Regulation of mitochondrial calcium in plants versus animals

Abstract

Ca²⁺ acts as an important cellular second messenger in eukaryotes. In both plants and animals, a wide variety of environmental and developmental stimuli trigger Ca²⁺ transients of a specific signature that can modulate gene expression and metabolism. In animals, mitochondrial energy metabolism has long been considered a hotspot of Ca²⁺ regulation, with a range of pathophysiology linked to altered Ca²⁺ control. Recently, several molecular players involved in mitochondrial Ca²⁺ signalling have been identified, including those of the mitochondrial Ca²⁺ uniporter. Despite strong evidence for sophisticated Ca²⁺ regulation in plant mitochondria, the picture has remained much less clear. This is currently changing aided by live imaging and genetic approaches which allow dissection of subcellular Ca²⁺ dynamics and identification of the proteins involved. We provide an update on our current understanding in the regulation of mitochondrial Ca²⁺ and signalling by comparing work in plants and animals. The significance of mitochondrial Ca²⁺ control is discussed in the light of the specific metabolic and energetic needs of plant and animal cells.

Bioenergetics, Ca²⁺, ion channels, ion transporters, mitochondria.

Introduction

Plant and animal cells share a need to adjust their physiology rapidly. This is particularly important to deal with changes in their environment, but also to support developmental programmes. Intracellular Ca²⁺ transients are an essential and universal signalling mechanism for mediating physiological flexibility in both the short and the long term, for instance in muscle contraction or plant pathogen defence (Escobar et al., 1994; Cannell et al., 1995; Thor and Peiter, 2014; Keinath et al., 2015). Animal and plant cells maintain free cytosolic Ca²⁺ at much lower concentrations than most other intracellular inorganic ions, such as K⁺, Cl⁻, and Mg²⁺. Active extrusion of Ca²⁺ from the cytosol to the extracellular space, the endoplasmic reticulum (ER) lumen, and the vacuole is necessary due to the chemical property of Ca²⁺ to bind anionic cellular compounds, such as organic carboxylates, phosphates, DNA, and RNA, and to interfere with their function. This results in steep Ca²⁺ gradients across the plasma membrane, the ER membrane, and the tonoplast membrane. Since those membranes are equipped with selective transport systems for Ca²⁺, rapid and well-defined changes in intracellular concentration can be evoked through activation and inhibition of the transporters. Changes in Ca²⁺ concentration can then regulate cellular processes through hundreds of cellular proteins that change their function in response to Ca²⁺, for example by binding Ca²⁺.

Mitochondria act as intracellular conductors of intracellular Ca²⁺ regulation, shaping, remodelling, relaying, and decoding Ca²⁺ signals, due their ability to accumulate Ca²⁺ rapidly and transiently (Thayer and Miller, 1990; Friel and Tsien, 1994; Drago et al., 2012). In the cellular response to environmental and endogenous stimuli, mitochondria play an integral part that goes beyond acting as passive supporters by providing the ATP required for cellular readjustment. Instead they take an active role in Ca²⁺ regulation and signalling, controlling central life processes within the organelles themselves as well as the entire cell (Chalmers et al., 2007; Colombatti et al., 2014). Despite the great interest in identifying the molecular players of the mitochondrial Ca²⁺-handling machinery, significant advances have been achieved only during the last decade. This is currently opening new doors towards a mechanistic understanding of organellar Ca²⁺ signalling. Although the plant community has been at the forefront of the study of the regulation of mitochondrial Ca²⁺, the most recent burst of interest, which was sparked by the identification of the molecular components of the mitochondrial Ca²⁺ uniporter, has predominantly involved animal systems. Here we review and contrast the current insights into the regulation of mitochondrial Ca²⁺ in plants and animals side by side, to distil general principles and specific differences, and to sketch out a conceptual picture of the physiological relationship between mitochondria and Ca²⁺ in plants and animals.

Ca²⁺ import into mitochondria

Outer mitochondrial membrane

Similar to other small molecules, Ca²⁺ is thought to pass the outer mitochondrial membrane (OMM) freely through VDACs (voltage-dependent anion channels, also called porins; Fig. 1). VDACs allow flux of metabolites and ions including Ca²⁺, for which mammalian VDAC1 also possesses binding sites, as demonstrated both in vitro and in vivo (Gincel et al., 2001; Rapizzi et al., 2002; Báthori et al., 2006; Israelson et al., 2007; Rizzuto et al., 2009; Shoshan-Barmatz et al., 2010; De Stefani et al., 2012). In mammals three and in Arabidopsis four functionally distinct protein isoforms have been found in the OMM (for recent reviews, see Shoshan-Barmatz et al., 2010; Szabo and Zoratti, 2014; Takahashi and Tateda, 2013). Direct electrophysiological or genetic evidence for Ca²⁺ uptake into mitochondria through plant VDACs is still missing, however, even though several studies described the electrophysiological properties of plant VDACs. Interestingly, yeast two-hybrid assays have suggested that VDAC1 from Arabidopsis (AtVDAC1) interacts with the EF-hand Ca²⁺-sensor protein CBL1 (Li et al., 2013).

Fig. 1.

Inventories of mitochondrial Ca²⁺ transport in animals and plants. (A) Mammalian proteins and protein complexes at both mitochondrial membranes and their proposed impact on import or export of Ca²⁺. Dotted grey arrows represent effects on Ca²⁺ transport which may be either direct or indirect. (B) Plant candidate proteins for the modulation of mitochondrial Ca²⁺ as hypothesized from their presence in plants and the proposed function of their mammalian homologues. Experimental evidence for an involvement of the plant proteins in handling mitochondrial Ca²⁺ is currently lacking, except for MCUC and GLR3.5. See main text for a detailed discussion. (This figure is available in colour at JXB online.)

