A whole mitoplast patch clamp method was used to measure whole IMM current to assess MCU_cx function (Fieni et al., 2012; Garg and Kirichok, 2019; Kirichok et al., 2004) and determine how the subunits contribute to the measured MCU_cx current. Mitochondria were isolated from DRP1 knockout (KO) mouse embryonic fibroblasts (MEFs) (Ishihara et al., 2009). DRP1 is encoded by the Dnm1l. The DRP1-KO MEFs were used to prepare mitoplasts using a French Press. This cell line was chosen as the source for many experiments because it provided a significantly higher proportion of large isolated mitoplasts and enabled the recording of stable MCU_cx currents with a favorable signal-to-noise ratio. We confirmed that this cell line expresses all principal subunits of the MCU complex (Figure 1—figure supplement 1). Importantly, the MCU_cx was intact in isolated mitoplasts, and its composition was the same as in intact mitochondria (Figure 1—figure supplement 2F). We also generated gene knockouts for all principal subunits of the MCU complex (MCU, EMRE, and MICU1−3) using CRISPR-Cas9 in this cell line (Figure 1—figure supplement 1). MCU, EMRE, and MICU1−3 are encoded by the Mcu, Smdt1 and Micu1−3 genes, respectively.

Figure 1A shows the [Ca²⁺]_i dependence of the MCU_cx current in mitoplasts from WT DRP1-KO MEFs, and shows the absence of Ca²⁺ current (I_Ca) in MCU-KO or EMRE-KO. Additionally, it shows an important feature of the MCU_cx; in the absence of extramitochondrial Ca²⁺ (control trace), there is outward current at positive potentials resulting from the efflux of Na⁺ through the MCU_cx due to the 110 mM Na⁺ gluconate in the matrix from the patch pipette (Figure 1A, Figure 1—figure supplement 2D and E). When extramitochondrial Ca²⁺ is present, Ca²⁺ enters the selectivity filter of the MCU_cx channel to block Na⁺ permeation (Fieni et al., 2012; Garg and Kirichok, 2019; Kirichok et al., 2004) and no outward current is seen. Importantly, MCU_cx currents can be rescued by the ectopic expression of the MCU and EMRE subunits in their corresponding knockout cell lines (Figure 1B and C, Figure 1—figure supplement 2G and H). From these results, the DRP1-KO MEFs recapitulate key findings in previous publications (Chaudhuri et al., 2013; Fieni et al., 2012; Kirichok et al., 2004; Sancak et al., 2013). We show an additional novel observation, important to our later experiments, that when Na⁺ is used to replace Ca²⁺ in the cytosolic compartment (i.e. the bath solution), an MCU-mediated Na⁺ current (I_Na) is observed, and this current also depends on the presence of MCU and EMRE (Figure 1D and E, Figure 1—figure supplement 2D and E).

In intact WT cells, the [Ca²⁺]_i increase (elicited by SERCA inhibitor thapsigargin) was followed, after a short delay, by [Ca²⁺]_m elevation as detected by a genetically-encoded Ca²⁺ indicator Cepia targeted to mitochondria (Suzuki et al., 2014; Figure 1—figure supplement 2A). However, as expected, in MCU-KO or EMRE-KO cell lines that have no functional MCU_cx (Baughman et al., 2011; De Stefani et al., 2011; Sancak et al., 2013), no significant [Ca²⁺]_m elevation was observed (Figure 1—figure supplement 2B and C).

One of the controversial elements in previously published experiments is the explanation of the cause of the ‘threshold’ of the MCU_cx Ca²⁺ influx into the matrix (Csordás et al., 2013; Tomar et al., 2019; Foskett and Madesh, 2014; Hoffman et al., 2013; Mallilankaraman et al., 2012b; Perocchi et al., 2010; Tufi et al., 2019). It was noted initially (Mallilankaraman et al., 2012b) that there is a cytosolic concentration of Ca²⁺ ([Ca²⁺]_i) below which there is no MCU_cx-medicated Ca²⁺ influx. Figure 1F shows that such a threshold for DRP1-KO WT MEF is indeed found at around 400 nM Ca²⁺. This threshold is largely gone in MICU1-KO cells (Figure 1G). In MICU2-KO and MICU3-KO cells, the changes in the threshold levels are shown in Figure 1H, and Figure 1—figure supplement 3A–D. Past studies of other investigators have used similar information from their MICU1-KO cells to argue that MICU1 forms a ‘plug’ or an occlusion in the channel pore. This implies that in MICU1-KO cells, the MCU_cx conductance should be greater due to the removal of the plug. This conclusion, however, is challenged by the Ca²⁺ conductance studies shown here in isolated mitochondria from WT MEFs shown in Figure 1I and Figure 1—figure supplement 3E. While the Figure shows increased conductance of MCU_cx at low [Ca²⁺]_i in the MICU1-KO mitochondria consistent with the removal of a putative MCU_cx‘plug’, there is decreased conductance at high [Ca²⁺]_i (~8 μM or higher), an observation that is inconsistent with the plug hypothesis (Mallilankaraman et al., 2012b). Additional recent modifications of this hypothesis add the prediction that allosteric actions of MICU1 on MCU_cx account for any inconsistencies or contradictions of the plug hypothesis (Csordás et al., 2013). Moreover, new findings suggest that there may be broad actions of MICU1 on non-MCU_cx targets within the mitochondria (Tomar et al., 2019; Gottschalk et al., 2019; Tufi et al., 2019). These findings and the unrefined and untested modifications of the plug hypothesis motivate additional investigations. We have carried out new quantitative experiments and analysis that may help us better understand how MICU1 works in the context of the MCU_cx as is shown in Figure 2.

To investigate how MICU1 works, I_Ca was measured in mitoplasts at five extramitochondrial [Ca²⁺]_i levels, 10 μM, 100 μM, 1 mM, 5 mM, and 25 mM (Figure 2A and B, and Figure 2—figure supplement 1A–C). WT mitoplasts show I_Ca records similar to the current density measurements from MICU2-KO and MICU3-KO mitoplasts. In contrast, WT mitoplasts have I_Ca current densities that are roughly twice the size of the current densities from the MICU1-KO mitoplasts. This finding is like the conductance measurements at elevated [Ca²⁺]_i in Figure 1I and thus inconsistent with the plug hypothesis that posits that MICU1 is an obstructing plug of MCU_cx.

