Slow activation of fast mitochondrial Ca2+ uptake by cytosolic Ca2+

Mitochondrial Ca2+ uptake through the mitochondrial Ca2+ uniporter (MCU) is a tightly controlled process that sustains cell functions mainly by fine-tuning oxidative metabolism to cellular needs. The kinetics of Ca2+ fluxes across the mitochondrial membranes have been studied both in vitro and in vivo for many years, and the discovery of the molecular components of the MCU has further clarified that this Ca2+ uptake mechanism is based on a complex system subject to elaborate layers of controls. Alterations in the speed or capacity of the in-and-out pathways can have detrimental consequences for both the organelle and the cell, impairing cellular metabolism and ultimately causing cell death. Here, we report that pretreatment of deenergized mitochondria with low-micromolar Ca2+ concentrations for a few minutes markedly increases the speed of mitochondrial Ca2+ uptake upon re-addition of an oxidizable substrate. We found that this phenomenon is sensitive to alterations in the level of the MCU modulator proteins mitochondrial calcium uptake 1 (MICU1) and 2 (MICU2), and is accompanied by changes in the association of MICU1–MICU2 complexes with MCU. This increased Ca2+ uptake capacity, occurring under conditions mimicking those during ischemia/reperfusion in vivo, could lead to a massive amount of Ca2+ entering the mitochondrial matrix even at relatively low levels of cytosolic Ca2+. We conclude that the phenomenon uncovered here represents a potential threat of mitochondrial Ca2+ overload to the cell.

(Յ100 nM) (6,7). Increases in cytosolic Ca 2ϩ are mirrored in the mitochondria, where the main entry pathway is the calcium uniporter (MCU), 3 an electrogenic channel that exploits the driving force of the steep electrochemical gradient present across the inner mitochondrial membrane (IMM). Two main efflux routes, namely the Na ϩ /Ca 2ϩ and H ϩ /Ca 2ϩ exchangers, control the accumulation of free Ca 2ϩ in the mitochondrial matrix and guard against Ca 2ϩ overload (8,9). Ca 2ϩ , in turn, controls mitochondrial metabolism. The increase of free Ca 2ϩ within the matrix activates mitochondrial dehydrogenases (pyruvate dehydrogenase, isocitrate dehydrogenase, oxoglutarate dehydrogenase), and intermembrane Ca 2ϩ concentration modulates the activity of carriers such as citrin and aralar (10); the net results are the increase of the respiratory rate, H ϩ extrusion, and ATP synthesis. Tightly controlled Ca 2ϩ fluxes enable mitochondria to fine-tune energy production to the necessities of the cell. Unregulated Ca 2ϩ handling, on the other hand, can cause alterations in mitochondrial morphology and major organelle dysfunctions that can have dire consequences for cells and the whole organism (11)(12)(13). Mitochondrial Ca 2ϩ uptake was studied for more than half a century (14,15) before the molecular nature of this process was unraveled; its properties have been investigated thoroughly, and its characteristic features, such as high capacity, selectivity (16), and cooperativity (17), have been clarified. The major components of the uniporter have been recently identified, including the poreforming subunit (MCU) (19,20), a dominant-negative form (MCUb) (21), a regulatory subunit (EMRE) (22,23), and Ca 2ϩsensitive regulators (MICU1 (24 -27) and MICU2 (28 -30)). MICU1 and MICU2 complexes (homo-or heterodimers) have been suggested to regulate Ca 2ϩ uptake, determining both the threshold and cooperative activation of MCU (28,29,31), although the specific role of each component is still debated; for recent reviews, see also De Stefani et al. (32), Pendin et al. (33), and De Stefani et al. (34).
It is well-established that the kinetics of Ca 2ϩ transport are governed principally by the inner membrane potential differ-ence (limited by the speed of the respiratory chain complexes) (18) and by the extramitochondrial Ca 2ϩ concentration. An unusual phenomenon was described in the late 1980s: a slow (complete in several minutes) activation of the uniporter with the transformation of the classical cooperative kinetics into an almost hyperbolic one, even at low-micromolar [Ca 2ϩ ] (17,35). This potent activation occurs after exposure of isolated, deenergized mitochondria to micromolar Ca 2ϩ levels for a few minutes (36). This process, however, has not been further investigated, and its mechanism and role have not been clarified.
Based on the new knowledge of the Ca 2ϩ uniporter complex components, we re-examined the accelerated Ca 2ϩ uptake mode of the uniporter to investigate whether and, if so, which of the components might be responsible for this effect. The ultimate aim is to unravel strategies for modulating the uniporter activity that may have potential therapeutic applications.

