The mitochondrial calcium uniporter compensates for Complex I dysfunction


 Calcium (Ca2+) entering mitochondria potently stimulates ATP synthesis. Increases in Ca2+ preserve energy synthesis in cardiomyopathies caused by mitochondrial dysfunction, and occur due to enhanced activity of the mitochondrial Ca2+ uniporter channel. The signaling mechanism that mediates this compensatory increase remains unknown. Here, we find that increases in the uniporter are due to impairment in Complex I of the electron transport chain (ETC). In normal physiology, Complex I promotes uniporter degradation via an interaction with the uniporter pore-forming subunit, a process we term Complex I-induced protein turnover (CLIPT). When Complex I dysfunction ensues, contact with the uniporter is inhibited, preventing degradation, and leading to a build-up in functional channels. Preventing uniporter activity leads to early demise in Complex I-deficient animals. Conversely, enhancing uniporter stability rescues survival and function in Complex I deficiency. Taken together, our data identify a fundamental pathway producing compensatory increases in Ca2+ influx during Complex I impairment.


INTRODUCTION
Ca 2+ is a potent regulator of metabolism, acting on multiple enzymes in mitochondria. Within the matrix, moderate Ca 2+ elevations double ATP synthesis rates, helping match energetic supply to demand 1,2 . Pathological failure to meet demand is a common feature across in cardiomyopathies. In fact, energetic failure can be a primary cause of cardiomyopathy in mitochondrial diseases. Such diseases involve deficient oxidative phosphorylation, and arise from mutations in mitochondrial proteins encoded by either the nuclear or the mitochondrial genome (mtDNA), with Complex I of the ETC most affected 3,4 . Despite often severe pathology, children with mitochondrial cardiomyopathies may survive prolonged periods, suggesting mechanisms exist to compensate for ETC dysfunction. Identifying such pathways offers new opportunities for broad therapeutic intervention, as ETC impairment is a fundamental feature of many common cardiac and neurological diseases.
There are limited prior investigations of mitochondrial Ca 2+ signaling during ETC deficiency [5][6][7][8][9][10][11] . The typical finding is reduced or unchanged Ca 2+ uptake in the presence of a diminished membrane voltage gradient (ΔΨ), where the diminished gradient correlated with severity of ETC deficiency. Because Ca 2+ influx through the uniporter is driven by this voltage gradient, a change to either ΔΨ or uniporter activity can alter the size of Ca 2+ influx. In ETC-deficient mitochondria, precisely such diminished ΔΨ may mask any compensatory increases in uniporter activity using typical in vitro imaging assays. In fact, we noted an interesting phenotype in mice with a cardiacspecific deletion of the transcription factor for mtDNA, Tfam (transcription factor A, mitochondrial). This is a well-established model for mitochondrial cardiomyopathies, featuring a dilated cardiomyopathy caused by impaired transcription of core mtDNA-encoded ETC subunits, with Complex I function most severely affected 12,13 . Notably, as cardiomyopathy develops in the Tfam knockout, cardiac mitochondria become extremely Ca 2+ avid, which may preserve ATP synthesis 12 . The increase in mitochondrial Ca 2+ is mediated by a multi-subunit channel known as the mitochondrial Ca 2+ uniporter [14][15][16][17][18][19][20] . This channel resides in the inner membrane of mitochondria, is activated by cytoplasmic Ca 2+ , and is the main portal for Ca 2+ entry into the matrix. In animal models, cardiac impairment of this channel leads to energetic supply-demand mismatch, leaving open the question of whether increased uniporter activity is an essential compensatory mechanism during ETC impairment.
Here, we unravel the mechanism for the enhancement in uniporter activity during ETC dysfunction and show that it is essential for survival. We find that this phenomenon depends on impairment in Complex I and is widespread, occurring in a variety of cell types, and across species. Under normal Complex I activity, uniporter turnover is accelerated by an interaction between the Nterminal domain (NTD) of the pore-forming subunit of the uniporter (MCU) and Complex I, a mechanism we term Complex I-induced protein turnover (CLIPT). When Complex I becomes dysfunctional, CLIPT is abrogated, leading to slower MCU turnover and a buildup of uniporter channels. This mechanism is evident in hearts from Tfam knockout mice, and its disruption leads to their faster demise. Similarly, in Drosophila, Ca 2+ signaling through the uniporter is essential for survival during Complex I dysfunction. Enhancing this mechanism, by overexpression of MCU or its NTD, in Complex I deficient flies rescues both functional impairments and survival.

