Coupled transmembrane mechanisms control MCU-mediated mitochondrial Ca²⁺ uptake

Significance

Control of the level of calcium inside mitochondria is important because mitochondrial calcium regulates metabolism, cell death, and cellular signaling. The main pathway for mitochondrial calcium uptake is a calcium-selective ion channel complex in the inner mitochondrial membrane called the mitochondrial calcium uniporter (MCU). How the activity of the MCU ion channel is regulated is controversial. Here we employed electrophysiology of single isolated mitochondria to record MCU calcium and sodium ionic currents. We found that MCU ion channel activity is controlled by distinct Ca²⁺-regulated mechanisms on both sides of the inner mitochondrial membrane that are coupled to each other. These mechanisms allow for enhanced cellular regulation of mitochondrial calcium homeostasis.

Abstract

Ca²⁺ uptake by mitochondria regulates bioenergetics, apoptosis, and Ca²⁺ signaling. The primary pathway for mitochondrial Ca²⁺ uptake is the mitochondrial calcium uniporter (MCU), a Ca²⁺-selective ion channel in the inner mitochondrial membrane. MCU-mediated Ca²⁺ uptake is driven by the sizable inner-membrane potential generated by the electron-transport chain. Despite the large thermodynamic driving force, mitochondrial Ca²⁺ uptake is tightly regulated to maintain low matrix [Ca²⁺] and prevent opening of the permeability transition pore and cell death, while meeting dynamic cellular energy demands. How this is accomplished is controversial. Here we define a regulatory mechanism of MCU-channel activity in which cytoplasmic Ca²⁺ regulation of intermembrane space-localized MICU1/2 is controlled by Ca²⁺-regulatory mechanisms localized across the membrane in the mitochondrial matrix. Ca²⁺ that permeates through the channel pore regulates Ca²⁺ affinities of coupled inhibitory and activating sensors in the matrix. Ca²⁺ binding to the inhibitory sensor within the MCU amino terminus closes the channel despite Ca²⁺ binding to MICU1/2. Conversely, disruption of the interaction of MICU1/2 with the MCU complex disables matrix Ca²⁺ regulation of channel activity. Our results demonstrate how Ca²⁺ influx into mitochondria is tuned by coupled Ca²⁺-regulatory mechanisms on both sides of the inner mitochondrial membrane.

The mitochondrial Ca²⁺ uniporter (MCU) is a Ca²⁺-selective ion channel in the inner mitochondrial membrane (IMM) (1) that represents the major mechanism by which mitochondria take up Ca²⁺ into the matrix to regulate bioenergetics, cell death pathways, and cytoplasmic Ca²⁺ signaling. MCU-mediated Ca²⁺ uptake is driven by the large (−150 to −180 mV) inner-membrane voltage (ΔΨ_m) generated by proton pumping into the intermembrane space (IMS) by the electron-transport chain during oxidative phosphorylation. Despite the considerable driving force for Ca²⁺ entry into the mitochondrial matrix, mitochondrial matrix free [Ca²⁺] ([Ca²⁺]_m) at rest is maintained at ∼100 nM, similar to the resting cytoplasmic free [Ca²⁺] ([Ca²⁺]_i). Whereas transient elevations of [Ca²⁺]_m promote enhanced adenosine 5′-triphosphate production to support cellular energy demands, failure to maintain low matrix [Ca²⁺] may trigger ΔΨ_m dissipation and opening of the permeability transition pore (PTP) that can result in cell death. Thus, mitochondrial Ca²⁺ uptake must be tightly controlled, but how mitochondrial Ca²⁺ uptake through MCU is regulated is controversial (2–4). In metazoans, the MCU channel exists as a protein complex composed of MCU as the tetrameric channel pore-forming subunit (5, 6), EMRE, a single-pass transmembrane protein that is essential for channel activity (7–9), and IMS-localized (but see ref. 10) Ca²⁺-binding EF hand domain-containing MICU1/2 heterodimers (11–14) that associate with the channel complex by interactions with the carboxyl terminus of EMRE (15) and with MCU (16, 17). Despite early controversies regarding the location of MICU proteins and the nature of the MICU-dependent regulation (see ref. 4), in the most widely accepted current model of MCU regulation by MICU proteins (but see refs. 2 and 4), MICU1/2 promotes cooperative channel activation in response to Ca²⁺ binding by their EF hands, whereas apo-MICU1/2 imposes a threshold [Ca²⁺]_i below which MCU channel activity is inhibited, so-called channel gatekeeping (14, 15, 18–25).

Two fundamental observations suggest that this model of MCU regulation is inadequate. First, it cannot account for the observation that the channel can be open with the MICU1/2 EF hands in their apo state under experimental conditions in which [Ca²⁺]_i is reduced to low levels that permit Na⁺ permeation (1, 26, 27). Second, by mitoplast electrophysiology of the IMM with pipette solutions containing defined free [Ca²⁺], we previously reported that matrix [Ca²⁺] also regulates MCU-channel gating, with strong inhibition at normal resting [Ca²⁺]_m and maximal channel inhibition ∼80% at ∼400 nM (27). In this condition, the channel was strongly inhibited by MICU1/2 despite their being fully Ca²⁺-liganded by Ca²⁺ present in the bathing solution, an observation that also cannot be accounted for by the current model of MCU regulation.

Here we set out to understand the matrix [Ca²⁺] regulation of MCU-channel activity and the mechanisms that regulate Na⁺ permeation. By varying mitochondrial Ca²⁺ buffering capacity and [Ca²⁺]_m in mitoplast patch-clamp electrophysiology experiments, we determined that Ca²⁺ that permeates through the channel pore regulates Ca²⁺ affinities of coupled inhibitory and activating sensors in the matrix. We localized the inhibitory sensor within the MCU amino terminus which, upon binding Ca²⁺, inhibits channel activity despite Ca²⁺ binding to MICU1/2. Disruption of this inhibitory Ca²⁺-binding site significantly promoted PTP opening in intact mitochondria in response to Ca²⁺ additions. Conversely, disruption of the interaction of MICU1/2 with the MCU complex abolished matrix Ca²⁺ regulation of channel activity. A model of coupled transmembrane Ca²⁺-regulatory mechanisms was developed to account for these data, including the observation that the channel can be open with the MICU1/2 EF hands in their apo state under conditions in which Na⁺ is the permeant ion (1, 26, 27).