Inner mitochondrial membrane

History of Ca²⁺ uptake

The inner mitochondrial membrane (IMM) is tightly sealed for Ca²⁺, and passage strictly requires channels/transporters. Uptake into plant mitochondria has been studied for >50 years after Hodges and Hanson (1965) observed Ca²⁺ accumulation by corn mitochondria (Fig. 2A). Since then, studies using isolated mitochondria from different plant species and tissues have generated a complex and, in parts, contradictory picture. While most mitochondrial preparations take up Ca²⁺ (Dieter and Marmé, 1980; Akerman and Moore, 1983), others do not (Moore and Bonner, 1977; Martins and Vercesi, 1985). Uptake strictly requires energization and does not take place in the presence of respiratory chain inhibitors such as antimycin A, KCN, and NaN₃ (Dieter and Marmé, 1980). Ca²⁺ import in most (Hodges and Hanson, 1965; Chen and Lehninger, 1973) but not all (Zottini and Zannoni, 1993) cases requires inorganic phosphate (P_i), and this has been interpreted as symport of Ca²⁺ with P_i (Day et al., 1978; Silva et al., 1992) or as a consequence of Ca–phosphate precipitates in the mitochondrial matrix that decrease free matrix Ca²⁺ and lead to a continuous drain from the extramitochondrial space (Akerman and Moore, 1983). Ruthenium red, an inhibitor of a general inhibitor of calcium-permeable ion channels, appears to block transport in some (Dieter and Marmé, 1980) but not all (Akerman and Moore, 1983) instances. Conflicting findings may result from studying Ca²⁺ uptake outside living cells with limited means to quantify free Ca²⁺ inside mitochondria (see ‘Measuring and sensing of mitochondrial Ca²⁺’ below) but could also suggest that different uptake systems are present depending on cell and tissue type. Although suffering from similar technical constraints, early Ca²⁺ uptake studies with mammalian mitochondria provided a clearer picture. First studied in the 1960s (Fig. 2A), transport required respiration (DeLuca and Engstrom, 1961; Vasington and Murphy, 1962) and was accompanied by P_i transport (Greenawalt et al., 1964). In light of Peter Mitchell’s chemiosmotic hypothesis (Mitchell, 1961), the underlying transporter was proposed to be an electrophoretic Ca²⁺ uniporter that does not require ATP hydrolysis but makes use of the steep electrochemical gradient across the IMM (Rottenberg and Scarpa, 1974). Chemiosmotic coupling also necessitates high selectivity of any mitochondrial Ca²⁺ transporter preventing energy dissipation by uncontrolled H⁺ influx. Once mitochondrial Ca²⁺ uptake could be monitored directly in intact mammalian cells (see ‘Measuring and sensing of mitochondrial Ca²⁺’ below), it became evident that free matrix Ca²⁺ could transiently reach high micromolar concentrations in specific cell types (Montero et al., 2000) and that the speed and amplitude of Ca²⁺ uptake exceeded the values that had been predicted from classical bioenergetic experiments in isolated mitochondria. Subsequent work in mammalian cells suggested an interaction of mitochondria with microdomains of high Ca²⁺ concentrations (Fig. 2A) generated by localized release from the ER and the extracellular space, allowing highly efficient uptake (Rizzuto et al., 2012).

Fig. 2.

The composition of the mitochondrial calcium uniporter complex in animals and plants. (A) Number of publications per year as listed in the PubMed literature database (http://www.ncbi.nlm.nih.gov/pubmed) as queried for ‘mitochondrial calcium uptake’. (B) Presence of MCUC core components (MCU, MCUb, MICU, and EMRE) and CCX family proteins with homology to mammalian NCLX in key eukaryotic model organisms. The taxonomic relationship of organisms is indicated by a schematic phylogenetic tree. (C) MCUC core components as identified in human/mouse as compared with their plant homologues based on their presence or absence in the Arabidopsis genome. The presence of Ca²⁺-binding EF-hand motifs is schematically indicated in the MICU proteins. (This figure is available in colour at JXB online.)

Key properties of the Ca²⁺ uniporter and its identification

The functional characteristics of the uniporter have since been investigated in fine detail. A membrane potential of –180 mV (negative inside) generated by the respiratory chain would theoretically lead to a 1 000 000-fold accumulation of matrix Ca²⁺ if electrophoretic Ca²⁺ passage was unrestricted. Accordingly, protonophores such as CCCP (carbonyl cyanide m-chlorophenyl hydrazone) not only trigger collapse of the membrane potential, but also inhibit Ca²⁺ transport (Selwyn et al., 1970). Ca²⁺ uptake into mammalian mitochondria is additionally blocked by low concentrations of ruthenium red and Ru360, which lead to a direct inhibition of the uniporter (Moore, 1971; Vasington et al., 1972; Reed and Bygrave, 1974). The finding that a highly Ca²⁺-selective ion channel, displaying a very small conductance of only 5 pS in 100mM Ca²⁺in vitro recapitulated the key characteristics observed for the mammalian mitochondrial uniporter in classical bioenergetic experiments, represented a milestone toward the molecular identification of the uniporter (Kirichok et al., 2004; Figs 1, 2). In a next step, a regulatory protein, mitochondrial calcium uptake 1 (MICU1), was identified by a combination of comparative physiology, evolutionary genomics, and organelle proteomics (Perocchi et al., 2010). Instrumental for this approach was the MitoCarta database, containing >1000 mitochondrial proteins as identified by subtractive proteomics and green fluorescent protein (GFP) fusion localization studies (Pagliarini et al., 2008). This delivered the basis for the identification of several mitochondrial calcium uniporter complex (MCUC) components in mammals, including the central pore-forming protein MCU (mitochondrial calcium uniporter; Baughman et al., 2011; De Stefani et al., 2011). At the current stage, the mammalian MCUC appears to consist of at least the pore-forming protein MCU, an MCU paralogue (MCUb), the essential MCU regulator (EMRE), the regulatory MICU proteins, and, possibly, the mitochondrial calcium uniport regulator 1 (MCUR1; Fig. 2B, C).

Multiple proteins constitute and regulate the Ca²⁺ uniporter

MCU and MCUb

MCU is a 40kDa protein that is inserted into the IMM via two transmembrane domains, and oligomerizes into tetramers to form a pore that allows Ca²⁺ entry into the mitochondrial matrix driven by the electrical membrane gradient (Baughman et al., 2011; De Stefani et al., 2011; Raffaello et al., 2013; Fig. 2C). Recombinant MCU protein, when incorporated into an artificial membrane, mediates Ca²⁺-permeable activity, resembling the electrophysiological characteristics of the mitochondrial uniporter (Kirichok et al., 2004; De Stefani et al., 2011). Mammalian MCU activity can be regulated through its paralogue MCUb. MCU and MCUb share 50% sequence similarity, and both proteins physically interact. MCUb carries two conserved amino acid exchanges in the intermembrane space (IMS)-exposed loop of MCU which is necessary to permit Ca²⁺ transport through MCU in lipid bilayer experiments (Raffaello et al., 2013). In cultured cells, MCUb forms hetero-oligomers with MCU (Fig. 2C) and constitutes a dominant-negative regulator of MCU transport activity.