Ca 2؉ -dependent activation of mitochondrial Ca 2؉ uptake (CAMCU)
We employed a standard Ca 2ϩ uptake protocol to monitor the effect on the rate of matrix Ca 2ϩ accumulation (substrate oxidation-driven) when nonrespiring mitochondria have stood prolonged preincubation time with fixed extramitochondrial [Ca 2ϩ ]. Briefly, isolated mouse liver mitochondria were incubated in the absence of substrates and supplemented with rotenone (to block the creation of membrane potential via the oxidation of endogenous substrates) and with oligomycin (to prevent the build up of the mitochondrial membrane potential by reverse ATPase activity) in a medium devoid of added Ca 2ϩ , supplied with 5 M EGTA, and containing the fluorescent Ca 2ϩ indicator Calcium Green-5N (Fig. 1A). Where indicated, 8 M CaCl 2 (3 M free Ca 2ϩ ) was added together with succinate to activate the electron flow in the respiratory chain. Alternatively, succinate was added 1, 2, 5, or 10 min after the addition of CaCl 2 . The inset in Fig. 1A shows representative traces of mitochondrial Ca 2ϩ uptake following the different incubation times. The rate of Ca 2ϩ decrease in the medium (dependent on the accumulation of the cation by mitochondria) was clearly slower when Ca 2ϩ and succinate were added together (time 0) compared with mitochondria exposed to 3 M Ca 2ϩ for increasing periods of time before the active uptake took place. Ca 2ϩ uptake speed was measured after the addition of the substrate by calculating the slope of the trace during its initial linear phase. The rate of Ca 2ϩ accumulation appears to reach a maximum at 5 min of preincubation, and it does not significantly increase further after 10 min of incubation. Fig. 1B shows the measurements (plus calculation of mean and S.D.) of Ca 2ϩ uptake rates at different incubation times, whereas Fig. 1C shows the values normalized to the Ca 2ϩ uptake rate measured after 10 min of preincubation. On average, we observed about a 3-fold increase in the rate of mitochondrial Ca 2ϩ uptake between 0 and 5 min of incubation.

[Ca 2؉ ] dependence and reversibility of CAMCU
The mitochondrial Ca 2ϩ uptake rate increases with the increase of the external Ca 2ϩ concentration; we analyzed the dependence of the activation of mitochondrial Ca 2ϩ uptake on the concentration of external Ca 2ϩ during a fixed preincubation period. Fig. 2A shows that, as expected, the initial rate of Ca 2ϩ accumulation increased as Ca 2ϩ concentration rose, but the acceleration due to the preincubation period was observed up to 7 M free Ca 2ϩ . Above this value, the rate of mitochondrial Ca 2ϩ uptake was no longer influenced by the preincubation period. As an example, the inset table presents the values of Ca 2ϩ uptake speed obtained with 7 and 10 M external free Ca 2ϩ ; at 10 M, the difference between the 0-and 2-min incubation times is no longer significant. These results obtained with isolated mitochondria confirm and extend the observation made by H. Kröner (17,35) demonstrating that a process requiring several minutes (up to 5 min to reach the maximum) is activated if nonrespiring mitochondria are first exposed to low-micromolar Ca 2ϩ , and afterward, active Ca 2ϩ uptake is triggered with the addition of an oxidizable substrate. Henceforth, this phenomenon (i.e. the Ca 2ϩ -dependent activation of mitochondrial Ca 2ϩ uptake) will be referred to as CAMCU. CAMCU is not substrate-specific, because it can be observed with succinate (substrate feeding reducing equivalents at the level of complex II of the respiratory chain) or with ascorbate plus TMPD (artificial substrate that feeds electrons at the level of complex IV) as the energy source (results not shown). It is more difficult to observe CAMCU in isolated mitochondria utilizing substrates for complex I, because the presence of endogenous substrates carried over during mitochondria extraction prevents the possibility of controlling Ca 2ϩ uptake, unless mitochondria are kept for several hours on ice until endogenous substrates are spontaneously consumed. Moreover, CAMCU does not depend on the tissue from which mitochondria are isolated, as the phenomenon was observed in mitochondria from mouse liver and heart (not shown).
The question then arises whether CAMCU is reversible or irreversible. To address this issue, nonrespiring mitochondria were first allowed 5 min in the presence of 3 M free Ca 2ϩ , followed by the addition of succinate (Fig. 2B). When the steady state was reached, 4 M EGTA was added, causing the almost instantaneous drop of free Ca 2ϩ in the medium to ϳ0.3 M. The addition of 7 M CaCl 2 about 40 s after EGTA (to reach 3 M free Ca 2ϩ ) resulted in a rate of Ca 2ϩ uptake that was similar to that of cells not preincubated with 3 M free Ca 2ϩ (Fig. 2B, representative trace (left) and quantification (right)).