Results

Matrix [Ca²⁺] Regulation of MCU Ca²⁺-Channel Activity.

We previously determined that matrix [Ca²⁺] regulated MCU-channel activity with a biphasic concentration dependence, with strong inhibition at normal resting [Ca²⁺]_m (∼100 nM) and maximal channel inhibition ∼80% at ∼400 nM (27). We first confirmed this matrix [Ca²⁺] regulation by recording MCU Ca²⁺ currents in mitoplasts isolated from HEK293 cells, with the patch-clamp pipette solution that filled the mitochondrial matrix containing defined [Ca²⁺] buffered with 1.5 mM EGTA. Ca²⁺ currents were large with pipette solutions containing 1.5 mM EGTA with either no added Ca²⁺ or with free [Ca²⁺] set to 1 μM, whereas they were inhibited by ∼80% at 400 nM Ca²⁺, in excellent agreement with the previous observations (Fig. 1 A and B). This regulation occurred while MICU1/2 was in its Ca²⁺-liganded state due to the presence of high bath [Ca²⁺] necessary to record MCU Ca²⁺ currents. Accordingly, in the simplest model to account for MCU channel regulation, the channel complex possesses two matrix-localized Ca²⁺-sensing sites, inhibitory (S_i) and activating (S_a), whose Ca²⁺ occupancies control the functionality of MICU1/2 on the other side of the membrane to regulate channel gating.

Fig. 1.

MCU activity is regulated by matrix [Ca²⁺]. (A) Representative families of MCU Ca²⁺ currents, normalized to mitoplast capacitance, in response to voltage ramps in wild-type mitoplasts, with pipette free [Ca²⁺] in 1.5 mM EGTA indicated under each panel. Colors denoting bath conditions are used throughout. (B) Biphasic matrix [Ca²⁺] regulation of MCU Ca²⁺-current density in wild-type mitoplasts at −160 mV with 1 mM bath [Ca²⁺]. Data from the current study shown in blue. Data from ref. 27 shown in light gray for reference. Peak inhibition is observed with matrix free [Ca²⁺] = 400 nM. Numbers indicate number of mitoplast recordings. Bars denote SEM. (C) Representative MCU Ca²⁺ currents with different pipette free [Ca²⁺] (shown below each set of traces) buffered with either 1.5 mM BAPTA (left group of three panels) or 5 mM BAPTA (right group of three panels). Normal inhibition with matrix free [Ca²⁺] = 400 nM buffered with 1.5 mM EGTA is shifted to 10-fold lower concentration in presence of 1.5 mM BAPTA (center in the left group). In 5 mM BAPTA, inhibition is shifted further to lower matrix free [Ca²⁺] (left in the right group). (D) MCU Ca²⁺-current density over range of pipette free [Ca²⁺] and buffering capacity. Black symbol marks the minimum reported in B. Numbers indicate numbers of mitoplasts recorded. Biphasic dependence of Ca²⁺-current density on matrix free [Ca²⁺] is shifted to lower concentrations with stronger buffering. (E) Relationship between the maximally inhibited MCU Ca²⁺-current density recorded [from (D)] as a function of pipette [BAPTA]. (F) Inhibition of MCU activity observed with matrix Ca²⁺ = 400 nM buffered with 1.5 mM EGTA (as in A and B) (Left) or with the matrix solution containing 5 mM BAPTA with no added Ca²⁺ (as in C, Right and D) is abolished in the absence of matrix Mg²⁺. Bars: SEM, ***P < 0.0000009.

Matrix Ca²⁺ Buffering Regulates Matrix Ca²⁺ Regulation of MCU Ca²⁺-Channel Activity.

Whereas it was previously inferred that the EMRE carboxyl terminus could contribute to matrix Ca²⁺ sensing (27), its location in the IMS (8, 15, 25, 28) precludes a direct role. To identify matrix Ca²⁺-regulatory sites, we first considered whether MCU can be regulated by [Ca²⁺]_i in the nanodomain created by Ca²⁺ flux through the channel, as is the case for plasma-membrane Ca²⁺ channels (29). Inhibition of such regulation by the fast Ca²⁺ buffer BAPTA but not by a slow buffer such as EGTA is considered evidence for permeant Ca²⁺ ion-flux regulation of channel activity (29, 30). To determine whether Ca²⁺ flux through MCU contributes to matrix [Ca²⁺] regulation of its activity, we replaced 1.5 mM EGTA with the same concentration of BAPTA. We reasoned that faster buffering might either eliminate matrix [Ca²⁺] regulation or shift the inhibitory [Ca²⁺]_m to higher concentrations in the bulk solution, since we assumed that the local flux contribution would be reduced by faster buffering. Instead, BAPTA unexpectedly shifted the entire biphasic matrix [Ca²⁺] dependence to 10-fold lower concentrations (Fig. 1C). Conversely, reducing [BAPTA] from 1.5 mM to 0.5 mM shifted the biphasic dependence in the opposite direction (Fig. 1D). Incremental increases of [BAPTA] in the pipette solution shifted the biphasic matrix [Ca²⁺] dependence of MCU activity to even lower concentrations such that, at 5 mM BAPTA, only the maximal inhibition and recovery from inhibition could be observed (Fig. 1 C and D). A linear relationship in a semilog plot of [BAPTA] and the inhibitory [Ca²⁺]_m (Fig. 1E) suggests that matrix Ca²⁺ buffering regulates a single dominant process that controls matrix [Ca²⁺] regulation of channel activity. Physiologically, mitochondrial matrix Ca²⁺ buffering capacity is provided by phosphate (P_i) (31). Matrix [P_i] also modulated [Ca²⁺]_m regulation of MCU-channel activity (SI Appendix, Fig. S1).

Sensitivity of [Ca²⁺]_m regulation of MCU activity to matrix BAPTA and P_i supports the suggestion that Ca²⁺ flux through the channel plays an important role in this process. Whereas it was expected that BAPTA would desensitize the channel to inhibition by matrix [Ca²⁺], the opposite was observed. Notably, the biphasic nature of matrix [Ca²⁺] regulation remained intact under different buffer regimes (Fig. 1D). This implies that the inhibitory S_i and activating S_a Ca²⁺ sensors are coupled and jointly regulated indirectly by what we here term a low-affinity Ca²⁺-flux sensor S_f. The simplest model suggests that Ca²⁺ occupancy of S_f allosterically tunes the Ca²⁺ affinities of S_i and S_a. When S_f is occupied, for example when matrix Ca²⁺ buffering is low, the apparent Ca²⁺ affinities of S_i and S_a are low, resulting in the biphasic dependence shifted to higher matrix [Ca²⁺]. When S_f is unoccupied, for example when the matrix is buffered with 5 mM BAPTA, the apparent Ca²⁺ affinities of S_i and S_a are high, resulting in the biphasic dependence shifted to very low [Ca²⁺]_m.

