into the plasma membrane and remains within the ER where it specifically regulates CRAC channels (43). Knockdown of endogenous Orai1 reduced agonist-evoked Ba 2ϩ influx compared with control cells (supplemental Fig. 1E). Similar findings were observed when cells were stimulated with different concentrations of ATP (supplemental Fig. 1F), demonstrating that I CRAC is the dominant source of Ca 2ϩ entry over a range of stimulus intensities.
Mitochondria regulate CRAC channel activity in several cell types (2,32,33). In RBL cells co-expressing Orai1 and eYFP-STIM1, both store-operated Ca 2ϩ influx in intact cells (Fig. 1, A and B) and I CRAC (Fig. 1, C and D) were substantially reduced following mitochondrial depolarization by inhibition of complex III of the respiratory chain with antimycin A (together with oligomycin to prevent the F 1 F 0 -ATP synthase from running in reverse). Similar findings were seen when mitochondria were depolarized with the protonophore carbonyl cyanide p-trifluoromethoxyphenylhydrazone (I CRAC was reduced by 79 Ϯ 4%, data not shown). Oligomycin alone (15 min pretreatment) had no inhibitory effect (data not shown; see also Ref. 26). Importantly, the extent of Ca 2ϩ release from the stores was not compromised by mitochondrial depolarization (Fig. 1A) (30). Mitochondria therefore target a step distal to store depletion. Because STIM1 movement occurs after store depletion, we looked to see whether its migration toward the plasma membrane was affected by impairing mitochondrial activity. Analysis of eYFP-STIM1 movement using confocal microscopy on fixed cells showed that very little STIM1 was located at the cell periphery in resting cells (Fig. 1E, 1st panel, labeled Control). Store depletion with thapsigargin (applied in Ca 2ϩ -free solution) resulted in strong staining of the cell periphery together with the appearance of punctate-like structures in the cytoplasm (Fig. 1E, 2nd panel). Strikingly, mitochondrial depolarization dramatically reduced staining around the cell surface and increased the number of punctate-like structures in the cytoplasm (Fig. 1E, 3rd panel), presumably reflecting formation of intracellular STIM1 multimers that were unable to traffic to the plasma membrane. Inhibition of ATP synthesis with oligomycin did not affect eYFP-STIM1 movement induced by store depletion (Fig. 1E, 4th panel), consistent with a recent report that oligomycin does not impair puncta formation in a breast cancer cell line (27). Analysis of eYFP-STIM1 location together with mitochondria (identified with MitoTracker red) in the same cells revealed very little co-localization, ruling out significant location of STIM1 on mitochondria (Fig. 1F).

Mitochondria Do Not Migrate to the Plasma Membrane-In
T cells, mitochondria migrate to the cell periphery following Ca 2ϩ entry through CRAC channels (48). To quantify the distribution of mitochondria in more detail and to examine whether they also migrated to the cell periphery like STIM1 after store depletion, we compared mitochondrial location in resting and store-depleted RBL-1 cells using electron microscopy (supplemental Fig. 3). We measured the number of mitochondria as a function of distance from the plasma membrane, obtained from quantitative analysis of 100 serial sections, taken every 50 nm and thus spanning the entire cell thickness. Within 100 nm of the plasma membrane, there was a paucity of mitochondria. To see whether Ca 2ϩ influx through CRAC channels changed the pattern of mitochondrial distribution, we stimulated cells with thapsigargin for 10 min (in the presence of 2 mM external Ca 2ϩ ) and then cut serial sections. The pattern was similar to that seen in control cells, indicating that store depletion and subsequent CRAC channel activation did not change the mitochondrial profile near the plasma membrane, at least with our stimulation protocol. Most mitochondria were found around 500 nm to 1 m from the cell surface. Consistent with these structural findings, functional evidence against close apposition of mitochondria with the plasma membrane was provided by measuring the rate and extent of rapid inactivation of CRAC channels. Ca 2ϩ permeation through CRAC channels results in the buildup of a microdomain of elevated Ca 2ϩ that can feedback to partially inactivate the channels (49 -51). The Ca 2ϩ -binding site is thought to reside within 10 nm of the channel pore because it is reduced by the fast chelator BAPTA but not the slower EGTA. Although I CRAC was reduced by mitochondrial depolarization, the rate and extent of fast inactivation was unaffected (supplemental Fig. 3D; measured using Ca 2ϩ as the charge carrier and with matched peak amplitudes between control and antimycin/oligomycin-treated cells), suggesting that the organelle was not sufficiently close to the plasma membrane to impact upon the buildup of local microdomains emanating from open CRAC channels. Collectively, these findings demonstrate that mitochondria do not migrate with STIM1 to the cell periphery following store depletion. In addition, our results show that mitochondria can regulate CRAC channels without needing to be located very close to the plasma membrane. This is in good agreement with a recent report that found mitochondria were absent from plasma membrane regions containing active store-operated Ca 2ϩ channels (52).