The expression of EMRE protein (but not MCU) was significantly reduced in MICU1-KO (Figure 2—figure supplement 1D–F), as was also shown previously (Liu et al., 2016). However, the lower EMRE expression in MICU1-KO was not a limiting factor for I_Ca, because EMRE overexpression in MICU1-KO cells did not rescue the I_Ca reduction (Figure 2—figure supplement 1G–I). Therefore, the I_Ca reduction in MICU1-KO was not caused by reduction in MCU or EMRE.

To better understand the role played by the MICU1 subunit in the function of the MCU_cx, we used a novel test to characterize the MCU_cx channel properties independent of Ca²⁺ conductance. We used Na⁺ current via MCU_cx (I_Na) in the absence of both Ca²⁺ and Mg²⁺ (using EDTA) (Fieni et al., 2012; Garg and Kirichok, 2019; Kirichok et al., 2004), to calibrate I_Ca and characterize MCU_cx as shown in Figure 2C–D. Here, it is shown that the Na⁺ current through MCU_cx is indistinguishable in magnitude when it is measured in WT and MICU1-KO mitoplasts (Figure 2D). This provides evidence that the MCU_cx conductance pathway is the same in WT and MICU1-KO. Nevertheless, when Ca²⁺ permeates MCU_cx, the I_Ca in WT is roughly twice that of the current through MICU1-KO mitoplasts (Figure 2E). Also, the I_Ca/I_Na ratio as measured in the same mitoplast decreased approximately twice in MICU1-KO in comparison to WT (Figure 2F). Importantly, the reduction in I_Ca/I_Na ratio in MICU1-KO could not be explained by altered relative affinities for Ca²⁺ and Na⁺ binding in the selectivity filter, because I_Na was inhibited to the same extent by 2 nM [Ca²⁺]_i in both WT and MICU1-KO mitoplasts (Figure 2—figure supplement 2A and B). From this we conclude that the MICU1 subunit enhances the MCU current at high [Ca²⁺]_i and does not occlude the MCU_cx channel when [Ca²⁺]_i is low.

We also reproduced these results in MEFs with intact DRP1 (Dnm1l^+/+). In these cells, the amplitudes of I_Ca and I_Na were the same as in Dnm1l^-/- MEFs (Figure 2—figure supplement 3A and B). Similar to MICU1 knockout in Dnm1l^-/- MEFs, MICU1 knockout in Dnm1l^+/+ MEFs did not affect I_Na while markedly reduced I_Ca (Figure 2—figure supplement 3C–E). Additionally, MICU1-KO reduced the I_Ca/I_Na ratio, as measured in the same mitoplast, to the similar extent in Dnm1l^+/+ MEFs (Figure 2—figure supplement 3F). Thus, as expected, Drp1 presence or absence does not affect currents mediated by the MCU complex or the MICU1-KO phenotypes.

The lack of the MCU_cx occlusion by MICU1 at low [Ca²⁺]_i was further tested in intact isolated mitochondria as shown in Figure 2G–I. In these experiments, we found that depletion of Ca²⁺ and Mg²⁺ using EDTA enables a Na⁺ influx via MCU_cx that depolarizes ΔΨ_m (Figure 2G and H). As shown in Figure 2I, this influx depolarizes ΔΨ_m to the same extent whether MICU1 was expressed or not, again showing the lack of MCU_cx occlusion by MICUs.

Since in MICU1-KO, all MICUs are removed from the MCU_cx complex, we conclude that MICUs do not plug the MCU_cx channel when [Ca²⁺]_i is low. Instead, the function of MICUs is to potentiate MCU_cx activity at elevated [Ca²⁺]_i.

The Ca²⁺-dependent potentiation of MCU_cx imparted by the MICU subunits is likely to be mediated by Ca²⁺ binding to their EF hands. To test this hypothesis, we recombinantly expressed MICU1–3 or MICU1–3 with mutated EF hands (mut-EF-MICU, to disable Ca²⁺ binding Kamer et al., 2017) in their respective knockout cell lines [Figure 3A] and examined the changes in I_Ca. Expression levels of both the recombinant WT and mut-EF-MICU proteins were significantly higher as compared to endogenous MICUs expression in each case (Figure 3A).

In MICU1-KO, expression of MICU1 was able to restore I_Ca to the WT level, but mut-EF-MICU1 expression failed to do so (Figure 3B). This confirms our hypothesis that Ca²⁺ binding to the EF hands of MICU1 is indispensable for the I_Ca potentiation.

In MICU2-KO, I_Ca was not significantly affected (Figure 3C, and Figure 2A and B), because the loss of MICU2 appeared to be compensated with increased MICU1 expression and formation of MICU1 homodimers (Patron et al., 2014; Figure 3—figure supplement 1A). Therefore, overexpression of recombinant MICU2 in the knockout background only reverted the MICU1 homodimer back to heterodimer without any change in the I_Ca amplitude (Figure 3C). In contrast, mut-EF-MICU2 overexpression displaced MICU1 from the homodimers in favor of MICU1/mut-EF-MICU2 heterodimer, leading to a decrease in the total number of functional EF hands in the heterodimer. This results in a significant decrease in MICU-dependent I_Ca potentiation (Figure 3C). These functional data, combined with biochemical/structural evidence for preferential formation of MICU1/MICU2 heterodimers (Fan et al., 2020; Patron et al., 2014; Petrungaro et al., 2015; Wang et al., 2019; Wang et al., 2020a; Wang et al., 2020b; Xing et al., 2019; Zhuo et al., 2021), suggest that MICU2, along with MICU1, is responsible for allosteric potentiation of MCU upon binding of cytosolic Ca²⁺ to their EF hands.

The composition of MICU dimers can also be affected by MICU3 that, similar to MICU2, was proposed to interact and form heterodimers with MICU1 (Patron et al., 2019; Plovanich et al., 2013). MICU3 is a minor protein as compared to MICU1 and 2 in the majority of tissues and cell lines (Patron et al., 2019), which also appears to be the case in our system (Figure 3A). Accordingly, I_Ca was not affected in MICU3-KO mitoplasts, and overexpression of recombinant MICU3 or mut-EF-MICU3 in MICU3-KO also had no effect on I_Ca (Figure 3D). MICU3 is profoundly expressed in neurons where it was shown to increase the efficiency of mitochondrial Ca²⁺ uptake in axons (Ashrafi et al., 2020).