Homologues of MCU were identified in genomes of several plant species, including maize and Arabidopsis where six homologues are present in each (Stael et al., 2012; Meng et al., 2015). The first proteomic evidence from Arabidopsis and potato suggests the presence of specific MCU homologues in mitochondrial fractions at low relative abundance, which may be expected for an organellar ion channel (Wagner et al., 2015a). Prediction algorithms such as TargetP (Emanuelsson et al., 2000) assign MCU proteins a high likelihood of mitochondrial targeting across plant species. Nevertheless, chloroplast localization reaches high prediction scores in several instances. This may be due to general similarities between mitochondrial and chloroplast targeting peptides, but could also have biological meaning, emphasizing the need for further experimental validation (Emanuelsson et al., 2007; Briesemeister et al., 2010). The diversification of MCU genes in plants may thus provide regulatory flexibility on the different levels of gene expression, including transcription, translation, and post-translational organization. This is supported by differential expression of MCU genes in Arabidopsis and maize tissues (Stael et al., 2012; Meng et al., 2015). It appears tempting to speculate about hetero-oligomerization of different plant MCUs to form pores of different Ca²⁺ transport efficiency by analogy with the mammalian situation, where the MCU current varies between different tissues possibly due to the differential expression of MCU and MCUb (Fieni et al., 2012) and/or of regulatory components. Yet empirical evidence for plant MCU proteins to act as functional channels is still missing.

MICU

The mammalian MICU protein family consists of three members that share >40% sequence identity (Fig. 2B). MICU1, the first uniporter component identified (Perocchi et al., 2010), is a 50kDa protein with two functional and two pseudo EF-hands and resides in the mitochondrial IMS (Csordás et al., 2013; Hung et al., 2014; Patron et al., 2014; Wang et al., 2014; Petrungaro et al., 2015). It was soon referred to as the uniporter ‘gatekeeper’ that sets a threshold for mitochondrial Ca²⁺ uptake through MCU at low extramitochondrial Ca²⁺ concentrations but activates the channel when surrounding Ca²⁺ concentrations are high (Mallilankaraman et al., 2012b; Csordás et al., 2013). Recent evidence that elevations in cytosolic Ca²⁺ are sufficient (EC₅₀ of 4.4 μM) to induce rearrangement of MICU1 multimers and to trigger activation of mitochondrial Ca²⁺ uptake are in agreement with this concept (Waldeck-Weiermair et al., 2015). The initial model for MICU function was further refined after the identification of two additional MICU isoforms, MICU2 and MICU3 (Plovanich et al., 2013). As MICU3 was found to be almost exclusively expressed in neural tissues (Plovanich et al., 2013), functional characterization focused on ubiquitously expressed MICU2. MICU2 forms a heterodimer with MICU1 through an intermolecular disulphide bond and closes the channel at low extramitochondrial Ca²⁺ concentrations (Patron et al., 2014; Petrungaro et al., 2015). The stability of MICU2 depends on MICU1 (Plovanich et al., 2013; Patron et al., 2014), and loss of MICU2 in MICU1-silenced cells complicates assignment of individual MICU1 and MICU2 functions. However, in electrophysiological experiments, MICU2 inhibits the channel activity, while MICU1 does the opposite in the presence of Ca²⁺, in accordance with the proposed model of MICU1 and MICU2 being activator and gatekeeper, respectively (Patron et al., 2014). Currently, two models co-exist that find MICU1 (i) to act as a uniporter activator at high cytosolic Ca²⁺ concentrations (Patron et al., 2014) or (ii) to disinhibit the uniporter gradually with increasing Ca²⁺ concentrations in the cytosol (Csordás et al., 2013). MICU is conserved in plants, where typically one or two homologues can be found depending on the species (Wagner et al., 2015a). Arabidopsis possesses only a single MICU gene (Fig. 2B, C), and knockout strongly affects mitochondrial Ca²⁺ dynamics, providing molecular evidence for a functional uniporter system in plants (Wagner et al., 2015a; Fig. 1). Arabidopsis MICU contains an additional, third canonical EF-hand motif, which is conserved amongst plants and protists but absent in mammalian MICU, and may open up additional degrees of freedom for Ca²⁺ regulation of MCUC activity. Interestingly, one of those three EF-hands is absent in a second splicing variant of Arabidopsis MICU, which is expressed at much lower abundance, and may thereby add to MICU-based fine regulation in plants. Mitochondrial Ca²⁺ sensing in living roots of micu knockout plants has suggested an inhibitory, rather than an activation, effect of Arabidopsis MICU on plant mitochondrial Ca²⁺ uptake, which implies that it represents a functional homologue of mammalian MICU2.

EMRE

Another core component of the mammalian MCUC is EMRE, a 10kDa protein that spans the IMM with a single transmembrane motif. EMRE has been proposed to bridge MCU and its regulators MICU1/2 and to be indispensable for the activity of the mammalian uniporter in vivo (Sancak et al., 2013), although MCU alone is sufficient to form a functional channel in vitro (De Stefani et al., 2011). EMRE is metazoan specific and its essential role is supported by reconstitution experiments in budding yeast that lacks an endogenous mitochondrial Ca²⁺ uniporter: while expression of MCU from the slime mould Dictyostelium alone was sufficient to import Ca²⁺ into yeast mitochondria, human EMRE needed to be expressed alongside mammalian MCU to form an active Ca²⁺ uniporter system (Kovács-Bogdán et al., 2014). Recent evidence suggests that the C-terminus of EMRE can sense Ca²⁺ on the matrix side of the IMM to regulate Ca²⁺ uniport negatively (Vais et al., 2016). Acting in concert with MICU, this may give rise to a sophisticated sensing module that integrates information on Ca²⁺ concentration from both sides of the IMM to avoid both Ca²⁺ depletion and overload. Similar to Dictyostelium, plants possess a minimal genetic uniporter configuration that lacks MCUb and EMRE (Wagner et al., 2015a; Fig. 2B, C).