Whereas this model can account for most of the results, strong channel inhibition observed in 5 mM BAPTA at very low [Ca²⁺]_m (Fig. 1 C and D) was not predicted, since all matrix Ca²⁺-regulatory sites as well as the flux sensor should be unoccupied, and MICU1/2 was fully Ca²⁺-liganded under our recording conditions, which together would be expected to activate the channel. The pipette solutions in all of our Ca²⁺-current measurements contained 2 mM MgCl₂, used historically to increase mitoplast electrical stability (26). We therefore considered that Mg²⁺ binding to the inhibitory Ca²⁺ sensor S_i might underlie channel inhibition observed in high-[BAPTA]/low-[Ca²⁺]_m. Remarkably, not only did removal of Mg²⁺ from the pipette solution completely abolish MCU-channel inhibition observed in 5 mM BAPTA, it also eliminated MCU-channel inhibition normally observed in 400 nM Ca²⁺/1.5 mM EGTA (Fig. 1F).

Identification of the Matrix Ca²⁺ Inhibition Sensor.

Previously, a structure of the isolated amino-terminal domain of human MCU, located in the mitochondrial matrix in the full-length channel (25, 28), revealed an acidic patch with a bound Mg²⁺ that was postulated to be a low-affinity divalent cation-binding site (32). We speculated that this site might be the structural basis for matrix [Ca²⁺] inhibition in full-length MCU. To test this, we individually mutated two key aspartic-acid residues important for divalent-cation binding (32)—D131 and D147—to alanines (D131A and D147A) and recorded Ca²⁺ currents through the mutant MCU channels following stable expression in MCU-knockout (KO) HEK cells overexpressing MICU1 to ensure proper regulation of the recombinant channels (33). Both mutant MCU channels expressed at somewhat higher levels than that of rescue wild-type MCU by Western blot analyses of isolated mitochondrial lysates, with MICU1/2 dimer expression similar among the different cell lines (SI Appendix, Fig. S2). However, with 1.5 mM EGTA and no added Ca²⁺ in the pipette solutions, Ca²⁺-current densities were not different between the mutant- and wild-type-MCU–expressing cells (Fig. 2 A–C), suggesting that the functional expression of the mutant and wild-type channels was equivalent among the lines. Importantly, normal channel inhibition by 400 nM [Ca²⁺]_m (Fig. 1 A and B) was abolished in both MCU-mutant channels (Fig. 2 A–C). Furthermore, channel inhibition observed in high-[BAPTA]/low-[Ca²⁺]_m (Fig. 1 D and E) was also abolished (Fig. 2D). In contrast, charge neutralization of E117, located in a distinct acidic patch in the amino terminus (32), in E117Q-MCU channels, was without effect on normal matrix [Ca²⁺] inhibition (SI Appendix, Fig. S3A). These results indicate that the amino terminus of MCU contains a functional Ca²⁺-binding site encompassing residues D131 and D147 that accounts for matrix Ca²⁺ (and Mg²⁺) inhibition of channel activity. Occupancy of this site overrides normal cytoplasmic Ca²⁺-mediated MICU1/2 activation of channel activity to inhibit MCU. To independently test this, we measured intact-mitochondrial Ca²⁺ uptake in digitonin-permeabilized cells in response to incremental boluses of 5 to 7 μM Ca²⁺. Low bath [Ca²⁺] before addition of Ca²⁺ pulses remained unchanged in response to inhibition of NCLX, the major mitochondrial Ca²⁺ extrusion mechanism, indicating that MCU was inactive in the low-[Ca²⁺]_i regime (19). Thus, channel gatekeeping remained functional in D131A-, D147A-, and E117Q-MCU–expressing cells (Fig. 2E and SI Appendix, Fig. S4A), indicating that the MCU/EMRE/MICU1/2 channel complex was intact in the cells expressing mutant MCU. The resting levels of [Ca²⁺]_m in the mutant MCU-expressing cells were moderately elevated compared with that observed in the MCU-rescue cells (SI Appendix, Fig. S4C). The rates of Ca²⁺ uptake in response to the first Ca²⁺ bolus were not different among the different cell lines (SI Appendix, Fig. S4B), consistent with the electrophysiological data that indicated that the functional expressions of wild-type and of mutant MCU channels were equivalent. Whereas mitochondria in MCU-KO cells rescued with wild-type MCU rapidly took up Ca²⁺ in response to 9 to 12 successive boluses, those in cells rescued with either D131A-MCU or D147A-MCU exhibited partial failure to take up Ca²⁺ after only two Ca²⁺ pulses, with complete failure observed after three or four pulses associated with collapse of ΔΨ_m (Fig. 2E), indicating matrix-Ca²⁺ overload-induced activation of the PTP. In contrast, cells rescued with E117Q MCU behaved similarly to wild-type MCU rescue cells (SI Appendix, Fig. S3B). These results indicate that matrix Ca²⁺ regulation of MCU-channel activity observed in mitoplast electrophysiology is functional in intact mitochondria, with elevated matrix [Ca²⁺] feeding back to inhibit Ca²⁺ influx, mediated by the Ca²⁺-binding D131/D147 site in MCU. This feedback inhibition overrides cytoplasmic Ca²⁺-dependent MICU1/2 activation of MCU to prevent excessive mitochondrial Ca²⁺ uptake and PTP opening.

Fig. 2.