Ca²⁺ binding to the EF hands of MICU subunits and a subsequent conformational change that potentiates the MCU_cx activity require a finite time and may delay I_Ca activation/deactivation in response to rapid changes in [Ca²⁺]_i. Therefore, we examined I_Ca activation and deactivation kinetics in response to rapid changes in [Ca²⁺]_i and tested whether they depend on MICUs. I_Ca activation upon rapid elevation of [Ca²⁺]_i from virtually Ca²⁺-free to 1 mM was immediate, with kinetics comparable to the rate of solution exchange (τ ~0.4 ms) achieved by our fast application system (Figure 3E, and Figure 3—figure supplement 2). Importantly, the kinetics of the I_Ca rapid response was not altered in MICU1-KO (Figure 3E and F). The deactivation kinetics was similarly fast and not dependent on MICU1 (Figure 3E and F). The result of these experiments correspond to the previous observation that EF hands of calmodulin bind Ca²⁺ with a µs time constant (Faas et al., 2011). The conclusion from these experiments is that the kinetics of Ca²⁺ binding to the MICU’s EF hands, and the resultant conformational change in the MCU_cx, are fast enough that MICUs and mitochondria will rapidly track changes in [Ca²⁺]_i.

The MCU_cx is an inward rectifying Ca²⁺ channel (Kirichok et al., 2004). However, it remains unclear if the MICUs contribute to this feature. To examine this possibility, we measured I_Ca in the presence of 2 mM [Ca²⁺]_m (pipette solution). Under these conditions, no outward I_Ca was observed either before or after [Ca²⁺]_i elevation in either WT or MICU1-KO. However, as expected, 1 mM [Ca²⁺]_i induced a robust inward I_Ca (Figure 3G). Thus, the MICUs do not appear to be responsible for the inward rectification of MCU_cx, and the inward rectification is an inherent property of the pore proteins.

Recently published work suggested that MCU_cx might be inhibited by matrix [Ca²⁺] (Vais et al., 2016; Vais et al., 2020), specifically at [Ca²⁺]_m ~400 nM. However, as shown in Figure 3—figure supplement 1B–C, I_Ca amplitude is unchanged when [Ca²⁺]_m was set at either Ca²⁺-free, or 400 nM, or 400 µM. Thus, the MCU_cx is not regulated by matrix Ca²⁺, and MICUs only impart the regulation of the MCU_cx by cytosolic Ca²⁺.

Taken together, these data indicate that binding of cytosolic Ca²⁺ to EF hands of MICU subunits allosterically potentiates MCU_cx currents.

To investigate the mechanism by which Ca²⁺-bound MICU subunits potentiate I_Ca, we examined the activity of single MCU_cx channels in inside-out (matrix-side out) IMM patches (Figure 4). Because the unitary MCU_cx current (i_Ca via a single MCU_cx channel) is very small if measured in physiological [Ca²⁺]_i, it must be recorded at high [Ca²⁺]_i = 105 mM to enable proper resolution (Kirichok et al., 2004). As desired for this experiment, using this [Ca²⁺]_i, the EF hand domains of the MICU subunits are fully saturated with Ca²⁺.

MCU_cx exhibits multiple levels of single channel conductance (Figure 4A–C; Kirichok et al., 2004). The sub-conductances can be observed at all tested voltages (−40,–80, and −120 mV), but their resolution improves as the transmembrane voltage, and the amplitude of single-channel currents increase. It is clear that the open probability (P_o) of the MCU_cx is increased by the hyperpolarization of the IMM as was also shown previously (Kirichok et al., 2004). At −120 mV there are full-sized stochastic openings of MCU_cx as well as sub-conductance openings at ~80% and ~60% of the amplitude of the fully open i_Ca (Figure 4A–C). Because similar amplitude levels were observed in all the patches, we conclude that these events represent genuine sub-conductances in the MCU_cx channel.

There was no difference in the single channel amplitude between control and MICU1-KO mitoplasts (Figure 4C). However, we found that the single-channel open probability (P_o) was significantly decreased ~2–3 fold in MICU1-KO versus WT mitoplasts, depending on the transmembrane voltage (Figure 4D). As a result, the time-averaged current contributed by a single MCU_cx channel differs significantly between control and MICU1-KO mitoplasts (Figure 4E), thus mirroring and explaining the effect of MICU1 knockout on the amplitude of the whole mitoplast I_Ca (Figure 2B).

We next recorded MCU_cx single channel activity using Na⁺ as the permeating ion (i_Na), in nominally Ca²⁺-free conditions (MICUs in Ca²⁺-free state) (Figure 5A and B). Similar to our findings with Ca²⁺ as the permeant ion, there were multiple conductance states (i.e. sub-conductances) when Na⁺ was the permeant ion. These sub-conductance states were the same in WT and MICU1-KO (Figure 5C). However, in contrast to i_Ca, there were no significant differences in the open probability of i_Na between WT and MICU1-KO when Na⁺ was the permeating ion (Figure 5D). Accordingly, there was no difference in the time-averaged currents contributed by a single MCU_cx channel in control and MICU1-KO mitoplasts (Figure 5E). This correlates well with the absence of differences in amplitude of the whole mitoplast MCU currents when Na⁺ was the permeant ion for these two genotypes (Figure 2D).

These results demonstrate that in the Ca²⁺-bound state, the MICUs potentiate MCU current by increasing the open probability of the MCU/EMRE pore. In the absence of Ca²⁺, the MICUs do not appear to affect the pore activity.

While manganese (Mn²⁺) is essential for the proper function of several mitochondrial enzymes, its excessive accumulation inhibits oxidative phosphorylation and causes toxicity (Gunter and Pfeiffer, 1990). MCU_cx appears to be the primary entry pathway for Mn²⁺ entry into mitochondria (Gunter et al., 2010). Recently, it has been suggested that MICU1 is responsible for the relatively low permeability of MCU_cx for Mn²⁺ as compared to Ca²⁺, and when MICU1 deficiency or loss-of-function occurs, it can lead to excessive mitochondrial Mn²⁺ accumulation and cellular toxicity (Kamer et al., 2018; Wettmarshausen et al., 2018).

We recorded the current carried by Mn²⁺ through MCU_cx (I_Mn) in the presence of 5 mM [Mn²⁺]_i. I_Mn disappeared in MCU-KO and EMRE-KO, confirming that Mn²⁺ current was solely mediated by MCU_cx (Figure 6A–C). I_Mn was significantly smaller (~7-fold) than I_Ca via MCU_cx, as was also shown previously (Kirichok et al., 2004; Figure 6D and E). Interestingly, in MICU1-KO, I_Mn and I_Ca were reduced to a similar extent (Figure 6F–H). Moreover, even the ratio between I_Mn and I_Ca calculated from the same mitoplast (I_Mn/I_Ca) was not affected in MICU1-KO (Figure 6I). Two important conclusions follow from these observations. First, MICU1 does not differentially regulate I_Mn and I_Ca. Second, MICU1 potentiates MCU_cx in the presence of both Mn²⁺ and Ca²⁺. These results are in contrast to a popular model in which MICUs occlude MCU_cx, and that this occlusion is relieved only by Ca²⁺ but not by Mn²⁺ (Kamer et al., 2018; Wettmarshausen et al., 2018). However, our results are in agreement with important earlier studies which found that cytosolic Mn²⁺ allosterically stimulates mitochondrial Ca²⁺ uptake just like Ca²⁺ (Allshire et al., 1985; Hughes and Exton, 1983; Kröner, 1986; Vinogradov and Scarpa, 1973).