STIM1 Puncta Formation Is Regulated by Mitochondria-To characterize the role for mitochondria on STIM1 trafficking in a more quantitative way, we monitored eYFP-STIM1 movement with TIRF microscopy, which detects events restricted to within ϳ100 nm of the plasma membrane. Stimulation of RBL-1 cells with thapsigargin resulted in eYFP-STIM1 movement to the cell periphery ( Fig. 2A), and this was substantially reduced by mitochondrial depolarization with antimycin A and oligomycin ( Fig. 2A) or carbonyl cyanide p-trifluoromethoxyphenylhydrazone and oligomycin (supplemental Fig. 2B) but not oligomycin alone (supplemental Fig. 2B). Aggregate data comparing the total fluorescence rise in TIRF microscopy mode for cells stimulated with thapsigargin in the absence and presence of antimycin A and oligomycin are compared in Fig.  2B, and the half-times to peak of the eYFP-STIM1 movement are compared in Fig. 2C. Mitochondrial depolarization significantly reduced both the rate and extent of STIM1 trafficking to the plasma membrane in RBL cells.
No Role for Mitochondrial Ca 2ϩ Buffering in STIM1 Migration-We designed experiments to address the mechanism whereby depolarized mitochondria inhibit STIM1 migration. One important role for mitochondria is to buffer a rise in cytoplasmic Ca 2ϩ . However, several pieces of evidence militate against such a role here. First, with our protocol to deplete stores (thapsigargin in Ca 2ϩ -free solution for ϳ5 min), very little Ca 2ϩ is taken up by mitochondria (53). Consistent with this, the amplitude and time course of Ca 2ϩ release was unaltered following mitochondrial depolarization (Fig. 1A) (53). Second, increasing cytoplasmic Ca 2ϩ buffering by loading cells with the fast Ca 2ϩ chelator BAPTA virtually suppressed the cytoplasmic Ca 2ϩ rise evoked by thapsigargin in Ca 2ϩ -free solution compared with control non-BAPTA-loaded cells (Fig.  3A), but this did not affect migration of STIM1 to the cell periphery (compare middle panels of Fig. 3, B and C). Move-ment of STIM1 to the plasma membrane does not therefore require a cytoplasmic Ca 2ϩ rise. Importantly, the robust redistribution of STIM1 to the periphery in BAPTA-loaded cells was much less pronounced following mitochondrial depolarization (Fig. 3C). Therefore, mitochondrial depolarization impairs STIM1 trafficking even in the absence of a cytoplasmic Ca 2ϩ rise and therefore in the absence of the ability of the organelle to take up Ca 2ϩ . Finally, we overexpressed Orai1 and STIM1 and then measured the Na ϩ flux through the CRAC channels that occurs when cells are exposed to divalent-free external solution (51, 54 -56). Following dialysis with InsP 3 and 10 mM BAPTA, a large Na ϩ current developed (Fig. 3D), which showed the characteristics of Na ϩ flux through CRAC channels ( Fig. 3E; inwardly rectifying current-voltage relationship, reversal potential of approximately ϩ60 mV, low current noise). Mitochondrial depolarization substantially reduced the size of this Na ϩ current (Fig. 3, D-F). Similar results were seen when cells were dialyzed with BAPTA alone. Because cells were dialyzed with 10 mM BAPTA and no Ca 2ϩ was present outside, the involvement of mitochondria can clearly be separated from a Ca 2ϩ buffering role.