The MCU amino terminus mediates matrix [Ca²⁺] inhibition of MCU. (A) Representative MCU Ca²⁺-current traces recorded from MCU-KO cells rescued with wild-type MCU and MICU1 (blue label) with pipette buffer and [Ca²⁺] as indicated. (B, Left) Similar, in MCU-KO cells rescued with D131A-MCU (red label). (B, Right) Summary of MCU Ca²⁺-current densities recorded in wild-type MCU and D131A-MCU mitoplasts with pipette solutions containing 0 or 400 nM free Ca²⁺ buffered with 1.5 mM EGTA. Bars: SEM, ***P = 0.0005. (C) Similar, in MCU-KO cells rescued with D147A-MCU. Bars: SEM, ***P = 0.0005. (D) MCU channel inhibition in 0 Ca²⁺/5 mM BAPTA (Left; black symbols in summary panel at Right; see also Figs. 1C and 3C) is abolished in mitoplasts expressing D131A-MCU (Middle; red symbols in summary panel). Bars: SEM, ***P = 0.0005. (E) Mitochondrial Ca²⁺ uptake and ΔΨ_m in digitonin-permeabilized cells in response to successive additions of 5 to 7 μM Ca²⁺ boluses in MCU-KO cells rescued with wild-type (WT) MCU and MICU1 (green; n = 3) or with D131A- (red; n = 4) or D147A- (blue; n = 4) MCU and MICU1. Traces show mean (solid line) ± SEM (shaded). Mitochondria in WT MCU-rescue cells are able to rapidly take up Ca²⁺ in response to 10 to 12 boluses (only 5 shown for comparison), whereas mutant MCU-rescue cells fail after three or four boluses, associated with collapse of ΔΨ_m, indicative of matrix Ca²⁺-induced PTP opening.

Interaction of Matrix [Ca²⁺] and MICU1/2 Regulation of MCU Activity.

Strong inhibition of MCU-channel activity by matrix [Ca²⁺] occurred despite MICU1/2 being fully Ca²⁺-liganded by Ca²⁺ present in the bathing solution. To explore the relationship between matrix [Ca²⁺] regulation and MICU1/2-mediated cytoplasmic [Ca²⁺] regulation of MCU activity, we genetically knocked down/out MICU1 or MICU2, or inhibited their association with the channel complex by mutating the EMRE carboxyl terminus (15). We previously demonstrated that knockdown (KD) of either MICU1 or MICU2 or expression of mutant EMRE have no effects on MCU expression (19). Furthermore, Ca²⁺-current densities measured with 0 or 1 μM Ca²⁺ in the pipette solution were comparable in MICU KO and mutant EMRE-expressing cells to those in wild-type cells (27). However, each of these genetic interventions that abolish [Ca²⁺]_i regulation of MCU channel activity also abolished matrix [Ca²⁺] regulation (Fig. 3 A and B and see also ref. 27). Reexpression of wild-type EMRE, MICU1, or MICU2 rescued normal matrix [Ca²⁺] regulation (Fig. 3B). Matrix [Ca²⁺] inhibition of channel activity observed with the pipette solution containing 5 mM BAPTA/0-Ca²⁺ (Fig. 1 C and D) was also fully abolished by genetic deletion of MICU1 or MICU2 (Fig. 3C). This result suggests that uncoupling of matrix [Ca²⁺] and MICU1/2 regulation could not account for channel inhibition observed in this high-[BAPTA]/low-[Ca²⁺]_m condition. Together these results demonstrate that matrix [Ca²⁺] regulation of MCU activity depends on MICU1/2 association with the channel on the opposite side of the membrane, suggesting that matrix [Ca²⁺] and MICU1/2 regulation of MCU activity are tightly coupled.

Fig. 3.

Matrix [Ca²⁺] regulation of MCU requires the MICU1/2 heterodimer. (A) Representative MCU Ca²⁺ currents recorded with pipette free [Ca²⁺] = 400 nM in mitoplasts from cells lacking functional MICU1/2 regulation due to expression in EMRE KO cells of EMRE with its acidic carboxyl terminus deleted (EMRE-ΔC) (Left) or in MICU1- (Middle) or MICU2- (Right) KO cells. (B) Summary of MCU Ca²⁺-current densities recorded with matrix free [Ca²⁺] = 400 nM in mitoplasts from wild-type cells and EMRE-ΔC, MICU1- (Middle), or MICU2- (Right) KO cells, and their respective rescue cells. Loss of normal inhibition observed with matrix free [Ca²⁺] = 400 nM in cells lacking MICU1/2 regulation is rescued by expression of wild-type EMRE, MICU1, and MICU2. Bars: SEM, ***P < 0.00003. See also ref. 27. (C) Matrix [Ca²⁺] inhibition of MCU Ca²⁺-channel activity observed with pipettes containing 5 mM BAPTA and no added Ca²⁺ (Left; see also B) requires MICU1 (Middle) and MICU2 (Right; representative families of current recordings). (Far Right) Summary of Ca²⁺-current densities in mitoplasts recorded at −160 mV with 1 mM bath [Ca²⁺] with each point representing a single mitoplast. Bars: SEM, ***P = 0.0007 (MICU1) and P = 0.00009 (MICU2).

Regulation of MCU Na⁺ Currents.

Whereas apo-MICU1/2 normally inhibits MCU-mediated Ca²⁺ influx in low [Ca²⁺]_i (14, 18–22), it fails to inhibit MCU Na⁺ currents observed in the absence of Ca²⁺ as permeant ion (1, 26, 27). Furthermore, Na⁺ currents observed with matrix [Ca²⁺] buffered with 1.5 mM EGTA are independent of MICU1/2 association with the channel complex (27). We considered that Na⁺ currents are observed because they are recorded with pipette solutions lacking Mg²⁺ (1, 26, 27). However, addition of 2 mM Mg²⁺ to the pipette solutions did not inhibit the inward Na⁺ currents (Fig. 4A). Furthermore, whereas 300 to 400 nM Ca²⁺_m buffered with 1.5 mM EGTA inhibited Ca²⁺ currents (Fig. 1 A and B), it was without effect on Na⁺ currents (Fig. 4A). In both conditions, the inhibitory D131/D147 Ca²⁺-sensor site was liganded, but it was no longer inhibitory with Na⁺ as the current carrier. Thus, under conditions in which Na⁺ is the current carrier, matrix [Ca²⁺]- and MICU1/2-regulation of MCU gating appear to be uncoupled. To explore this further, we examined the effects of high-[BAPTA]/low-[Ca²⁺]_m/0-Mg²⁺ pipette solutions. Notably, Na⁺ currents were strongly inhibited in this condition (Fig. 4B). Thus, with sufficiently strong Ca²⁺ buffering to ensure that all matrix Ca²⁺ sensors are unoccupied, normal apo-MICU1/2 inhibition of MCU activity is similarly observed for Ca²⁺ influx and Na⁺ currents.

Fig. 4.

