Structure and function of the Mitochondrial Calcium Uniporter complex

Abstract

The Mitochondrial Calcium Uniporter (MCU) is the critical protein of the inner mitochondrial membrane mediating the electrophoretic Ca²⁺ uptake into the matrix. It plays a fundamental role in the shaping of global calcium signaling and in the control of aerobic metabolism as well as apoptosis. Two features of mitochondrial calcium signaling have been known for a long time: i) mitochondrial Ca²⁺ uptake widely varies among cells and tissues, and ii) channel opening strongly relies on the extramitochondrial Ca²⁺ concentration, with low activity at resting [Ca²⁺] and high capacity as soon as calcium signaling is activated. Such complexity requires a specialized molecular machinery, with several primary components can be variably gathered together in order to match energy demands and protect from toxic stimuli. In line with this, MCU is now recognized to be part of a macromolecular complex known as the MCU complex. Our understanding of the structure and function of the MCU complex is now growing promptly, revealing an unexpected complexity that highlights the pleiotropic role of mitochondrial Ca²⁺ signals.

Introduction

Back into 1961 and 1962, two seminal papers unequivocally described for the first time that energized mitochondria could accumulate large amounts of Ca²⁺[1,2]. However, despite the immediate interest and the high-quality scientific works on the topic, this notion soon faded into oblivion. Indeed, due the relative low affinity of their Ca²⁺ transport systems, mitochondria became simple bystanders in the exploding field of global Ca²⁺ signaling, where plasma membrane and endoplasmic reticulum Ca²⁺ channels hit the scene. Mitochondrial Ca²⁺ uptake had to wait decades for a complete payback. The first step forward into the renaissance of mitochondrial Ca²⁺ was the development of high-specific, genetically-encoded, mitochondria-targeted Ca²⁺ probes, that lead up to the demonstration that mitochondria can rapidly and efficiently take up Ca²⁺ in living cells whenever global Ca²⁺ signaling is activated[3–6]. The next step forward was the direct, in situ measurement of the Ca²⁺-selective channel of the inner mitochondrial membrane, the so-called Mitochondrial Calcium Uniporter (MCU) [7]. Finally, in 2011, we and the group of Mootha identified a still uncharacterized protein, known as CCDC109A, as necessary and sufficient for mitochondrial Ca²⁺ uptake in vitro and in vivo, recapitulating most of the features of the MCU [8,9]. Thus, the discovery of the molecular identity of the channel was the last piece of the puzzle needed for the definitive revival of mitochondrial Ca²⁺ signaling field. Indeed, in the last four years we observed the proliferation of a number of high impact works related to MCU, its regulators and its physio-pathological role. This review will try to summarize the current knowledge on what we now call “the MCU complex”, i.e. the macromolecular structure that composes the mitochondrial Ca²⁺ uptake machinery. For the sake of clarity, we divided the MCU complex in i) core components of the membrane pore, ii) MCU-associated regulators and iii) physiological role of mitochondrial Ca²⁺ uptake. It is however clear that the whole picture is incomplete and rapidly evolving, since many of the findings are new, they still need independent validation by other labs, and additional components are likely to be not yet discovered.

The core components of the MCU complex

To date, three different membrane proteins have been shown to be part of the Ca²⁺ permeant pore, namely MCU, MCUb and EMRE.