We further sought to explain why I_Mn via MCU_cx is smaller than I_Ca. In the presence of Mn²⁺, I_Ca was decreased (Figure 6E). This decrease was the same in MICU1-KO demonstrating that it was a pore effect (Figure 6F and J). This suggests that Mn²⁺ slows down Ca²⁺ permeation simply because it dwells in the pore longer than Ca²⁺ due to tighter binding (Lansman et al., 1986). The higher affinity of Mn²⁺ to the pore and the longer dwell time also explains why I_Mn is smaller than I_Ca.

Thus, the I_Ca and I_Mn phenotypes of MICU1-KO are the same, and MICU1 does not determine the preference of MCU_cx for Ca²⁺ over Mn²⁺. Permeation of both Ca²⁺ and Mn²⁺ is enhanced, rather than inhibited by MICU1.

Mg²⁺, the most abundant cytosolic divalent ion, is an important negative regulator of MCU_cx-mediated mitochondrial Ca²⁺ uptake (Gunter et al., 2010; Hutson et al., 1976). However, the mechanism of the inhibitory action of Mg²⁺ on MCU_cx is poorly understood. Our previous study suggested that in Ca²⁺-free conditions, Mg²⁺ occludes the MCU_cx pore for Na⁺ permeation (Kirichok et al., 2004). Here, we investigate how Mg²⁺ affects Ca²⁺ conduction through MCU_cx.

We first studied how cytosolic Mg²⁺ affects activation of I_Ca by [Ca²⁺]_i (Figure 7A and B). The results of these experiments clearly demonstrated that I_Ca is inhibited in the presence of Mg²⁺, and that this inhibition was primarily prominent in the lower range of micromolar [Ca²⁺]_i (Figure 7B). Thus, we specifically tested the effect of [Mg²⁺]_i in this range of [Ca²⁺]_i by recording I_Ca at 30 μM [Ca²⁺]. In these experiments, we found that I_Ca in the WT MCU_cx remains about double the I_Ca in MICU1-KO whether [Mg²⁺] is 200 μM or 0 (Figure 7C). Thus, these data suggest that the potentiating effect that MICU1 exerts on MCU_cx does not depend on [Mg²⁺]. Furthermore, as shown in Figure 7D, Mg²⁺ inhibits I_Ca with similar IC₅₀ in WT (149 ± 20 μM) and MICU1-KO (156 ± 20 μM). Taken together, these findings indicate that Mg²⁺ exerted its inhibitory effect at the MCU_cx pore (Figure 7C and D) and not through the MICUs. The pronounced competitive nature of the Mg²⁺ inhibition (Figure 7B) suggests that Mg²⁺ binds within the MCU_cx selectivity filter formed by Asp and Glu residues (Baradaran et al., 2018; Fan et al., 2018; Nguyen et al., 2018; Yoo et al., 2018). Mg²⁺ is a smaller divalent cation than Ca²⁺, and it is more difficult for Mg²⁺ to shed its hydration shell to fit into a narrow high-affinity Ca²⁺ binding site formed by Glu (site 2). However, Mg²⁺ could bind to the outermost and wider Asp binding site (site 1) of the selectivity filter even with a hydration shell. Such Mg²⁺ binding would not allow Mg²⁺ permeation, but would occlude the pore.

In conclusion, Mg²⁺ is an MCU_cx pore blocker that at the resting [Ca²⁺]_i, would strongly compete with Ca²⁺ for binding to the selectivity filter, limiting Ca²⁺ permeation. The Mg²⁺ occlusion of the MCU_cx pore can at least partially explain low mitochondrial Ca²⁺ uptake at resting [Ca²⁺]_i.

Assuming that K_d for Ca²⁺ binding to MICU EF hands is ~600 nM (Kamer et al., 2017), MICUs would exert their potentiating effect over a broad range of physiological [Ca²⁺]_i that range from resting to low micromolar. By doing so, MICUs can help the MCU_cx to overcome the mitochondrial Ca²⁺ efflux machinery and to elevate [Ca²⁺]_m to achieve adequate stimulation of the mitochondrial ATP production. In previous reports, there has been significant inconsistency as to the proposed effect of MICU1 on mitochondrial Ca²⁺ uptake at high [Ca²⁺]_i (Supplementary file 1a and b). For example, mitochondrial Ca²⁺ uptake rate appears to be decreased in isolated liver mitochondria from mice following siRNA-mediated knockdown of MICU1 or MICU2 in an earlier study (Plovanich et al., 2013). Another study (Csordás et al., 2013) also showed decreased uptake rates when MICU1 was knocked down in HeLa cells and hepatocytes leading authors to propose that MICU1 contributes to the cooperative activation of MCU_cx, but the effect was mild and it disappeared in the absence of Mg²⁺. On the other hand, many reports (Kamer et al., 2018; Logan et al., 2014; Mallilankaraman et al., 2012a; Vais et al., 2016) showed no change in MCU activity upon MICU1 loss at high [Ca²⁺]_i but only alteration in the threshold for mitochondrial Ca²⁺ uptake. However, a recent study again showed a decrease in Ca²⁺ uptake in isolated liver mitochondria in MICU1-KO mouse relative to WT (Liu et al., 2016). This correlates well with our data showing direct potentiation of the MCU_cx activity by MICU1. Thus, the potentiating effect of MICUs on the MCU_cx was discernible in the previous research but was largely rejected due to the predominant view that the primary function of MICUs is to occlude the MCU_cx pore.

Our data is incompatible with the model in which MICUs occlude the MCU pore at low [Ca²⁺]_i and impart a [Ca²⁺]_i activation threshold on the MCU_cx. This model explains the increase of mitochondrial Ca²⁺ accumulation at resting [Ca²⁺]_i in MICU1-KO to ‘unblocking’ of MCU_cx. However, in this concentration range, the MCU_cx activity is so slow that the mitochondrial Ca²⁺ accumulation is definitely influenced by many other factors. Among other possibilities, one can speculate that mitochondrial Ca²⁺ efflux, matrix Ca²⁺ buffering, matrix pH and permeability of the outer mitochondrial membrane might be altered in MICU1-KO to facilitate mitochondrial Ca²⁺ accumulation at low [Ca²⁺]_i.