Complex I dysfunction leads to enhanced mitochondrial calcium uniporter levels.
Measuring Ca 2+ uptake using typical Ca 2+ fluorescence assays poorly captures uniporter activity during ETC dysfunction, because such dysfunction also alters other parameters controlling Ca 2+ transport, including ΔΨ, pH, morphology, and Ca 2+ buffering 8,21,22 . To more precisely examine uniporter-mediated Ca 2+ uptake, we used whole-mitoplast voltage-clamp electrophysiology 17 . In this assay, micropipettes are attached to individual spherical mitoplasts (mitochondria stripped of their outer membranes) to record ionic currents through a voltage-clamp feedback electrode. This allows full control over ΔΨ, matrix and external solutions, eliminating uncontrolled variation in the factors listed, and allowing direct measurement of uniporter Ca 2+ currents (IMiCa).
The complexes with Tfam-dependent subunits (Complexes I, III, and IV) may integrate into a supercomplex known as the respirasome. Therefore, we first tested whether enhanced IMiCa depended on disrupting the respirasome or a specific ETC complex. We examined HEK293T cells because long-term culture with drugs is possible, gene-edited lines exist for both uniporter 23 and Complex I analysis 24 , and uniporter currents are robust 15 . Individual ETC Complexes were inhibited pharmacologically for 2-3 days in HEK293T cells. To maintain viability, cells were cultured with 0.4 mM uridine to maintain pyrimidine biosynthesis and 2 mM pyruvate to regenerate NAD + 25, 26 . Whereas chronic inhibition with Complex III or IV antagonists (1 µM antimycin A, 200 µM sodium azide) produced no change in IMiCa (Fig. S1a, b), disruption of Complex I with rotenone produced a dose-dependent increase in IMiCa (Fig. 1a, b). Rotenone did not alter channel kinetics, and failed to increase IMiCa when added during mitoplast recordings (Fig. S1c, d). This suggests that rotenone did not enhance IMiCa directly, but through its effect on Complex I, by inhibiting electron transfer to ubiquinone. Similar to the effect in Tfam knockout hearts 12 , protein levels for uniporter subunits MCU and EMRE were increased (Fig. 1c), though gene expression was not (relative gene expression MCU, 0.89±0.18; EMRE, 1.37±0.4; MICU1, 1.03±0.15, n = 3).
Having established that the increase in uniporter currents observed in Tfam knockout cardiomyocytes could be reproduced by isolated Complex I inhibition, we tested the robustness of this phenomenon. To confirm the increase in IMiCa was due to Complex I and not an off-target rotenone effect, we measured IMiCa in HEK293T cells featuring deletion of the Complex I accessory subunit NDUFB10 or late assembly factor FOXRED1 (NDUFB10 KO , FOXRED1 KO , Fig. S1e) 24,27 . In these, IMiCa was also substantially increased over controls (Fig. 1d, e). We saw a similar increase in IMiCa after disruption of accessory subunit NDUFS4, which produces much milder Complex I deficiency (Fig. S1f) 10,24 , suggesting that even limited Complex I impairment can activate this mechanism. To see if IMiCa enhancement was prevalent in Complex I-mediated disease, we obtained fibroblasts from an infant with fatal lactic acidosis and cardiomyopathy due to NDUFB10 deficiency (NDUFB10 -/C107S ) 28 . Induced pluripotent stem cells (IPSCs) derived from these fibroblasts retained the compound heterozygous mutations and had essentially absent NDUFB10 expression ( Fig. S1g-i), validating their use for gauging uniporter activity in Complex I-deficient disease. IMiCa in these cells was also increased, and this effect was rescued by re-expression of wild-type NDUFB10 (Fig. 1f). Next, to see if this effect was evolutionarily conserved, we examined Drosophila with Complex I dysfunction in flight muscle, generated by the expression of short hairpin RNA (shRNA) targeting NDUFB10 via MHC-Gal4 (NDUFB10 RNAi ) 29 . Here too Complex I dysfunction was associated with an increase in IMiCa (Fig. 1g). Of note, matrix free [Ca 2+ ] levels were higher in the Complex I deficient cells (Fig. S1j), though there was substantial overlap with controls. Thus, one main purpose of enhancing IMiCa may be to maintain mitochondrial Ca 2+ levels when dysfunctional Complex I leads to a depolarized mitochondrial membrane potential and blunted Ca 2+ uptake through individual channels. Finally, we confirmed that the increase in current was specific to the uniporter, as the chloride current carried by the ubiquitous inner membrane anion channel showed no change after rotenone incubation in HEK293T cells (Fig. S1k). In summary, we show that disrupting Complex I leads to an approximately 2-3 fold increase in IMiCa. This effect can be triggered by a variety of insults to Complex I (Tfam knockout, rotenone, gene editing, shRNA, and congenital mutations), and occurs across species (mouse, Drosophila, human), cell types (cardiomyocytes, flight muscle, cultured lines, and IPSCs), and in human disease.
As with the Tfam knockout mice 12 and the rotenone-treated HEK293T cells, NDUFB10 KO , FOXRED1 KO , and NDUFB10 -/C107S cells also exhibited increased MCU protein, suggesting that the enhancement in IMiCa reflects an increase in channels (Fig. 1c). MCU levels increased in these different lines despite the absence of a corresponding mRNA upregulation (MCU expression versus control: 0.95, NDUFB10 KO ; 0.68, FOXRED1 KO ; 0.88, NDUFB10 -/C107S , n = 2). The discordance between MCU protein and mRNA levels may be explained by a post-transcriptional mechanism. To further verify this, we expressed carboxy-terminal Flag-tagged MCU in MCU KO HEK293T cells from a plasmid lacking the native promoter, introns, and other untranslated regions, disrupting endogenous transcriptional regulation 23 . Here too we found that IMiCa increased after rotenone incubation (Fig. 1h). Thus, our results suggest Complex I regulation of the uniporter is posttranscriptional.
Next, we investigated the signal that leads to IMiCa enhancement. When Complex I becomes dysfunctional, its NADH:ubiquinone oxidoreductase activity is impaired, leading to an increased NADH:NAD + ratio and greater superoxide production, which persist until excess ROS leads to selfinactivation and Complex I disassembly (Fig. 2a) 30,31 . We found evidence for increases in superoxide production using the mitochondrially-targeted sensor MitoSOX, and increased NADH:NAD + ratio using a mitochondrially-targeted version of the genetically-encoded SoNar sensor (Fig. 2b, c) 32 . To test if these were key signals for IMiCa enhancement, we turned again to whole-mitoplast electrophysiology. To blunt changes in the NADH:NAD + ratio, cell lines were engineered to stably express a mitochondrially-targeted water-forming NADH oxidase from Lactobacillus brevis (LbNOX, Fig. S2a). This approach was previously shown to reduce NADH/NAD + ratios during Complex I inhibition 25 . IMiCa magnitudes from LbNOX-expressing cells were no longer increased following rotenone treatment, suggesting that abnormal NADH oxidation contributes to uniporter enhancement (Fig. 2d).
Employing a similar strategy to blunt the increase in superoxide, we generated cells stably overexpressing mitochondrial superoxide dismutase 2 (SOD2, Fig. S2b). In these cells, baseline and rotenone-induced superoxide production was blunted (Fig. S2d). IMiCa also failed to increase following rotenone treatment (Fig. 2e). We then tested the reciprocal hypothesis, that producing ROS would be sufficient to induce IMiCa enhancement. A cell line was created stably expressing a mitochondria-targeted version of mini-singlet oxygen generator (mt-miniSOG, Fig. S2c). This fluorescent flavoprotein generates singlet oxygen when excited by blue light, and substantially increased production of superoxide ( Fig. S2e) 33 . In cells exposed to blue light 2-3 days before recording, IMiCa size was ~3x greater than in unexposed cells (Fig. 2f).
Mito-LbNOX has been used to rescue cell survival by preserving NADH:NAD + homeostasis during dysfunction induced by the Complex I inhibitor piericidin 25 . Thus, to determine whether maintaining NADH:NAD + homeostasis during Complex I impairment required the uniporter, we expressed mito-LbNOX in control and MCU KO HEK293T and exposed them to 500 nM piericidin. In control cells, mito-LbNOX expression did not alter cell survival and partially rescued the cell proliferation defect caused by piericidin, as expected (Fig. S2f). In contrast, expressing mito-LbNOX in MCU KO cells both impaired baseline cell survival and failed to rescue cell death after piericidin treatment, suggesting that the uniporter boosting of NADH:NAD + is required for survival during Complex I dysfunction. Taken together, these data indicate that aberrant NADH oxidation and ROS generation are critical signals for uniporter enhancement during Complex I dysfunction.

The MCU N-terminal domain is necessary for Complex-I mediated enhancement.
In a prior report, MCU was shown to be sensitive to matrix redox status and oxidative stress via S-glutathionylation at a conserved cysteine residue in its N-terminal domain (NTD) 34 . Though such regulation was not specific to Complex I, it offered a potential mechanism to explain IMiCa enhancement. Therefore, we expressed an MCU construct where all five cysteines were mutated (Cysteine-free MCU, CF-MCU) in MCU KO cells. As in the prior report, CF-MCU conducted IMiCa, with no obvious effect on basal channel function. Unexpectedly, rotenone-mediated IMiCa enhancement persisted in these cells (Fig. S2g). These data indicate that redox sensation by MCU cysteines fails to explain uniporter enhancement observed during Complex I dysfunction.
Nevertheless, the MCU NTD itself remained an interesting target to examine further. This structure is evolutionarily conserved, forming an independent domain within the matrix [35][36][37] . The NTD has been implicated in channel dimerization, though it is not essential for channel activity 35,36 . Given its highly-conserved structure, we hypothesized that the NTD may be responsible for Complex I-mediated enhancement. MCU lacking the NTD (ΔNTD-MCU) was stably expressed in MCU KO cells. Consistent with our hypothesis, IMiCa in these cells failed to increase in response to rotenone (Fig. 2g). Thus, an intact MCU NTD is necessary for the enhancement in current seen during Complex I dysfunction.