CAMCU in permeabilized cells
CAMCU is a phenomenon observable not only with isolated mitochondria. It can also be detected in permeabilized cells. We used mainly the human immortalized cell line HeLa; as proof of principle, some experiments were also repeated with SHSY5Y cells, yielding the same results (not shown). Fig. 3A shows representative traces from cells in culture. The cells were transfected with aequorin targeted to the mitochondrial matrix. Accordingly, in these experiments, we measured the rate of intramitochondrial Ca 2ϩ increase and not that of the decrease of medium [Ca 2ϩ ]. Intact cells were first perfused with a high-[KCl] medium, with no added Ca 2ϩ and in the presence of 50 M EGTA; 50 M digitonin was added for 1 min, which caused the permeabilization of the plasma membrane while leaving the mitochondrial membranes intact. After permeabilization, the Time-dependent Ca 2؉ modulation of mitochondrial Ca 2؉ uptake

Time-dependent Ca 2؉ modulation of mitochondrial Ca 2؉ uptake
cells were perfused with the KCl medium without added Ca 2ϩ (in the presence of oligomycin and rotenone, again to prevent the creation of membrane potential via the oxidation of endogenous substrates or hydrolysis of ATP), and after 2 min, 3 M CaCl 2 together with succinate were added (trace a; 0 min). As in the experiments with isolated mitochondria, presented in Fig.  1, the addition of succinate was delayed by 1 min (trace b) or 5 min (trace c) after that of 3 M free CaCl 2 (Ca 2ϩ and substrate additions not represented in the figure). The acceleration in the maximal rate of Ca 2ϩ accumulation produced by prolonging the preincubation time with 3 M CaCl 2 was even more evident in permeabilized cells than with isolated mitochondria. Indeed, the increase in the rate of mitochondrial [Ca 2ϩ ] in some experiments was, after 5 min of preincubation, as large as 20-fold and on average 15-fold. As in the case of isolated mitochondria, CAMCU was not dependent on the type of substrate oxidized, as it was observed also with glutamate malate (in the absence of rotenone; results not shown). The inset in Fig. 3A shows a quantification of multiple measurements performed.
The use of cell lines offers the opportunity to genetically manipulate the mitochondrial Ca 2ϩ uptake machinery and, thus, to determine whether and, if so, which of the known components of the MCU complex is involved in CAMCU. In Fig.  3B, the cells were transfected with MCU itself or with two of the known uniporter's partners: MICU1 and MICU2. On average, the amount of each overexpressed protein was about 4-fold that of the native one. The rate of Ca 2ϩ accumulation when 3 M Ca 2ϩ was added to the medium together with succinate was slightly faster in cells overexpressing MCU compared with mock-transfected controls, but the time required to reach the maximal rate of Ca 2ϩ accumulation and the percentage increase at 5-min incubation were not significantly different from those observed in controls. On the contrary, in cells overexpressing MICU1, the rate of Ca 2ϩ accumulation at time 0 (3 M CaCl 2 added together with succinate) was notably accelerated compared with controls, and the time necessary to reach the maximal rate of uptake was significantly shorter than in controls; in fact, after 1 min of incubation, the Ca 2ϩ uptake rate

Time-dependent Ca 2؉ modulation of mitochondrial Ca 2؉ uptake
had reached almost its maximal value. The rate of Ca 2ϩ increase in MICU1-overexpressing cells was in fact increased also after 5 min of preincubation, compared with controls, and the accelerated rate was evident also when the values were normalized to the maximal uptake rate. Finally, overexpression of MICU2 reduced the rate of Ca 2ϩ accumulation at all of the preincubation times examined, and the kinetics required to reach the maximal rate, when normalized, were slightly slower than in controls. It is noteworthy that, in these and all of the following experiments, we verified that the maximal Ca 2ϩ influx rate elicited by CAMCU activation at the extracellular [Ca 2ϩ ] of 3 M was lower than the V max obtainable at higher [Ca 2ϩ ] (e.g. at 20 M), meaning that in our experimental conditions, it was not limited by the rate of oxygen consumption.

Effects of mutated MICU1 on CAMCU
Given that the overexpression of MICU1 resulted in a significant alteration of the process of Ca 2ϩ -dependent activation, we next tested (Fig. 4)  where the mutations abolish the Ca 2ϩ -binding properties of the EF-hand domains within the protein). In cells expressing MICU1 C465A , the extent of activation was similar to that observed for the cells overexpressing MICU1, whereas the kinetics were notably downsized, because maximum activation could be observed at 5 min of incubation and not at shorter incubation time. In cells expressing MICU1 EFmut , on the contrary, the presence of the mutant abolished the stimulatory effect observed with WT MICU1, and the mutated protein seems to behave as a dominant negative isoform that quashes the effect of the native WT protein.

Contribution of the different uniporter components to CAMCU
To better understand the effect that each of the uniporter complex components has on CAMCU, in the next series of experiments, each one of them was independently down-regulated by siRNA. As expected, reducing the level of MCU by about 51 Ϯ 5% (average of at least 5 measurements) reduced the rate of Ca 2ϩ accumulation at all preincubation times, but no significant difference in the kinetics and percentage of activation was observed between controls and MCU-down-regulated cells (Fig. 5A). Down-regulation of either MICU1 (Fig. 5B) or

Time-dependent Ca 2؉ modulation of mitochondrial Ca 2؉ uptake
CAMCU activation was much faster in MICU-down-regulated cells compared with controls. This effect was particularly striking in cells where MICU1 was down-regulated; in fact, after 1 min of preincubation, the rate of Ca 2ϩ uptake was about 75% of the rate achieved at 5 min. By comparing the results shown in Figs. 3 and 5, it is clear that the down-regulation of MICU1 and -2 and the overexpression of MICU1 qualitatively have the same effect both on the rate of Ca 2ϩ accumulation at time 0 and on the speed of CAMCU activation. These paradoxical results (see "Discussion") might derive from the fact that down-regulation of MICU1 drastically reduces also the expression of its partner MICU2 (28,29). As shown in Fig. S1, when MICU1 is down-regulated, as in HeLa cells via siRNA, or completely absent, as in HEK-293T KO cells (22), not only is the highmolecular weight MICU1-MICU2 complex markedly diminished or absent, but also, the MICU2 monomer is drastically reduced.