No Role for Mitochondrially Derived ATP-Another important function of mitochondria is ATP production. However, several arguments can be raised against a role for mitochondrially derived ATP in STIM1 trafficking. First, mast and RBL-1 cells are glycolytically competent, and depolarization of mitochondria with antimycin A and oligomycin does not reduce cellular ATP levels provided glucose is available (57). In our experiments, we always had 10 mM glucose present. We nevertheless measured cytoplasmic ATP levels in single cells using cytoplasmic Mg 2ϩ as an indicator of Mg-ATP (58,59). As Mg-ATP is consumed, free Mg 2ϩ rises because the hydrolytic product ADP has significantly lower affinity for Mg 2ϩ . In cells loaded with mag-fura, treatment with antimycin A and oligomycin failed to generate a clear rise in Mg 2ϩ levels in the presence of glucose (Fig. 3G). However, replacement of glucose with the nonmetabolizable analog 2-deoxyglucose together with iodoacetate (an inhibitor of glycolysis; used at 1 mM) resulted in a substantial rise in cytoplasmic Mg 2ϩ , consistent with depletion of Mg-ATP. Although mag-fura can also bind Ca 2ϩ , it does so with an affinity of ϳ50 M. Experiments with fura 2 (affinity of ϳ200 nM) revealed a small and inconsistent rise in cytoplasmic Ca 2ϩ following exposure to antimycin A and oligomycin, which was ϳ200 nM in amplitude. This is therefore too small to impact on the mag-fura signals of Fig. 3G. Although these global measurements fail to demonstrate a fall in ATP upon mitochondrial depolarization provided glycolysis is intact, we were concerned that ATP below the plasma membrane might have been reduced, especially as this is the site for STIM1 translocation. To test this possibility, we stimulated cells with thapsigargin in Ca 2ϩ -free solution and measured the time course of decay of the Ca 2ϩ signal (Fig. 3H), because the latter reflects extrusion of Ca 2ϩ by the plasma membrane Ca 2ϩ ATPase. The time constant of decay was unaffected by antimycin A and oligomycin (Fig. 3I), arguing against significant depletion of subplasmalemmal ATP. Finally, block of the mitochondrial F 1 F 0 -ATP synthase with oligomycin failed to affect eYFP-STIM1 trafficking to the cell periphery (Fig. 1E). Collectively, these results show that the inhibition of STIM1 trafficking to the cell periphery upon mitochondrial depolarization was not due to cytoplasmic ATP depletion. STIM1 translocation to ER-plasma membrane junctions has been reported to require the C-terminal polybasic motif, which, in other proteins, binds polyphosphoinositides in the plasma membrane (12). We do not think mitochondrial depolarization depletes the plasma membrane phosphoinositide pool because agonist-evoked phosphatidylinositol 4,5-bisphosphate hydrolysis and subsequent InsP 3 -dependent Ca 2ϩ release are unaffected by antimycin A and oligomycin (30). Furthermore, the amplitude of the inward rectifier, a K ϩ current regulated by phosphoinositides, was unaffected by mitochondrial depolarization (supplemental Fig. 4).
We also considered that mitochondrial depolarization might alter the distribution of the ER, the organelle in which STIM1 is embedded. We used a red fluorescent protein construct targeted to the ER (ER-red fluorescent protein) to monitor ER distribution. ER-red fluorescent protein labeling revealed a reticular network (supplemental Fig. 5), and this was unaffected by store depletion with thapsigargin. Mitochondrial depolarization with antimycin A and oligomycin for 15 min also had no discernible effect on ER distribution (supplemental Fig. 5).
Strong Overexpression of STIM1 Rescues CRAC Channel Activity-In T cells, two-thirds of the specialized ER-PM junctions where STIM1 accumulates upon store depletion are thought to be already pre-formed in resting cells (14). We reasoned that extensive overexpression of STIM1 might therefore result in an increased probability of STIM1 occupying some of these pre-existing ER-PM junctions, thus overcoming the inhibitory effects of mitochondria on STIM1 trafficking and store-operated influx. HEK293 cells are an ideal system for testing this idea because of their high transfection efficiency and ability to overexpress recombinant proteins massively. Indeed, we find absolute eYFP-STIM1 fluorescence in HEK293 cells to be at least 10 times greater than in RBL-1 cells, consistent with the very large size of I CRAC seen in HEK293 cells compared with RBL-1 cells following overexpression of Orai1 and STIM1 (HEK current approximately Ϫ100 to Ϫ500 pA/picofarads, see also Refs. 42, 43; RBL approximately Ϫ10 to Ϫ12 pA/picofarads). We first confirmed that store-operated Ca 2ϩ entry in HEK293 cells was reduced by mitochondrial depolarization. As shown in Fig. 4A, readmission of external Ca 2ϩ to store-depleted wild type cells (exposed to thapsigargin in Ca 2ϩ -free external solution at t ϭ 0) resulted in robust Ca 2ϩ entry, which was significantly reduced by mitochondrial depolarization. Ca 2ϩ influx in cells with depolarized mitochondria has been normalized to the corresponding control influx in this and subsequent panels. Overexpression of eYFP-STIM1 alone partially rescued Ca 2ϩ influx in cells with depolarized mitochondria (Fig. 4B). Co-expression of Orai1 and eYFP-STIM1 resulted in more substantial recovery of store-operated Ca 2ϩ influx (Fig.  4C), and the rate and extent of Ca 2ϩ influx in cells expressing Orai1 and STIM1 with depolarized mitochondria were larger than in wild type nontransfected cells. One interpretation of these results is that strong overexpression of STIM1 in HEK cells rescues the inhibition of Ca 2ϩ entry seen when mitochondria are compromised. We were nevertheless concerned that the Ca 2ϩ signals, particularly after overexpression of Orai1 and STIM1, were so large that they saturated the dye fura 2. Hence the apparent recovery of the Ca 2ϩ signal could be misleading, because the control store-dependent Ca 2ϩ responses (following expression of either STIM1 or STIM1 together with Orai1) have been underestimated because of dye saturation. We therefore used the lower affinity dye fura 5F to test this. Although store-operated Ca 2ϩ entry recovered somewhat in cells with depolarized mitochondria following overexpression of STIM and Orai1 (Fig. 4D), the extent of rescue was less than that detected with fura 2 as the Ca 2ϩ -sensitive dye (Fig. 4C). These results serve as a salutary warning in quantifying cytoplasmic Ca 2ϩ signals following overexpression of STIM1/Orai1 based on measurements using high affinity dyes such as fura 2.