Uniporter enhancement depends on an interaction between MCU and Complex I
The importance of ROS in IMiCa enhancement revealed an intriguing discrepancy. Complex III inhibition with antimycin A produces abundant ROS within the matrix and intermembrane space 38 , yet it did not lead to IMiCa enhancement. We surmised that aberrant ROS may be primarily disrupting Complex I, and Complex I dysfunction subsequently altering uniporter behavior, which predicts close proximity between MCU and Complex I. Evidence for such an interaction was found incidentally in a proteomic screen for Complex I assembly factors, where MCU (annotated as CCDC109A) bound Complex I without affecting its assembly 39 . Similarly, a more recent compendium of mitochondrial protein-protein interactions revealed close proximity between MCU and several subunits of Complex I 40 . We therefore tested for a Complex I-MCU interaction. Immunoprecipitation of Flag-tagged MCU in 1% digitonin was followed by mass spectrometric analysis of co-precipitating proteins, identifying NDUFA3, NDUFA8, and NDUFA13 as potential interactors (Fig. S3a). All three of these are closely apposed on the Complex I structure. We confirmed this interaction by co-immunoprecipitating NDUFA13 with MCU-Flag but not Flagtagged succinate dehydrogenase complex subunit B (SDHB-Flag) (Fig. 3a).
To confirm the interaction within live cells with intact mitochondria, we turned to Förster energy resonance transfer (FRET) assays. For the mVenus-mCerulean pair used here, the Förster radius for 50% FRET efficiency is ~5 nm 41 . We took advantage of prior studies that showed that several Complex I subunits, typically those that had carboxy-termini exposed at the Complex I surface, are unaffected by carboxy-terminal fusion with fluorescent proteins 42 . MCU can similarly be linked at its C-terminus with fluorescent proteins 34 . To sample various portions of Complex I, we tagged eight NDUF subunits with mVenus, while MCU was fused to mCerulean. FRET was detected from co-transfected constructs in HEK293T cells using flow cytometry, which allows us to measure interactions over a wide range of protein expression levels 43 . NDUFA2 and NDUFA5 failed to target mitochondria when tagged with mVenus, and reassuringly showed no FRET with MCU-mCerulean (Fig. 3b). Expressing mitochondria-targeted mVenus with MCU-mCerulean revealed that some FRET was detected at high mVenus concentrations. Such concentration-dependent effects likely reflect the small volume of the matrix relative to cytoplasm. This FRET level served as the bound for spurious interactions, and a similar value was seen for NDUFA7, NDUFB6, NDUFS6, and NDUFV2, suggesting that these either were distant from MCU or had fluorophore orientations that minimized FRET. NDUFA10 and NDUFS3, however, demonstrated robust interaction with MCU-mCerulean, implying close physical proximity.
Although our biochemical and FRET assays showed MCU-Complex I interaction, these required heterologous expression of tagged proteins. To confirm such interaction between endogenous molecules, we used the Duolink proximity ligation system, which stochastically produces bright fluorescent spots when target proteins less than ~40 nm apart are co-immunolabeled 44 . We used anti-MCU and anti-NDUFS2 monoclonal antibodies to label the uniporter and Complex I respectively in HEK293T cells and IPSCs. Controls with either antibody alone, or in MCU KO cells, displayed no Duolink spots (Fig. S3b, c). When both antibodies were used, we saw robust Duolink labeling (3.6 ± 0.3 spots/cell, Fig. 3c). To confirm the interaction, we took advantage of the more lenient distance threshold detected by the Duolink system and used an antibody targeting Complex IV (MTCO1), since this is also part of the respirasome. This antibody also produced substantial, though less robust, Duolink signal, possibly because it is further from the putative MCU interaction site (2.8 ± 0.5 spots/cell, Fig. 3c, Fig. S3d). To show that the interaction was specific, we also performed the Duolink assay targeting Complex V (ATP5A), which is not part of the respirasome, and found substantially reduced labeling (0.7 ± 0.2 spots/cell).
Having established that endogenous Complex I interacts with MCU, we investigated changes produced by Complex I dysfunction. Treating HEK293T cells with rotenone eliminated NDUFA13-MCU co-immunoprecipitation and markedly reduced Duolink targeting (0.6 ± 0.1 spots/cell, Fig.  3a, c). Similarly, the NDUFB10 -/C107S IPSCs had diminished labeling compared to control (Control: 3.5 ± 0.1 spots/cell, NDUFB10 -/C107S : 1.4 ± 0.1 spots/cell), despite preserved NDUFS2 and MCU (Fig. 3d, Fig. S3e). In sum, based on immunoprecipitation, FRET, and proximity ligation, in both HEK293T cells and IPSCs, we find that MCU interacts with Complex I, but becomes decoupled when Complex I dysfunction ensues. Interestingly, an unexpected result from these assays was that the MCU-interacting NDUF subunits all clustered on the lateral surface of Complex I (Fig. S3f, g). This is consistent with the architecture of the respirasome, since MCU would not be hindered by the Complex III dimer nor Complex IV, which reside on the opposite sides. Taken together, we establish an interaction between MCU and Complex I that is disrupted during Complex I dysfunction.

Complex I-dependent protein turnover (CLIPT) controls MCU degradation
A common finding in the multiple systems examined was an increase in uniporter protein consistent with enhanced IMiCa (Fig. 1c). Such increases could be mediated by either enhanced synthesis or diminished degradation. The absence of mRNA upregulation and the ability to enhance IMiCa in heterologously-expressed channels suggested the effect was likely not from greater synthesis. To evaluate MCU degradation, we designed a tetracycline-repressible MCU-Flag construct and expressed it in MCU KO cells 45 . Cells were grown in 3% fetal bovine serumsupplemented media, to minimize proliferation. Addition of 1 µg/mL doxycycline suppresses MCU-Flag transcription, leading to depletion within two days in control cells. In contrast, upon treatment with rotenone, MCU-Flag expression persisted for the four-day experimental timeline (Fig. 4a), confirming that reduced degradation of the uniporter was the primary mechanism for this effect. Inhibition of Complex III or IV produced no such effect on MCU. To confirm that stabilization was specific to MCU, we looked at another mitochondrial transmembrane protein regulated by ROS, ROMO1, and found that its lifetime failed to enhance after Complex I inhibition ( Fig. S4a).
At this stage, we considered two potential hypotheses linking ROS production in Complex I to stabilization of MCU (Fig. S4b). The first, analogous to the cysteine mechanism described previously 34 , would involve post-translational modification (PTM) of a specific NTD residue that increases MCU stability (PTM hypothesis). Although simple, concerns about this hypothesis include the promiscuous modifications ROS can induce on target peptides. Many of these are irreversible, and tend to damage rather than enhance protein activity. Moreover, while ROS leaks from Complex I under physiological conditions, excess ROS produced during Complex I dysfunction induces self-inactivation 30,31 . Excess ROS produced by mito-miniSOG is also more likely to non-specifically damage Complex I, which has large matrix components and is quite sensitive to indiscriminate ROS production, than produce a specific modification on MCU. Finally, Complex I impairment decouples MCU, and it is unclear how the ROS signal would modify channels no longer bound. Thus, we considered an alternate hypothesis, termed Complex I induced protein turnover (CLIPT). Here, under normal conditions MCU, via its NTD, interacts with Complex I to serve as a "ROS buffer", being turned over by quality-control proteases as oxidative damage from basal ROS leakage impairs MCU function. When Complex I becomes dysfunctional from excess ROS, MCU can no longer interact and buffer ROS, becoming more stable.
To distinguish the PTM and CLIPT mechanisms, we first assayed how depleting quality-control proteases would affect MCU stability. Under the PTM hypothesis, this should not affect MCU stability, as there should be little ROS-induced damage. Conversely, if MCU is buffering Complex I ROS, impairing quality control will stabilize MCU. We depleted several quality-control proteases using shRNA, and found that LONP1 depletion led to an increase in MCU stability ( Fig. S4c; % knockdown by qPCR: 0.88, AFG3L2; 0.89, CLPP; 0.93 LONP1; 0.88, SPG7, n = 2). LONP1 performs quality control of matrix proteins, and this effect is consistent with modifications of matrixresident NTD leading to quality control, rather than transmembrane domain quality-control proteases AFG3L2 and SPG7. Moreover, a close interaction between LONP1 and MCU was also detected in the recently published mitochondrial protein-protein compendium 40 . Next, we examined how the NTD affected stability. Expressing the NTD fragment by itself should have no effect on MCU stability under basal conditions for the PTM hypothesis, whereas excess NTD should disrupt the MCU-Complex I interaction and stabilize the channel during CLIPT. We expressed an HA-tagged NTD fragment in cells also containing doxycycline-repressible MCU-Flag, and found that MCU stability increased, consistent with CLIPT (Fig. 4b).
To further confirm CLIPT, we designed the drug-induced dimerization experiment outlined in Fig.  4c. The FK506-binding protein (FKBP) binds the FKBP-rapamycin-binding domain (FRB) of MTOR only in the presence of rapamycin 46 , and using this system allows us to determine if MCU-Complex I interaction is the key determinant of MCU stability, without having to create Complex I dysfunction. The MCU NTD was replaced with the similarly-sized FRB fragment, to generate a FRB-MCU fusion construct (Fig. 4d). For Complex I, we fused FKBP to NDUFA10 (NDUFA10-FKBP), the Complex I subunit showing strong FRET with MCU (Fig. 3b). As a control, we created a mitochondrially-targeted FKBP (mito-FKBP). First, we established that the system was functional. In MCU KO cells, FRB-MCU was able to confer Ca 2+ uptake (Fig. S4d), showing this construct formed functional channels. In the absence of rapamycin, FRB-MCU and the FKBP constructs failed to interact (Fig. 4e), confirming the importance of the NTD in MCU-Complex I binding, whereas adding 100 nM rapamycin induced robust co-immunoprecipitation.
Having established functional rapamycin-induced interaction, we used the tetracyclinerepressible system to examine if MCU turnover depended on its Complex I interaction (Fig. 4f). 100 nM rapamycin was added to cell culture dishes one day prior to adding doxycycline to repress FRB-MCU transcription. When co-expressed with mito-FKBP, FRB-MCU showed minimal degradation in the presence or absence of rapamycin. Similarly, when co-expressed with NDUFA10-FKBP, FRB-MCU protein was stable in the absence of rapamycin. Remarkably, however, when rapamycin was added in this condition, FRB-MCU was rapidly degraded. Four implications arise from these results. First, the changes in MCU turnover are due to its interaction with Complex I, and not an off-target rotenone effect. Second, whereas a large fraction of full-length MCU degraded over 2 days under control conditions (Fig. 4a), removing the NTD conferred stability on the channel (Fig. 4f). Third, MCU turnover due to Complex I is not dependent on a specific modification of any particular NTD residue, since we could alter turnover in channels with the NTD replaced. Finally, MCU turnover appears to be tunable. Whereas enhancing the MCU-Complex I interaction with the FKBP-FRB-rapamycin system led to rapid MCU degradation, overexpressing the NTD alone stabilized it. Taken together, our results identify a new mechanism, Complex I induced protein turnover (CLIPT), critical for controlling uniporter levels and activity.