MCU

The MCU gene is well conserved in all eukaryotes except for yeasts and encodes for a protein composed by two coiled-coil domains and two transmembrane domains separated by a short loop enriched in acidic residues (EYSWDIMEP). Initially, MCU alone appeared to be necessary and sufficient for an efficient mitochondrial Ca²⁺ uptake [8]. Indeed, on one hand its silencing invariably leads to a dampen of mitochondrial calcium transients; on the other hand, its overexpression causes a significant increase of mitochondrial transients (at least twice than the control) and the recombinant MCU is sufficient per se to form a Ca²⁺-selective channel in planar lipid bilayer. Channel activity is inhibited by Ruthenium Red and Gd³⁺ and abrogated by site-specific mutagenesis of two residues of the loop region. The current is similar but not identical to the one recorded in patch clamp experiments of isolated mitoplasts [7], especially in terms of open probability (much lower in planar lipid bilayer). This could be in principle due to i) different lipid environments, ii) post translational modification that are missed in the recombinant protein or iii) the lack of some endogenous regulators. In the last four years, the role of MCU as a necessary component for organelle calcium uptake was confirmed in different systems. Indeed, in vivo siRNA delivery abrogates mitochondrial calcium uptake in the liver [9]. Similar finding have been obtained in neonatal rat cardiomyocytes [10], heart [11], pancreatic β cells [12], neurons [13], and breast epithelial cells. Finally, analysis of isolated organelle from different tissues of MCU knock out mice revealed the lack of any mitochondrial Ca²⁺ uptake, thus demonstrating the absolute requirements of the MCU gene [14]. However, it was immediately clear that MCU alone could not ensure the fine tuning needed to guarantee the pleiotropic role that mitochondrial calcium signals play within the cell. Indeed, the complexity of MCU channel increased significantly with the identification of two components of the channel itself, namely MCUb [15] and EMRE [16].

MCUb

Genomic analysis revealed a gene closely related to MCU, originally reported as CCDC109B and now known as MCUb. The encoded protein shares 50% similarity to MCU, possesses two coiled-coil domains and two transmembrane domains separated by a short loop that slightly differs from MCU (VYSWDIMEP). MCUb is conserved in most of the vertebrates and in many plants but absent in other organisms where MCU is present. We found that MCU forms a multimer (likely to be a tetramer) in which MCU can self-oligomerize or hetero-oligomerize with MCUb. More importantly, this isoform has a crucial amino acid substitution in the loop region (E256V) that could in principle have an impact on channeling properties by removing one critical negative charge. Indeed, molecular dynamics simulations predicted that Ca²⁺ permeation through MCUb is greatly impaired [15]. This was confirmed in both HeLa cells and planar lipid bilayers. On one hand, MCUb overexpression decreased agonists-evoked mitochondrial Ca²⁺ transient in living cells. On the other hand, recombinant MCUb inserted in artificial membranes showed no current when Ca²⁺ was used as permeating ion. In addition, concomitant expression of MCU and MCUb greatly decreased the open probability in planar lipid bilayer, even when MCUb is present in low amount, thus supporting the notion that the insertion of a small fraction of MCUb within the oligomer can efficiently inhibit Ca²⁺ channeling activity. Finally, silencing of MCUb in HeLa cells causes a significant increase of mitochondrial Ca²⁺ uptake, thus underlying the inhibitory role that MCUb exerts on channel activity. Hence, in our model MCUb is the endogenous dominant negative subunit of the MCU complex. As to its potential physiological meaning, we envisaged that he ratio between MCU/MCUb varies a lot among different tissues. This is in line with the recent demonstration that also the overall activity of the MCU complex is highly variable among tissues [17]. Most importantly, a nice correlation between MCU/MCUb expression ratio and the recorded mitochondrial Ca²⁺-selective currents is apparent. This opens the possibility that this ratio set the overall mitochondrial Ca²⁺ carrying capacity of different tissues. In conclusion, MCUb appears to mediate a novel regulatory mechanism of the pore formation; by the inclusion of different moieties of the dominant negative subunit in the multimeric channel, different tissues can set the maximal mitochondrial Ca²⁺ currents.

EMRE

As discussed above, MCU is sufficient per se to form a Ca²⁺ channel in planar lipid bilayer, but what happens in vivo is unclear. EMRE is a broadly expressed 10KDa protein that spans the IMM and possesses a highly conserve C-terminus rich in aspartate residues, with a still unclear membrane topology [16]. Mootha and colleagues propose that this protein is required for Ca²⁺ channeling activity and keep the MICU1/MICU2 dimer attached to the MCU complex. In the contrast to its essential role, EMRE homologs are not present in plants, fungi or protozoa in with MCU and MICU1 are highly conserved. Downregulation or knockout of EMRE totally abolishes mitochondrial Ca²⁺ uptake, even when MCU is overexpressed. At protein level, EMRE stability is impaired in the absence of MCU, suggesting that other MCU complex components appears to be require in vivo for its correct function. It must be noted that he putative role of EMRE in mediating the binding between MCU and MICU1 is in contrast with the clear positive effect of MICU1 on MCU in planar lipid bilayer, where no other components are present. However, it is also clear that in absence of EMRE the MCU complex became smaller, as revealed by blue-native PAGE [16]. This open the possibility that EMRE could be an essential protein for efficient assembly of the MCU complex. In line with this, it was recently shown that, in a heterologous system such as yeast, EMRE is required for the formation of a functional channel only with the mammalian MCU, but not with MCU derived from fungi [18]. We think that additional experimental work is needed to solve the remaining issues.