We confirmed that overexpressed STIM1 was able to traffic to the cell periphery in HEK293 cells even after mitochondrial depolarization by monitoring puncta formation with TIRF microscopy. Compared with nonstimulated cells (Fig. 4E), store depletion with thapsigargin led to numerous puncta being formed (Fig. 4F), and these were still prominent when cells were pretreated with antimycin A and oligomycin prior to thapsigargin (Fig. 4G), although these STIM1 puncta appeared to coalesce into larger structures after mitochondrial depolarization (Fig. 4G). Fewer, larger puncta were observed in the presence of antimycin A and oligomycin and thapsigargin compared with the numerous smaller ones seen in thapsigargin-treated cells. We analyzed puncta intensity by measuring eYFP-STIM1 fluorescence in 1-m 2 sections spanning the entire cell footprint. Whereas intensity was largely a normal distribution in cells exposed to thapsigargin (Fig. 4H,  upper panel), the pattern changed dramatically following mitochondrial depolarization (Fig. 4H, lower panel). Now, numerous high intensity regions formed, indicative of the merging of . Store-operated Ca 2؉ entry can be partially rescued by overexpressing STIM1 in HEK293 cells with depolarized mitochondria. A-C, HEK293 cells were stimulated with thapsigargin in Ca 2ϩ -free solution (at t ϭ 0) and then 2 mM external Ca 2ϩ was readmitted as shown. Only the Ca 2ϩ entry component is shown for simplicity. The extent of Ca 2ϩ release was similar for the different conditions. A, store-operated entry in wild type cells is inhibited by antimycin A and oligomycin. B, overexpression of eYFP-STIM1 results in modest recovery of store-operated Ca 2ϩ entry in cells with depolarized mitochondria. C, overexpression of eYFP-STIM1 and Orai1 results in substantial rescue of store-operated Ca 2ϩ entry in the presence of antimycin A and oligomycin. For each graph, the response in antimycin A and oligomycin has been normalized to the corresponding control response. Control response has been re-calculated as ((R Ϫ R 0 )/R peak )⅐100%, where R is the measured ratio (356/380) at any time point following Ca 2ϩ readmission; R peak is the peak ratio measured in the cell (maximal response); and R 0 is the basal ratio (resting Ca 2ϩ ). D, Ca 2ϩ influx following overexpression of Orai1 and STIM1 is compared before and after mitochondrial depolarization in cells loaded with fura 5F. E-H, TIRF images comparing puncta formation in a resting HEK cell shown in (E), and then after store depletion with thapsigargin (Thap) for 4 min (F) and after store depletion following exposure to antimycin A (Anti) and oligomycin (oligo) for 15 min (G). H, histograms plot the absolute fluorescence intensity measured in 1-m 2 segments against the frequency of occurrence. Upper panel represents cells exposed to thapsigargin, and the lower panel analyses cells pre-exposed to antimycin A and oligomycin prior to thapsigargin challenge.