Uniporter stabilization during ETC impairment prolongs survival in mitochondrial cardiomyopathies
Multiple studies over two decades have revealed that cardiac Tfam deletion in mice produces many of the same clinical, biochemical, and ultrastructural features found in human mitochondrial cardiomyopathies [47][48][49][50] . Therefore, to examine whether uniporter stability is enhanced in vivo during disease, we turned to this model. For cardiac-specific deletion, we use the Myh6-Cre recombinase driver, which begins expressing embryonically 51 . When crossed with Tfam loxP/loxP animals, loss of myocardial TFAM leads to a 75% reduction in Complex I activity, with 30-35% inhibition of Complex III and IV, and early death between 3-6 weeks of age 12 (Fig. 4a). 10-14 day old mouse hearts were processed by Western blotting. Notably, despite embryonic initation of Mcu deletion, we could detect substantial levels of uniporter subunit proteins in Tfam-Mcu DKO animals, approaching those seen in wild-type animals, compared with much lower levels in Mcu KO hearts (Fig. 5a). This result suggests that, although its transcription has been disrupted, the MCU protein already present becomes much more stable in the ETC-deficient animals, compared to those with a functional ETC. This was not due to deficient Cre activity in the Tfam-Mcu DKO, as protein levels of TFAM were equally reduced between Tfam KO and Tfam-Mcu DKO hearts (Fig.  5a), Tfam mRNA transcripts were equally reduced between Tfam KO and Tfam-Mcu DKO hearts, and Mcu mRNA transcripts were equally reduced between Mcu KO and Tfam-Mcu DKO hearts (Fig. 5b). Moreover, this was also not due to MCU protein from excess infiltration of fibroblasts or other non-cardiac cells into the myocardium. Non-cardiomyocyte cell quantities are lowest in juvenile hearts, and their mitochondrial mass per cell is far smaller than in cardiomyocytes, demonstrated by the trivial amount of TFAM left after its deletion (Fig. 5a). A fibroblast marker, vimentin, was no different across the different genotypes (Fig. 5a). We also directly quantified cell amounts in histological slices. CellProfiler software was used to count the number of cellular nuclei in tissue slices from Masson's trichrome-stained hearts. Mcu KO and Tfam-Mcu DKO hearts had similar numbers of nuclei per mm 2 of tissue, these tended to be lower compared to wild-type animals, and no obvious excess non-cardiomyocyte infiltrates were noted (Fig. S5a, b). Therefore, as in cultured cells, disruption of the ETC in vivo in mouse myocardium led to increased stability of the uniporter.
To determine if these persistent uniporter channels were functional, we measured Ca 2+ uptake in mitochondrial fractions isolated from mouse hearts. We used fluorescent Ca 2+ imaging, as this allows an integrative assessment of Ca 2+ uptake in the context of TFAM deletion. Decreasing Oregon Green BAPTA 6F fluorescence following a Ca 2+ pulse indicated mitochondrial Ca 2+ uptake (Fig. 5c). Because TFAM deletion alters the driving force for Ca 2+ uptake, we quantified the steadystate ability of mitochondria to take up Ca 2+ rather than uptake rates (Fig. 5d). Wild-type mice took up the Ca 2+ pulse rapidly, whereas Mcu KO mice were unable to take up Ca 2+ , consistent with loss of functional uniporter. Most of the Tfam-Mcu DKO mice, however, had persistent Ca 2+ uptake, revealing preservation of functional uniporter channels, despite deletion of the Mcu gene.
Next, we assessed if loss of MCU impaired the health of Tfam-Mcu DKO mice. Of note, though MCU levels and activity persisted in the juvenile mice, as noted above, these were not at wild-type levels (Fig. 5a, d). Moreover, during the second postnatal week there appeared to be a reduction in the persistent uniporter channels, as a subset of Tfam-Mcu DKO mice were no longer capable of cardiac mitochondrial Ca 2+ uptake (lower points in Fig. 5d). Thus, we were still able to assess how partial depletion of MCU altered cardiac status. The Tfam-Mcu DKO developed a cardiomyopathy similar to Tfam KO mice. They had enlarged hearts (Fig. 5e), with reduced contractile function, dilated ventricles, and thinned walls on echocardiography (Fig. S5c-f). Remarkably, even partial loss of uniporter channels proved fatal, as a steep decline in survival occurred during the second postnatal week (Fig. 5f)

