MICU1 controls cristae junction and spatially anchors mitochondrial Ca2+ uniporter complex

Recently identified core proteins (MICU1, MCU, EMRE) forming the mitochondrial Ca2+ uniporter complex propelled investigations into its physiological workings. Here, we apply structured illumination microscopy to visualize and localize these proteins in living cells. Our data show that MICU1 localizes at the inner boundary membrane (IBM) due to electrostatic interaction of its polybasic domain. Moreover, this exclusive localization of MICU1 is important for the stability of cristae junctions (CJ), cytochrome c release and mitochondrial membrane potential. In contrast to MICU1, MCU and EMRE are homogeneously distributed at the inner mitochondrial membrane under resting conditions. However, upon Ca2+ elevation MCU and EMRE dynamically accumulate at the IBM in a MICU1-dependent manner. Eventually, our findings unveil an essential function of MICU1 in CJ stabilization and provide mechanistic insights of how sophistically MICU1 controls the MCU-Complex while maintaining the structural mitochondrial membrane framework.


Results
MICU1 is exclusively localized in the IBM. Applying superresolution SIM, we first determined the sub-mitochondrial localization of MICU1. HeLa cells expressing MICU1 tagged with yellow fluorescent protein (YFP) were labeled with Mitotracker Red FM (MTR). Super-resolution images under resting conditions and after stimulation with histamine were acquired. These experiments show that MICU1 exclusively localizes to the IBM 15 , and that this distribution does not change upon histamine stimulation (Fig. 1a, b). Morphological analysis of mitochondria showed no effect of MICU1-YFP expression ( Supplementary   Fig. 1). To further establish that MICU1 is confined to the IBM, we co-transfected cells with MICU1-YFP and mCherry-TOM22. By analyzing the cross-section intensity profiles of the two fluorophores based on a fit of two superimposed Gaussian functions, we estimated a distance of 21 ± 13 nm (mean ± SD, n = 105) between the MICU1 and TOM22 (Fig. 1c-e; Supplementary Fig. 2 and 3), which is consistent with electron microscopy measurements 22 . Hence, by co-expression of MICU1-mCherry and MICU1-YFP a difference of −5 ± 11 nm (mean ± SD, n = 80) was found. To validate these results, we measured diffraction-limited 100-nm-sized Tetraspec beads to analyze x/y and z chromatic abbreviation of our setup. For x/y, we determined a relative mean local chromatic abbreviation of 1.05% ( Supplementary Fig. 4) while for z-axes a 45-nm shift was determined ( Supplementary Fig. 5). The z-shift of 45 nm as well as dispersion measured with MTR-and MTG-labeled mitochondria in live cells only have a very minor effect of~1.6 nm with regard to the x/y measurements (Supplementary Figs. 6 and 7). Other proteins of the MCU-Complex (i.e., MCU, EMRE and UCP2) co-localized with Mitotracker Green (MTG) as IMM marker and were clearly separated from TOM22 ( Supplementary  Fig. 8).
MICU1 localization requires mitochondrial membrane potential. The CJ enables the generation of high proton gradients across the CM 23,24 . By its poly-basic domain 10,20 MICU1 associates potentially with negatively-charged phospholipids (e.g. phosphatidylserine, cardiolipin 25 ). Therefore, we examined the importance of a high proton gradient for MICU1 distribution by gradually reducing the mitochondrial membrane (IMM) potential (ψ mito ) with oligomycin A and antimycin A. For analyzing the sub-mitochondrial localization of MICU1-YFP a new quantification method based on a robust spatial thresholding approach was used (IBM association index) ( Supplementary Fig. 9). The higher the IBM association index, the more the respective protein is localized to the IBM. This approach revealed that during depolarization MICU1 but not EMRE rapidly redistributed from the IBM into the entire IMM (Fig. 2b, c). Measurements of ψ mito using the potentiometric dye tetramethylrhodamine methyl ester perchlorate (TMRM) revealed that even partial depolarization caused diffusion of MICU1 into the CM (Fig. 2d, e). Because no correlated structural changes in mitochondrial morphology were observed ( Supplementary Fig. 10), and MICU1 is not undergoing proteolytic cleavage even after 10 min of oligomycin A and antimycin A treatment ( Supplementary Fig. 11), it is likely that the exclusive localization of MICU1 to the IBM is controlled by the ψ mito .
IBM localization of MICU1 is dependent on its poly-basic domain. To examine the importance of the poly-basic domain of MICU1 for its IBM localization, the MICU1 1-140 and MICU1 1-70 mutants tagged to YFP (Fig. 2a) were transiently overexpressed in HeLa cells that were subsequently labeled with MTR. While MICU1 1-140 mimicked the IBM localization of wild-type MICU1 (Fig. 2f), MICU1 1-70 was located in the entire IMM (Fig. 2f). For   b Representative images of SIM time-lapse experiment using HeLa cells transiently transfected with MICU1-YFP (green) and EMRE-mCherry (magenta) after T = 0 and T = 270 s. Cells were pre-incubated with oligomycin A (2 µM) and antimycin A (4 µM) was added at T = 90 s. c Line plots of MICU1 (green line) and EMRE (magenta line) of regions in between the white arrows in b confirm the redistribution of MICU1 into the IMM after addition of antimycin A. Differences between oligomycin A and oligomycin A plus antimycin A traces were measured from T = 50 s. d Time course of the MICU1 IBM association index over time after pre-incubation with oligomycin A with (green) or without (black) exposure to antimycin A starting at T = 90 s (n Control = 7; n Oligomycin/Antimycin = 9. Data are shown as the mean ± SEM. e Corresponding tetramethylrhodamine methyl ester perchlorate (TMRM) fluorescence, which indicates ψ mito . Cells were pre-incubated with oligomycin A and exposed to antimycin A (green) or DMSO (black) at T = 90 s (n Antimycin = 8; n Antimycin/ Oligomycin = 9). A significant drop in ψ mito after addition of antimycin A correlates with the drop in MICU1 IBM association index. Data are shown as the mean ± SEM. f Representative dual-SIM images of HeLa cells stained with MTR (magenta) and expressing MICU1 1-140 -cpEGFP (green, upper panel) or MICU1 1-70 -cpEGFP (green, lower panel). g Respective statistical analyses of the IBM association index of MICU1 1-140 and MICU1  . The submitochondrial distribution of MICU1 1-140 and MICU1  in the IBM and the CM was determined with the IBM association index. Horizontal lines represent the median, the lower and upper hinge show respectively first quartile and third quartile, and lower and upper whisker encompass minimal and maximal values. Images and analyses were obtained from n = 9 independent experiments, which assessed a total of 45 cells. *P < 0.05 vs. respective control conditions carried out with unpaired double-sided T-test. Source data are provided as a Source Data file quantitative and statistical analyses the IBM association index was calculated (Fig. 2g). This different sub-mitochondrial localizations of both MICU1 mutants point to an important role of the protein's poly-lysine domain for its spatial location to the IBM. Recent reports shown that MICU1 is closely associated with cardiolipin 26 . Knockdown of taffazin (TAZ), an enzyme responsible for cardiolipin maturation 27 , led to a rearrangement of MICU1-YFP into the entire IMM and slightly reduced mitochondrial form factor ( Supplementary Fig. 12). These results show the correlation between the abundance of cardiolipin and the spatial distribution of MICU1.