MCUR

Not considered core components of the MCUC, other IMM proteins have been proposed to regulate uniport activity. MCUR1 (mitochondrial calcium uniporter regulator 1)/CCDC90A is a 39kDa protein with two predicted transmembrane domains that is thought to interact with MCU (Mallilankaraman et al., 2012a), although later studies were unable to find support for this interaction (Sancak et al., 2013; Paupe et al., 2015). Paupe et al. (2015) provided evidence that MCUR1 is in fact an assembly factor of cytochrome c oxidase and argued that genetic manipulation modulates mitochondrial membrane potential, imposing only a secondary effect on Ca²⁺ transport. In support of that, MCUR1 has an orthologue in budding yeast which lacks core MCUC components. Although Vais et al. (2015) recently showed that MCUR1 affects MCU activity in patch-clamp experiments, direct regulation of Ca²⁺ uniport through MCUR1 is still debated. Arabidopsis possesses two MCUR1 homologues that lack functional characterization. Interestingly one of them has been identified as a plant-specific subunit of complex IV by proteome analysis (Millar et al., 2004; Klodmann et al., 2011).

Additional components contribute to Ca²⁺ import

APCs

Small Ca²⁺-binding mitochondrial carrier protein 3 (SCaMC3 or SLC25A23) is an EF-hand-containing protein that belongs to the family of mitochondrial carriers. Ca²⁺-binding mitochondrial carriers (CaMCs) are further subdivided into two classes: aspartate/glutamate carriers (AGCs) and ATP/P_i carriers (APCs/SCaMCs/SLCs; Del Arco et al., 2000; Del Arco and Satrustegui, 2004; Satrustegui et al., 2007). SCaMC3 has been shown to reduce mitochondrial Ca²⁺ uptake upon knockdown in cultured mammalian cells (Hoffman et al., 2014; Fig. 1). The same was not observed for its paralogues SCaMC1 and 2 (SLC25A24 and SLC25A25, respectively). As P_i in the mitochondrial matrix is critical for free Ca²⁺ buffering, it is not fully resolved whether this is a direct or indirect effect on uniporter activity (Seifert et al., 2015). Similar to MCUR1, independent studies for SCaMC3 and MCU interaction in different cell lines have delivered contradictory results (Sancak et al., 2013; Hoffman et al., 2014). The Arabidopsis genome codes for three SCaMC homologues, ATP/phosphate carriers (APC) 1–3 that all reside in mitochondria and bind Ca²⁺ (Stael et al., 2011). Reconstituted in liposomes, they transport phosphate and adenosine nucleotides, and are regulated by Ca²⁺ (Monné et al., 2015). Intriguingly, Arabidopsis APC2 has recently been shown to transport ATP-Ca instead of ATP-Mg in vitro (Lorenz et al., 2015). Considering the physiological baseline concentrations of free Ca²⁺ and Mg²⁺ in the plant cytosol (100nM free Ca²⁺ versus 200–250 µM free Mg²⁺; Igamberdiev and Kleczkowski, 2001; Logan and Knight, 2003; Gout et al., 2014), it remains questionable whether this ATP-Ca²⁺ transport can also take place in the living plant.

GLR3.5

Another recent study found a member of the glutamate receptor family, AtGLR3.5, in the mitochondria of Arabidopsis (Fig. 1). Although there is currently no direct evidence indicating that the subfamily 3 member AtGLR3.5 functions as a Ca²⁺-permeable ion channel, a close homologue, AtGLR3.4, as well as AtGLR1.4 and AtGLR1.1 behave as Ca²⁺-permeable cation channels when expressed in heterologous systems (Tapken and Hollmann, 2008; Vincill et al., 2012; Tapken et al., 2013). AtGLR1.4 was found to be permeable to Ca²⁺ in a physiological concentration range even in the presence of a physiological concentration of K⁺. Whether these channels preferentially permit the flux of Ca²⁺ over Na⁺ and K⁺in vivo is still under investigation, but studies employing knockout plants lacking some members of the subfamily 3 indicate that glutamate-induced Ca²⁺ uptake correlates with the presence of the channel (Qi et al., 2006). In vivo measurements of mitochondrial Ca²⁺ dynamics in plants lacking AtGLR3.5 indicated a contribution of the protein to Ca²⁺ uptake upon wounding, which may be direct or indirect (Teardo et al., 2015). Absence of more pronounced alterations of mitochondrial Ca²⁺ dynamics may be attributed to redundancy in mitochondrial Ca²⁺ uptake processes. Plants lacking AtGLR3.5 harbour mitochondria with a strongly altered ultrastructure. Increased AtGLR3.5 transcript abundance in older leaves together with an early senescence phenotype in mutant plants makes it tempting to speculate about a developmental stage-specific role for the putative Ca²⁺ transport activity of the protein (Teardo et al., 2015). Such a hypothesis is consistent with the observation that activation of a related mitochondrially localized NMDA (N-methyl-d-aspartate) inotropic glutamate receptor increased the matrix Ca²⁺ level in mammalian neurons (Korde and Maragos, 2012).

UCPs

Preceding the molecular identification of MCUC components, additional Ca²⁺ uptake mechanisms in mammalian mitochondria were proposed. Uncoupling proteins 2 and 3 (UCP2/3) have been deemed essential components of mitochondrial Ca²⁺ uniport (Trenker et al., 2007; Fig. 1). This view was challenged (Brookes et al., 2008), and indirect effects on Ca²⁺ uptake into mitochondria have been proposed (De Marchi et al., 2011). Despite a convincing case against a direct role for UCPs as Ca²⁺ transporters, the discussion of UCP2/3 function is still ongoing (Bondarenko et al., 2015). Although an MCU homologue of tomato is up-regulated when the plants overexpress an Arabidopsis UCP protein (Barreto et al., 2014), potential functional interplay has not yet been investigated in plants.