NTD overexpression improves survival and function in Complex I-impaired Drosophila
Complex I deficiency is the most common cause of monogenic mitochondrial disorders, and is frequently implicated in neurological and cardiac disease 52 . To further explore the physiological relevance of the MCU-Complex I interaction. Drosophila is an ideal system, as models for Complex I disease exist 29 , the mitochondrial Ca 2+ uptake machinery is closely conserved 53 , and crosses can be rapidly generated. For these analyses, Complex I was inhibited in Drosophila flight muscle, which possesses sarcomeric organization and mitochondrial Ca 2+ uptake that mimics mammalian cardiomyocytes 54 . Complex I dysfunction in NDUFB10 RNAi led to mild developmental lethality and weak flies (Fig. 6a, b). To test flight muscle, the time it takes Drosophila to fly off a platform sitting in water is measured (Supplemental Videos). On this island assay, NDUFB10 RNAi flies took longer to escape compared to controls. We also analyzed the recently-described Drosophila whole-body MCU knockout (MCU 1 ) 53 . These flies have no ruthenium-red sensitive IMiCa (Fig. S6a), and show neither developmental lethality nor flight weakness 53 . When crossed with NDUFB10 RNAi , however, there was a clear genetic interaction. Double-mutant flies suffered severe developmental lethality and flight muscle weakness (Fig. 6a, b). Notably, these impairments were entirely rescued, reverting NDUFB10 RNAi flies to near wild-type function, by re-expressing a fulllength MCU 53 . To test the importance of the NTD, we created a Drosophila ΔNTD-MCU transgene, which targeted mitochondria and was functional in muscle (Fig. S6d-f). This construct, however, entirely failed at rescue, reinforcing the importance of the MCU-Complex I interaction. To confirm the importance of Ca 2+ uptake through the uniporter, we also tested mutant flies that express a flight muscle-restricted pore mutant of MCU (MCU DQEQ ) that inhibits Ca 2+ transport in a dominantnegative fashion 55 . These Drosophila developed in expected numbers and had no flight impairment compared to controls. Here too, a clear genetic interaction was noted between MCU DQEQ and NDUFB10 RNAi , with double-mutant Drosophila having substantial developmental lethality and the survivors being impaired in flight (Fig. 6c, d). This genetic interaction was not specific to NDUFB10 RNAi nor was there a threshold for Complex I impairment necessary, as we saw a similar, milder pattern with NDUFA13 RNAi , which produces a much weaker level of Complex I deficiency (Fig. S6b, c) 29 . Therefore, in Drosophila as in Tfam KO mice, uniporter activity was necessary to preserve cellular function when Complex I is impaired.
The NTD is critical for the functional, biochemical, and genetic interaction between MCU and Complex I. Moreover, expressing the NTD fragment alone stabilized the uniporter (Fig. 4b), which appears necessary for maintaining homeostasis during Complex I impairment. Reasoning that this strategy may alleviate the phenotype caused by Complex I dysfunction, we generated a construct encoding the isolated Drosophila NTD fragment, which also targeted mitochondria (Fig.  S6d-f). When expressed in NDUFB10 RNAi flies, the NTD improved both survival (male flies) and flight, though the rescue was not as complete as expressing full-length MCU (Fig. 6a). Taken together, these results imply that targeting the MCU NTD may be a novel strategy for treating Complex I impairment.

DISCUSSION
In this report, we identify an essential functional, biochemical, and genetic interaction between Complex I and the uniporter, necessary for survival when Complex I becomes impaired. In deciphering how Complex I dysfunction enhances uniporter levels, we identified CLIPT as a mechanism for MCU protein turnover (Fig. 7), and show that it may be exploited to preserve organismal function during Complex I deficiency, a pathology common to varied diseases.
In deciphering how Complex I dysfunction enhances uniporter levels, we revealed the Complex I interaction is the key determinant for MCU protein turnover, which we term CLIPT. In this process, MCU binding to Complex I leads to uniporter degradation, most likely via oxidative damage from physiological ROS generated in Complex I. Unlike other forms of post-translational MCU regulation, there does not appear to be a specific residue modification driving CLIPT. Rather, as long as MCU and Complex I interact, CLIPT occurs. Aberrant ROS production during Complex I dysfunction may either prevent the MCU-Complex I interaction by modifying the binding site, or lead to Complex I inactivation and disassembly. In either case, enhancement of uniporter levels occurs due to reduced MCU degradation. It remains to be seen if CLIPT controls the turnover of other mitochondrial proteins as well.       (A, B), except the dominant-negative pore mutant MCU DQEQ was expressed with MHC-GAL4. E, F. As in (A, B), except the isolated NTD fragment was expressed with MHC-GAL4. * p < 0.05, ** p < 0.01, *** p < 0.001.

Figure 7. Complex I ROS controls uniporter turnover.
Top, under physiological conditions, MCU interacts with Complex I and is oxidized by the mild ROS leak produced by Complex I. Such oxidized MCU becomes damaged and degraded by LONP1 or other quality-control proteases, leaving Complex I available to interact with additional channels. Bottom, when Complex I becomes impaired or misassembled, it produces excessive ROS and self-inactivates. Such dysfunctional Complex I can no longer interact with MCU, nor damage it with basal ROS leak, and is cleared by housekeeping proteases CLPP and LONP1. Thus, functional MCU levels build up, and additional Ca 2+ influx through these channels maintains energetic homeostasis.

Plasmids for stable expression, transient transfection, and short hairpin RNA
Newly-derived plasmids were created using the NEB HiFi DNA Assembly Cloning Kit (Ipswich, MA), and will be deposited at Addgene. Plasmids were verified by Sanger sequencing. Mito-LbNOX was expanded from pUC57-mitoLbNOX and cloned into pLenti-CMV-puro for lentiviral generation. GCaMP6m was cloned into pLenti-CMV-puro with an N-terminal Cox8 mitochondrial targeting sequence and a C-terminal mCherry tag. Human SOD2 was expanded from a HEK293T cDNA library and cloned into pLenti-CMV-puro with a C-terminal HA tag. The mitochondrial targeting sequence from ABCB10 (aa 1-193) was expanded from a HEK293T cDNA library and cloned in-frame upstream of SoNar to create mitochondrially-targeted MitoSoNar. MCU-Flag was shuttled from pLYS1-MCU-Flag into pCW57.1-MAT2A to create pCW57.1-MCU-Flag. The MCU NTD (aa 1-185) was expanded from pLYS1-MCU and cloned into pLenti with an HA-tag at its C-terminus. FRB-MCU was generated by incorporating the FRB fragment immediately downstream of the mitochondrial targeting sequence within ΔNTD-MCU-Flag, and placing this in the pCW57.1-MAT2A plasmid. For mito-FKBP, the FKBP fragment was placed between a 4-fold repeat of the COX8A mitochondrially targeting sequence and an HA tag in pLenti-CMV-puro. NDUFA10-FKBP was created similarly, except the full NDUFA10 sequence was used instead of the mitochondrial targeting sequence. Drosophila MCU lacking amino acids 56-182 was cloned (UniProt Q8IQ70), along with a C-terminal HA tag, from pAC5.1 59 into pUASg.attB to create ΔNTD-MCU-pUASg.attB. A similar strategy was used to incorporate the Drosophila MCU NTD fragment (amino acids 1-182), creating NTD-pUASg.attB.
For FRET experiments, mitochondria-targeted mVenus (mt-mVenus) and mCerulean (mt-mCerulean) were created by adding four copies of the human COX8 mitochondrial targeting sequence to the N-terminal of mVenus in the pLenti-CMV-puro backbone. The sequence for MCU was expanded from pLYS1-MCU-Flag and cloned into the mCerulean-pLenti vector in place of the mitochondrial targeting sequence. Sequences for NDUFA2, NDUFA5, NDUFA7, NDUFA10, NDUFS3, NDUFS6, NDUFB6, and NDUFV2 were expanded from a HEK293T cDNA library via polymerase chain reaction and cloned N-terminal to mVenus in a pEGFP vector. For FRET calibration, we also cloned mitochondrially-targeted mVenus-mCerulean dimers separated by linkers of 5, 43, and 236 amino acids into pLenti-CMV-puro to create mito-C5V, mito-C43V, and mito-CTV.
Short hairpin RNA lentiviral constructs were obtained from The RNAi Consortium (Sigma). TRC IDs for shRNA plasmids are TRCN0000412514 (AFG3L2), TRCN0000046859 (CLPP), TRCN0000310154 (LONP1), TRCN0000046793 (LONP1), TRCN0000063825 (SPG7). The control was shGFP (SHC005). HEK293T genetic fingerprinting validation is via short tandem repeat analysis, performed at the University of Utah DNA Sequencing Core. Clones that failed to grow in galactose had genomic DNA Sanger sequenced for evidence of editing, and clones were subject to Western blot to confirm FOXRED1 deletion, and one of these that appeared to grow well was further studied.