Regulation of MCU Na⁺ currents and a model for coupled matrix- and cytoplasmic-Ca²⁺ regulation. (A) MCU Na⁺ currents are not inhibited by matrix Ca²⁺ or Mg²⁺. First two panels: representative recordings of MCU-mediated Na⁺ currents in absence of added pipette Ca²⁺ (Left) and presence of 300 nM matrix Ca²⁺ (Right) with 2 mM Mg²⁺. Normal Ca²⁺-current inhibition by 300 nM matrix Ca²⁺ buffered with 1.5 mM EGTA (Fig. 1 A and B) is not observed for MCU Na⁺ currents. Inward rectification of Na⁺ currents caused by Mg²⁺ in the pipette solution. Third panel: representative recordings of MCU-mediated Na⁺ currents in absence of both matrix Ca²⁺ (1.5 mM EGTA) and Mg²⁺. Lack of Mg²⁺ results in linear Na⁺ currents. Fourth panel: summary data with pipette solutions containing 1.5 mM EGTA with 0 (blue) or 300 to 400 nM (red) Ca²⁺ with and without 2 mM Mg²⁺. Data for 0 Mg²⁺/400 nM Ca²⁺ from ref. 27. Bars: SEM. No significant differences among groups. (B, Left) Representative Na⁺ current recordings with pipette solution containing 0 Mg²⁺ and 0 Ca²⁺/5BAPTA. (B, Right) Summary demonstrating inhibition of MCU Na⁺ currents by strong Ca²⁺ buffering in absence of matrix Mg²⁺. Bars: SEM, ***P = 4 E-10. (C) Cartoon models of MCU-channel regulation by coupled MICU1/2 and matrix [Ca²⁺] mechanisms. Numbers 1 through 4 depict physiological cytoplasmic- and matrix [Ca²⁺] regulation of MCU-channel open probability. Numbers 5 through 7 depict regulation of experimental MCU-mediated Na⁺ currents. MICU1 and MICU2 are depicted as a single entity (MICU1/2) facing the IMS with 4 Ca²⁺-binding sites. Distinct functional conformations of MICU1/2 and MCU are depicted, with comparable conformations in Ca²⁺- and Na⁺-permeation experiments aligned vertically. Pore block depictions in 1, 3, and 5 are meant to reflect inhibited MCU activity rather than a particular molecular mechanism. Arrows within MCU denote conformational changes in MCU associated with Ca²⁺ binding to inhibitory (S_i), activating (S_a), and flux (S_f) sensor sites. See text for details.

Discussion

Previous studies of intact mitochondria in permeabilized cells (11, 12, 14, 18–22) and in mice (23, 24, 34) and humans (35–37) have suggested important roles for IMS-localized MICU1/2 in MCU regulation by cytoplasmic [Ca²⁺]. Our results indicate that cytoplasmic MICU1/2 regulation of MCU is not an independent channel regulatory mechanism but is instead regulated by Ca²⁺ sensors located in the matrix, at least one of which is an integral component of the channel pore-forming subunit. A model that can account for our data is schematized in Fig. 4C. In the low-[Ca²⁺]_i regime characteristic of resting conditions in the cytoplasm, MICU1/2 in its apo conformation maintains MCU in a closed-channel conformation, either by inhibiting channel gating or, more likely, by physically occluding the permeation pathway (25) (Fig. 4 C, 1). Maintenance of this inhibited channel conformation may also be facilitated by matrix Ca²⁺, since we observed relatively strong inhibition of MCU Ca²⁺ currents even at 100 nM, the physiological resting matrix [Ca²⁺]. In the presence of elevated cytoplasmic [Ca²⁺], either physiologically or when recording Ca²⁺ currents in mitoplast electrophysiology, Ca²⁺ binding to MICU1/2 induces a conformational switch (25, 38–41) that relieves channel inhibition imposed by the apo conformation, and mitochondrial Ca²⁺ uptake (11–14, 18–22) and Ca²⁺ currents (1, 26, 27, 42) can be recorded (Fig. 4 C, 2). Ca²⁺ influx promotes Ca²⁺ binding to a flux sensor, which in turn tunes the affinities of other, coupled inhibitory and activating Ca²⁺-sensor sites located in the matrix. The evidence for this was revealed by the biphasic matrix [Ca²⁺] dependence of MCU Ca²⁺ current densities, and the effects of the fast Ca²⁺ buffer BAPTA in the matrix on MCU channel activity. The biphasic nature of the relationship remained intact as BAPTA concentration was increased, suggesting that the inhibitory and activating sites are coupled and jointly regulated by a sensor that is sensitive to bulk matrix Ca²⁺-buffering capacity. Because BAPTA at the same concentration as the slower Ca²⁺ buffer EGTA also shifted the matrix [Ca²⁺] dependence in a manner similar to that of simply increasing buffer capacity, we assume that the sensor that regulates the coupled inhibitory and activating sites is located within a region of high Ca²⁺ concentration that cannot be well buffered by EGTA. The source of this Ca²⁺ is most likely the Ca²⁺ that permeates through the channel pore. Accordingly, we have termed this the flux sensor. Its physical location remains to be determined, but we speculate that it could be an internal Ca²⁺-binding site associated within the channel selectivity filter (43) [we note that Ca²⁺ remained bound in the selectivity filter in the MCU/EMRE/MICU1/2 complex structure solved in 5 mM EGTA with no added Ca²⁺ (25)] or localized at a putative inner gate at the bottom of the cavity below the pore (28). With the flux sensor and the inhibitory site, which we localized to the MCU amino terminus, fully Ca²⁺-liganded, MCU channel activity is strongly reduced even with MICU1/2 fully Ca²⁺-liganded (Fig. 4 C, 3). In the presence of 1.5 mM EGTA, maximal inhibition of ∼80% is observed when matrix [Ca²⁺] is ∼400 nM. Two lines of evidence suggest that these observations in mitoplasts with artificial buffers are physiologically relevant. First, we demonstrated that changes in the matrix [P_i], the major mitochondrial matrix Ca²⁺ buffer in situ, also affected matrix Ca²⁺ regulation of MCU channel open probability, although the experiments were complicated by the low Ca²⁺ affinity of P_i. Second, we demonstrated the existence and importance of matrix [Ca²⁺] regulation of MCU channel activity in situ in intact mitochondria by Ca²⁺-uptake assays in permeabilized cells expressing mutant MCU channels that lacked matrix Ca²⁺ inhibition of mitoplast Ca²⁺ currents. These mitochondria underwent Ca²⁺-induced PTP opening associated with a collapse of the ΔΨ_m in response to Ca²⁺ transients that were without detrimental effects on mitochondria expressing wild-type MCU (with matrix Ca²⁺ inhibition intact). Importantly, inhibition of MCU-channel activity by matrix Ca²⁺ depends upon the presence of MICU1/2 within the channel complex. The evidence for this was the lack of matrix [Ca²⁺] inhibition of MCU Ca²⁺ currents in cells lacking MICU1 or MICU2, and also in cells in which the heterodimer MICU1/2 was not functional due to the absence of its EMRE-mediated tethering to the channel complex. This suggests that Ca²⁺ binding to the matrix inhibitory site in the amino terminus of MCU results in a conformational change in MCU that enables Ca²⁺-bound MICU1/2 to inhibit its activity (Fig. 4 C, 3), by pore block or inhibiting channel gating. In contrast, the data suggest that matrix Ca²⁺ binding to the lower-affinity activating site induces a different channel conformation that no longer accommodates Ca²⁺-liganded MICU1/2, enabling Ca²⁺ permeation (Fig. 4 C, 4). Failure to account for the fundamental modulation by matrix [Ca²⁺] of MICU1/2 regulation may result in an inability to observe MICU1/2 regulation of Ca²⁺ currents by patch-clamp electrophysiology if the pipette [Ca²⁺] is either too high or too low, depending upon the matrix Ca²⁺-buffering properties (1, 44). For example, Hoffman et al. (10) reported enhanced Ca²⁺ currents in MICU1-KD mitoplasts compared with those observed in wild-type mitoplasts, whereas we have observed no difference (27). Whereas both studies employed similar patch-pipette solutions containing 1.5 mM EGTA, Hoffman et al. (10) challenged mitoplasts with 5 mM bath Ca²⁺ compared with 1 mM Ca²⁺ in our studies. It is likely that the approximately fivefold greater flux was not adequately buffered, leading to an increase in the matrix [Ca²⁺] and consequent inhibition of MCU activity recorded from wild-type mitoplasts. Because [Ca²⁺]_m regulation of MCU channel activity is absent in MICU1-KD mitoplasts, the MCU Ca²⁺ currents would be larger than in the wild type, as observed. MICU1/2 regulation of Na⁺ currents was only revealed under conditions that strongly minimized occupancy of all matrix Ca²⁺ sensors (Fig. 4 C, 5). Because this has not been previously considered (Fig. 4 C, 6 and 7), it can explain why the channel has been observed to be constitutively active when Na⁺ currents are measured (1, 26, 27, 44).