EF-hands and the methylation site do not contribute to MICU1 localization. Next, we evaluated the role of the two EF-hand motifs of MICU1 on its IBM localization. Sub-mitochondrial localization of MICU1-ΔEF (Fig. 2a) was measured and the IBM association index was calculated. Disabling both EF-hands of MICU1 did only slightly reduce the proteins localization in the IBM ( Supplementary Fig. 13). Because the apparent Ca 2+ binding affinity of MICU1 in response to R455 methylation by protein arginine methyltransferase 1 (PRMT1) is strongly attenuated 21 , we examined whether R455 methylation may affect the IBM localization of MICU1. Therefore the distributions of the two respective mutations MICU1-R455F-YFP and MICU1-R455K-YFP 21 (Fig. 2a) were assessed and the IBM association index was calculated ( Supplementary Fig. 13). Inhibition of PRMT1mediated methylation (MICU1-R455K-YFP) and mimicking its stable methylation (MICU1-R455F-YFP) did not affect the proteins localization in the IBM ( Supplementary Fig. 13). The general mitochondrial morphology regarding form factor, aspect ratio and size was also not changed by MICU1-R455F-YFP, MICU1-R455K-YFP or MICU1-ΔEF-YFP ( Supplementary Fig. 14).
MICU1 di-/multi-merization is necessary for IBM localization. We next examined whether the quaternary structure of MICU1 12 contributes to its distribution. The sub-mitochondrial localization of the oligomerization site-lacking ΔC-MICU1-YFP mutant 20 was assessed (Fig. 2a). Notably, in cells with constitutive wild-type MICU1 ΔC-MICU1 was confined to the IBM (Fig. 3a). Depletion of endogenous MICU1 ( Supplementary Fig. 15) yielded change of the distribution of the ΔC-mutant to the entire IMM (Fig. 3a). Mitochondrial morphology was not affected by ΔC-MICU1-YFP expression or knockdown of endogenous MICU1 ( Supplementary  Fig. 16).
An intact cristae junction is essential for the IBM location of ΔC-MICU1. The CJ is stabilized by several proteins, including OPA1, that forms pore-like structures 16,17 . To test whether the stability of the CJ is another important molecular determinant for MICU1 localization, OPA1 expression was reduced by using respective siRNA. Depletion of OPA1 that induced slight mitochondrial swelling (Fig. 3b, Supplementary Fig. 16) did not influence the distribution of wild-type MICU1 but allowed the ΔC-MICU1 mutant to localize in the CM even in the presence of endogenous MICU1 (Fig. 3b). Hence, the contribution of OPA1 on the CJ integrity was verified with transmission electron microscopy (TEM) that revealed a widening of the CJ by OPA1 knockdown (Fig. 4a, b, d). These data indicate that the exclusive IBM location of the ΔC-mutant of MICU1 requires a stable CJ.
MICU1 contributes to the structural integrity of the CJ. Our findings, showing that wild-type MICU1 did not enter the CM despite the knockdown of OPA1 and the dependency of the location of ΔC-MICU1 on the presence of wild-type MICU1 gave us reason to believe that MICU1 may be involved in CJ stabilization. This was investigated with TEM where the effect of MICU1 depletion on the width of the CJ was obtained. Remarkably, similar to the knockdown of OPA1, silencing of MICU1 yielded strong widening of the CJ (Fig. 4a-d) (Control: 17.1 ± 0.8 nm; OPA1 knockdown: 26.9 ± 1.3 nm; MICU1 knockdown: 27.9 ± 1.3 nm; mean ± SEM, n Control si = 117; n OPA1 si = 97; n MICU1 si = 106). Cristae widening by knockdown of OPA1 or MICU1 was largest at the interface to the IBM and got gradually smaller, while beyond 60 nm from the CJ no significant effect was found (Fig. 4d). To test for a contribution of mitochondrial Ca 2+ signaling to the CJ stabilization by MICU1, MCU and EMRE alone or together with MICU1 were silenced. MCU/EMRE knockdown did not change the CJ morphology and silencing of MCU/EMRE and MICU1 changed CJ morphology similar to the knockdown of MICU1 alone ( Supplementary Fig. 17).
CJ function of MICU1 for mitochondrial functional morphology. We recently described a correlation between the CM movement and the mitochondria-ER proximity, the presence of OPA1 but not matrix Ca 2+ levels 28 . Accordingly, we analyzed the impact of silencing MICU1 on CM kinetics. Similar to the depletion of OPA1 28 , knockdown of MICU1 yielded significant reduction of CM kinetics in the entire mitochondria (Supplementary Fig. 18).
Morphological analyses of mitochondria showed a reduction in mitochondrial branching and mitochondrial size upon depletion of MICU1. Using co-localization of MTG stained mitochondria and an ER-targeted RFP construct, we found that the number of mitochondria associated membranes (MAMs) remained unchanged ( Supplementary Fig. 19). Cytochrome c release is a hallmark of apoptosis induction 29,30 . Considering the large widening of the CJ upon knockdown of either MICU1 or OPA1, the role of MICU1/OPA1 for cytochrome c localization was evaluated. Therefore, HeLa cells were depleted from MICU1 or OPA1, stained with Mitotracker Red CMXROS (MTR-CMX), fixed with paraformaldehyde (PFA) and immunostained for cytochrome c. Knockdown of either MICU1 or OPA1 yielded translocation of cytochrome c from the cristae into the IMS ( Fig. 4e-g). Notably, cell viability and caspase-3/7 activity remained unchanged by knockdown of MICU1 or OPA1 ( Supplementary Fig. 20).
To estimate the physiological consequences of CJ widening by knockdown of OPA1 or MICU1 we measured mitochondrial cristae Ca 2+ and ψ mito . Depletion of MICU1 or OPA1 but not MCU or EMRE reduced the ψ mito (Fig. 5a, Supplementary  Fig. 21). Depolarization by knockdown of OPA1 or MICU1 was recovered by an overexpression of MICU1-YFP. In contrast to its wild-type form, expression of ΔC-MICU1-YFP was not able to reestablish ψ mito in MICU1/OPA1-depleted cells (Fig. 5a). These data suggest that silencing of OPA1 or MICU1 leads to structural loss in CJ integrity resulting in a leakage of H 3 O + from cristae lumen (CL) and, thus, a reduction of ψ mito (Fig. 5b).