Patient-derived induced pluripotent stem cells
Fibroblasts obtained post-mortem from an infant with NDUFB10 deficiency (NDUFB10 -/C107S ) 28 were reprogrammed into IPSCs using CytoTune-iPS Sendai Reprogramming vector (A16517, ThermoFisher), and cultured in Essential 8 stem cell media (A1517001, ThermoFisher) supplemented with 2 mM pyruvate, 0.4 mM uridine, and 1 mM N-acetylcysteine. At around Day 20 post-transduction, newly formed IPSC colonies were transferred to new plates and expanded for another month. For quality control of IPSCs we performed four standard assays to determine (1) pluripotent status, (2) genetic fingerprinting, (3) karyotype and (4) absence of bacterial or mycoplasma contamination. Pluripotent status was determined by immunofluorescence against Oct4, Sox2, and ZO-1, and counterstained with Hoechst dye prior to imaging on a fluorescent EVOS Cell Imaging System (ThermoFisher). IPSC genetic fingerprinting is via short tandem repeat analysis. Molecular karyotyping of genomic DNA is via an nCounter Human Karyotype Assay (Nanostring Technologies Inc.). Contamination testing is via an e-Myco PLUS Mycoplasma PCR Detection Kit (iNtRON Biotechnology) and an PCR Bacteria Test Kit (PromoCell GmbH).

Mouse strains and survival
All animal procedures have been reviewed and approved by the Institutional Animal Care and Use Committee at the University of Utah and Boston Children's Hospital. Tfam loxP/loxP mice were developed by Nils-Göran Larsson 13 and obtained from Ronald Kahn 63 . Mcu loxP/loxP mice were obtained from John Elrod 18 . Myh6-Cre transgenic mice were obtained from the Jackson Laboratory (Bar Harbor, ME, stock # 011038) 64 . Animals were kept on a C57BL/6J background. Animals were housed under standard conditions and allowed free access to food and water. Animals used for experiments were 10-14 days old. Heart and body weights were recorded at time of euthanasia. Survival analysis was performed in OriginPro 9 (OriginLab). A Kaplan-Meier curve was constructed, censoring animals used for experiments. Statistical comparison was via a log-rank test.  53 . PhiC31 integrase-mediated transgenic flies were generated by BestGene (Chino Hills, CA) from NTD-pUASg.attB. Integration site was at attP40, as with UAS-MCU WT . Genomic PCR confirmation of insertion and subsequent balancing were performed by BestGene. Flies were reared at 25 ºC in a 12:12 light/dark cycle. Fly genotypes are listed in Supplemental Table 1.

Mitochondrial and mitoplast isolation
Mitochondria were isolated from cultured cells or mouse hearts by differential centrifugation, and mitoplasts prepared as previously described 12,15 . For Drosophila mitochondrial isolation, lines were raised at 20 o C. Adult flies >3 days of age after eclosion were used. Mitochondria were prepared from thoraces (devoid of legs and wings) of >50 flies. Thoraces were homogenized with a Potter-Elvehjem grinder set to 250 rpm, and the remainder of the protocol was as previously described 15 .

Electrophysiology
Whole-mitoplast electrophysiology was performed as described previously 15 . For Drosophila, we used MHC-Gal4 driven expression to measure changes in flight muscle IMiCa. To confirm with electrophysiology that transgenes were expressing, we used the global Da-Gal4 driver for ΔNTD-MCU IMiCa in MCU 1 flies, and isolated MCU NTD in wild-type flies. Borosilicate glass pipettes with a resistance of 15-20 MΩ were used. Whole-mitoplasts currents were acquired at 5 kHz and filtered at 1 kHz using an Axopatch 200B amplifier (Molecular Devices, San Jose, CA). Mitoplasts had a capacitance 0.2-3.7 pF. Pipette solution was composed of (mM): Na-Gluconate 150, HEPES 10, EDTA 1, EGTA 1, pH 7.2, brought up to 320-340 mOsm with D-Mannitol. Bath solution for inner membrane anion channel (mM): 150 KCl, 10 HEPES, 1 EGTA. Bath solution for mitochondrial calcium uniporter (mM): Na-Gluconate 150, HEPES 10, 5 CaCl2, pH 7.2. A ~12 mV junction potential was cancelled on switching between solutions. Ruthenium red (1 µM) was added to block uniporter-specific currents. For display purposes, capacitance transients caused by changing levels of solutions in the bath have been removed and the Simplify filter from Adobe Illustrator has been used to reduce the number of points, without altering the shape of the traces. Analysis was performed using pClamp v10 (Molecular Devices), Excel (Microsoft), and OriginPro 9. For statistical comparisons, we used two-sided Mann-Whitney U tests, except for the rotenone concentration-response curve, for which we used ANOVA with post-hoc Bonferroni-corrected means comparisons, and the IPSC KO and rescue experiment, for which we used a Kruskal-Wallis test to establish overall p value, followed by Bonferroni-corrected Dunn's test for pairwise comparisons.

Quantitative reverse transcription-polymerase chain reaction (qPCR) expression analysis
Quantitative PCR from IPSC, HEK293T cells, and mouse hearts was performed as described previously, using Power SYBR Green PCR Master Mix (ThermoFisher) 12 . Quantification of gene expression was performed on a 96-well BioRad CFX Connect Real-Time PCR Detection System (BioRad, CA, USA). Analysis was performed by using the 2 -ΔΔCt method. Primers were obtained via NCBI Primer-BLAST, Primerbank, or prior reports 65,66 .

Western blots
Cells or tissue were lysed in RIPA buffer containing Halt protease and phosphatase inhibitor cocktail (78440, ThermoFisher). After lysate clarification, protein concentration was determined by BCA assay (23227, Thermo Fisher). 5-20 µg of protein from cell lysates were loaded on polyacrylamide gels and immunoblotted as described previously 59 .

Blue light assay for miniSOG
HEK293T cells expressing mito-miniSOG at >70% confluence were exposed to blue light from an LED array for 10 min at RT. Longer periods of blue light exposure would often lead to substantial cell death. Mitoplast isolation as described above was performed 48-72 hours later. Cells with the same construct without exposure to blue light were used as control.