Electron microscopy provides the highest spatial resolution for resolving protein location. We identified the position of eYFP-STIM1 in serial ultra-thin sections of HEK293 cells prepared for electron microscopy using the ABC-coupled horseradish peroxidase system (see "Materials and Methods" for more details). Following expression of eYFP-STIM1, cells were fixed and then exposed to anti-GFP primary antibody followed by incubation with biotinylated secondary antibody. We then applied avidin-coupled horseradish peroxidase, which bound to the secondary antibody. The complex was visualized using the 3,3Ј-diaminobenzidine chromagen, which is oxidized by peroxidase to form an electron-dense precipitate at the site of the reaction, thereby revealing the location of STIM1. In micrographs from nonstimulated HEK293 cells expressing eYFP-STIM1, we saw very little protein near the plasma membrane. Instead, most dark deposits were seen around the nuclear membrane and ER-like tubular structures (Fig. 5A, box indicated in the left-hand panel has been magnified in the righthand panel). However, store depletion with thapsigargin led to the formation of electron-dense deposits at the cell periphery (Fig. 5B), with a loss of deposit from the nuclear membrane. Store depletion after pretreatment with antimycin A and oligomycin resulted in some labeling at the cell surface, although this was less extensive (Fig. 5C). We compared the fraction of the cell periphery that stained for STIM1 between nonstimulated cells, cells stimulated with thapsigargin, and cells pretreated with antimycin A and oligomycin prior to thapsigargin exposure (Fig. 5D). Whereas very little staining was detectable in control cells, a substantial portion of the periphery contained STIM1 after store depletion. The extent of this was reduced by antimycin A and oligomycin, although staining was still prominent. Because of lateral diffusion, the DABS reaction product is not restricted just to the site of HRP. Hence, we were unable to quantify the lateral extent of puncta in electron micrographs between thapsigargin-treated cells and those first exposed to antimycin A and oligomycin to see how this related to the changes observed in the TIRF microscopy experiments (Fig.  4G). Collectively, these results confirm that a fraction of the STIM1 pool migrates up to the plasma membrane following store depletion after mitochondrial depolarization in HEK cells, when STIM1 is strongly overexpressed.

Ca 2ϩ Influx to a STIM1 Mutant That Accumulates in ER-PM Junctions in Nonstimulated Cells Is Insensitive to Mitochondrial
Depolarization-An explanation of the results so far is that mitochondrial depolarization impairs STIM1 trafficking to the plasma membrane but not the events that arise once STIM1 is at ER-PM junctions. To test this more directly, we expressed the mutant D76A STIM1, which localizes to ER-PM junctions even when stores are full. This construct has a point mutation in the Ca 2ϩ -binding EF-hand domain, so that the expressed protein forms punctate-like structures in cells with replete Ca 2ϩ stores, resulting in constitutive store-operated Ca 2ϩ influx (9, and presence (C) of antimycin A/oligomycin (Anti/oligo). STIM1 location was identified using the 3,3Ј-diaminobenzidine reaction, which is visible as dark deposits. D, fraction of the cell periphery stained for STIM1-eYFP is compared for control cells, cells stimulated with thapsigargin, and cells pretreated for 15 min with antimycin A and oligomycin prior to exposure to thapsigargin. E, mutant STIM1 localizes to the cell periphery in HEK293 cells in the absence of store depletion, and this is not altered by mitochondrial depolarization. F, constitutive store-operated Ca 2ϩ entry, revealed by transiently removing and then readmitting 2 mM external Ca 2ϩ , is unaffected by mitochondrial depolarization. G, aggregate data, measuring the rate of Ca 2ϩ influx from experiments as in F, are summarized. Each bar denotes Ͼ50 cells.
11, 60). Expression of D76A STIM1 in HEK293 cells resulted in formation of puncta close to the cell periphery (Fig. 5E, lefthand panel), and these structures were unaffected by mitochondrial depolarization (Fig. 5E, right-hand panel). Constitutive Ca 2ϩ influx was revealed by briefly exposing cells to Ca 2ϩ -free external solution for 90 s and then readmitting external Ca 2ϩ . The subsequent store-operated Ca 2ϩ influx was unaffected by mitochondrial depolarization (Fig. 5F; aggregate data is summarized in Fig. 5G), at least over the duration of mitochondrial depolarization we have used (10 -15 min in this study).
In aggregate, these results have two important implications. First, mitochondria are not involved in the late stages of CRAC channel activation because of the following: (i) constitutive store-operated Ca 2ϩ influx after D76A mutant STIM1 expression was unaffected by impairing mitochondria (Fig. 5, E-G), and (ii) overexpression of eYFP-STIM1 and Orai1 could partially overcome the block by mitochondrial depolarization (Fig.  4D). Presumably, the interaction between STIM1 and Orai1 at the ER-PM junctions (20) can activate CRAC channels without a requirement for mitochondria. Second, the tools used for inducing mitochondrial depolarization do not interfere with store-operated Ca 2ϩ entry nonspecifically or with the CRAC channels themselves, otherwise neither rescue by Orai1 and STIM1 nor constitutive Ca 2ϩ influx to the mutant STIM1 would have been seen in cells with depolarized mitochondria.