MICU1 was reported to serve as a gatekeeper to prevent mitochondrial matrix Ca 2+ overload 31,32 . Accordingly, we evaluated whether the CJ alteration under OPA1/MICU1 knockdown have implications on cristae and IMS Ca 2+ levels. For this purpose we exploited recently developed Ca 2+ sensors referred to as CL-GEMGeCO1 and IMS-GEMGeCO1 (M.W.-W., B.G., C.T. M.-S., C.K., H.B., E.E., R.M., W.F.G., manuscript submitted) to measure the Ca 2+ concentrations in the CL and the IMS. MICU1 and OPA1 knockdown increased basal Ca 2+ levels in the CL (Fig. 5c, Supplementary Fig. 22) but not IMS (Fig. 5d, Supplementary Fig. 22). However, the Δ-ratio values after cell stimulation with histamine remained unchanged in the CL (Fig. 5c, Supplementary Fig. 22) and the IMS (Fig. 5d, Supplementary Fig. 22). In line with elevated basal cristae Ca 2+ , basal mitochondrial matrix Ca 2+ was significantly higher in OPA1-and MICU1-depleted cells ( Supplementary Fig. 23). These results point to an increased CJ permeability for Ca 2+ in cells lacking OPA1 or MICU1 (Fig. 5e).   . b Schematic representation of the impact of MICU1 or OPA1 depletion on ψ mito caused by enlarged cristae junction. c HeLa cells were transfected with the cristae Ca 2+ sensor CL-GEMGeCO1 and Control (black), OPA1 (blue), or MICU1 (magenta) siRNA. Cells were challenged with 100 µM histamine to induce ER Ca 2+ release (n Control si = 9; n OPA1 si = 10; n MICU1 si = 10). d HeLa cells were transfected with the IMS Ca 2+ sensor IMS-GEMGeCO1 and Control (black), OPA1 (blue), or MICU1 (magenta) siRNA. Cells were challenged with histamine (n Control si = 8; n OPA1 si = 8; n MICU1 si = 7). e Schematic representation of the impact of loss of MICU1 or OPA1 on the sub-mitochondrial Ca 2+ redistribution reasoned by enlarged cristae junction. f HeLa cells were co-transfected with green IMS-or CL-GEMGeCO1 and red CARGeCO1 or mt-CARGeCO1 for a temporal correlation of Ca 2+ elevations upon stimulation with 100 µM histamine in the absence of extracellular Ca 2+ . Average curves of dynamic Ca 2+ increases in the cytosol (blue curve), the IMS (green curve), the CL (orange curve) or the mitochondrial matrix (red curve) in comparison with a representative sigmoidal fitted MICU1-FRET curve obtained from published data (black) 12  Spatiotemporal correlation of sub-mitochondrial Ca 2+ entry. The kinetics of sub-mitochondrial Ca 2+ signaling upon histamine were analyzed using the respective Ca 2+ sensor ( Supplementary  Fig. 24). As shown in Fig. 5f, cytosolic Ca 2+ elevation was instantly followed by Ca 2+ elevations in the IMS. The Ca 2+ signal in the CL developed slowly but reached maximum in the same time than the cytosol and the IMS (Fig. 5g). This delayed onset of CL Ca 2+ elevation correlated with the rearrangement of MICU1 heteromers (MICU1-FRET data are reused from a previous publication 12 ). Elevation of matrix Ca 2+ was delayed compared with all other regions and showed the highest diversity within cells (Fig. 5f, h, Supplementary Fig. 24).
Upon Ca 2+ mobilization MICU1 anchors MCU/EMRE at the IBM. The discovery of MICU1 10 was followed by the identification of MCU 5,6 that co-immunoprecipitate with each other 5 . We next determined the distribution of MCU within the IMM. Using MCU-mCherry together with MICU1-YFP we found that, in contrast to MICU1, MCU is uniformly distributed throughout the entire IMM in resting cells (Fig. 6a, b). However, after stimulation with histamine, MCU joins MICU1 to predominantly localize in the IBM. Ca 2+ -induced redistributions of MCU were not observed in cells exclusively expressing the ΔC-mutant of MICU1 that does not bind MCU 20 (Fig. 6a, b, Supplementary Fig. 25). Excitingly, mitochondrial matrix Ca 2+ elevation upon histamine strictly correlates with the shift in the distribution of MCU to the IBM (Fig. 6c, d). Our results additionally revealed that the redistribution of MCU is independent of extracellular Ca 2+ (Supplementary Fig. 26) and requires MICU1 ( Supplementary  Fig. 27). We further performed experiments to validate the MCUshuttling process and mitochondrial morphology in HEK, A549 and MCF-7 cells (Supplementary Fig. 28). All cell types showed the same phenomena of Ca 2+ -induced MCU shuttling to MICU1 in the IBM. Notably, upon histamine, HeLa and HEK cells undergo mitochondrial fission, MCF-7 cells show slight swelling but no fragmentation, and A549 cells did not change their morphology ( Supplementary Fig. 29). These findings indicate that cytosolic Ca 2+ triggers anchoring of MCU to MICU1 in the IBM in non-excitable cell lines. Next, we mapped the distribution of the second pore-forming protein of the MCU-Complex, EMRE 6 . An EMRE-mCherry construct was co-expressed with MICU1-YFP. Under basal conditions, EMRE was distributed throughout the entire IMM and this distribution did not change upon stimulation with histamine ( Fig. 7a, b). Since EMRE-MCU stoichiometry could contribute to this result, we pursued experiments involving the co-expression of MCU, MICU1-YFP, and EMRE-mCherry. Under these conditions, EMRE accumulated in the IBM upon Ca 2+ mobilization (Fig. 7c, d). The physiological functionality of EMRE-mCherry was validated by recovery experiments in EMRE-KO cells ( Supplementary Fig. 30). Furthermore, coimmunoprecipitation of EMRE-mCherry-HIS and MCU-mCherry-HIS with MICU1-FLAG showed direct interaction of these constructs (Supplementary Fig. 30). Notably, MICU1-YFP expression did not change the arrangement of the IMM-marker MTR 33 , indicating that the integrity of the IMM was maintained throughout our experimental conditions ( Supplementary Fig. 31).
Our next experiments examined the distribution of UCP2-mCherry in Mitotracker Green (MTG)-stained cells. For this purpose we used HeLa cells that have high levels of methylated MICU1 and therefore require UCP2 for MCU-Complex activation 21 . Under basal conditions UCP2 localizes in the entire IMM and accumulates in the IBM upon histamine (Fig. 8b, c). Depletion of either MICU1 or PRMT1 prevented this Ca 2+ -evoked redistribution of UCP2 (Fig. 8b, c), pointing to a dynamic interaction of methylated MICU1 with UCP2. Overexpression of PRMT1 did not change the shuttling characteristics of UCP2 or MCU in HeLa or UCP2-KO cells ( Supplementary  Figs. 33 and34). In agreement with our previous work 21 , depletion of PRMT1 in cells expressing the methylation mimicking MICU1-R455F-YFP mutant did not affect the accumulation of UCP2 upon histamine, while an expression of the methylation-resistant MICU1-R455K-YFP did not show any UCP2 translocation to the IBM ( Supplementary Fig. 35). Neither UCP2-mCherry nor PRMT1 knockdown affected mitochondrial morphology in WT-MICU1-YFP, MICU1-R455F-YFP or MICU1-R455K-YFP expressing cells ( Supplementary Fig. 36).