Flow cytometric measurements
Data was collected at the Flow Cytometry Core Facility at the University of Utah. For mitochondrial ROS measurements, cells were incubated with 5 µM MitoSOX (M36008, ThermoFisher) at 37°C for 15-30 minutes. For mitochondrial NADH/NAD + measurement, cells were transfected 24-48 hours prior to analysis with the MitoSoNar construct. Cells were analyzed with a BD FACSCanto Analyzer running FACSDiva 6 software (both BD Biosciences, San Jose, CA). For mitochondrial Ca 2+ , mito-GCaMP6m was measured using 488 nm excitation laser and 530/30 nm emission filters and mCherry was measured using a 561 nm laser and 585/15 nm emission filter. For MitoSOX, we used a 488 nm excitation laser and 585/15 nm emission filter. For MitoSoNar, we used the ratio of 525/50 nm emissions when excited by either 405 nm or 488 nm laser lines after background correction in untransfected cells. For analysis, live, single cells were selected using forward and side-scatter parameters. Analysis was performed on FlowJo (v10.6, BD Life Sciences, Ashland, OR) and OriginPro 9. Statistical test was via ANOVA with post-hoc Bonferronicorrected means comparisons.

Confocal imaging and immunocytochemistry
To determine targeting of mCerulean-and mVenus-tagged constructs we used live cell confocal imaging. Cells were washed with phosphate buffered saline and incubated with 100 nM MitoTracker Orange CMTMRos (M7510, ThermoFisher) or MitoView Green (70054, Biotium, Fremont, CA).
For immunocytochemistry, cells were grown on poly-L-lysine-coated glass coverslips. After phosphate buffered saline washes, cells were fixed in 10% neutral buffered formalin (VWR), permeabilized with 1% Triton X-100, and blocked with fetal goat serum. Primary antibodies were incubated at the indicated dilutions, and when needed secondary antibodies were goat antimouse Alexa Fluor 555 or goat anti-rabbit Alexa Fluor 488.
For Drosophila imaging, flies were embedded in Tissue-Tex O.C.T. Compound (Sakura), sliced to 7 µm thickness and mounted on glass coverslips. The sections were fixed using Formalin (Fisher Healthcare), permeabilized using 0.5% Triton X-100, and blocked using goat serum. Staining with primary antibodies occurred overnight in the presence of Drosophila larvae homogenate in 0.2% Triton X-100. Secondary antibodies were goat anti-mouse Alexa Fluor 555 or goat anti-rabbit Alexa Fluor 488.
Image acquisition was at room temperature. Imaging was performed on a Leica TCS SPE confocal microscope (Buffalo Grove, IL) using Leica Application Suite X v3.5 software. Acquisition was for 1024 x 1024 pixels at an 8-bit depth. Images have been enhanced for contrast uniformly, without alteration of gamma, and pseudocolored from grayscale.

Affinity Purification and Co-Immunoprecipitation
HEK293T cells stably expressing MCU-Flag were grown to confluency, while wild-type cells and cells stably expressing SDHB-Flag served as controls. After washing in phosphate-buffered saline, cells were lysed in 1% digitonin, 50 mM Tris, 150 mM NaCl and Halt protease and phosphatase inhibitors. Lysates were centrifuged at 4 °C for 10 minutes at 16000g. 200-500 µg of cleared lysates were incubated with 10 μl of FLAG antibody conjugated beads overnight at 4 °C. For co-immunoprecipitation studies, after extensive washing, the beads were heated to 80°C in 50 μl of sample loading buffer and then used for Western blotting as described above.
For proteomic analysis, cells were processed as above but eluted in 2% SDS in mass spectrometry-grade water. We analyzed six independent MCU-Flag samples and four control nontransduced samples.

Liquid chromatography-tandem mass spectrometry
Sample preparation for mass spectrometry was performed as described previously 67,68 . Briefly, samples were diluted with UA buffer, reduced, alkylated and proteins were digested with trypsin overnight at 37°C, then acidified using 1% formic acid. Peptides were analyzed with an Orbitrap Velos Pro mass spectrometer (Thermo) interfaced with an EZ nLC-1000 UPLC and outfitted with a PicoFrit reversed phase column (15 cm x 75 μm inner diameter, 3 μm particle size, 120 Å pore diameter, New Objective). Spectra were acquired in a data-dependent mode with dynamic exclusion. MS1 spectra were acquired at a resolution of 30,000 and the top 20 peaks were fragmented using CID fragmentation and analyzed in ion trap. The top twenty MS1 peaks were analyzed at a resolution of 30,000. Samples were run in duplicate.
The resulting spectra were analyzed using MaxQuant 1.6.0.16 against the UniprotKB human database. Database search engine parameters were as follows: trypsin digestion, two missed cleavages, precursor mass tolerance of 20 ppm, fragment mass tolerance of 0.5 Da, and dynamic acetyl (Protein N-term), and oxidation (M), The false discovery rate (FDR) was 1% and modified peptides had a minimum Andromeda score of 40. The proteins identified were further filtered to only include those identified in at least 50% of all FLAG pull downs as confident interactors. For relative quantification, peptide abundance was log2 transformed and missing values were imputed with Perseus 1.6.5.0. Normalized log2 intensities of peptides were used in statistical comparisons of groups: Student's two sample t-test was used for comparisons between the two sample groups and volcano plots were generated with an FDR of 0.1. Statistical analyses were performed in Perseus 1.6.5.0.

Flow cytometric Förster resonance energy transfer
Data was collected at the Flow Cytometry Core Facility at the University of Utah. We followed the protocol described previously to quantify FRET using flow cytometry 43 . Cells in 6-well dishes were transfected with 5 µg plasmid DNA using Lipofectamine 2000 (Thermo) 1-3 days prior to analysis. On the day of cytometry, cells were incubated with 100 µM cycloheximide for ~2 hours prior to analysis, to prevent artifacts from incomplete protein synthesis 43 . Flow cytometry was performed on a BD FACSCanto Analyzer. Signals were recorded using (1) 408 nm laser excitation and 450/50 nm emission (mCerulean), (2) 408 nm excitation and 525/50 emission (FRET), and (3) 488 nm excitation and 530/30 nm emission (mVenus). In all cases, we subset populations for live fractions of single cells using forward and side-scatter parameters. We obtained signals from untransfected cells for background correction. Background-corrected signals are designated SVen, SCer, and SFRET for the excitation/emission pairs 488/530, 408/450, and 408/525, respectively. Signals from mt-mVenus transfected alone were used to calculate the cross-talk ratio between SVen and SFRET (RA1), while signals from mt-mCerulean transfected alone were used to calculate the cross-talk ratios between SCer and SFRET (RD1) and between SCer and SVen(RD2). These ratios were then used to correct for cross-talk signals within each analyzed cell co-transfected with MCU-mCerulean and either mt-mVenus or each of the mVenus-tagged NDUF constructs (NDUFA2, NDUFA5, NDUFA7, NDUFA10, NDUFS3, NDUFS6, NDUFB6, and NDUFV2). The correction calculations are: where (3) is the FRET signal corrected for cross-talk from fluorescence due directly to mVenus and mCerulean, calculated using the prior ratios and equations (1)- (2). Next, to calibrate the FRET signal, we make use of the mitochondria-targeted mVenus-mCerulean dimers with varying linkers. Because of the one-to-one relationship between FRET donor (mCerulean) and acceptor (mVenus) in these constructs, the following equation holds true where instrument-and fluorophore-related constants for excitation and emission are captured in the g and f terms, respectively. Because this is a linear equation, by plotting against for the three mVenus-mCerulean pairs we can obtain the intercept from the best-fit regression line. Our value of 1.62 was in close agreement with the 1.65 value obtained in the prior publication, even with different instruments and mitochondria-targeted fluorophores. Finally, we calculated donor FRET efficiency for each cell by using the equation and plotting against the mVenus fluorescence of that cell.
Analysis was carried out in FlowJo and Excel. Data was then exported to OriginPro 9 for display as density plots. For display, we fit a straight line to those constructs that demonstrated no FRET or concentration-dependent spurious FRET, or an exponential of the form = �1 − − � for the constructs that demonstrated strong FRET. For the summary graph in Fig. 3b, we used the FRET efficiencies in the fluorescence window of 150,000-200,000 for mVenus expression. Statistical comparisons were in OriginPro 9 via ANOVA with post-hoc Bonferroni-corrected means comparisons.