Pharmacological activation of Ca 2؉ uptake
A few years ago, Montero et al. (37) showed that a series of MAPK inhibitors (among which the most potent were kaempferol and SB202190) are able to strongly activate the rate of mitochondrial Ca 2ϩ uptake in both intact and permeabilized cells. They concluded, however, that this effect was not dependent on the inhibition of MAPKs, but it was rather a side effect of the drugs. They also showed that these drugs increased the apparent affinity for Ca 2ϩ of the Ca 2ϩ uptake mechanism of mitochondria. To investigate whether these MAPK inhibitors and the Ca 2ϩ preincubation protocol affected the same target, in the experiment presented in Fig. 6 (A and B), permeabilized cells were perfused with 10 or 40 M kaempferol (a dietary flavonoid and phyto-estrogen) or, alternatively, with 10 M SB202190, either together with the 3 M Ca 2ϩ buffer or after different preincubation time periods with Ca 2ϩ . The rate of Ca 2ϩ uptake at time 0 was increased by about 4-fold by pretreatment with the drug without measurable delay, whereas after 1 or 5 min of Ca 2ϩ preincubation, the effect of the drug was almost negligible. The observed acceleration was more pronounced in control cells, as compared with cells overexpressing MICU1, suggesting that kaempferol, MICU1, and the preincubation period affect the same target(s). As shown in Fig. 6B, in control cells, SB202190 (without preincubation with Ca 2ϩ ) was slightly more potent than kaempferol, but the difference observed is not statistically significant. To better understand whether kaempferol interacts with either MICU in promoting its effect on the activation of the Ca 2ϩ uptake rate, we took advantage of the availability of MICU1 KO HEK-293T (22) cells. In this case, no MICU1 was expressed, and a possible confounding effect of protein residual expression (as in the case of siRNA treatment) was abolished. It is noteworthy that, in MICU1 KO cells, a substantial reduction in MICU2 was observed, as in the case of HeLa cells treated with MICU1 siRNA. The rate of Ca 2ϩ accumulation in digitonin-permeabilized HEK cells was extremely variable; for this reason, the next experiments were carried out in intact cells. As in the previous experiments, the cells were transfected with mitochondrially targeted aequorin, and mitochondrial Ca 2ϩ uptake was elicited by treatment with 500 M charbacol in the presence or absence of 50 M kaempferol. The results in Fig. 7A show that in the total absence of MICU1, the Ca 2ϩ uptake rate produced by kaempferol treatment was clearly reduced from a 2.5 Ϯ 1.5-fold increase (control HEK-293T cells) to a 1.4 Ϯ 0.3-fold increase (MICU1 KO HEK-293T cells). Given that we did not have MICU2 KO available, in the experiments presented in Fig. 7B, HeLa cells treated with MICU2 siRNA were used. The mitochondria Ca 2ϩ uptake rate elicited by 10 M histamine in the presence of 50 M kaempferol in this case was stimulated by a 2.6 Ϯ 1.3-fold increase in controls compared with a 1.9 Ϯ 0.4fold increase in siRNA MICU2 cells.

Ca 2؉ affects the composition of the mitochondrial Ca 2؉ uptake machinery
Given the sensitivity of CAMCU to changes in MCU complex composition, we next investigated whether Ca 2ϩ preincubation affects the protein composition of the complex itself. To this end, we performed protein extraction from isolated mouse liver mitochondria or HeLa cells overexpressing MCU, MICU1, and MICU2, using mild detergent conditions, which should preserve protein interactions, in the presence of 500 M EGTA, 3 M free Ca 2ϩ , or 50 M kaempferol, respectively. In the experiment presented in Fig. 8A, HeLa cells overexpressing MCU-Myc (or MCU-FLAG as control), MICU1-HA, and MICU2-FLAG were solubilized using mild detergent conditions, and the extracts were immunoprecipitated using magnetic beads