Our results indicate that Ca²⁺ occupancy of matrix Ca²⁺ sensors regulates MCU channel-conformational states that determine the efficacy of MICU1/2 regulation. Previously, regulation of MCU channel activity was defined by cytoplasmic Ca²⁺ modulation of MICU1/2 that sets a [Ca²⁺]_i threshold for MCU channel activity and its cooperative activation. Our results indicate that this regulation is set not only by the Ca²⁺ affinities of MCU1/2 EF hands but also by matrix [Ca²⁺] and buffering capacity, allowing for enhanced cellular regulation of mitochondrial Ca²⁺ homeostasis. The importance of this enhanced regulation is reflected by the different behaviors of mitochondria expressing D131A or D147A MCU mutants. These mutations abolished the matrix Ca²⁺ regulation of mitoplast Ca²⁺ currents, leaving only the cytoplasmic MICU1/2-dependent gatekeeping mechanism to protect mitochondria from Ca²⁺ overload. Although this mechanism was sufficient to protect these mitochondria at resting cytoplasmic Ca²⁺ levels (SI Appendix, Fig. S4C), it was insufficient in the absence of matrix Ca²⁺ regulation, as evidenced by PTP opening when the mitochondria were challenged with Ca²⁺ boluses (Fig. 2E).

Recent studies demonstrated a direct biochemical interaction of MICU1/2 with the MCU selectivity filter (16, 17). Because mutations in either MCU or MICU1 that disrupt this interaction lead to a constitutively open channel, it was suggested that MICU1/2 could possibly function as a classic channel pore blocker (17). Recent structural data strongly support this model (25). Our results here indicate that structural changes that affect this interaction are localized not only in MICU1/2 (38–41), regulated by cytoplasmic Ca²⁺, but also in the MCU channel itself, regulated by Ca²⁺ sensors localized on the opposite side of the IMM. We suggest that the structural plasticity of the MCU–MICU interface enables MCU-channel regulation to be dependent upon Ca²⁺ levels in both the cytoplasm and mitochondrial matrix. This mechanism is reminiscent of Ca²⁺-dependent inactivation of plasma membrane voltage-gated Ca²⁺ channels (VGCC). There, Ca²⁺ flux through the channel pore interacts with channel elements in the cytoplasm that results in structural changes in the selectivity filter on the other side of the membrane that close the channel (45). Similarly, voltage-dependent inactivation of VGCC involves elements in the cytoplasmic aspect of the channel that drive conformational changes near to the ion-conducting pore (45, 46). Alternately, it has been suggested that MCU may possess an internal gate at the base of the internal central cavity below the selectivity filter (28). It is possible that structural changes regulated by MICU1/2 and matrix Ca²⁺ sensors regulate this gate. An important goal now is to understand how Ca²⁺ binding to mitochondrial matrix sensors is transduced into conformational changes of the MCU channel that affect MCU interactions with MICU1/2 to regulate Ca²⁺ permeation.

The importance of the mechanisms described here for other cell types remains to be determined. Quantitative differences of MCU activity exist among mitochondria in various tissues (26), and MCU regulation in the heart may be different from that in other tissues by lack of a MICU1/2 “gatekeeping” mechanism, employing instead low MCU-channel density to enable Ca²⁺ intake to be effectively compensated for by the extrusion activity of NCLX (4, 47). In a preliminary electrophysiology study of mitoplasts from mouse embryonic fibroblasts lacking DRP1, it was suggested that MICU1/2 inhibition of MCU channel activity (“gatekeeping”) does not exist, and that matrix [Ca²⁺] regulation of MCU activity is absent (44). With regard to the former, a cryo-electron microscopic analysis of the human MCU/EMRE/MICU1/2 complex revealed a structural basis for channel “gatekeeping” by observing that apo-MICU1/2 physically occludes the channel pore (25), strongly supporting numerous studies that have demonstrated the role of apo-MICU1/2 in maintaining the MCU channel in an off state in resting cytoplasmic [Ca²⁺] conditions. With regard to the latter, we point out that [Ca²⁺] in the pipette solutions used was not measured, only one calculated [Ca²⁺] was tested, and the possible role of matrix [Mg²⁺] was not considered. Future studies will be necessary to clarify these controversies.