Mitofusin 2 (Mfn 2) Regulates STIM1 Trafficking-Mitochondria can be tethered to the ER through the mitochondrial dynamin-related protein Mfn 2 (36). Mfn 2 is found mainly in the outer mitochondrial membrane, with a small fraction in the ER. It is particularly abundant within the contact sites between the mitochondria and ER, where it forms transorganellar homotypic and heterotypic interactions between mitofusin 1 or 2 on mitochondria and Mfn 2 on the ER (36). In mouse embryonic fibroblasts (MEFs) lacking Mfn 2, mitochondria are physically uncoupled from the ER, and the spatial distance between them is increased (35,36). We considered the possibility that mitochondrial tethering to the ER might hinder STIM1 movement to the cell periphery. To test this, we compared STIM1 migration and store-operated Ca 2ϩ influx between control MEFs and those in which Mfn 2 had been knocked out (10). Stimulation with thapsigargin in control cells resulted in prominent STIM1 puncta formation (Fig. 6A). Store-operated Ca 2ϩ entry was also present (Fig. 6, C and E). Both STIM1 puncta formation (Fig.  6A) and store-operated Ca 2ϩ entry (Fig. 6C) in wild type cells were suppressed by antimycin A and oligomycin pretreatment (aggregate data is summarized in Fig. 6E). Wild type mouse embryonic fibroblasts therefore behave in a manner similar to RBL and HEK293 cells. In mouse embryonic fibroblasts lacking Mfn 2, numerous STIM1 puncta formed after exposure to thapsigargin (Fig. 6B), and this was followed by store-operated Ca 2ϩ entry (Fig. 6D). The rate of Ca 2ϩ entry was ϳ2-fold faster than in wild type cells (Fig. 6, D and E). Importantly, in Mfn 2-deficient cells, STIM1 puncta formation (Fig. 6B) and store-operated Ca 2ϩ entry (Fig. 6, D and E) were unaffected by mitochondrial depolarization. Hence, STIM1 puncta formation and FIGURE 6. STIM1 puncta formation and subsequent store-operated Ca 2؉ entry are unaffected by mitochondrial depolarization in mitofusin 2-deficient cells. A, EYFP-STIM1 distribution is compared between control (resting) mouse embryonic fibroblasts and those stimulated with thapsigargin (Thap), in the absence and then presence of antimycin A (anti) plus oligomycin (oligo). WT above the images denotes wild type fibroblasts. B, experiments were conducted as in A but now using mitofusin 2-deficient (Mfn2 Ϫ/Ϫ ) mouse embryonic fibroblasts. C, after depleting Ca 2ϩ stores with thapsigargin (2 M) in fura 2-loaded cells, readmission of external Ca 2ϩ (2 mM) resulted in a rapid increase in cytosolic Ca 2ϩ concentration (control) that was suppressed by depolarizing mitochondria. D, Ca 2ϩ influx was prominent in mitofusin 2-deficient cells and was unaffected by impairing mitochondria. E, bar chart compares the rate of cytosolic Ca 2ϩ rise upon Ca 2ϩ readmission in control and mitofusin 2-deficient cells, in the absence and presence of antimycin A plus oligomycin. Mitochondrial depolarization reduces the rate of Ca 2ϩ entry in WT, but not in mitofusin 2-deficient cells. The rate of Ca 2ϩ influx in wild type fibroblasts has been taken as 100%, for comparative purposes.
store-operated entry are insensitive to mitochondrial depolarization when Mfn 2 is absent.
It is important to note that the extent of Ca 2ϩ release and the rate of recovery of the Ca 2ϩ signal (to thapsigargin in Ca 2ϩ -free solution) were similar between wild type MEF cells and those lacking Mfn 2. Hence, stores are loaded with Ca 2ϩ , and sufficient ATP is in the cytoplasm to support Ca 2ϩ ATPase activity in Mfn 2 Ϫ/Ϫ cells.