Because these multiple factors can be affected by MICU1-KO differently in different experimental system and conditions, the change in the threshold can vary from one system to another. In Trypanosoma cruzi, where the composition of the MCU complex (including EMRE and MICU1) is similar to mammals, MICU1 knockout causes an increase in the Ca²⁺ uptake ‘threshold’ and a marked decrease in Ca²⁺ uptake capacity at all [Ca²⁺]_i (Bertolini et al., 2019). A recent report also found no apparent [Ca²⁺]_i threshold for MCU in isolated mitochondria from heart and skeletal muscle (Wescott et al., 2019). Similarly sympathetic neurons lack a discernible threshold and mitochondria were shown to accumulate small amount of Ca²⁺ at [Ca²⁺]_i levels as low as 200 nM (Colegrove et al., 2000a; Colegrove et al., 2000b). Finally, interpretation of the MICU1-KO phenotypes can be further complicated by possible functional roles of MICU1 outside of the MCU complex (Gottschalk et al., 2019; Tomar et al., 2019; Tufi et al., 2019). In particular, MICU1 was proposed to interact with the MICOS complex, affecting the cristae structure (Tomar et al., 2019).

To circumvent the problem of isolating the MCU_cx activity from the other mitochondrial Ca²⁺ homeostatic mechanisms, we leveraged the capacity of MCU_cx to conduct Na⁺. The Na⁺ permeation via MCU_cx is observed not only in isolated mitoplasts but also in intact mitochondria. Na⁺ permeates via MCU_cx because Ca²⁺ and Na⁺ ions have the same ionic diameter, and Na⁺ can pass through Ca²⁺ channels when Ca²⁺ is removed from the selectivity filter (Hess and Tsien, 1984; Hess et al., 1986; Tang et al., 2014). Ca²⁺ binds to the MCU_cx selectivity filter with an exceptionally high affinity (K_d ≤2 nM) (Kirichok et al., 2004), and re-addition of cytosolic Ca²⁺ leads to I_Na inhibition upon Ca²⁺ binding to this site. Importantly, MICUs are not involved in this inhibition, as it is not affected by MICU1-KO (Figure 2—figure supplement 2). In intact mitochondria, Na⁺ uptake via WT or MICU1-deficient MCU_cx causes the same ΔΨ depolarization, demonstrating that at low [Ca²⁺]_i MCU_cx is not occluded regardless of its association with MICUs. Moreover, using mitochondrial patch-clamp under Ca²⁺-free conditions, we recorded a robust I_Na via MCU, and the amplitude of this I_Na was not affected in MICU1-KO. This demonstrates that MCU pore is not occluded by MICUs at low [Ca²⁺]_i.

The complexity of the MCU_cx presents formidable difficulties even for the direct electrophysiological analysis. The electrophysiological phenotypes observed may be associated with altered expression levels of other subunits rather than the loss of MICU1. We not only analyzed the expression levels of all MCU complex subunits (Figure 2—figure supplement 1D–F) but also measured I_Ca/I_Na ratio in the same mitoplast (Figure 2C–F) to ensure that we correctly capture the role of MICU1 in the MCU complex.

Recently, cryo-EM structures of the MCU_cx holocomplex reveal that it is composed of two conjoined MCU/EMRE pores with a MICU1/MICU2 heterodimer attached to each of them (Fan et al., 2020; Wang et al., 2020b; Zhuo et al., 2021). Based on molecular weight of the MCU_cx as observed on the Blue native PAGE (Sancak et al., 2013), this structure likely represents the complete MCU_cx of intact mitochondria. This is the primary structural form of MCU_cx in the mitochondrial inner membrane.

However, in addition to this complete MCU_cx, the cryo-EM analysis also reported structures of a monomeric MCU/EMRE pore with a single MICU1/MICU2 heterodimer. Although this does not appear to be the predominant state of the native MCU_cx, it was primarily used for the analysis of the interaction between the pore and the MICU1/MICU2 dimer. Based on this analysis, a conclusion was made that in the absence of Ca²⁺, MICU1/MICU2 dimer occluded the MCU_cx pore. Wang et al. (Wang et al., 2020b) observed occlusion of the dimeric complex, but such MICU1-occluded dimeric complex represented only ~10% of the total number of analyzed particles in the absence of Ca²⁺. The low prevalence of the dimeric complexes can be purely due to the experimental limitations, but a more thorough analysis of occlusion in this native form of MCU_cx is warranted. At the same time, another work that analyzed the complete MCU_cx with two pores and two MICU1/MICU2 heterodimers, did not show any occlusion (Zhuo et al., 2021). In this structure, the two MICU1/MICU2 heterodimers interact between each other, leaving both pores unoccluded (Zhuo et al., 2021). As more MCU_cx structural data appear in the future, it is important to take into consideration the completeness of the complex and the new functional data on the MICU subunits presented here.

We are at the very beginning of the structural characterization of MCU_cx. As this crucial work continues, selection of the experimental conditions is key to understanding how the channel works under physiological conditions. As we emphasize here, the direct inhibitory effect of Mg²⁺ on the MCU pore is important but has not yet been taken into consideration. In fact, Mg²⁺ was omitted in all structural studies of MCU_cx holocomplex (Fan et al., 2020; Wang et al., 2020b; Zhuo et al., 2021). As we discussed before, Mg²⁺ is likely to bind to the same Asp ring of the selectivity filter where MICU1 was predicted to bind and, may therefore significantly affect conformational states of MCU_cx. Thus, to determine how MICU1 interacts with the MCU_cx pore, structural studies should be performed under conditions that include physiological concentrations of Mg²⁺. Furthermore, occlusion of the MCU/EMRE pore by MICUs in the absence of Ca²⁺ was observed when the cryoEM particles were obtained in a solution that had ionic strength ~three times lower than physiological (Fan et al., 2020; Wang et al., 2020b). In contrast, the unoccluded state was observed at the physiological ionic strength (Zhuo et al., 2021). This difference is critical, because the occlusion of the MCU pore by MICU1 strongly depends on ionic interactions (Fan et al., 2020; Wang et al., 2020b). Another important consideration is the presence of cardiolipin in MCU_cx structures, likely affected by the type of detergent used during purification. Interestingly, the occluded state was observed in the structures that contained no cardiolipin (Fan et al., 2020; Wang et al., 2020b), while the unoccluded MCU_cx structure contained cardiolipin (Zhuo et al., 2021). This point is important as cardiolipin interacts with the MCU complex (Ghosh et al., 2020; Zhuo et al., 2021) and may regulate MICU function (Kamer and Mootha, 2014).