Materials and Methods

Cell Lines, Constructs, and Antibodies.

Wild-type HEK-293T and KO cell lines were grown in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum, 100 U/mL penicillin, and 100 µg/mL streptomycin at 37 °C and 5% CO₂. MCU-KO, MICU1-KO, and MICU2-KO HEK-293T cells were a generous gift from Vamsi Mootha, Harvard Medical School, Boston, MA. Full-length MICU1, MICU2, and EMRE complementary DNAs (cDNAs) were purchased from Origene Technologies USA. EMRE-ΔC-V5 constructs were generated by a PCR-based cloning strategy as described (27). All constructs were sequence-verified and stably transfected into KO backgrounds. D131A-MCU-Flag and D147A-MCU-Flag cDNAs in pCMV6-A-BSD were a gift from Muniswamy Madesh, University of Texas Health Science Center at San Antonio, San Antonio, TX. E117Q-MCU-V5-His in pCMV6-A-Puro was generated using the QuikChange II XL site-directed mutagenesis kit and XL10-Gold ultracompetent Escherichia coli. All MCU constructs were sequence-verified and stably transfected in the MCU-KO backgrounds together with MICU1. MCU WT-rescue or mutant cells were grown in complete DMEM supplemented either with blasticidin (5 μg/mL), puromycin (2 μg/mL), or both, until stable expression of each plasmid was achieved. Clones were isolated by limiting dilution for each condition and maintained in complete medium with antibiotics. Exclusive mitochondrial localization was verified by immunofluorescence. Only clones expressing recombinant proteins in all mitochondria in all cells were selected for experimentation.

Uniporter Electrophysiology.

Mitoplast electrophysiology was performed as described (1, 26, 27). Patch pipettes had resistances of 20 to 60 MΩ when filled with (in millimolar): 120 TMA-OH, 120 Hepes, 80 d-gluconic acid, 10 glutathione, and 2 MgCl₂, pH = 7.0, with d-gluconic acid; osmolarity was 390 to 450 mOsm/kg. EGTA or BAPTA were used, as specified, to buffer correspondent amounts of CaCl₂ for [Ca²⁺]_free ≤400 nM (confirmed by Ca²⁺-sensitive dye fluorimetry). The free [Mg²⁺] ([Mg²⁺]_free, assessed using MaxChelator software (https://somapp.ucdmc.ucdavis.edu/pharmacology/bers/maxchelator/webmaxc/webmaxcE.htm), was not significantly affected by the different Ca²⁺ chelators used in our study: In 5 mM and 0.5 mM BAPTA, [Mg²⁺]_free = 1.8 mM and 2.0 mM, respectively. Mitoplasts were initially bathed in (in millimolar): 150 KCl, 10 Hepes, and 1 EGTA, pH = 7.2; osmolarity was 300 mOsm/kg (“KCl-DVF” solution). Voltage pulses (350 to 500 mV for 15 to 50 ms) were delivered by the PClamp-10 (Molecular Devices) program to obtain the “whole-mitoplast” configuration. Access resistance (30 to 90 MΩ) and mitoplast capacitance C_m (0.2 to 1 pF) were determined using the membrane test protocol of the PClamp-10 software. After the whole-mitoplast configuration was obtained, the KCl-DVF bath solution was exchanged with Hepes-EGTA (in millimolar: 150 Hepes and 1.5 EGTA, pH = 7.0, with Tris base) for baseline measurements and then replaced with Hepes-0-EGTA solutions with 0.1, 0.3, or 1 mM CaCl₂ successively. Finally, a Hepes-based solution with 1 mM CaCl₂ and 200 nM ruthenium red (RuR) was perfused into the bath to record the final baseline (= I_RuR) after block of the MCU currents. Osmolarities of bath solutions were 297 to 305 mOsm/kg, adjusted with sucrose. The voltage protocol, delivered by the PClamp-10 software with a DigiData-1550 interface (Molecular Devices), consisted in stepping the command membrane voltage V_m from 0 mV to −160 mV for 20 ms, followed by ramping to 80 mV (rate of 279 mV/s), dwelling at 80 mV for 20 ms, and returning to 0 mV. Currents were recorded using an Axopatch 200-B amplifier at room temperature with a sampling rate of 50 kHz and antialiasing-filtered at 1 kHz. Data analysis was performed with the PClamp-10 software. For quantitative comparisons, current densities are defined as: I_MCU/C_m = (I_Ca – I_RuR)/C_m, with I_Ca and I_RuR measured at V_m = −160 mV in the presence of 1 mM Ca²⁺ in the bath.

Simultaneous Measurements of Mitochondrial Ca²⁺ Uptake and Mitochondrial Membrane Potential in Permeabilized Cells.

Concurrent measurements of mitochondrial Ca²⁺ uptake and IMM potential (ΔΨ_m) in permeabilized HEK-293T cells were performed as described (18). Cells were grown in 10-cm tissue culture-coated dishes for 48 h prior to each experiment; 6 to 8 × 10⁶ cells were trypsinized, counted, and washed in a Ca²⁺-free extracellular-like medium, centrifuged, suspended in 1.5 mL of intracellular-like medium (ICM, in millimolar: 120 KCl, 10 NaCl, 1 KH₂PO₄, 20 Hepes, and 5 succinate [pH 7.2]) that had been treated with BT Chelex 100 resin before use to attain an initial free [Ca²⁺] of ∼20 nM, and transferred to a cuvette. The cuvette was placed in a temperature-controlled (37 °C) experimental compartment of a multiwavelength-excitation dual wavelength-emission high-speed spectrofluorometer (Delta RAM; Photon Technology International). Membrane-impermeable Fura2 K⁺ salt (dissocation constant [K_D] = 140 nM) or FuraFF K⁺ salt (K_D = 5.5 µM) (final concentration 1 µM) and tetramethylrhodamine ethyl ester (TMRE) (final concentration 1 µM) were added at time t = 25 s from the start, to measure bath [Ca²⁺]_c and ΔΨ_m concurrently. Fluorescence at 549-nm excitation/595-nm emission for TMRE, along with 340-nm and 380-nm excitation/535-nm emission for Fura2/FuraFF, was measured at 5 Hz; 40 µg/mL digitonin was added at t = 50 s to permeabilize cells and allow the cytoplasm to equilibrate with the bath solution such that the degree of quenching of TMRE fluorescence reports the relative ΔΨ_m. Fura2/FuraFF fluorescence under these conditions is related to the cytoplasmic [Ca²⁺] ([Ca²⁺]_c) according to