MCU-associated regulators

The road to the discovery of the MCU actually started with the identification of one of its key regulators, MICU1 [19]. Other proteins that exert a regulatory role on the channel have been identified so far, and these include the whole MICU’s family [20], MCUR1 [21] and SLC25A23 [22].

The MICU’s family

One key feature of the mitochondrial Ca²⁺ uptake machinery is the sigmoidal response to extra mitochondrial (i.e. cytosolic) [Ca²⁺], with very low rate at resting cytoplasmic [Ca²⁺] levels (thus preventing mitochondrial Ca²⁺ overload and ion vicious cycling), and a very large Ca²⁺ carrying capacity at higher [Ca²⁺], that ensures prompt responses to cell stimulation [5,6,23,24]. This property could be in principle due to the channel per se or to a set of different regulators that keep the channel close at resting condition and activate it at high Ca²⁺ concentration. However, MCU exposes in the intermembrane space only a small loop, with the vast majority of the protein residing within the matrix [25]. Hence, this sigmoidicity is unlikely to be a property of the channel per se, but it is likely conferred to MCU by different regulators located in the intermembrane space. In the last four years different groups showed that MICU’s family has a role in the sigmoidal response of MCU to external Ca²⁺ levels [26-28]. This hypothesis appeared plausible especially because these proteins have two conserves EF hand domains that confers Ca²⁺ sensitivity [20].

MICU1 was the first member of the MICU’s family discovered in 2010 by Mootha’s group [19]. In the beginning, it was considered necessary for mitochondrial Ca²⁺ uptake, but unlike to be the channel per se since it has only one (eventually wrongly) predicted transmembrane domain. The actual localization of this protein has been matter of debate [26] but both recent proteomic data [29,30] and the resolution of its function [28] strongly indicate that the MICU1, as well as the other members of the family, are soluble proteins of the intermembrane space. However, later another group reported that MICU1 acts as a MCU gatekeeper, i.e. keep MCU close when the extra mitochondrial Ca²⁺ concentration is low [26]. Indeed, Madesh and coworkers showed that mitochondria from MICU1 silenced-cells are constitutively overloaded with Ca²⁺, thus uncovering the gatekeeping role of MICU1. Soon after, Hajnoczky added another level of complexity to the function of MICU1. On one hand they confirmed the MICU1 as MCU gatekeeper, but on the other hand they elegantly showed that, in the absence of MICU1, mitochondrial Ca²⁺ uptake is less efficient [27]. According to their model, MICU1 not only controls the threshold of MCU opening but also cooperate to activate the channel open state at high Ca²⁺ concentration. These data thus indicated MICU1 as the only determinant of the sigmoidal response of organelle Ca²⁺ uptake to the extramitochondrial [Ca²⁺]. Subsequently, we extended this model by taking into account the other members of the MICU’s family.

Indeed, similarly to MCU, MICU1 (formerly known as EFHA3) has two different isoforms, MICU2 (formerly known as EFHA1) and MICU3 (formerly known as EFHA2). They are likely to be located in the intermembrane space as well [30] and their specific role is still under debate. We showed that MICU2 forms an obligate heterodimer with MICU1 that interact with MCU in the DIME loop, thus facing the intermembrane space [28]. We demonstrated that MICU2 has a genuine gatekeeping function at low Ca²⁺ level both in living cells and in electrophysiological recordings carried out in planar lipid bilayer. Its overexpression indeed is able to decrease the agonist evoke stimulus in HeLa cells and decrease the open probability of the reconstitute MCU channel in planar lipid bilayer with no effect at high Ca²⁺ concentration. As already reported [20], we confirmed that the stability of MICU2 is dependent on the presence of MICU1 at protein level. Indeed, MICU1 silenced cells have a drastic reduction also in MICU2 protein level, despite no effect on MICU2 mRNA amount, thus pointing to a post-translational mechanism. Hence, we think that the previously reported loss of MCU gatekeeping in the absence of MICU1 [26,27], is likely to be due to the concomitant loss of MICU2.