Identified uniporter components shed light on Ca²⁺ uptake modes

Other mitochondrial Ca²⁺ uptake modes [e.g. Ca²⁺-selective conductance (mCa) 2 and rapid mode of uptake (RaM)] that have been observed in animals have currently no matching molecular identities. These uptake modes were proposed to differ from MCUC-mediated Ca²⁺ uptake in terms of Ca²⁺ affinity, uptake kinetics, and pharmacology (Sparagna et al., 1995; Michels et al., 2009). Although the latter report remains controversial, it is tempting to interpret those observations in the light of the molecular complexity of the MCUC that has been emerging since. Different Ca²⁺ uptake modes may be accommodated by MCUC existing and operating in different functional states set by MCU–MCUb stoichiometry, MICU regulation, and other interacting proteins. MCU knockdown efficiently abolishes Ca²⁺ transients in mammalian cell culture (Bondarenko et al., 2014; Baughman et al., 2011; De Stefani et al., 2011), indicating that the MCUC can have a dominating role amongst uptake mechanisms. In accordance with this, Ca²⁺ uptake into mitochondria was almost completely abolished in the liver of mcu animals (Pan et al., 2013). On the other hand, this does not rule out the possibility that other mechanisms make major contributions to Ca²⁺ uptake, particularly in specialized tissues. Potential candidates include the TRPC3 channel (Feng et al., 2013; L. Wang et al., 2015) and the mitochondrial ryanodine receptor (mRyR1). A low level of RyR1 is detectable in heart mitochondria and provides rapid transport of Ca²⁺ that is insensitive to ruthenium red (Beutner et al., 2001, 2005).

Both TRPC3 and RyR1 have no obvious homologues in plants (Fig. 1). The availability of several animal model systems in which MCU is genetically knocked out should help to test the hypothesis of MCUC being responsible for different uptake modes and clarify the presence and kinetics of co-existing uptake mechanisms. In plants, a similar rationale is currently hampered by multiple MCU homologues with unclear and possibly redundant function.

Matrix Ca²⁺ buffering

Once inside the mitochondrial matrix, Ca²⁺ predominantly exists as insoluble Ca–phosphate precipitate but is also bound to proteins and inorganic acids. This sequestration allows isolated mitochondria to accumulate large amounts of total Ca²⁺ up to a concentration of 1M, with bound Ca²⁺ exceeding free Ca²⁺ by 150 000-fold (Chalmers and Nicholls, 2003). In neurons in the resting state, the ratio between the bound and free form reaches values of ~4000 in the mitochondrial matrix compared with values of ~100 in the cytosol (Neher, 1995; Babcock et al., 1997). Yet, concentrations of free Ca²⁺ in the resting state are similar between the cytosol and the mitochondrial matrix {animals: [Ca²⁺]_m and [Ca²⁺]_c=100–200nM (Rizzuto et al., 1992; Babcock et al., 1997); Arabidopsis: [Ca²⁺]_m=200nM, [Ca²⁺]_c=100nM (Logan and Knight, 2003)}. The exact chemical states of bound Ca²⁺ inside the matrix of the living cell and the relative contributions of proteins, metabolites, and P_i are largely unclear in both plants and animals.

Mitochondrial Ca²⁺ export

Described as the mitochondrial ‘Ca²⁺ cycle’ (Carafoli, 1979), Ca²⁺ can be extruded from mitochondria by an antiport mechanism, to regulate matrix Ca²⁺ concentrations and to avoid overload, which can be deleterious for mitochondrial function (see below when discussing PTP). In the late 1970s, two Ca²⁺ export systems were discussed: a Na⁺/Ca²⁺ exchanger (Crompton et al., 1977, 1978) and a H⁺/Ca²⁺ exchanger (Akerman, 1978; Fiskum and Lehninger, 1979).

NCLX exports Ca²⁺ in exchange for Na⁺

While the molecular identity of the latter remains unclear, the mammalian protein NCLX (Na⁺/Ca²⁺/Li⁺ exchanger; Palty et al., 2010) has been proposed as the underlying molecular entity of electrogenic transport of one Ca²⁺ against three Na⁺ (Fig. 1). De Marchi et al. (2014) have recently made a strong case for NCLX to represent the long-sought mediator of Ca²⁺ export from the mitochondrial matrix. Arabidopsis possesses five homologues that belong to the cation/Ca²⁺ exchanger (CCX) family (Fig. 2B; Emery et al., 2012). However, these proteins reach lower prediction scores for mitochondrial targeting than for localization in the secretory pathway [based on The SubCellular Proteomic Database SUBA3 (Tanz et al., 2013) and Aramemnon (Schwacke et al., 2003)]. In agreement with this, a GFP fusion of CCX3 was found in the endomembrane system, where it was suggested to mediate H⁺/K⁺ exchange (Morris et al., 2008). On a physiological level, the involvement of Na⁺ raises further questions about a corresponding antiport situation in plants, for which, in contrast to animals, Na⁺ is not essential (Blumwald, 2000).

LETM proteins as exporters for matrix Ca²⁺?

Mammalian LETM1 was thought to act as an a K⁺/H⁺ exchanger (Nowikovsky et al., 2004; Dimmer et al., 2008) before a genome-wide RNAi screen for proteins mediating mitochondrial Ca²⁺ dynamics identified LETM1 as a Ca²⁺/H⁺ antiporter with Ca²⁺ affinity in the physiologically meaningful range (~200nM; Fig. 1; Jiang et al., 2009). Follow-up work pointed to a function in mitochondrial Ca²⁺ uptake or export dependent on the balance between intra- and extramitochondrial Ca²⁺ concentrations (Waldeck-Weiermair et al., 2011; Doonan et al., 2014; Tsai et al., 2014), in agreement with electroneutral antiporter activity found in vitro (Tsai et al., 2014). This model has been challenged by Nowikovsky and Bernardi (2014) who put forward strong arguments in favour of K⁺/H⁺ exchange through LETM1: budding yeast mitochondria lacking a uniporter complex for rapid Ca²⁺ uptake possess a LETM1 homologue (Mdm38; Nowikovsky et al., 2004). Inactivation of Mdm38 leads to mitochondrial swelling, which points to abnormal accumulation of K⁺ in the mitochondrial matrix (Lodish et al., 2000; Rodriguez-Navarro, 2000). This phenomenon is also associated with LETM1-like proteins from other species (Hasegawa and van der Bliek, 2007; McQuibban et al., 2010; Hashimi et al., 2013) and can be reverted by the ionophore nigericin that specifically mediates K⁺/H⁺ exchange (Nowikovsky et al., 2004). Intriguingly, the yeast Mdm38 lacks Ca²⁺-binding EF-hands, and a dual function for LETM1-like proteins in antiport of H⁺ against both K⁺ and Ca²⁺ cannot be fully excluded. Recent results suggest that both mitochondrial Ca²⁺ influx and efflux rates are impaired in LETM1 knockdown that did not affect the expression level of MCUC proteins. Expression of the ΔEF-hand LETM1 mutant largely prevented Ca²⁺ uptake (Doonan et al., 2014). Yet, such observations generally need to be interpreted with a systems view on mitochondrial physiology, considering indirect effects of LETM1 removal on Ca²⁺ homeostasis.