Discussion
Considering the Ca 2+ -induced rearrangement of MICU1 20 as an essential step in the activation of the MCU-Complex 12 , the concept that the MCU-Complex consists as a rather stable complex awaiting its activation by elevation of Ca 2+ in the IMS is challenged 9 . Notably, our recent work, introducing superresolution SIM on sub-organelle membrane dynamics, described a unique regulation of cristae dynamics by spatial Ca 2+ inside the MAMs 28 . In this work, we utilized live-cell SIM and combined it with transmission electron microscopy 34 and newly designed tools for spatial Ca 2+ measurements to investigate the spatial dynamics of the MCU-Complex and the organization of the IMM under resting conditions and upon cell stimulation.
Applying dual-color SIM, a surprising shell-like distribution of MICU1 but not of the other MCU-Complex components MCU, EMRE and UCP2 was found that markedly differed from the IMM structure. The localization of MICU1 was confirmed lately by Proteinase K and carbon extraction assays to be at the outer leaflet of the IMM facing the IMS 35 . However, the comparison with OMM marker TOM22 revealed a distance of 21 ± 13 nm (mean ± SD) between this protein and MICU1, which is consistent with reported measures using electron microscopy 22 . These data indicate that MICU1 20 is exclusively localized in non-cristae structure of the IMM, referred as IBM 13 . Because MICU1 remained in the IBM even upon intracellular Ca 2+ release that triggers rearrangement of MICU1 to dimers 12 , it is tempting to speculate that the specific location of MICU1 in the IBM does not depend on the status of oligomerization but the protein's structure.
Because depolarization of the IMM by oligomycin A and antimycin A yielded a redistribution of MICU1 into the entire IMM, we speculate that MICU1 because of its positively charged poly-lysine domain may not enter the cristae which is positively charged by the accumulation of H 3 O + inside the cristae due to the activity of the respiration chain [13][14][15] . Our findings that all MICU1 variants (i.e. MICU 1-140 , MICU1-F, MICU1-K, MICU1-ΔEF, ΔC-MICU1) that contain the poly-lysine domain (Fig. 2a) are located to the IBM while the poly-lysine domain lacking MICU1 1-70 mutant was found throughout the entire IMM, indicate that sub-mitochondrial localization of MICU1 depends on the poly-lysine domain and its ability of electrostatic repulsive power. A potential interaction partner of MICU1 for spatial electrostatic interaction might be cardiolipin 25 27 . The high sensitivity of MICU1 to depolarization might explain the ubiquitous IMM distribution of MICU1 in APEX2-based electron microscopy, in which depolarization during fixation might have influenced the sub-organellar localization of MICU1 (ref. 37 ).
MICU1 exists either as dimer or as hexamer via its C-terminal oligomerization site 20 . We found that ΔC-MICU1, which does not form multimers 20 , is restricted to the IBM in the presence of endogenous MICU1 but distributed through the entire IMM in cells depleted of endogenous MICU1. These data indicate that the oligomeric MICU1 affects the distribution of the ΔC-MICU1 mutant by stabilizing the CJs. The CJ is known to be established by several proteins, including OPA1/OMA1 16,17 . Our findings that knockdown of OPA1 did not change the exclusive IBM location of wild-type MICU1 but facilitated the redistribution of ΔC-MICU1 into the CM even in the presence of endogenous MICU1 suggest a model in which the sub-mitochondrial localization of MICU1 to the IBM is dependent on the stability of the CJ. In turn, MICU1 regulates the CJ stability pointing to an IBM self-containing function of MICU1. This assumption on a new feature of MICU1 is further supported by our electron microscopy data that revealed a MCU/EMRE-independent widening of the CJ by~58% under conditions of MICU1 knockdown, similar to the depletion of OPA1 38,39 .
The contribution of MICU1 to CJ integrity is further supported by a series of functional assessments and respective reports in the literature: first, similar to our recent findings that widening of the CJ by knockdown of OPA1 yields reduction in CM-kinetic 28 , MICU1 knockdown also reduced CM kinetics. Second, diminution of MICU1 resulted in higher mitochondrial fragmentation and lower branching, an effect that also occurs upon depletion of OPA1 28,40 . Third, our data reveal that similar to a knockdown of OPA1, depletion of MICU1 yields relocalization of cytochrome c from the cristae into the IMS (Fig. 4g). These findings are coherent with reports showing that knockdown of OPA1 and MICU1 enhance susceptibility for the induction of apoptosis 17,38,41,42 . Fourth, we demonstrate that knockdown of either OPA1 or MICU1 results in mitochondrial membrane depolarization. The known localization of ATPase and proteins of the electron transport chain within the CM 14,15 yields a proton accumulation within the CL 24 . Accordingly, IMM depolarization upon knockdown of OPA1 or MICU1 may be a consequence of the CJ widening and a subsequent dilution of the CM's proton gradient (Fig. 5b). Fifth, our findings that the depolarization upon MICU1/OPA1 knockdown is prevented by expression of MICU1-YFP, but not by the C-terminal lacking ΔC-MICU1-YFP mutant support the assumption of the contribution of MICU1 to CJ   (Fig. 5e).
Our data demonstrate that, in contrast to the exclusive IBM distribution of MICU1, MCU and EMRE localize to the entire IMM. However, after stimulation with histamine MCU colocalized with MICU1 in the IBM, thus, indicating a dynamic assembly of the MCU-Complex upon cytosolic/IMS Ca 2+ elevation exclusively in the IBM. Our findings that EMRE-mCherry only relocalizes into the IBM upon histamine stimulation if MCU was simultaneously overexpressed, points to a strict stoichiometry for these proteins. Co-immunoprecipitation experiments show a direct interaction of MICU1 and EMRE, potentially by electrostatic interaction of the poly-basic domain of MICU1 with the Cterminus of EMRE 45,46 . That disagrees with our finding that EMRE is only shuttling to the IBM in a MICU1 and MCUdependent manner. Vais et al. showed that the C-terminal tail of EMRE is oriented into the mitochondrial matrix, not able to interact with IMS localized MICU1 47 . Thus, we speculate that after the lysis process for protein extraction, integrity of membrane structures is affected, thus, misleadingly allow interactions of proteins that do not interact under physiological conditions due to their trans-membrane orientations. In case of EMRE/ MICU1, this possibility is even more likely since both proteins are part of the MCU-Complex and are in close proximity. Furthermore, the poly-basic domain of MICU1 is potentially crucial for the interaction of MICU1 with IMM phospholipids (e.g., cardiolipin 25 ) and its IBM localization.