Time-dependent Ca 2؉ modulation of mitochondrial Ca 2؉ uptake
conjugated to antibodies against Myc. The samples were eluted from the beads with nonreducing sample buffer and subjected to 10% SDS-PAGE. In these experimental conditions, a MICU1-MICU2 protein complex of about 100 kDa co-immunoprecipitates with MCU. In the presence of 3 M Ca 2ϩ or 50 M kaempferol, the amount of the complex was reduced by 20 Ϯ 8% or 30 Ϯ 5% (mean of 3 independent immunoprecipitations), respectively, when compared with the complex immunoprecipitated in the presence of EGTA. Fig. 8B shows a similar experiment performed with isolated mouse liver mitochondria protein extracts. In this case, there was no overexpression of the uniporter complex components, and the proteins were immunoprecipitated using an antibody directed against endogenous MCU. Elution conditions and SDS-PAGE were the same as with HeLa cell extracts. Also, in this case, we observed about a 30% decrease in the amount of a 100 kDa band, co-immunoprecipitated with MCU when the incubation was performed in the presence of 3 M free Ca 2ϩ .
To determine whether uniporter high-molecular weight complexes were influenced by the presence or absence of Ca 2ϩ , we performed blue native experiments. Isolated mouse liver mitochondria were incubated in the presence of about 3 M free Ca 2ϩ or 50 M EGTA, solubilized with 6% digitonin, and sub-sequently centrifuged at 100,000 ϫ g for 30 min. The supernatants were loaded onto 3-12% bis-tris native gel and separated by electrophoresis. Part of the gel was stained with Coomassie Blue, and the rest was transferred onto a polyvinylidene difluoride membrane, which was probed with anti-MCU antibody and subsequently with antibody against complex II for proteinloading comparison. As shown in Fig. 8C, we observed an MCU-positive, high-molecular weight complex at about 480 kDa, and the amount of the complex extracted in the presence of Ca 2ϩ was about 55 Ϯ 18% of the complex extracted in the presence of EGTA (mean of 4 independent extractions).

Discussion
Ca 2ϩ transport across the inner mitochondrial membrane is a tightly regulated phenomenon supervised by at least two independent pathways: one for Ca 2ϩ uptake, represented mainly by the Ca 2ϩ uniporter, and one for the efflux carried out mostly by the sodium calcium exchanger.
The existence of CAMCU was described over 30 years ago by H. Kröner (17,35) when he defined a condition able to modify the kinetics of mitochondrial Ca 2ϩ uptake at low physiological Ca 2ϩ concentrations, by means of an apparent increased Ca 2ϩ affinity. The remarkable features of CAMCU are its time course (a few minutes to be completed) and its efficacy: more than 10-fold increase (in permeabilized cells) of the rate of Ca 2ϩ uptake for extramitochondrial Ca 2ϩ levels in the low-micromolar range. To the best of our knowledge, no other pathophysiological event or genetic treatment can modify in such a dramatic way the speed of mitochondrial Ca 2ϩ accumulation, and the slow time necessary for completion of its effect strongly argues against a classical conformational modification of some protein involved in the MCU complex. In fact, protein conformational changes usually occur in a time scale of micro-or milliseconds and not of minutes. Accordingly, the molecular mechanism of CAMCU most likely involves some more complex phenomena, such as protein-protein interactions or enzyme-dependent covalent modifications. As to enzymatic activity, the present data exclude the possibility of a kinase as the trigger, given that the phenomenon occurs in the total absence of added ATP and in the presence of oligomycin, an inhibitor of the mitochondrial ATPase. In addition, CAMCU is reversible upon removal of extramitochondrial Ca 2ϩ in dozens of seconds, again in the absence of ATP, thus excluding the possibility of rephosphorylation as the mechanism for CAMCU inhibition under these conditions. The effect of extramitochondrial Ca 2ϩ (and thus of its concentration in the intermembrane space) appears, on the contrary, consistent with an effect on MCU or the proteins interacting with it. In fact, both MICU1 and MICU2 possess EF-hand Ca 2ϩ -binding sites that are exposed to the intermembrane space, and the expression of a MICU1 mutant that abolishes the EF hands inhibits CAMCU. The overexpression of MICU1 strongly accelerates CAMCU, whereas the overexpression of MICU2 reduces the Ca 2ϩ uptake rate of the uniporter (consistent with their proposed role as uniporter's gatekeepers) (29). The effects of MICU1 and MICU2 down-regulation by siRNA are apparently contradictory, as there is an increase in the rate of Ca 2ϩ uptake at time 0 and 1 min in both cases. The increase in Ca 2ϩ uptake by a