In conclusion, we find no evidence for the occlusion of the MCU_cx pore by MICUs at resting [Ca²⁺]_i. Instead, under these conditions MCU_cx-mediated uptake is limited by the low concentration of the conducting ion and by the Mg²⁺ block. A slow mitochondrial Ca²⁺ uptake at resting [Ca²⁺]_i is likely at balance with the Ca²⁺ efflux mechanisms, which prevents mitochondrial Ca²⁺ overload (Nicholls, 2005). On the other hand, the phenomenon of allosteric potentiation of MCU_cx has been known for many decades (Gunter et al., 2010; Kröner, 1986), and this work provides its clear mechanistic explanation.

Further information and requests for reagents may be directed to and will be fulfilled by Lead Contact Yuriy Kirichok (yuriy.kirichok@ucsf.edu).

Mice were maintained on a standard rodent chow diet under 12 hr light and dark cycles. All animal experiments were performed with male mice (2–5 month old) according to procedures approved by the UCSF Institutional Animal Care and Use Committee and adhered to NIH standards. C57BL/6J were obtained from the Jackson laboratory. MCU-KO mice were obtained from Dr. Torren Finkel and have been used previously (Pan et al., 2013).

Gene expression analysis (qRT-PCR) qPCR was performed by Syd Labs (Natick, MA, USA). Total RNA was isolated from cells using the RNAeasy Minikit (QIAGEN), and reverse transcribed using the First Strand cDNA Synthesis Kit (Syd Labs). qPCR reactions were performed with the following gene-specific primers (generated by Integrated DNA Technologies):

cDNA (100 ng, calculated from initial RNA) samples were pre-amplified for 12 cycles using ABsolute qPCR SYBR Green Low ROX Mix (ThermoFisher). qPCR reactions were performed using an Agilent MX3000 (Fluidigm) with 40 cycles of amplification (15 s at 95°C, 5 s at 70°C, and 60 s at 60°C). Ct values were calculated by the Real-Time PCR Analysis Software (Fluidigm). Relative gene expression was determined by the ΔCt method. Hprt was selected as the reference gene.

Knockout MEF cell lines were generated using the CRISPR/Cas9 method (Ran et al., 2013). All knockouts (except the MCU-KO line) were generated by Alstem LLC (Richmond, CA, USA). Either one sgRNA or a pair of two adjacent sgRNAs were used to create a point indel or a truncate indel, respectively (Figure S1).

MEFs were transfected with the Cas9 gRNA vector (Addgene: PX459) via electroporation (Invitrogen Neon transfection system) using the following parameters: 1×10⁶ cells and 1 µg of two different gRNA-Cas9 plasmids. Puromycin was used for enrichment of transfected cells, and serial dilution was performed to select single-cell clones. A stable homozygous knockout cell line was confirmed by PCR amplification of the targeted region, cloning into a pUC19 vector, and sequencing showing that either a frameshift or large deletion had occurred in the targeted region of the gene (Figure 1—figure supplement 1). All knockout clones were further validated by western blotting (Figure 1—figure supplement 1). The primers used for amplification of genomic sites and cloning into pUC19 sequencing vector were as follows:

Mitoplasts were isolated from MEFs using methodology previously described (Garg and Kirichok, 2019). Briefly, MEFs were homogenized in ice-cold medium (Initial medium) containing 250 mM sucrose, 10 mM HEPES, 1 mM EGTA, and 0.1% bovine serum albumin (BSA) (pH adjusted to 7.2 with Trizma base) using a glass grinder with six slow strokes of a Teflon pestle rotating at 280 rpm. The homogenate was centrifuged at 700× g for 10 min to create a pellet of nuclei and unbroken cells. The first nuclear pellet was resuspended in the fresh Initial medium and homogenized again to increase the mitochondrial yield. Mitochondria were collected by centrifugation of the supernatant at 8500× g for 10 min.

Mitoplasts were produced from mitochondria using a French press. Mitochondria were suspended in a hypertonic solution containing 140 mM sucrose, 440 mM D-mannitol, 5 mM HEPES, and 1 mM EGTA (pH adjusted to 7.2 with Trizma base) and then subjected to a French press at 1200–2000 psi to rupture the outer membrane. Mitoplasts were pelleted at 10,500× g for 15 min and resuspended for storage in 0.5–1 ml of solution containing 750 mM KCl, 100 mM HEPES, and 1 mM EGTA (pH adjusted to 7.2 with Trizma base). Mitoplasts prepared and stored with this method contained the same amount of auxiliary MICU1 and MICU2 subunits as compared to intact mitochondria (Figure 1—figure supplement 2F, see TCo-immunoprecipitation section below).

Mitochondria and mitoplasts were prepared at 0–4°C and stored on ice for up to 5 hr. Immediately before the electrophysiological experiments, 15–50 μl of the mitoplast suspension was added to 500 μl solution containing 150 mM KCl, 10 mM HEPES, and 1 mM EGTA (pH adjusted to 7.0 with Trizma base) plating on 5 mm coverslips pretreated with 0.1% gelatin to reduce mitoplast adhesion.

Whole mitoplast currents were measured as described previously (Garg and Kirichok, 2019). Gigaohm seals with mitoplasts were formed in the bath solution containing 150 mM KCl, 10 mM HEPES and 1 mM EGTA, pH 7.2 (adjusted with KOH). Voltage steps of 350–500 mV for 2–8 ms were applied to rupture the IMM and obtain the whole-mitoplast conFiguration. Typically, pipettes had resistances of 20–40 MΩ, and the access resistance was 35–65 MΩ. The membrane capacitances of mitoplasts range from 0.2 to 0.6 pF.

All indicated voltages are on the matrix side of the IMM (pipette solution), relative to the cytosolic side (bath solution, Figure 1—figure supplement 2D; Figure 1—figure supplement 2E; Garg and Kirichok, 2019). Currents were normally induced by a voltage ramp from −160 mV to +80 mV (interval between pulses was 5 s) to cover all physiological voltages across the IMM, but other voltage protocols were also used as indicated in the Figures. All whole-IMM recordings were performed under continuous perfusion of the bath solution. Currents were normalized per membrane capacitance to obtain current densities (pA/pF). Currents flowing into mitochondria are shown as negative, while those flowing out are positive. Membrane capacitance transients observed upon application of voltage steps were removed from current traces.

Typically, pipettes were filled with one of the following three solutions (Garg and Kirichok, 2019) (tonicity was adjusted to ∼350 mmol/kg with sucrose).