The specificity of MCU/EMRE redistribution is indicated by our findings that MCU/EMRE do not relocalize in the exclusive presence of ΔC-MICU1 that does not bind MCU 20 . The dynamic and reversible assembly of MCU-Complex to harbor MCU by the MICU1 dimers in the IBM is further demonstrated by our results on the strict correlation of mitochondrial matrix Ca 2+ elevation with the IBM localization of MCU. Consistent with our findings, a recent report describes that the permanent activation of MCU by hydrogen peroxide also causes its accumulation in the IBM 48 . Our results also show that the redistribution of MCU/EMRE is independent of extracellular Ca 2+ but requires a rise in cytosolic/ IMS Ca 2+ and the presence of MICU1. Our data demonstrate that IMS Ca 2+ increases trigger the dissociation of MICU1 12,20 that, subsequently serves as diffusion trap for MCU that, subsequently, anchors EMRE. While the shuttling process of MCU was confirmed by us in different non-excitable cells (HeLa, HEK, A549, or MCF-7 cells), excitable cells like neurons or cardiomyocytes might rely on different regulatory mechanisms for mitochondrial Ca 2+ uptake (e.g. various stoichiometries of the MCU-Complex members) 49 , posttranslational modifications (phosphorylation, methylation, and oxidation) 21,35,48 , or the abundance of cardiolipin 3 .
Our recent findings demonstrate that PRMT1-mediated asymmetric methylation of MICU1 decreases its Ca 2+ binding affinity yielding reduced mitochondrial Ca 2+ uptake 21 . However, UCP2 specifically binds to methylated MICU1, and, consequently, restores its Ca 2+ sensitivity and, thus, MCU-Complex activity 21 . In the present work using HeLa cells with high PRMT1 activity 21 , no MCU shuttling was found in UCP2-KO HeLa cells after cell stimulation. However, knockdown of PRMT1 recovered MCU shuttling in UCP2-KO cells. These data confirm our previous results and further demonstrate that mitochondrial Ca 2+ uniporter function depends on UCP2 in cell types with elevated PRMT1 activity 21 . The sub-mitochondrial localization of UCP2 has not been previously explored. While UCP4 can localize to the IBM 24 , our data revealed that UCP2 is in the entire IMM under resting conditions and translocates upon cell stimulation into the IBM in a MICU1-and PRMT1-dependent manner.
In accordance to our previous work 21 , we examined the role of MICU1 methylation for UCP2 translocation using MICU1-R455F and MICU1-R455K mutants that either mimic or prevent methylation of R455, respectively 21 . Our findings that UCP2 translocation to the IBM upon ER-Ca 2+ release was independent of PRMT1 when MICU1-R455F was co-expressed but disabled in the presence of MICU1-R455K, indicate that methylation of MICU1 at R455 is prerequisite for the relocation of UCP2 to the IBM. They also confirm a fundamental role for UCP2 in the assembly and activation of the MCU-Complex under conditions of elevated PRMT1 activity 21,50 .
Overall, our present data change the current model of organization, localization and processes of stimulation of the MCU-Complex (Fig. 8d). In particular, our findings demonstrate that MICU1 is exclusively localized in the IBM due to its hexamerization and poly-lysine domain. Hence, MICU1 exhibits essential function by contributing to CJ stability. By this newly discovered function of MICU1, the protein is involved in maintaining ψ mito , protection against matrix Ca 2+ overload, and restriction of cytochrome c to cristae. Therefore, a new model of MCU-Complex activation is presented in which MICU1 dissociates into dimers serving as diffusion trap for MCU in the IBM. The IBMharbored MCU subsequently anchors EMRE, thereby establishing functional MCU-Complexes exclusively at this location. PRMT1 activity and MICU1 methylation prevent MICU1 dimerization and MCU-Complex assembly, but UCP2 that interacts with methylated MICU1 facilitates MCU-Complex assembly in the IBM. This new model seamlessly meshes existing data with new findings and provides an advanced perspective on the dynamic organization of mitochondrial ion regulation and how MICU1 contributes to CJ integrity.

Methods
Single and dual camera SIM imaging. The SIM setup used is composed of a 405-, 488-, 515-, 532-and a 561-nm excitation laser introduced at the back focal plane inside the SIM box with a multimodal optical fiber. For super-resolution, a CFI SR Apochromat TIRF ×100-oil (NA 1.49) objective was mounted on a Nikon-Structured Illumination Microscopy (N-SIM) System with standard wide field and SIM filter sets and equipped with two Andor iXon3 EMCCD cameras mounted to a Two Camera Imaging Adapter (Nikon Austria, Vienna, Austria). For calibration and reconstruction of SIM images the Nikon software Nis-Elements was used. Reconstruction was always done with the same robust setting to avoid artifact generation and ensures reproducibility with a small loss of resolution of 10% compared to most sensitive and rigorous reconstruction settings. Prior to each measurement, laser adjustment was checked by projecting the laser beam through the objective at the top cover of the bright field arm of the microscope. Lasers were aligned and focus using the interlock system screws at the N-SIM box to ensure appropriate illumination of the sample. 100 nm Tetraspec beads (Invitrogen, Thermo Fischer Scientific, Vienna, Austria) were diluted 1:200 in 1 ml 0.01% poly-L-lysine and incubated on the 1.5 H high-precision glass cover slips (Marienfeld-Superior, Lauda Königshofen, Germany) for 20 min. Afterwards, the plate was washed once and transferred to the live-cell chamber containing 1 ml of buffer (2CaB) containing in mM: 2 CaCl 2 , 140 NaCl, 5 KCl, 1 MgCl 2 , 1 HEPES and 10 D-Glucose, pH adjusted to 7.4 (all buffer salts were obtained from Roth, Graz, Austria). Ring correction was done as follows: 3D stacks of beads were acquired to verify the point spread function of the system. Potential asymmetric pointspread function was corrected using the objective correction ring. Subsequently, grating block adjustment was performed to find the perfect focus for the laser beam interference at the focal plane. This process was automatically run by the Nikon software. To align both channels for parallel dual-color experiments NIS-Elements Two-CAM registration was used taking the same Tetraspec bead samples. Error estimation in x, y and z-axes caused by chromatic abbreviations was done by imaging multiple z-stacks of Tetraspec beads. The exact localizations of the beads in both colors were determined with sub-pixel accuracy using the imageJ plugin Quickpalm. For each bead, the x and y abbreviation was calculated and defined as a vector. The relative error of chromatic aberration was determined by the vector product of each bead with its nearest neighbor normalized to the distance of the beads. The alignment of both channels along the z-axes was done by selecting each bead in a 320 × 320 nm square, projecting it on the x-z axes, and determining the maximum intensity in both channels with sub-pixel accuracy using a Gaussian-fit in ImageJ (Supplementary fig. 5). Cells were transiently transfected with the respective constructs or siRNAs and stained, if necessary, with Mitotracker Green/ FM (MTG), Mitotracker Red/FM (MTR), or Mitotracker Red CMXROS (MTR-CMX) (Invitrogen, Thermo Fischer Scientific, Vienna, Austria). Simultaneous dual channel 3D-SIM images were acquired 90 s after placing cells in 2CaB. Experiments with MTG-stained mitochondria were not possible with the dual bandpass filter setup because of fluorescence bleed through into the red channel. Sequential dualcolor 3D-SIM images were performed instead. To analyze for z-axes chromatic dispersion Hela cells were stained with MTR and MTG simultaneous and z-stacks with 60 nm increments using the dual-SIM system were acquired. The AUC (area under the curve) was measured in z-projections of single mitochondria and using linear interpolation for both labels the z-position was determined at which the AUC/2 is reached.