Time-dependent Ca 2؉ modulation of mitochondrial Ca 2؉ uptake
reduction in MICU2 levels is expected and consistent with the inhibitory action of this protein on the MCU complex activity (29) and with the effect of its overexpression (Fig. 3). Less obvious is why MICU1 down-regulation results in an acceleration of CAMCU. We propose that the speed of CAMCU activation depends on the MICU1/MICU2 ratio. This explanation is consistent with the effect of MICU1 overexpression, as observed in the experiments presented in Fig. 3, and also with the paradoxical result that MICU1 down-regulation has the same effect on CAMCU as MICU1 overexpression. It has been observed, in fact, that reduction in MICU1 levels by a specific siRNA causes an even more dramatic decrease in MICU2 protein levels, without any effect on MICU2 mRNA. It has been proposed that a reduced level of MICU1 impairs the stability of MICU2 and thus causes a drastic reduction of the expression also of the latter protein (28,29). Thus, paradoxically, the treatment with a MICU1-specific siRNA results in a reduction in the levels of both proteins but could produce a net increase in the MICU1/ MICU2 ratio and thus an acceleration of the CAMCU process.
Clearly, given the slow development of the process, CAMCU cannot be due solely to the conformational change of individual proteins, but rather to a more complex mechanism. Recently, Petrungaro et al. (31), while studying the effect of the oxidoreductase MIA-40 on the formation of MICU1-MICU2 dimers, demonstrated that these latter two proteins are capable of forming a complex with MCU in a Ca 2ϩ -dependent way. In particular, they showed that the MICU1-MICU2 complex coprecipitates with MCU at low Ca 2ϩ levels, whereas it dissociates at high Ca 2ϩ concentrations. Multiple experimental data suggest that MICU1 and/or MICU2 act as gatekeepers of MCU and that they control the affinity of the channel for Ca 2ϩ . This process occurs rapidly, as activation of mitochondrial Ca 2ϩ influx follows by a few milliseconds the increase in cytosolic Ca 2ϩ . CAMCU, on the contrary, depends on the medium (cytosolic) Ca 2ϩ levels, but it takes minutes to reach completion. It is therefore compatible with a slow dissociation of the complex from MCU, as described (31). Our data confirm and extend the observations of Petrungaro et al. (31), as we observed a reduction in the association of MICU1 and -2 with MCU in the presence of 3 M Ca 2ϩ and a reduction in the high-molecular weight complexes in the presence of Ca 2ϩ . In other words, the simplest explanation regarding the molecular mechanism of CAMCU is that the presence of micromolar levels of Ca 2ϩ for a few minutes in the intermembrane space results in the binding of Ca 2ϩ to the EF-hand domains of the MICUs that causes a conformational change of the two proteins. This, in turn, might cause a slow dissociation of the proteins from the MCU, with a consequent increase in the affinity of the channel for Ca 2ϩ . The effect of drugs such as kaempferol and SB202190 (previously shown to increase the apparent affinity for Ca 2ϩ of the uniporter (i.e. similar to the effect observed in CAMCU)) is particularly interesting. Indeed, these drugs and CAMCU appear not to be additive, as the effect of kaempferol is strong at time 0, it is reduced after 2 min of incubation with Ca 2ϩ , and it is null at 5 min when CAMCU has reached its maximum. The stimulatory effect of kaempferol on Ca 2ϩ uptake rate is markedly reduced in MICU1-KO cells or, although less effectively, when MICU2 is down-regulated. Both drugs increase the apparent Ca 2ϩ affinity of the Ca 2ϩ uptake system, and in cells incubated with kaempferol (in the absence of Ca 2ϩ ), we observe a reduction of the MICU1-MICU2 complex associated with MCU. In other words, kaempferol, and possibly SB202190, mimic both functionally and at the molecular level the effects of CAMCU. It remains unclear whether the target(s) of these drugs is MICU1 and/or MICU2 or MCU itself. The latter possibility appears unlikely, given the strong inhibition by MICUs down-regulation on kaempferol activation of Ca 2ϩ uptake rate.
The key and final question concerns the physiological role of CAMCU. It is easy to speculate that CAMCU will be maximally activated by ischemic conditions, where nonrespiring mitochondria will be exposed for prolonged periods (minutes to hours) to high Ca 2ϩ . Upon reperfusion, and thus reactivation of the respiratory chain activity and repolarization of membrane potential, CAMCU should be maximally activated and may thus contribute to the mitochondrial Ca 2ϩ overload characteristic of ischemia-reperfusion in different tissues, the heart in particular. Prolonged, smaller increases in cytosolic Ca 2ϩ are known to occur in several pathophysiological conditions, and CAMCU can therefore be activated under those conditions and contribute to increasing the efficacy of mitochondrial Ca 2ϩ uptake. Clearly, CAMCU is a double-faced tool. For relatively small increases in [Ca 2ϩ ], it can be beneficial, priming the mitochondria Ca 2ϩ uptake machinery and thus making it more efficient. On the other hand, however, it can be disastrous; it can increase the probability of mitochondrial Ca 2ϩ overload, leading to cell death.

Experimental procedures
Procedures involving the use of animals were carried out in strict adherence to the Italian regulations on animal protection and care and with the explicit approval of the Italian animal welfare regulations: authorization number 287/2015-PR from  1, 2, 5, and 6), 3 M free Ca 2ϩ (lanes 3 and 7), and 50 M kaempferol (lanes 4 and 8), respectively. Protein extracts were immunoprecipitated (IP) using magnetic agarose beads conjugated to anti-Myc mouse antibody. Proteins bound to the beads were eluted with nonreducing Laemmli sample buffer. Lanes 1-4, input; lanes 5-8, immunoprecipitates. Membranes were immunoblotted (IB) with antibody against MICU1 or antibody against MCU, respectively. B, mouse liver mitochondria protein extract treated as described for the HeLa cell extract except for kaempferol. The protein extracts were incubated with an antibody against WT MCU or with an irrelevant antibody produced in rabbit as a control; the immunoprecipitates were eluted using a nonreducing sample buffer.