Endogenous Mfn 2 has a patchy distribution throughout the cytoplasm in resting MEF cells (Fig. 7A) (36). Stimulation with thapsigargin did not alter the pattern of Mfn 2, and unlike the case with STIM1, no clear punctate-like structures were formed (Fig. 7A). We also transfected cells with a Mfn 2-GFP construct. However, the distribution of Mfn 2-GFP was drastically different from endogenous Mfn 2 in that it formed aggregates around the nucleus (data not shown). This is in agreement with other studies that have found that overexpression of tagged Mfn 2 leads to a nonphysiological distribution (61)(62)(63).
Re-expression of untagged Mfn 2 into MEF cells resulted in an accelerated decline of the thapsigargin-evoked Ca 2ϩ signal, reflecting a loss of store-operated Ca 2ϩ entry (Fig. 7B). As with mitochondrial depolarization, the inhibitory effect of Mfn 2 could be overcome by transfecting STIM1 together with Mfn 2 in the MEF cells (Fig. 7B). Mfn 2 is located on both mitochondria and ER. This prompted us to ask whether mitochondrial Mfn 2, ER Mfn 2, or both were involved in regulating store-operated Ca 2ϩ entry. To address this, we expressed Mfn 2 constructs that are selectively expressed in either the ER (Mfn 2-IYFFT) (36) or mitochondria (Mfn 2-ActA) (36) in Mfn 2 Ϫ/Ϫ cells and then looked to see whether either construct could re-introduce sensitivity to mitochondrial depolarization. Cells were stimulated with thapsigargin in Ca 2ϩ -free solution in the presence of antimycin A and oligomycin, and Ca 2ϩ was then readmitted. Whereas Mfn 2 Ϫ/Ϫ cells responded by generating robust Ca 2ϩ influx in the presence of depolarized mitochondria, Ca 2ϩ entry was substantially reduced by expression of the Mfn 2-ActA construct (Fig.  7C, panels i and ii). Ca 2ϩ release was only slightly reduced, and in several cells it was indistinguishable from control release yet was followed by much less Ca 2ϩ entry. However, store-operated Ca 2ϩ entry was largely unaffected by mitochondrial depolarization following expression of Mfn 2-IYFFT (Fig. 7D, panels i and ii). These results reveal that Mfn 2 restricted to mitochondria renders store-operated entry susceptible to mitochondrial depolarization, whereas an ER-resident Mfn 2 is less effective.

DISCUSSION
Activation of the ubiquitous store-operated Ca 2ϩ influx pathway by the ER Ca 2ϩ sensor STIM1 is a three-step process (12,13). First, upon store depletion STIM1 monomers come together to form multimers in the ER membrane. Second, the oligomers then migrate to specialized ER-PM junctions, resulting in punctate-like structures at the cell periphery. Finally, STIM1 activates the plasma membrane CRAC channels by binding to the N and C termini of Orai1 (20,21). Although recent studies have provided insight into how STIM1 senses store depletion and interacts with Orai1 channels, little is known about mechanisms that regulate STIM1 migration to the cell periphery. Our new results reveal that mitochondrial depolarization selectively regulates trafficking of STIM1 multimers to the plasma membrane and in a manner dependent on the mitochondrial protein Mfn 2.
A substantial body of evidence has established that mitochondrial depolarization inhibits store-operated Ca 2ϩ entry through CRAC channels and that energized mitochondria increase the size of the CRAC current (15,16,20,21). A major factor contributing to this effect involves mitochondrial buffering of cytoplasmic Ca 2ϩ (42,43). Mitochondrial Ca 2ϩ uptake reduces the extent of Ca 2ϩ -dependent inactivation of CRAC channels and thus leads to an enhanced Ca 2ϩ entry. Our new findings add a further component to mitochondrial gating of CRAC channels that is independent of the Ca 2ϩ -buffering action and involves an action on STIM1 trafficking.