Ca²⁺ uptake by electroneutral exchange for protons, as observed in vitro (Ca²⁺ in; 2 H⁺ out; Tsai et al., 2014), is thermodynamically implausible in an actively respiring mitochondrion considering a proton gradient of up to 10-fold (i.e. 1 pH unit) as part of the inner membrane electrochemical gradient. Only at very high cytosolic/IMS free Ca²⁺ concentrations could the exchanger mechanism allow uptake. A conclusive physiological case is currently missing and the thermodynamic argument appears striking enough for all genetic or biochemical evidence to be interpreted in its light. In contrast, a H⁺-driven Ca²⁺ export function for LETM1 is thermodynamically plausible, which is a necessary but not a sufficient argument for export to be mediated by LETM1 under physiological conditions. Indeed a recent study has provided genetic evidence against such a role by showing that overexpression of LETM1 did not increase Ca²⁺ export rates, while overexpression of NCLX did (De Marchi et al., 2014). Although some of the proposed activities appear unlikely in vivo, the mechanism accounting for the impact that LETM1 has on mitochondrial Ca²⁺ homeostasis remains unknown.

The Arabidopsis genome contains two genes with homology to LETM1 (Fig. 1), and a double knockout is not viable (Zhang et al., 2012). Both Arabidopsis proteins, LETM1 and LETM2, reside in the IMM, contain EF-hands, but lack, like yeast Mdm38, the leucine zipper domain of animal LETM1-like proteins (Zhang et al., 2012). Partial depletion of LETM in a letm1-1(−/−) LETM2-1(+/−) line does not compromise mitochondrial morphology but rather mitochondrial protein translation (Zhang et al., 2012). Such an effect is also associated with absence of LETM1 in yeast (Frazier et al., 2006; Bauerschmitt et al., 2010) where dysfunctional mitochondrial translation was proposed to be a secondary effect of disrupted K⁺ homeostasis (Hashimi et al., 2013), based on the observation that nigericin rescued the translation phenotype in cultured cells.

Opening of the mitochondrial permeability transition pore for extrusion of matrix Ca²⁺?

Transient opening of the mitochondrial permeability transition pore (PTP) has been proposed to cause release of Ca²⁺ from mammalian mitochondria (Bernardi and von Stockum, 2012; Fig. 1). Such a mechanism appears attractive to counteract matrix Ca²⁺ overload. Under specific conditions, plant mitochondria have also been observed to undergo permeability transition (Arpagaus et al., 2002; Petrussa et al., 2004; Vianello et al., 2012; for a recent review, see Zancani et al., 2015). Yet, the physiological consequences that a Ca²⁺ release function of the PTP implicate appear drastic. Extrusion of Ca²⁺ through a PTP-like pore would need to rely on Ca²⁺ outflow that is thermodynamically plausible; that is, the gradient between free Ca²⁺ levels in the matrix and the cytosol/IMS would need to exceed the electrical potential. The electrical potential can be expected to be, at least partially, dissipated in the first place by the transient PTP opening, leaving the gradient of free Ca²⁺ as a main driver. On partial loss of membrane potential only a large Ca²⁺ gradient, as expected at Ca²⁺ overload, would allow Ca²⁺ extrusion. Yet, active export coupled to the electrochemical gradient may be more effective and offer better control. More importantly, a partially or fully dissipated electrochemical gradient would not only allow Ca²⁺ extrusion, but would also severely interfere with matrix physiology, including ATP/ADP exchange, P_i uptake, metabolite shuttling, and also Ca²⁺ extrusion via the NCLX, which strictly depend on the proton motive force. Transient variations in membrane potential that have been observed in both plants and animal cells and occasionally been interpreted as transient PTP opening have been shown to coincide with an increase in the pH gradient, implying that the underlying mechanism does not involve an unselective pore and that the proton motive force overall remains intact during the transients (Schwarzländer et al., 2012a, b; Santo-Domingo et al., 2013). The question of whether the drastic situation of opening a large unspecific pore, that impacts severely, albeit transiently, on the characteristic physiological makeup of the mitochondrion, can fulfil a physiological housekeeping function like Ca²⁺ export, or rather is reserved for extreme pathological situations, remains to be thoroughly tested and validated.

Measuring and sensing of mitochondrial Ca²⁺

Ca²⁺ dyes

Our understanding of mitochondrial Ca²⁺ dynamics relies on the development and optimization of Ca²⁺-sensing tools. Early uptake studies relied on radiolabelled ⁴⁵Ca²⁺ (DeLuca and Engstrom, 1961), but deduction of kinetic parameters came with technical pitfalls (Borle, 1981). The first Ca²⁺ dyes used in the 1960s and 1970s, such as murexide (Mela and Chance, 1968), partially overcame this issue, but many of them lacked the properties to quantify Ca²⁺ specifically. A more sophisticated generation of Ca²⁺ dyes was introduced in the late 1970s (Tsien, 1980). These 1,2-bis(o-aminophenoxy)ethane-N,N,N',N'-tetraacetic acid (BAPTA)-related dyes were developed to be membrane permeable and trappable in cells (Tsien, 1981). However, only few chemical dyes are cell compartment specific, limiting their applicability for in vivo Ca²⁺ measurements in mitochondria. Rhod-2 that accumulates in the mitochondrial matrix constitutes an important exception and has been extensively used.