Isolated mitochondria calcium uptake
Mitochondria were isolated from the liver of C57Bl/6 mice by standard differential centrifugation (38). Protein concentration was measured using the bicinchoninic acid assay.
Mitochondrial calcium uptake was measured in 2-ml stirred cuvettes with a PerkinElmer Life Sciences LS50B spectrofluorometer (excitation and emission wavelengths 505 and 535 nm, respectively). Mitochondria were incubated for 2 min in the medium without substrates and Ca 2ϩ to equilibrate, and then 8 M Ca 2ϩ was added together with substrate specific for the respiratory complex examined; alternatively, substrate addition was delayed 1, 2, 5, or up to 10 min. Free-Ca 2ϩ concentration was determined with MaxChelator version 2.1 (39,40). Calcium Green-5N fluorescence was converted into [Ca 2ϩ ] after careful titration with Ca 2ϩ additions of known concentration.

Cell culture and transfection/silencing
HeLa cells (ATCC CCL-2 TM ) and HEK-293T cells (for MICU1 KO HEK-293T cell generation; see Sancak et al. (22)) were grown in Dulbecco's modified Eagle's medium supplemented with 5% fetal bovine serum, 100 IU/ml streptomycin plus penicillin. Cells were maintained in an incubator at 37°C and 5% CO 2 and controlled humidity. After 1 month in culture or alternatively 10 division passages, the culture was discarded, and a new batch of cells was thawed from stocks in liquid nitrogen. Care was taken that morphological and functional characteristics of the cells in terms of Ca 2ϩ signaling did not change significantly with the time in culture.

Transfection
HeLa or HEK-293T (0.6 ϫ 10 5 ) cells were plated on 13-mm round glass coverslips (coverslips were treated with 0.1 mg/ml polylysine to promote the adhesion of HEK-293T cells). At ϳ60% confluence, 24 h after seeding, cells were transfected using TransIT-LT1 (Mirus Bio) as per the manufacturer's instructions. Cells were transfected with low-affinity mitochondrial aequorin probe together with plasmid for the overexpression of the protein specified in the figure legend or with plasmid pcDNA3.1 used as control. Ca 2ϩ uptake measurements were performed 24 h after transfection. For RNAi experiments, the growth medium was replaced 1 h before transfection with antibiotic-free medium. Cells were transfected using Lipofectamine RNAiMAX transfection reagent (Thermo Fisher Scientific); control siRNA (MISSION siRNA Universal Negative Control #1 SIC001) or MCU SASI Hs01 002334628, or MICU1 SASI Hs01 00070248, or MICU2 SASII Hs01 00106267 (Sigma-Aldrich) was added to the transfection mix to a final concentration of 5 nM. After 24 h, the medium contain-ing siRNA plus transfection agent was replaced with fresh medium, and cells were transfected with low-affinity mitochondrial aequorin probe as described. Ca 2ϩ uptake measurements were performed 48 h after silencing.

Ca 2؉ uptake measurements
The coverslip with the cells was incubated with 5 M coelenterazine for 1-2 h in Krebs-Ringer-modified buffer (KRB: 135 mM NaCl, 5 mM KCl, 0.4 mM KH 2 PO 4 , 1 mM MgSO 4 , 1 mM MgCl 2 , 5.5 mM glucose, 20 mM HEPES, pH 7.4, at 37°C) supplemented with 1 mM CaCl 2 , and then transferred to the perfusion chamber. After a 3-min equilibration with KRB, cells were permeabilized using a buffer mimicking the cytosolic ionic composition: 130 mM KCl, 10 mM NaCl, 1 mM KH 2 PO 4 , 1 mM MgCl 2 , 20 mM Hepes, supplemented with 50 M EGTA and 50 M digitonin. Cells were perfused for 1 min with digitonin and washed for 2 min with cytosolic-like buffer; subsequently, cells were perfused with cytosolic-like buffer devoid of EGTA and supplemented with 3 M Ca 2ϩ together with 5 mM glutamate, 2.5 mM malate, 1 M oligomycin, or 5 mM succinate plus oligomycin and 1 M rotenone; alternatively, the addition of the respiratory chain substrates followed 1 or 5 min after Ca 2ϩ addition. The experiments were terminated by lysing the cells with 100 M digitonin in a hypotonic Ca 2ϩ -rich solution (10 mM CaCl 2 in H 2 O), thus discharging the remaining aequorin pool. The light signal was collected and calibrated into [Ca 2ϩ ] values by an algorithm based on the Ca 2ϩ response curve of aequorin at physiological conditions of pH, [Mg 2ϩ ], and ionic strength, as described previously (41). Mitochondrial Ca 2ϩ uptake speed was calculated as the first derivative by using the Origin differentiate function and averaging for three time points.