On other hand, MICU1 is the cooperative activator of the channel, since it increase agonist-evoked mitochondrial Ca²⁺ transients and increase the MCU open probability in planar lipid bilayer with no effect in Ca²⁺-free medium. According to our model, at low [Ca²⁺], the prevailing inhibitory effect of MICU2 ensures minimal Ca²⁺ accumulation in the presence of a very large driving force for cation accumulation, thus preventing the deleterious effects of Ca²⁺ cycling and matrix overload. As soon as extramitochondrial [Ca²⁺] increases, Ca²⁺-dependent MICU2 inhibition and MICU1 activation guarantees the prompt initiation of rapid mitochondrial Ca²⁺ accumulation, thus stimulating aerobic metabolism and increasing ATP production [28] (see Figure 1). Finally, MICU3 has probably a minor role in this process, since it appears to be predominantly expresses in the CNS [20]. However, the exact functions of MICU3 in MCU complex regulation need to be investigated.

Figure 1.

Schematic representation of the MCU complex: in resting condition (see left) mitochondrial calcium uptake is controlled by a multiprotein complex that can be composed by MCU and MCUb (the channel forming subunits) together with EMRE, MICU1, MICU2 (other components are omitted for the sake of clarity). In these conditions, MICU1/MICU2 heterodimers act as MCU gatekeeper, thanks to the prevailing inhibitory effect of MICU2, thus preventing vicious calcium cycles and energy sink. As soon as calcium signaling is activated, the increase in extramitochondrial [Ca²⁺] (see right) induces a conformational change in the whole dimer that releases MICU2-dependent inhibition and triggers MICU1-mediated enhancement of MCU channeling activity.

MCUR1

In order to identify other components of the Ca²⁺ uptake machinery, Madesh and colleagues performed a direct human RNAi screen of 45 mitochondrial membrane proteins in HEK293T cells predicted to be part of the inner mitochondrial membrane [21]. They identified two proteins with a modest (SLC25A23) and a strong (CCDC90A) effect on mitochondrial Ca²⁺ uptake. The latter is now known as MCUR1, a 40kDa protein of the inner mitochondrial membrane with one predicted transmembrane domain, one coiled-coil region, the N-termini facing the intermembrane space and the major part of the protein exposed to the matrix. Madesh and coworkers demonstrated that MCUR1 silencing not only largely inhibited agonist-induced mitochondrial Ca²⁺ uptake, but also caused a striking decrease of basal mitochondrial matrix [Ca²⁺]. They also suggested that different MCU complexes exist. While both MCUR1 and MICU1 could co-immunoprecipitate with MCU, MCUR1 did not interact with MICU1, thus indicating at least two qualitatively different MCU containing complexes [21]. This would imply that at least a pool of the MCU would work without a gatekeeper, thus allowing Ca²⁺ uptake into the matrix even at resting cytosolic [Ca²⁺]. Interestingly, also MCUR1 appears to have an isoform conserved in most vertebrates and named CCDC90B, whose function is still uncharacterized. Recently, a paper by Shoubridge and colleagues seriously questions the direct involvement of MCUR1 in the modulation of the MCU complex. They demonstrated that the silencing of MCUR1 causes a drop of mitochondrial membrane potential that correlates with a decrease of complex IV assembly and activity [31]. Considering that mitochondrial membrane potential is the main driving force guiding the entrance of all cations inside organelle matrix, this could explain the decrease of mitochondrial [Ca²⁺] and most of the data reported by Mallilankaraman et al., where anyhow differences in ΔΨ_m were measured but not detected. In addition, MCUR1 is supposed to have a homologue in Saccharomyces cerevisiae (Fmp32), an organism lacking mitochondrial Ca²⁺ uptake. However, Fmp32 and MCUR1 share only partial similarities and their functional homology is still to be conclusively demonstrated. Overall, the contribution of MCUR1 to the regulation of the MCU complex still need further investigation.