Genetically encoded Ca²⁺ sensors

Compartment-specific studies across species (Davies and Terhzaz, 2009; Laude and Simpson, 2009; Stael et al., 2012) largely benefited from the development of genetically encoded, protein-based Ca²⁺ probes. The luminescent protein aequorin binds Ca²⁺ via EF-hands and undergoes an irreversible conformation change upon Ca²⁺ binding triggering the emission of a photon (Johnson and Shimomura, 1972). While the protein initially had to be isolated from the jellyfish Aequorea in the absence of Ca²⁺ and injected into the cells to be imaged, cloning of its cDNA (Inouye et al., 1985; Prasher et al., 1985) allowed recombinant expression and targeting to subcellular compartments, including the mitochondria (Rizzuto et al., 1992). Native and modified aequorins cover a wide range of free Ca²⁺ concentrations, from ~100nM to the millimolar range and, compared with Ca²⁺ dyes, introduce a very low Ca²⁺ buffering themselves (Bonora et al., 2013). As luminescence yield can be calibrated to deduce absolute Ca²⁺ concentration (Bonora et al., 2013), aequorin has allowed measurement of mitochondrial Ca²⁺ concentrations in animals (Rizzuto et al., 1992) and plants (Logan and Knight, 2003; Mehlmer et al., 2012).

The use of aequorin is constrained, however, by low light emission limiting microscopic applications, such as in a specific cell or mitochondrion, in particular. A variety of fluorescent protein-based sensors has been developed to overcome this limitation. The cameleon family of FRET-based Ca²⁺ sensors, for instance, was introduced by Miyawaki et al. (1997) and has been optimized since then (Miyawaki et al., 1999; Griesbeck et al., 2001; Nagai et al., 2004; Horikawa et al., 2010). The Yellow Cameleon (YC) 3.6, a particulary popular variant, consists of calmodulin (CaM) and a CaM-binding M13 peptide, both of which are inserted between the sequences on an enhanced cyan fluorescent protein (ECFP) and a cpVenus FRET pair (Fig. 3A). The CaM can bind Ca²⁺ through three competent EF-hands which induces binding to the M13 peptide, triggering a conformational change that increases FRET, which is measureable as a change in the relative intensity of both fluorescent proteins (Fig. 3B). Pericams (Nagai et al., 2001), GCaMPs (Nakai et al., 2001), and GECOs (Zhao et al., 2011) use a circularly permuted fluorescent protein inserted in between Ca²⁺-binding CaM and the M13 peptide. Here the Ca²⁺-induced conformational change alters the chemical environment of the chromophore, its protonation state, and its fluorescent properties in turn. Criteria to select the most suitable sensor depend on the specific question and have been extensively reviewed (Palmer et al., 2011; Perez Koldenkova and Nagai, 2013). The pH stability of the sensor is of particular concern for mitochondria where matrix pH can naturally fluctuate, particularly during Ca²⁺ transients (Santo-Domingo and Demaurex, 2012; Marland et al., 2015). pH stability together with a Ca²⁺ dissociation constant close to the resting matrix concentration (K_d=250nM; Nagai et al., 2004) has made YC3.6 a particularly powerful sensor for matrix Ca²⁺ dynamics in animals (Yi et al., 2011) and the current standard in plants (Fig. 3C; Loro et al., 2012; Behera et al., 2013; Teardo et al., 2015; Wagner et al., 2015a, b). R-GECO1, the most recent genetic Ca²⁺ sensor to be introduced into Arabidopsis for cytosolic measurements, allows for higher sensitivity and multiplexing with other blue, green, and yellow fluorescent sensors, due to its large spectroscopic response range and its red colour (Keinath et al., 2015; Y. Wang et al., 2015). The intensiometric, rather than ratiometric, readout of the sensor and its pronounced pH sensitivity will require careful optimization, however, before exploiting it also for mitochondrial measurements.

Fig. 3.

Yellow Cameleon (YC) 3.6 as an in vivo sensor for mitochondrial Ca²⁺ dynamics. (A) Hypothetical model of YC3.6 based on the Protein Data Bank (PDB) entries 1huy (for cpVenus), 1cv7 (for CFP), 2bbm (for CaM–M13), and 1cfd (for CaM in its Ca²⁺-unbound state). Linker segments between proteins and the Ca²⁺-free M13 peptide were added manually. YC3 proteins lack one of four Ca²⁺-binding sites in wild-type CaM (Nagai et al., 2004). Upon Ca²⁺ binding, the CaM–M13 fusion undergoes a pronounced conformational change that re-orientates the fluorescent proteins and amplifies Förster resonance energy transfer (FRET) from CFP to cpVenus. (B) Ratiometric behaviour of YC3.6. Dynamically changing Ca²⁺ concentrations determine the degree of FRET, which is low in the Ca²⁺-unbound state and high in the Ca²⁺-bound state. This manifests in changes of the relative emission intensities of cpVenus and CFP, and their ratio in turn. (C) Expression of YC3.6 targeted to the mitochondrial matrix of Arabidopsis leaf epidermal cells. Scale bar=10 µm. (This figure is available in colour at JXB online.)

Mitochondrial re-modelling of Ca²⁺ transients

Mitochondria shaping cytosolic Ca²⁺ transients

In animal cells, mitochondria were the first intracellular organelle to be associated with Ca²⁺ handling. Their ability to sense Ca²⁺ signals rapidly and to act as localized Ca²⁺ capacitors has long been recognized. By changing the Ca²⁺ concentration in its direct vicinity, mitochondrial Ca²⁺ uptake can influence the frequency and amplitude of cytosolic Ca²⁺ transients, which depend on release channels that are regulated by a Ca²⁺-mediated feedback mechanism. For example, Ca²⁺ flux across both the calcium release activated channel CRAC (Orai1/Stim1) on the plasma membrane/ER and the inositol-1,4,5-trisphosphate receptor on the ER are influenced by the physical proximity of mitochondria. This proximity, sustained by specific mitochondria-associated membrane (MAM) contacts via chaperones, such as sigma receptor 1, has been reported to set the extent and duration of mitochondrial Ca²⁺ increase. In addition, recruitment of mitochondria to specific regions has been suggested to constrain Ca²⁺ signals to defined cell domains, which may be particularly relevant in large cells. In support of those concepts, mitochondrial Ca²⁺ uptake has been shown to be associated with numerous pathophysiological processes including insulin secretion, neuronal excitotoxicity, cardiomyocyte function, and tumorigenesis (see recent reviews by Rizzuto et al., 2012; Foskett and Philipson, 2015).