Solution A was used to measure Ca²⁺ currents and contained: 110 mM Na-gluconate, 40 mM HEPES, 10 mM EGTA and 2 mM MgCl₂ (pH 7.0 with NaOH).

Solution B was used to measure Na⁺ or Mn²⁺ currents and contained: 110 Na-gluconate, 40 HEPES, 1 EGTA, 5 EDTA, and 2 mM NaCl (pH 7.0 with Tris base).

Solution C was used to measure outward Ca²⁺ currents (the MCU rectification experiments) and contained: 130 mM tetramethylammonium hydroxide (TMA), 100 mM HEPES and 2 mM CaCl₂ (pH 7.0 with D-gluconic acid).

To measure whole-mitoplast Ca²⁺ currents, the bath solution was formulated to contain only 150 mM HEPES (pH 7.0 with Tris base, tonicity ∼300 mmol/kg with sucrose) and different dilutions of CaCl₂ from a 1 M stock (Sigma) (Kirichok et al., 2004). The control solution contained: 150 mM HEPES, 80 mM sucrose and 1 mM EGTA (pH 7.0 with Tris base, tonicity ∼300 mmol/kg with sucrose). The bath solution used for measuring Na⁺ current contained: 110 mM Na-gluconate, 40 mM HEPES, 1 mM EGTA and 5 mM EDTA (pH 7.0 with Tris base, tonicity ∼300 mmol/kg with sucrose). The bath solution for measuring inhibition of Na⁺ current by cytosolic Ca²⁺ contained: 110 mM Na-gluconate, 40 mM HEPES, and 10 mM EDTA (pH 8.0 with Tris, tonicity ∼380 mmol/kg with sucrose) and varying amounts of CaCl₂ were added to the bath solution to achieve the free [Ca²⁺] calculated using the MaxChelator program (C. Patton, Stanford University).

A rapid exchange of [Ca²⁺]_i from virtual zero (control solution) to 1 mM was achieved using a commercially available fast solution exchange system (Warner Instruments, SF-77B perfusion fast step system). It was interfaced with our pClamp acquisition software in order to precisely time the steps during solution change. The timing (τ ~0.4 ms) for solution exchange was judged by the current changes because of a junction potential difference using solutions with different ionic strengths.

Currents were recorded using an Axopatch 200B amplifier (Molecular Devices). Data acquisition and analyses were performed using PClamp 10 (Molecular Devices) and Origin 9.6 (OriginLab). All data were acquired at 10 kHz and filtered at 1 kHz.

All single-channel data were acquired from inside–out patches excised from isolated mitoplasts (Kirichok et al., 2004). For Ca²⁺ single channel (i_Ca) recordings, patches were excised in a bath solution containing 150 mM KCl, 10 mM HEPES and 1 mM EGTA, pH 7.2 (adjusted with KOH). Recordings were performed under symmetrical conditions (the same bath and pipette solutions): 105 mM CaCl₂ and 40 mM HEPES, pH 7.0 with Tris base. Signals were sampled at 50 kHz and low-pass filtered at 1 kHz. Fire-polished, borosilicate pipettes (Sutter QF-150–75) coated with Silguard (Dow Corning Corp., Midland, MI) and having a tip resistance of 50–70 MΩ were used for low noise recordings.

For Na⁺ single channel (i_Na) recordings, patches were excised in a bath solution containing 150 mM Na-gluconate, 10 mM HEPES, 1 mM EGTA and 1 mM MgCl₂, pH 7.2 (adjusted with NaOH). Pipette solution contained 150 mM Na-gluconate, 10 mM HEPES, 1 mM EGTA, 1 mM EDTA, and 2 mM NaCl, pH 7.2 (adjusted with NaOH). Signals were sampled at 50 kHz and low-pass filtered at 1 kHz.

To characterize the single-channel conductance and subconductance levels and their occupancy probabilities, we used the MLab version of the QuB software, freely available from the Milescu lab at: https://milesculabs.biology.missouri.edu/QuB_Downloads.html. The data were first resampled at 2.5 kHz and then were idealized with the Baum-Welch and Viterbi algorithms, as implemented in QuB, which classify each point in the data to a conductance level and produce estimates of current amplitudes and occupancy probabilities. The time-averaged single-channel current can be calculated as the product between occupancy probability and current amplitude, summated over all conductance levels (main open state and substates).

For imaging experiments, MEFs were plated on collagen type-I-coated glass-bottom 35 mm dishes (P35G-1.5–14 C, Matek), 48–72 hr before imaging. Cells were imaged at the interval of 3 s on a Nikon Ti-E microscope using a 40× objective (NA 1.30, oil, CFI Plan Fluor, Nikon), Lambda 421 LED light source (Sutter) and ORCA Flash 4.0 CMOS camera (Hamamatsu Photonics) at room temperature (25°C). The following excitation/emission filter settings were used: 340±13/525±25 nm and 389±19/510±40 nm for cytosolic Ca²⁺ imaging using fura-2 (K_d=224 nM) and 480±40/525±15 nm for mitochondrially targeted cepia2 (CEPIA2mt, K_d=160 nM (Suzuki et al., 2014), cloned into a lentiviral vector). Cells were loaded with 3 μM fura-2 AM (Life Tech., USA) in DMEM/FBS at room temperature for 30 min. After three washes with physiological salt solution (PSS) containing (in mM) 150 NaCl, 4 KCl, 2 CaCl₂, 1 MgCl₂, 5.6 glucose, and 25 HEPES (pH 7.4), each dish was placed on the stage for imaging. Imaging was performed in PSS within 1 hr of dye staining. Baseline fluorescence was taken for 1–2 min after which thapsigargin (Tg) (final [Tg] = 300 nM) was added while imaging was continued for another 10–15 min.

Briefly, mitochondria were isolated from MEF cells using differential centrifugation as described above. The mitochondrial pellet was resuspended in resuspension buffer (RB) supplemented with 2 mM EGTA and 2 μM of Fura-2-acetoxymethyl ester (Fura-2 AM) and kept at room temperature for 10 min to allow loading of Fura-2 into the mitochondrial matrix. The RB buffer contained: 100 mM KCl, 50 mM MOPS, 1 mM MgCl₂. Mitochondria were pelleted at 3200 g, and further incubated on ice for 50 min to allow de-esterification of Fura-2 AM in RB supplemented with 2 mM EGTA. Mitochondria were pelleted at 3200 g and resuspended in RB supplemented with 10 μM EGTA. Mitochondria were further pelleted and resuspended twice in RB supplemented with 40 μM Fluo-4 pentapotassium salt (for measurements carried out below 3 μM [Ca²⁺]_i), or RB supplemented with 40 μM EGTA (for measurements carried out above 3 μM [Ca²⁺]_i). After final centrifugation step at 3200 g, protein concentration was determined by Lowry assay.