Cloning of constructs. For generating mCherry or YFP-tagged constructs, the respective coding sequences were subcloned into backbones already containing the sequences for the FPs 51 . Tom22 was fused to N-terminal mCherry (mCherry-Tom22), and, cloned into a pcDNA3.1(−) backbone flanked by ClaI and HindIII restriction sites. All other constructs were fused to C-terminal mCherry (EMRE-mCherry, UCP2-mCherry, MCU-mCherry) and subcloned into a pcDNA3.1 (+) backbone. EMRE and MCU are flanked by HindIII and EcoRI, UCP2 by KpnI and EcoRI and ΔMICU1 by BamHI and EcoRI. For SIM measurements the plasmids encoding for either wild-type MICU1, MICU1 R455K (MICU1-K), MICU1 R455F (MICU1-F) or EF hand mutated MICU1 (MICU1-EF) were C-terminally fused to a citrine (YFP) as described before 12,21 . MICU1 lacking its C-terminal domain (MICU1ΔC) was fused in frame to the same YFP via BamHI and EcoRI restriction sites in a pcDNA3(+) vector using the primers MICU1 pos. 1 forward 5′-CCCGGATCCATGTTTCGTCTGAACTCACT-3′ and MICU1 pos. 1332 reverse 5′-GGTTGAATTCCATCAGCCGTTGCTTCATGAT-3′. MICU1 1-70 and MICU1 1-140 were inserted in frame via XhoI/EcoRI in a pcDNA3.1(−) vector containing a circular permutated enhanced green fluorescent protein (cpEGFP) 52 with the primers as follows: MICU1 pos. 1 forward; MICU1 pos. 210 reverse; MICU1 pos. 420 reverse. For sub-organelle-specific Ca 2+ measurements within the mitochondrial cristae (MC) a mutated version of reactive oxygen species modulator 1 (ROMO1) was synthesized by exchanging potential ROS modulating cysteines at pos. 15, 27 and 42 to serines from General Biosystems Inc. (Morrisville, NC, USA). ROMO1 and the IMS specific targeting sequence of MICU1 1-140 were then fused to a ratiometric GEMGeCO1 (Addgene; Cambridge, MA, USA) via XhoI/BamHI in a pcDNA3.1(−) with the primers ROMO1 pos. 1 forward; ROMO1 position 255 reverse; MICU1 pos. 1 forward and MICU1 position 420 reverse. For immunohistochemistry of MCU-mCherry and EMRE-mCherry in frame with a myc-HIStag we constructed the myc-HIS-tag to the C-terminal end of the mCherry in a series of PCR amplifications by using overlapping reverse primers. The resulting mCherry-myc-His-tag fragment was then subcloned in fusion with either MCU or EMRE via EcoRI and HindIII restriction sites. All respective primers can be found in the Supplementary Table 1.
Generation of UCP-KO and EMRE-KO cells. For generating UCP2-KO or EMRE-KO cells we ordered pcDNA3.3-Cas9-2A-eGFP and pGS-U6-gUCP2/pGS-U6-gEMRE from Genscript (Thermo Fischer Scientific). The gRNA sequence of UCP2 is 5′-CCCAGTACCGCGGTGTGAT-3′ and for EMRE 5′-GCTAGTATTGG-CACCCGTC-3′. For homologous recombination in case of UCP2 we cloned an 800-bp upstream fragment (HindIII/EcoRI) and an 800-bp downstream fragment (Bsm I) of the genomic UCP2 cutting side by Cas9 into pcDNA3.1 flanking neomycin cassette. All three plasmids were co-transfected into HeLa cells in a ratio of 1:1:1. After 48 h, GFP-positive cells were single sorted into 96-well plates. Clones were further cultivated and selected with G-418 (Sigma-Aldrich) for two weeks in 12-well plates. Nearly 50 clones were analyzed with PCR. To check if each allele of 5 clones was mutated, we made bacterial sub-clones of amplified PCR fragments and sequenced them using the Sanger method (Microsynth, Balgach, Switzerland).
For generation of EMRE-KO, HeLa cells were cotransfected with pcDNA3.3-Cas9-2A-eGFP and pGS-U6-gEMRE. Next day we made single-cell sorting of GFPpositive cells into 96-well plates. Clones were further cultivated and analyzed with PCR. To check if each allele was mutated, we made bacterial sub-clones of amplified PCR fragments and sequenced them with the Sanger method (Microsynth, Balgach, Switzerland).
Transfection and siRNA treatment. HeLa cells were grown under standard culture conditions until 50% confluence was reached, transfected in DMEM (without FCS and antibiotics) with 1.5 µg/ml plasmids or 100 nM siRNA using 2.5 µg/ml TransFast transfection reagent (Promega, Madison, WI, USA) 10 . After 24 h, the medium was replaced with DMEM containing 10% FCS and 1% penicillin/streptomycin and kept for further 24 h before experiments. The specific siRNAs (Microsynth) used in this study are listed in Supplementary Table 1.
Cytochrome c staining. After staining the cells with MTR-CMX cells were washed with PBS, fixed with 3.7% paraformaldehyde in PBS for 15 min and permeabilized with 0.25% (v/v) Triton X-100/PBS for 15 min at RT. Following permeabilization, cells were washed twice with PBS and nonspecific absorption was blocked using UltraVision blocking reagent (Thermo Fisher Scientific) for 10 min at RT. Then, cells were incubated at 4°C overnight with anti-Cytochrome c antibody (Cell Signaling, # 12963; 1:100 in Antibody Diluent, Dako). After washing with PBS, Alexa488-labeled goat-anti-mouse IgG (Invitrogen, # A11001; 1:300 in Antibody Diluent) was used as secondary antibody (60 min, RT). Slides were stored in 1% (w/ v) BSA/PBS until further use.