SLC25A23

The other protein identified by Madesh and coworkers is SLC25A23 [22]. This protein belongs to a family of solute carriers that transport Mg-ATP/Pi across the inner mitochondrial membrane. It apparently participates in mitochondrial Ca²⁺ uptake probably due to its interaction with MCU and MICU1. The effect of this EF-hand containing mitochondrial protein is likely to depend on the local [Ca²⁺]. Indeed, mutation of these Ca²⁺-binding sites exhibited a dominant-negative effect, i.e. it reduced mitochondrial Ca²⁺ transients. Until now, the mechanism is not clear, but it was proposed that SLC25A23 could act by sequestering MICU1 in order to increase the MCU-mediated Ca²⁺ uptake. Further work is thus needed to clarify this mechanism.

MCU-independent mitochondrial Ca²⁺ uptake?

One still unresolved question in the field is whether only one mechanism for mitochondrial Ca²⁺ uptake exist. Indeed, few but consistent works describe at least two organelle Ca²⁺ uptake pathways, i) one activated by relatively high extramitochondrial [Ca²⁺], highly sensitive to Ruthenium Red and capable of sequestering huge amount of Ca²⁺; ii) the other one can be activated by lower extramitochondrial [Ca²⁺], it is inhibited by higher Ruthenium red concentrations and enables the rapid and transient accumulation of short Ca²⁺ pulses. While there is a general consensus that the first mechanism is mediated by MCU, much less is known about the latter. Historically, it was first described by Gunter and coworkers [32,33] in isolated mitochondria and named “Rapid uptake Mode” (RaM). This notion was then supported by direct electrophysiological channel recordings in isolated mitoplasts, where two or more Ca²⁺-selective currents have been recorder. Hoppe and colleagues identified the so-called mCa1 (showing the well-known features of the classical MCU-dependent Ca²⁺ uptake) and mCa2 (with many similarities to the RaM) [34]. More recently, Graier and coworkers identified three apparently different Ca²⁺ currents, the predominant MCU-mediated (here named i-MCC), and two others, known as xl-MCC and b-MCC, both characterized by a lower sensitivity to Ruthenium Red, thus reminding the RaM [35]. However, several issues still need to be resolved. On one hand, it is indeed still unclear whether these different mitochondrial Ca²⁺ uptake modes take place in vivo, and in which conditions. More importantly, it is not known yet what are the consequences of these two modes in terms of changes of mitochondrial [Ca²⁺]. One intriguing hypothesis on this point has been proposed by O’Rourke and coworkers. By measuring free matrix [Ca²⁺] in isolated mitochondria, they showed that different Ca²⁺ uptake modes (in this case, the RaM-like mode is named MCU_mode1, while the classical mode is MCU_mode2) can activate different intramitochondrial Ca²⁺-buffering systems, thus leading to significant differences in free matrix [Ca²⁺][36]. Finally, the remaining issue regards the molecular identity of these different Ca²⁺ uptake systems. Are these modes all MCU-dependent or are they mediated by different Ca²⁺ channels? Our biased opinion is that there is enough complexity within the MCU complex that could account for different mitochondrial Ca²⁺ uptake modes. As an example, the RaM could be in principle mediated by the MCU alone without its regulators MICU1 and MICU2 (i.e. a Ca²⁺-selective channel with a lower activation threshold but with an overall lower Ca²⁺ carrying capacity). The further elucidation of the genuine role of the different MCU complex components will help to solve these questions.