Measurements of mitochondrial Ca²⁺ influx were carried out using a BMG LABTECH CLARIOstar plate reader as described before (Wescott et al., 2019). Experiments were carried out with mitochondria (0.5 mg/ml) in an uptake assay buffer (uAB) that contained: 130 mM KCl, 20 mM HEPES, 1 mM MgCl₂, 1 mM K₂HPO₄, pH 7.2 (with KOH), supplemented with energetic substrates (glutamate, malate and succinate, each 5 mM), and 1 μM TMRM. The uAB and all the stock solutions were made with analytical-grade deionized water (OmniSolv LC–MS, Sigma Aldrich) and contained less than 50 nM of residual [Ca²⁺] (measured daily). After 3 min of incubation with substrates, assays were initiated by injection of 100 μl of Ca²⁺ stock to bring the final volume to 200 μl. TMRM (ex: 546 ± 4 nm and 573 ± 5 nm, em: 619 ± 15 nm) and Fura-2 (ex: 335 ± 6 nm and 380 ± 6 nm, em: 490 ± 15 nm) fluorescence were measured, along with Fluo-4 or Fluo4-FF (ex: 485 nm, em: 520–542 nm) within the same well for 35 s. To measure MCU Ca²⁺ flux (J_MCU), two protocols were used. Protocol 1 ([Ca²⁺] range ≤ 3 μM): here, Fluo-4 (3 μM) is the single significant buffer of extra-mitochondrial Ca²⁺ (i.e., [Ca²⁺]_i). Protocol 2 ([Ca²⁺] range from 4 μM to 25 μM): here mitochondria were suspended in 40 μM EGTA and 1 μM Fluo4-FF was used. Total Ca²⁺ influx (J) is taken as the first derivative of the linear fit to the measured total extramitochondrial [Ca²⁺] over the first 20 s of each experiment. The total Ca²⁺ conductance of the IMM (G) was obtained from the simultaneous measurements of J, [Ca²⁺]_i, [Ca²⁺]_m and ΔΨ_m according to the typical Hodgkin–Huxley model (Wescott et al., 2019):

I = G (ΔΨ_m – E_Ca²⁺) where E_Ca²⁺ is the Nernst reversal potential for Ca²⁺ obtained from simultaneously measured [Ca²⁺]_i, and [Ca²⁺]_m. Measured J was converted to I using the Faraday constant (Wescott et al., 2019).

Membrane potentials in intact mitochondria were evaluated with TMRE using previously described method (Scaduto and Grotyohann, 1999). Mitochondria isolated from mouse liver or MEF cells were suspended in ice-cold initial medium. These mitochondria were mixed in 50–100 μl of an uptake assay buffer (liver: 150 mM NaCl, 10 mM HEPES, 1 mM EGTA, 2 mM glutamate, 2 mM malate, and 2 mM succinate, pH 7.2 with Trizma base; MEF: 30 mM NaCl, 120 mM TrisCl, 10 mM HEPES, 1 mM EGTA, 1 μM MgCl₂, 5 mM glutamate, 5 mM malate, and 5 mM succinate, pH 7.2 with Trizma base) with 200 nM TMRE. Mitochondrial concentration in the assay buffer was 0.25 mg/ml. TMRE fluorescence were measured at 550/570 nm and 570/589 nm (excitation/emission, 9 nm band width) with using a Biotek Synergy H4 plate reader, and the fluorescence ratio between two fluorescence was calculated. After 5 min incubation of mitochondria in the assay medium, assays were initiated by injection of 0.5–1 μl of EDTA (5 mM final), RuR (1–3 μM final) or FCCP (1 μM final), and the ratio change within 5 min after the injection was evaluated. The ratio change induced by EDTA and/or RuR was normalized with that by FCCP.

Mitochondria or mitoplasts were isolated from MEFs deficient in the MCU subunit but stably expressing Flag-tagged MCU. Mitochondrial fraction from wild type cells (without MCU-FLAG) was used as negative control. Isolated mitoplasts (but not mitochondria) were incubated in 750 mM KCl for 30 min before solubilization. Briefly, 300 µg of protein lysate was solubilized with 500 µl of lysis buffer (50 mM HEPES pH 7.4, 150 mM NaCl, 1 mM EGTA, 0.2% DDM and Halt protease inhibitor cocktail [Thermo Fisher]) for 30 min at 4°C. Lysates were cleared by spinning at 20,000× g for 10 min at 4°C. Cleared lysates were incubated with anti-Flag M2 affinity gel (Sigma A2220) for 2 hr at 4°C. Immunoprecipitates were washed with 1 ml of lysis buffer three times and boiled in 20 μl of Laemmli buffer (without β-mercaptoethanol). One-third of the immunoprecipitate was loaded onto a 4–20% gradient SDS-PAGE gel for detection of the indicated proteins by Western blotting. Flow-through fraction was also collected and analyzed in the same gel.

For western blot analysis, MEFs or isolated mitochondria/mitoplasts were lysed in radioimmunoprecipitation assay (RIPA) buffer (1% IGEPAL, 0.1% sodium dodecyl sulfate, 0.5% sodium deoxycholate, 150 mM NaCl, 1 mM EDTA, 50 mM Tris-HCl (pH 7.4) and a cocktail of proteases inhibitors). Lysates were resolved by SDS-PAGE; transferred to PVDF membrane (Millipore); and probed with anti-MCU (Sigma, HPA016480, 1:2000), anti-EMRE (Santa Cruz, sc-86337, 1:200), anti-HSP60 (Santa Cruz, sc-1052, 1:3000), anti-VDAC (Abcam, ab15895, 1:2000), anti-MICU1 (Cell Signaling Technology, 12524S, 1:2000), anti-MICU2 (Bethyl, A300-BL19212, 1:500), anti-MICU3 (Sigma, HPA024779, 1:1000), and anti-TOM20 (Santa Cruz, sc-11415, 1:2000). Anti-MICU1 antibody produced a non-specific band near its monomeric molecular weight (~50 kDa), so samples were prepared in Laemmli buffer without β-mercaptoethanol to detect MICU1 homo- or heterodimers (~100 kDa).

Data are presented as mean ± standard error of the mean (SEM), as specified in the Figure legend. Statistical analysis was completed in Excel or Origin 9.6. All experiments were performed in triplicate or more. Statistical significance at an exact p-value was determined with the methods as indicated in the corresponding Figure legends.