Western blotting and immunoprecipitation
For protein extraction, about 10 6 HeLa cells were grown and transfected with the indicated constructs or siRNA as described above. 24 or 48 h after transfection/silencing, cells were solubilized in radioimmune precipitation buffer (50 mM Tris, 150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholic acid, 0.1% SDS, pH 7.5), with cOmplete TM , mini, EDTA-free protease inhibitor mixture (Roche Applied Science) and incubated on ice for 30 min. Crude extracts were centrifuged at 10,000 ϫ g for 20 min at 4°C to remove debris; proteins in the supernatant were quantified using the BCA protein assay kit (Pierce TM ). 40 g of proteins were dissolved in reducing Laemmli sample buffer and heated for 5 min at 85°C.
For native mild IPs, 10 6 HeLa cells grown on a 10-cm Petri dish were transiently transfected as described above with pcDNA3.1-MCU-FLAG, pcDNA3.1-MCU-Myc, pcDNA3.1-MICU2-FLAG, and pcDNA3.1-MICU1-HA. After 24 h of expression, cells were washed three times with cold PBS and then lysed with an appropriate volume of native mild lysis buffer (50 mM HEPES, pH 7.4, 150 mM NaCl, 0.2% n-dodecyl ␤-D-maltopyranoside (DDM), 1.7 mM MgCl 2 , and 0.5 mM EGTA, cOmplete TM , mini, EDTA-free protease inhibitor mixture (Roche Applied Science), in the presence or absence of 3 M free Ca 2ϩ or 50 M kaempferol, incubated for 10 min at 4°C, and centrifuged for 1 h at 25,000 ϫ g. The supernatants Time-dependent Ca 2؉ modulation of mitochondrial Ca 2؉ uptake were quantified for protein content and immunoprecipitated using magnetic beads conjugated to antibodies against Myc as per the manufacturer's instructions. The samples were washed three times using native mild lysis buffer and once with lysis buffer without DDM and then eluted from the beads with 2ϫ concentrated nonreducing sample buffer and subjected to 10% SDS-PAGE. Alternatively, 1 mg of mouse liver crude mitochondria was also used and lysed using the same protocol. The supernatants were precleared by adding 1 g of irrelevant rabbit IgG together with 20 l of protein A/G PLUS-agarose (sc-2003, Santa Cruz Biotechnology, Inc.) and incubated at 4°C for 1 h on a rotating mixer. Cleared extracts were then incubated overnight with an antibody against MCU (HPA016480, Sigma-Aldrich) and then adsorbed on 20 l of protein A/G PLUSagarose at 4°C for 3 h. Immunoprecipitates were collected by centrifugation at 2,000 rpm and washed three times using native mild lysis buffer and once with lysis buffer without DDM. Elution was performed with 40 l of 2ϫ nonreducing Laemmli sample buffer. Proteins were separated by SDS-PAGE and immunoblotted with the indicated antibody

Blue native gel electrophoresis
Isolated mouse liver mitochondria were resuspended at 1 mg of protein/ml of buffer (250 mM mannitol, 20 mM Hepes, 1 mM P i , 1 mM MgCl 2 , 2 M TPEN, 1 M rotenone, 1 g/ml oligomycin) in the presence of 3 M free Ca 2ϩ (measured) or 50 M EGTA; incubated for 5 min at room temperature; and subsequently centrifuged at 4°C for 15 min in a microcentrifuge at maximal speed. All of the following procedures were performed at 4°C. Pellets were resuspended with 80 l of extraction buffer: 50 mM NaCl, 50 mM imidazole, 2 mM 6-aminocaproic acid, 1 mM EDTA, pH 7. 6% digitonin was added to each suspension, which was then incubated for 30 min and centrifuged at 100,000 ϫ g for 30 min. Supernatants were treated with 5% glycerol (final concentration) and with 5% Coomassie G-250 to a final ratio of 8 g of detergent/1 g of dye. 100 g of protein were loaded onto each lane of a 3-12% bis-tris nativePAGE TM gel (Thermo Fisher Scientific) and run as per the manufacturer's instructions. Part of the gel was stained with Coomassie Brilliant Blue and examined at the UVITECH mini HD imaging system (Cleaver Scientific); the rest of the gel was transferred onto a polyvinylidene difluoride membrane and probed with anti-MCU antibody as described above.

Statistical analyses
Data were analyzed using Origin version 8 SR2 (OriginLab, Northampton, MA) or with GraphPad Prism software. Averages are expressed as mean Ϯ S.D. For samples with normal distribution, a two-sample t test was used. When comparing more than two samples, one-way analysis of variance and Bonferroni post hoc tests were used. For samples without normal distribution, nonparametric Wilcoxon-Mann-Whitney test and Kruskal-Wallis plus Dunn's multiple-comparison test were used. *, p Յ 0.05; **, p Յ 0.01; ***, p Յ 0.001.
Author contributions-E. B. and T. P. conceptualization; E. B. data curation; E. B., G. R., and A. E. Z. formal analysis; E. B. and T. P. supervision; E. B. and T. P. writing-original draft; E. B., A. E. Z., and T. P. writing-review and editing; E. B., G. R., and A. E. Z. investigation; T. P. funding acquisition; E. B. and T. P. methodology.