$$\left[C{a}^{2+}\right]=\left(\frac{R-R_{\min }}{R_{\max }-R}\right)\left(\frac{S{f}_2}{S{b}_2}\right)\ast K_D,$$

where (R) is the measured fluorescence ratio (340/380 nm), (R_min) is the fluorescence ratio under 0 [Ca²⁺], (R_max) is the fluorescence ratio under saturating [Ca²⁺], (Sf₂) and (Sb₂) are the absolute 380-nm excitation/535-nm emission fluorescence under 0 [Ca²⁺] and saturating [Ca²⁺], respectively, and (K_D) is the dissociation constant of the dye for Ca²⁺. Thapsigargin (2 µM) and CGP37157 (20 µM) were added at 100 s to inhibit Ca²⁺ uptake into the endoplasmic reticulum and NCLX-mediated mitochondrial Ca²⁺ extrusion, respectively. After [Ca²⁺]_c reached a steady state at ∼300 s, MCU-mediated Ca²⁺ uptake was initiated by adding a bolus of 5 µL of 5 mM CaCl₂ to the cuvette to achieve increases in [Ca²⁺]_c between ∼5 µM. The volume of CaCl₂ required to achieve the desired increase in [Ca²⁺]_c was calculated based on the activity coefficient for Ca²⁺ in ICM. After [Ca²⁺]_c was monitored for 150 s following Ca²⁺ addition, a second bolus of CaCl₂ was added. This procedure was repeated for a total of five to nine successive additions of Ca²⁺; 150 s after addition of the final bolus, CCCP (2 µM) was added to uncouple ΔΨ_m and allow unimpeded Ca²⁺ efflux from mitochondria as a measure of the total extent of uptake. Under our conditions, each Ca²⁺ bolus caused a transient depolarization as measured with TMRE that recovered after ∼50 to 60 s. Concurrent failure of Ca²⁺ uptake and loss of ΔΨ_m indicates opening of the mitochondrial PTP.

Initial Ca²⁺ uptake rates in permeabilized cells were determined for each experiment using single-exponential fits from the time of the first Ca²⁺ addition (t = 0 s) to achievement of a new steady state (t = 300 s) to obtain parameters A (extent of uptake) and τ (time constant) (IGOR Pro; Research Resource Identifier [RRID]: SCR_000325). The instantaneous rate of uptake (R) at t = 0 is equal to the first derivative of the fit, and SDs of rates were calculated from SDs of A (SD_A) and τ (SD_τ) as follows:

$$\mathrm{R}=\mathrm{A}/\mathrm{\tau }\hspace{0.17em}\hspace{0.17em}\mathrm{R}+\mathrm{SD}=\left(\mathrm{A+S}{\mathrm{D}}_{\mathrm{A}}\right)/\left(\mathrm{\tau -S}{\mathrm{D}}_{\mathrm{\tau }}\right)\hspace{0.17em}\mathrm{and}\hspace{0.17em}\mathrm{R}-\mathrm{SD}=\left(\mathrm{A}-\mathrm{S}{\mathrm{D}}_{\mathrm{A}}\right)/\left(\mathrm{\tau }+\mathrm{S}{\mathrm{D}}_{\mathrm{\tau }}\right).$$

For determination of the steady-state [Ca²⁺]_c after completion of uptake (gatekeeping threshold [Ca²⁺]_c), the calculated [Ca²⁺]_c 300 s after challenge with 0.1 to 10 µM Ca²⁺ (Y) was plotted as a function of the (peak) initial [Ca²⁺]_i achieved after bolus Ca²⁺ addition (X). Data were fitted using a one-phase association model where (Y₀) is the Y value when X is zero, (P) is the plateau, or Y value at infinite X, and (K) is the rate constant, expressed in reciprocal of the x axis:

$$Y=Y_0+\left(P-Y_0\right)\left(1-e^{\left(-KX\right)}\right).$$

The value of (P) represents to the steady-state [Ca²⁺]_c at which gatekeeping is reestablished ± the SE of (P) determined from the fit.

Sodium Dodecyl Sulfate Polyacrylamide Gel Electrophoresis Western Blotting.

For Western blots of mitochondrial proteins, mitochondria were isolated as described (6). Total protein concentrations were calculated using BCA Protein Assay kit (23225; Thermo Fisher Scientific), and samples for polyacrylamide gel electrophoresis (PAGE) were prepared in 2× Laemmli sample buffer (1610737; Bio-Rad) ± 2-mercaptoethanol (βME) (1610710; Bio-Rad). NuPAGE gels (4 to 12%, NP0321; Thermo Fisher Scientific) were transferred to Immobilon P poly(vinylidene difluoride) membranes (IPVH00010; Millipore-Sigma) and probed with various antibodies. Antibodies used were anti-MCU (14997S; Cell Signaling Technology), anti-V5 tag (13202S; Cell Signaling Technology), anti-MICU2 (ab101465; Abcam), anti-DYKDDDDK (Flag tag) (14793S; Cell Signaling Technology), anti-MICU1 (12524S; Cell Signaling Technology), anti–β-tubulin (32-2600; Invitrogen), and anti-Hsp60 (ab46798; Abcam). Membranes were blocked in 5% fat-free milk for 1 h at room temperature (RT), incubated overnight at 4 °C with primary antibody, and then for 1 h at RT with anti-rabbit immunoglobulin G (IgG)–horseradish peroxidase (HRP) (7074S; Cell Signaling Technology) or anti-mouse IgG–HRP (7076S; Cell Signaling Technology) secondary antibodies conjugated to HRP. Chemiluminescence detection was carried out using SuperSignal West Chemiluminescent Substrate (34580; Thermo Fisher Scientific). The mean pixel density of bands was quantified using the Fiji software package in ImageJ (RRID: SCR_003070), corrected for the intensity of the loading-control band (β-tubulin for whole-cell lysates or Hsp60 for isolated mitochondria), and normalized to the response in wild-type cells.

Data Availability.

All data relevant to the conclusions of the manuscript are included in the paper or in SI Appendix.