Sub-mitochondrial localization of UCP2, MCU, EMRE and TOM22. Cells transfected with EMRE-mCherry, MCU-mCherry, UCP2-mCherry or mCherry-TOM22 were stained with MTG and imaged using 3D-SIM microscopy. Images were background subtracted with an ImageJ Plugin (Mosaic Suite, background substractor, NIH). The width of mitochondria was measured using the full width at half maximum (FWHM). Line plots were taken for the images in ImageJ and the FWHM was determined.
Assigning MICU1 localization to inner mitochondrial membrane. Dual-color 3D-SIM images of MICU1-YFP and mCherry-TOM22 transfected cells in 2CaB were background subtracted with an ImageJ Plugin (Mosaic Suite, background substractor, NIH). Blind line plots of mitochondrial cross-sections within the mCherry-TOM22 label were done. The outer mitochondria membrane can be modeled as a hollow tube that, if stained, displays a double line structure in zprojection 53 . Intensity line plots of MICU1-YFP and mCherry-TOM22 were fitted to a double Gauss-function. The distance of the peak values of these fits can be interpreted as the diameter of the mitochondria and were determined for both channels. By bisecting the difference of peak distances of MICU1-YFP and mCherry-TOM22 staining the distance of both labels on one side can be determined. Using the hollow tube model and the Gauss fitting method, the precision for estimating the center localization of MICU1 and TOM22 distributions can be increased beyond the actual resolution of SIM microscopy as long as the mitochondrial double line structure in images or the double peak structure of mitochondrial cross-section intensity plots is present. As control cells were transfected with MICU1 YFP and -mCherry. For image analysis the freeware program ImageJ (NIH, MA, USA) was used. Analysis of protein sub-mitochondrial localization. Dual-color images were split into two separate channels, background corrected with an ImageJ Plugin (Mosaic Suite, background subtractor, NIH) and further described as object and reference channels. The reference channel was Otzu auto thresholded 54 and further single dilated (iterations = 1-4, count = 1) and single treated with an erosion algorithm (iterations = 2-6, count = 1). Erosion, dilation iterations and counts were changed according to the experiment to get the highest sensitivity possible for each experimental set. More detailed information about the parameters used and comparability between the experiments is listed elsewhere (Supplementary Table 2). Different settings of erosion and dilation parameters were necessary to extract the sub-mitochondrial localization using different reference markers with variable mitochondrial localization patterns. In case of MICU1-YFP as the reference after auto thresholding, a fill holes step was added. Pixel-wise subtraction of the erosion reference of the dilated reference image yields in a hollow structure. The hollow structure was further used as a template to determine the fluorescence intensity in the IBM while the erosion reference served as template for the bulk or cristae fluorescence intensity. The measured fluorescence intensities were normalized to the area covered by the cristae and IBM template. The ratio of IBM/CM mean intensity is a value to estimate changes of the object label distribution inside a mitochondrion and termed IBM association index. The higher the ratio value the higher the distribution of protein label in the IBM. For image analysis the freeware program ImageJ was used.
Electron microscopy and analysis of cristae junction width. Cells were washed in PBS, fixed with 2.5% glutardialdehyde and 2% formaldehyde in a buffered solution and postfixed in 1% osmium tetraoxide that had been reduced with 1% potassiumhexacyanoferrate 34 . The cells were dehydrated in an ascending ethanol series, embedded in TAAB embedding resin, and sectioned on a Leica Ultracut 7 ultramicrotome using a Diatome diamond knife. The sections were counter stained using platinum blue (IBIlabs) and lead citrate (Leica) and visualized in an FEI Tecnai 20 transmission electron microscope. They were photographed at ×27,000 magnification with a Gatan ultrascan 1000 camera. To quantitatively analyze the electron microscopically images a line was drawn in imageJ manually into the cristae starting with the cristae junction proceeding into the cristae volume as far as the image quality was sufficient for image quantification, the curvature of the cristae was not crossing the line or a maximum of 130-nm length was reached. Along the drawn line every 2 nm orthogonally line plots with a width of 10 nm were measured with an ImageJ macro starting from the CJ. Each line plot was halved and the position of the minimal intensity was determined for both sides. The distance of both minimal intensities was set as the distance between opposing sides of the cristae.
Morphological analysis of mitochondria imaged with SIM. Single 3D-SIM and time-lapsed images of EMRE-mCherry, MTR, MTG, or MICU1-YFP were used for morphological analysis over time. Images were background corrected with an ImageJ Plugin (Mosaic Suite, background substractor, NIH) and a binarization was done using a Yen auto threshold 55 . The ImageJ particle analyzer was used to extract the mitochondrial count (c), area (a), perimeter (p), minor (x) and major (y) axes of the mitochondria. Aspect ratio (AR) was determined as AR ¼ y x The form factor (FF) was determined as follows: FF ¼ p 2 4πÁa . For image analysis the freeware program ImageJ was used.
3D-morphological analysis of mitochondria. z-stacks of mitochondria with 0.2µm increments were imaged with a Zeiss Observer Z.1 inverted microscope equipped with a Yokogawa CSU-X1 Nipkow spinning disk system, a piezoelectric z-axes motorized stage (CRWG3-200; Nippon Thompson Co., Ltd., Tokyo, Japan), and a CoolSNAP HQ2 CCD Camera (Photometrics). The image stacks were deconvoluted using blind deconvolution (NIS-Elements, Nikon, Austria). Morphology parameters were measured automatically with a custom made ImageJ macro using the following procedure. An additional background subtraction using the rolling ball method was used to further increase contrast for later analysis. Both a global auto Otsu threshold using the stack histogram as well as a local auto Otsu threshold (radium of 640 nm) using the single slice histogram were applied to the stack and merged. The ImageJ plugin 3D manager was used to segment the binarized mitochondria. With the plugin 3D Geometrical Measure the mitochondrial volume and surface were determined. The Plugin 3D Ellipsoid Fitting generated an ellipsoid fit of the mitochondria to measure elongation and flatness parameters.
CM kinetics quantification. HeLa cells transfected with ER-RFP and stained with MTG were recorded with live dual-color SIM imaging. Analysis of the CM and IBM kinetics was done as described elsewhere 28 . In short, using a combination of local and global automated thresholding of the MTG staining the CM-membrane kinetics were quantified by measuring the CM alterations pre frame within the mitochondria. The overlap of ER-RFP and MTG staining was defined as MAMs.
The CM kinetics were determined inside the MAMs using the overlap of ER and mitochondrial staining as a mask.
Mitochondrial shape descriptors and co-localization. HeLa cells transfected with ER-RFP and stained with MTG were recorded with live dual-color SIM imaging. The mitochondrial branching was measured by determining the area weighted FF and mitochondrial size using ImageJ 28 . For co-localization studies the ImageJ plugin coloc2 was used to measure the Pearson coefficient.