How might mitochondrial depolarization be relayed to STIM1 proteins in the ER? The two organelles are held in close proximity to one another through interaction between proteins spanning the respective membranes with a major role for the dynamin-related protein Mfn 2 (36,37). Mfn 2 is abundant at contact sites between ER and mitochondria, where it forms transorganellar homotypic and heterotypic interactions. Knock-out of the Mfn 2 gene in mouse embryonic fibroblasts uncouples mitochondria from the ER and increases the dis-tance between them (36,37). By using Mfn 2-deficient cells, we found that STIM1 trafficking and store-operated Ca 2ϩ entry were no longer impaired by mitochondrial depolarization. This suggests that Mfn2 is required to confer sensitivity of storeoperated Ca 2ϩ entry to mitochondrial depolarization, at least in MEF cells. Mfn2 is expressed both in the mitochondria and ER. By expressing Mfn 2 constructs in either mitochondria or ER in Mfn 2-deficient MEF cells, we found that mitochondrial Mfn 2 rendered store-operated Ca 2ϩ entry sensitive to mitochondrial depolarization. It is unlikely that Mfn 2 senses mitochondrial depolarization directly because it is expressed in the outer mitochondrial membrane (64). However, it can interact with inner mitochondrial membrane proteins, including OPA1 (35,65). Hence, changes in mitochondrial potential could be relayed to the adjacent ER through protein-protein interactions with mitofusin 2 acting as a transducer. How does mitofusin 2 control STIM1 trafficking and thereby CRAC channel activation following mitochondrial depolarization? Strong overexpression of Mfn 2 leads to the formation of aggregates of mitochondria that are uncoupled from one another and clump around the nucleus (61,63). These mitochondria have a decreased membrane potential (63). De-energized mitochondria are less able to buffer cytosolic Ca 2ϩ , and this would result in stronger Ca 2ϩ -dependent inactivation of CRAC channels (29,31). Although this mechanism can help account for the reduced Ca 2ϩ influx seen upon over expression of Mfn 2, it fails to explain the following: (i) why STIM1 movement after store depletion is impaired by mitochondrial depolarization in the absence of Ca 2ϩ entry and when cytoplasmic Ca 2ϩ is buffered with BAPTA, conditions that would prevent Ca 2ϩ -dependent inactivation of CRAC channels; and (ii) why STIM1 overexpression, at least partially, rescues Ca 2ϩ entry in Mfn 2-overexpressing cells. An alternative mechanism involves a direct or indirect physical block on STIM1 movement by Mfn 2. This could be due either to steric hindrance whereby mitofusin 2 needs to be displaced from a site in order for STIM1 multimers to migrate toward ER-PM junctions or that a component of the mitochondrial tethering complex (which includes Mfn 2, voltage-dependent anion channel, Grp75, 1 receptor, and PACS-2 (35)) binds to STIM1, thus impeding its movement toward the periphery. In either case, mitochondrial depolarization would be predicted to stabilize this interaction. Loss of Mfn 2 might therefore be expected to increase the number of STIM1 multimers that successfully migrate to ER-PM junctions, resulting in increased Ca 2ϩ entry. Consistent with this, the rate of Ca 2ϩ influx was slightly higher in Mfn 2-deficient cells (Fig. 7E).
Although overexpression of STIM1 led to partial recovery of store-operated Ca 2ϩ influx in HEK293 cells with depolarized mitochondria, the pattern of puncta formation was strikingly different from that seen in normal cells. After mitochondrial depolarization, electron micrograph analysis revealed that ϳ40% less STIM1 migrated to the cell periphery after store depletion (Fig. 5D). TIRF measurements showed that although fewer puncta formed, they had a higher STIM1 intensity suggesting coalescence of individual puncta into larger structures. Despite such changes, store-operated Ca 2ϩ influx still developed, albeit to a lesser extent. Hence, it would appear that formation of puncta per se is sufficient to activate several Orai1 channels, in that numerous discrete puncta are only moderately (ϳ2-fold) more effective in evoking Ca 2ϩ entry that larger, merged structures. This would be consistent with the finding that disaggregation of microfilaments results in fewer, larger puncta being formed but without any effect on CRAC channel activity (66).
What might be the physiological relevance for Mfn 2 regulation of STIM1 movement? Mfn 2 is not essential for the storeoperated pathway because store-operated Ca 2ϩ entry in both MEF fibroblasts and RBL cells (data not shown) was prominent despite its knockdown. Therefore, Mfn 2 has a regulatory role but not an essential one. Rather, its effects are manifest only after mitochondrial depolarization. Our results suggest Mfn 2 might serve as a brake, inhibiting Ca 2ϩ entry only after mitochondria depolarize. Although significant fluctuations in the mitochondrial membrane potential have been reported in some intact cells, the sustained mitochondrial depolarization we have evoked is more typically seen following glutamate excitotoxicity in neurons. From a teleological standpoint, Mfn 2 regulation of STIM1 migration and subsequent CRAC channel opening might serve to oppose Ca 2ϩ overload under conditions where mitochondrial Ca 2ϩ buffering is compromised, due to collapse of the mitochondrial membrane potential. Precisely how Mfn 2 controls migration of STIM1, its impact, if any, on Ca 2ϩ entry under physiological levels of stimulation, and whether this mechanism contributes to mitofusin-related diseases such as Charcot-Marie tooth neuropathy await further study.