The physiological role of mitochondrial Ca²⁺ uptake

Even before the discovery of MCU, a significant amount of work has been carried out to uncover the physiological role of mitochondrial calcium homeostasis. Hundreds of papers undisputable indicate that in a plethora of cellular models mitochondrial [Ca²⁺] can regulate both ATP production and cell death through apoptosis. In particular, matrix [Ca²⁺]_mt acts (by either direct Ca²⁺ binding or via Ca²⁺-dependent dephosphorylation) as a positive allosteric regulator of three key dehydrogenases of the tricarboxylic acid (TCA) cycle [37,38] and increases the activity of the ETC complexes [39]. Thus, the [Ca²⁺]_mt increase evoked by a cytosolic [Ca²⁺] rise leads to a boost in ATP production. In parallel, unregulated and sustained organelle Ca²⁺ overload can also lead to the opening of the mitochondrial Permeability Transition Pore (mPTP) [40,41], with consequent dissipation of mitochondrial membrane potential (ΔΨ_mt), release of caspase cofactors into the cytoplasm and triggering of the apoptotic cascade [42]. In addition, recent work carried out after the molecular identification of MCU started to confirm these finding, at least at cellular level. As an example, in neonatal rat cardiomyocytes, MCU expression regulates the buffering of cytoplasmic Ca²⁺ during systole [10]. In pancreatic β cells, MCU regulates cellular glucose sensing capacity [12], despite the other candidates was previously proposed [43]. In neurons, MCU overexpression increases mitochondrial Ca²⁺ transient and NMDA-mediated excitotoxicity; in addition, synaptic activity repressed MCU transcription [13], thus suggesting that MCU level can be coupled with cellular demands. Along this line, MCU has also been shown to be target of transcriptional control through miR-25 [44]. In breast epithelial cells, MCU appears to regulate apoptosis in normal but not in transformed cells [45]. Therefore, mitochondrial Ca²⁺ uptake appeared to be a pleiotropic signal potentially regulating many aspects of organism physiology. In line with this, genetic manipulation of MCU in lower organisms resulted in major developmental and energetic defects, as demonstrated in zebrafish [46] and in Trypanosome brucei [47]. However, the recent publication of the first MCU knockout mouse model showed shocking results for any scientist in this field. Indeed, genetic ablation of MCU resulted in a very mild phenotype, i.e. normal mice with a minor defect in muscle strength after endurance training [14]. However, it must be stressed that MCU knockout in a pure C57/BL/6 inbred mouse strain results in embryonic lethality, as one would expect. In addition, although viable animal could be obtained in the CD1 background, the birth ratio was approximately the half of the expected [48]. Detailed and exhaustive discussions on this model have already been carried out elsewhere [49,50]. We think that some kind of adaptation must take place in these animals, most likely in some still unexplored signaling routes activated downstream to mitochondria calcium homeostasis. Conditional and inducible knockout models, as well as viral-based gene-delivery systems will be needed to rule out this possibility and to conclusively assess the real physiological impact of mitochondrial calcium homeostasis. In support of this view, two recent papers highlight the patho-physiological relevalnce of the impairment of mitochondrial Ca²⁺ uptake. On one hand, a family carrying a loss-of-function mutation of MICU1 has been identified. Homozygous individuals for this mutation are characterized by proximal myopathy, learning difficulties and a progressive extrapyramidal movement disorder, thus underlining the requirement of a proper MCU gating for multiple functions [51]. On the other hand, manipulation of MCU levels after birth (by using AAV-mediated gene transfer) clearly demonstrated that mitochondrial Ca²⁺ homeostasis contributes to regulate skeletal muscle tropism. Indeed, MCU overexpression and downregulation causes muscular hypertrophy and atrophy, respectively. This effect is apparently independent of the control of aerobic metabolism but rather relies on two major hypertrophic pathways of skeletal muscle, PGC-1α4 and IGF1-Akt/PKB [52]. These results also uncovers the existence of a Ca²⁺-dependent mitochondria-to-nucleus signaling route that links organelle physiology to the control of muscle mass.

Overall, the recent advances in the molecular control of mitochondrial Ca²⁺ uptake revealed a complex but fascinating picture that needs further examination in order to lead to novel understandings of major patho-physiological impact.

Highlights.

• MCU is the main Ca²⁺ channel of the inner mitochondrial membrane

• MCU is part of a macromolecular complex, called the MCU complex

• MCU is gated by calcium, through a mechanism involving MICU1 and MICU2

• The physiological role of mitochondrial Ca²⁺ uptake still needs to be elucidated