Duolink proximity ligation assay
Duolink proximity ligation (DUO92102, Sigma) assays were performed per the manufacturer's protocol. We used monoclonal mouse anti-NDUFS2, and monoclonal rabbit anti-MCU, both of which have been knockout validated. Mitochondrial counterstain was with anti-COX IV-Alexa Fluor 488. Cells were imaged on a confocal microscope as described above. Areas for imaging were selected based on uniform distribution of DAPI-stained nuclei with about 20-50 cells per field. Separate images were taken of each field for DAPI (nuclei, 405 nm laser), COX IV-Alexa Fluor 488 (mitochondria, 488 nm laser), and Duolink spots (561 laser). Analysis was performed on CellProfiler (v3.1) 69 . After median filtering DAPI-stained nuclei, the ExpandObjects module was used to dilate these until touching adjacent objects to define cell borders. Mitochondrial objects were identified in median-filtered COX IV stained images, and Duolink spots were identified in the Duolink images. To detect Duolink spots corresponding to mitochondria the RelateObjects module was used, and any non-mitochondrial Duolink spots filtered out. Spots were assigned to the cell object they were in to generate an estimate of spots per cell. Results were analyzed in Excel and OriginPro 9 using ANOVA with post-hoc Bonferroni-corrected means comparisons. For display purposes only, the size and contrast of the Duolink spots in the manuscript figures have been increased uniformly to allow visualization with the small figure size.

Doxycycline repression for determining MCU stability
Stable lines expressing pCW57.1-MCU-Flag (derived from pCW57.1-MAT2A 45 ) allowed repression of MCU transcription by doxycycline. The cells were grown using media supplemented with 3% fetal bovine serum, to minimize cell division. Day 0 cells were collected immediately prior to doxycycline addition, and the remaining cells were incubated with 1 µg/mL doxycycline for the indicated number of days. For Complex I inhibition, cells were incubated in 1 µM rotenone for 48 hours prior to doxycycline addition.

Rapamycin-induced dimerization
Stable cell lines were created using lentiviral transduction, expressing (1) FRB-MCU-Flag and mito-FKBP-HA, or (2) FRB-MCU-Flag and NDUFA10-FKBP-HA. Co-immunoprecipitation was performed as described above, except lysates were incubated with either DMSO or 100 nM rapamycin during affinity purification overnight at 4°C. For stability studies, cultured cells were incubated with DMSO or 100 nM rapamycin for one day prior to adding 1 µg/mL doxycycline, and cells were collected and processed for Western blotting at the indicated timepoints as described above.

Mitochondrial Ca 2+ Imaging
Imaging was performed in 96-well plates on a Cytation 5 microplate reader (Biotek, Winooski, VT). For mice, 100 µg of cardiac mitochondria were used per trial. For HEK293T, 10 5 cells were permeabilized with 0.005% digitonin and used per trial. Samples were incubated in 100 µL of solution containing (in mM): 125 KCL, 20 HEPES, 5 K2HPO4, 1 MgCl2, 5 L-glutamic acid, 5 L-malic acid, 0.01 EGTA, 0.1% BSA, and 1 µM Oregon Green BAPTA-6F (Thermo Fisher). pH was adjusted to 7.3 with KOH, and osmolality to 290-300 mOsm/L). Excitation and emission wavelengths were 485/510 nm. 10 µM CaCl2 was injected per trial. Percent Ca 2+ clearance was calculated as: where is the steady-state fluorescence level after the mitochondria have taken up the pulse, typically around 3-5 minutes following the Ca 2+ injection, is the initial fluorescence level, and is the maximal fluorescence level immediately following the Ca 2+ injection. Analysis was performed in OriginPro 9 using ANOVA with post-hoc Bonferroni-corrected means comparisons.

Histology and Nuclei Counting
Extracted hearts were incubated in a fixative solution containing 4% paraformaldehyde in PBS for 48 hours at 4°C, and then placed in a 70% ethanol solution. The samples were embedded in paraffin, cut, and stained with Masson's trichrome by the Research Histology core at the Huntsman Cancer Institute (University of Utah). Slides were imaged using a BX51WI microscope (Olympus, Center Valley, PA).
Nuclei density was calculated via automated analysis performed on CellProfiler (v3.1) 69 . The UnmixColors module was used to separate each image into two, one that highlighted nuclei (setting: hematoxylin), and one for cellular staining (setting: eosin). The eosin image was analyzed to determine how much of the image was occupied by cells, to correct for tissue processing and any blank spaces. The hematoxylin image was analyzed to count nuclei. Nuclei density was calculated by dividing the number of nuclei by the area of each image covered by cells. For the eosin image, we used IdentifyPrimaryObjects, MeasureImageAreaOccupied, and the image scale to calculate the area occupied by cells. For the hematoxylin image, to smooth over heterogeneity of staining within each nuclei, we sequentially applied the GaussianFilter, Threshold, and Opening modules. Subsequently, the IdentifyPrimaryObjects module was used to count nuclei. All images were processed identically. Statistical analysis was performed in R using a Kruskal-Wallis test to establish overall p value followed by Bonferroni-corrected Dunn's test for pairwise comparisons.

Echocardiography
Echocardiography was performed as described previously 12 . Mice were sedated with inhaled isoflurane, restrained in the supine position, and cleared of chest fur. M-mode images were recorded in short-axis at the level of the papillary muscles, using a Vevo 2100 ultrasound machine equipped with a 55-MHz probe (Visual Sonics, Toronto, Ontario, Canada). Statistical analysis was performed in R using a Kruskal-Wallis test to establish overall p value followed by Bonferronicorrected Dunn's test for pairwise comparisons.

Island Assay
Flight capabilities were determined using a modified island assay 70 . We used a rectangular ice pan that held a flypad in the middle. The bottom of the container was filled with cold soap water that reached the top of the flypad without covering it. Clear plastic wrap was placed over half of the container in order to prevent flies from escaping. Female flies were collected within 48 hours of eclosion, and placed on normal fly food in groups of 12. Two days later, flies were tapped to the bottom of the vial, the vial was immediately inverted and flies were tapped onto the flypad in the island. Vials were inspected for any flies remaining after inversion to determine the starting number of flies on the island. Videos were taken of each group, and were manually scored to determine how many flies remained on the flypad every 2 seconds until the 20 second mark was reached. Log-rank (Mantel-Cox) testing was performed in GraphPad Prism 8 with Bonferroni correction for multiple comparisons.

Drosophila viability assay
To determine viability, separate groups of male and female flies were counted from at least 3 crosses. Viability was calculated by dividing the number of flies without balancers (experimental group) by the number of flies with balancers (control group) and multiplying by 100 to obtain a percentage. Fischer's Exact Test was performed in GraphPad Prism 8 with Bonferroni correction for multiple comparisons.

Statistics
For flow cytometric assays, N refers to individual cells. For electrophysiological assays, N refers to individual mitoplasts. For Duolink imaging, N refers to cells calculated from nuclei staining. For cell culture qPCR, N refers to technical replicates. For animal studies, N refers to individual animals. We considered p < 0.05 statistically significant. Individual tests and software are described in the corresponding sections of the methods.