Cell viability and apoptosis assay. Twenty-four hours after transfection in 10 cm dishes, cells were seeded in 96-well plates at a density of 5000 cells per well. Cell viability was measured using CellTiter-Blue assay (Promega) and apoptotic caspase activity via Caspase-Glo3/7 assay (Promega) following the standard protocols 56 .
Live cell Fura-2, TMRM, and mitochondrial Ca 2+ measurements. Measurements were performed on an inverted and advanced fluorescent microscope with a motorized sample stage (Till Photonics, Graefling, Germany). The probes were excited via a polychrome V (Till Photonics), and emission was visualized using a ×40 objective (alpha Plan Fluar 40, Zeiss, Goettingen, Germany), and a chargecoupled device camera (AVT Stringray F145B, Allied Vision Technologies, Stadtroda, Germany). Fura-2 (Invitrogen) was alternately excited at 340 and 380 nm, and emissions were captured at 515 nm (495dcxru; Omega Optical, Brattleboro, VT, USA). TMRM (cat: T668; Invitrogen) was excited at 550 nm, and emissions were captured at 600 nm (59004; Chroma, Bellows Falls, VT, USA). Mitochondrial Ca 2+ FRET measurements with 4mtD3cpv were done using 430 nm and 480 nm excitation and emission was captured at 480 and 535 nm (69008; Chroma), respectively 12 . For control and acquisition, the live acquisition 2.0.0.12 software (Till Photonics, Munich, Germany) was used. For Fura-2 imaging cells were incubated in loading-buffer containing 3.3 µM Fura-2 for 40 min 56 . TMRM loading was done likewise for 40 min with a TMRM concentration of 100 nM 57 . For mitochondrial Ca 2+ measurement cells were transfected with the 4mtD3cpv-Plasmid 9 . All recordings were background subtracted using a background ROI and corrected for bleaching using an exponential decay fit. Mitochondrial targeted R-GECO was used to measure mitochondrial Ca 2+ using the single-channel SIMsetup in wide field mode. Time-lapse images were background subtracted using a background ROI and corrected for bleaching using an exponential decay fit.
Live cell measurement of IMS and CL Ca 2+ . Ratiometric green emitting genetically encoded Ca 2+ indicators for optical imaging (GEMGeCO1) fused to either ROMO1 or MICU1 1-140 was excited at 430 nm and emissions were collected by using a conventional CFP/YFP dichroic filter at 480 and 535 nm. The resulting ratio was calculated by the division of the fluorescence intensity at 480 nm through 535 nm. CARGeCO1-based Ca 2+ indicators (cyto-CARGeCO1 and mt-CAR-GeCO1) were excited at 575 nm and emitted at 600 nm. For simultaneous measurements, GEM-and CARGeCO1 targeted sensors were alternately excited for 400 ms each at 430 and 575 nm. Emissions derived from both sensors were taken in 3 s intervals. Alternatively, we used an ultrafast switching mode where both Ca 2+ indicators were excited for 150 ms in a 310 ms interval. All recordings were background subtracted using a background ROI and corrected for bleaching using an extrapolation of an exponential decay fit. The maximal slope values and duration of lag time evaluated from the onset of cytosolic Ca 2+ rise to the maximal slope were calculated.
mRNA Isolation and real-time PCR. Total RNA was isolated according to manufacturer's protocol using the PEQLAB total RNA isolation kit (Peqlab, Erlangen, Germany). cDNA was generated using a high-capacity cDNA reverse transcription kit (Applied Biosystems, Foster City, CA) and a thermal cycler (Peqlab). Knockdown and overexpression efficiency of MICU1 was examined by RT-PCR using a QuantiFast SYBR Green RT-PCR kit (Qiagen, Hilden, Germany) and a LightCycler 480 (Roche Diagnostics, Vienna, Austria). Relative expression of the gene of interest was normalized to GAPDH, as a reference gene. Primers for real-time PCR were designed by us and obtained from Invitrogen (Vienna, Austria). The respective primer sequences used can be found elsewhere (Supplementary Table. 1).
Co-immunoprecipitation. HeLa or HEK cells were transfected with MICU1-FLAG alone or in combination with MCU-mCherry-HIS or EMRE-mCherry-HIS constructs. The transfection process was repeated three times over a period of 3 days. Transfected cells were lysed in 0.7 ml lysis buffer (100 mM NaCl, 20 mM Tris, 1 mM EGTA, 5 mM n-Dodecyl β-D-maltoside, pH 7.5-HCl) supplemented with protease inhibitors (10 μM phenylmethylsulfonyl fluoride, and 1 μg/ml each aprotinin, leupeptin, and pepstatin). After a short freeze-thaw cycle lysates were centrifuged (12,000×g, 10 min) and 60 µl of supernatants were removed for protein determination using a BCA assay (Thermo-Fisher) and total cell lysate analyses (input). The rest of the lysates (~1.2 mg total protein of HeLa cells and~2 mg of HEK cells) were incubated with 30 µl of anti-Flag M2 affinity gel (Sigma-Aldrich; A2220-1ML; beads were prepared as recommended by the manufacturer) for 30 min on a rotary shaker. Then, beads were collected by centrifugation (8000 × g, 30 s) and washed three times with TBS. All steps were performed on ice or at 4°C. Elution of FLAG-fusion proteins was performed in 50 µl of 2× Lämmli sample buffer with 5% β-mercaptoethanol and boiling for 10 min. For immunoblotting, whole cell lysates (input, 120 µg of total protein) and Co-IP samples were subjected to SDS-PAGE. PVDF membranes were probed at 4°C overnight with primary monoclonal mouse anti-6x His-Tag antibody (Thermo Fisher; 1:1000; MA1-21315). To avoid interference from denatured heavy and light chain IgG, the HRPconjugated immunoblot reagent Veriblot (Abcam; 1:000; ab131366) was used for detection of co-immunoprecipitated proteins. For detection of protein expression in input samples, HRP-conjugated goat-anti-mouse secondary antibody (Thermo Fisher, 1:2500, 31430) was used. Immunoreactive bands were visualized using Immobilon Western HRP Substrate (Thermo Fisher) and the chemiluminescence detection system ChemiDoc (Bio-Rad).
Statistical analysis and reproducibility. Each exact n value and the number of independent experiments is indicated in each figure legend. Statistical analysis was performed using the GraphPad Prism software version 5.04 (GraphPad Software, San Diego, CA, USA) or Microsoft Excel (Microsoft Office 2013). Analysis of variance (ANOVA) with Bonferroni post hoc test and t-test were used for evaluation of the statistical significance. All Box-plots show minimum to maximum values. The central line is the median with boxes extending to 25 and 75% and the whiskers encompass all data. P < 0.05 was defined to be significant. At least three different experiments on three different days performed in at least triplicates have been performed for each experimental setup.
Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The data that support the findings of this study are available from the authors on reasonable request, see author contributions for specific data sets. All source data for the preparation of this manuscript are available from the authors on request. The source data underlying Figs