In plant cells, most research on how Ca²⁺ transients are generated and shaped has been performed with a focus on cytosolic signatures (Knight et al., 1991; Johnson et al., 1995; Knight et al., 1997; Wymer et al., 1997; Kiegle et al., 2000; Allen et al., 2001; Y. Wang et al., 2015). The interplay between influx, buffering, and export shapes the spatiotemporal properties of the transients which are thought to contribute specificity to intracellular Ca²⁺ signalling (Fig. 4A). This explains the large diversity of cytosolic Ca²⁺ signatures that have been observed. Yet, the Ca²⁺ transients inside the mitochondrial matrix take regulatory complexity of Ca²⁺ dynamics to another level.

Fig. 4.

Concepts of shaping mitochondrial Ca²⁺ dynamics. (A) Cytosolic Ca²⁺ transients are shaped by import and export from and to various Ca²⁺ stores as well as buffering. Cytosolic transients are typically reflected in the mitochondrial matrix but are re-modelled through the import and export systems at the IMM aided by the electrochemical potential, and Ca²⁺ buffering in the matrix. (B) In principle, matrix Ca²⁺ signatures may be evoked independently of the cytosol through (1) channelled import of Ca²⁺ from Ca²⁺ stores or (2) organelle autonomous Ca²⁺ uptake from the cytosol/IMS at baseline Ca²⁺ driven by the steep electrochemical gradient. (This figure is available in colour at JXB online.)

Controlling matrix Ca²⁺ dynamics at the inner mitochondrial membrane level

The available in vivo data suggest that, similarly to the situation in animal cells, matrix Ca²⁺ transients in plants generally follow transients in the cytosol. This may be seen as evidence that Ca²⁺ is first released into the cytosol to then be taken up from there in a secondary step (Fig. 4A). Alternatively, direct Ca²⁺ influx from extracellular, vacuolar, or ER stores may occur via contact sites, while Ca²⁺ may coincidently also be released into the cytosol (Fig. 4B). There is evidence for both scenarios in mammalian cells (Lawrie et al., 1996; Rizzuto et al., 1998; Csordás et al., 1999), which can be extended to plant cells in principle, where the ER can also form physical contacts with mitochondria (Stefano et al., 2014). Both scenarios have in common that Ca²⁺ needs to pass the mitochondrial membranes, via specific uptake machineries discussed above. It is not clear if propagation of transients across the OMM may modify the signature, but its impact is often assumed to be minor at most. Much control, regulation, and integration occur at the level of the IMM, which specifically choreographs the resulting matrix transient through its influx and efflux machineries. As such, the IMM acts as an intracellular integration platform that processes and re-shapes cytosolic/IMS Ca²⁺ signatures while passing them on into the matrix, which also impacts on the signature through its specific Ca²⁺-buffering environment that differs from that of the cytosol.

Experimental data from living plant and animal cells confirm this additional level of complexity and regulation. The data consistently show obvious differences between the spatiotemporal properties of cytosolic and matrix Ca²⁺ transients at a given stimulus, such as extracellular application of ATP (eATP), glutamate, or histamine (Loro et al., 2012, 2013; Logan et al., 2014; Waldeck-Weiermair et al., 2015). The first aequorin-based measurements in the cytosol and the mitochondrial matrix of Arabidopsis seedlings revealed slightly higher baseline Ca²⁺, slower onset, lower amplitude, and longer recovery times for matrix transients triggered by environmental stimuli as compared with their cytosolic counterparts (Logan and Knight, 2003). This pattern could be confirmed using cameleon sensors (Loro et al., 2012; Wagner et al., 2015a). Similar steady-state concentrations of free Ca²⁺ in the cytosol and the matrix in the presence of a steep electrochemical gradient are evidence of the remarkable degree of control through the interplay of a tightly sealed IMM with the necessity for Ca²⁺ activation of Ca²⁺ uptake (via MICU) by an otherwise low affinity channel (MCU), a high buffering potential for Ca²⁺ in the matrix, and efficient export against the electrical gradient with high affinity.

An increase in Ca²⁺-selective IMM permeability by activation of a transporter is required to generate a matrix transient. This level offers many options for integration and tuning, which is reflected in the makeup and composition of the uptake systems that appear to include multiple potential channel classes of variable relative abundances, affinities, conductivities, and regulators integrating different stimuli. At the level of the MCUC this plasticity is apparent, and far from fully understood.

Ca²⁺ regulation of mitochondrial Ca²⁺ uptake

The empirical observation of a slightly delayed Ca²⁺ increase in the matrix as compared with the cytosol supports the concept of elicitation of channel activity by Ca²⁺ itself, while lowered matrix amplitudes and delayed recovery to baseline are in general agreement with high matrix buffering and Ca²⁺ export driven by and against the electrochemical gradient. Interestingly, however, lower amplitudes of matrix Ca²⁺ appear not to hold true for all stimuli in Arabidopsis. Auxin application can stimulate much higher amplitudes in the matrix than in the cytosol (Wagner et al., 2015a). This could be interpreted as channelled Ca²⁺ flux from the store directly into the matrix with only minor involvement of the cytosol (see above; Fig. 4B). The assumption of channelling is not critically required, however, since it is thermodynamically plausible that free Ca²⁺ in the matrix accumulates to relatively high levels while cytosolic free Ca²⁺ remains low. Minor Ca²⁺ elevation in the cytosol may trigger transport, and the steep electrochemical gradient across the IMM can in turn drive uptake to much higher levels in the matrix than in the cytosol/IMS. This means that minor Ca²⁺ transients can be ‘magnified’ in the matrix, and it is intriguing to speculate how this may be harnessed by the cell to generate Ca²⁺ signatures that specifically act in mitochondria, but not in other cell compartments. For MICU as gatekeeper for the uniporter, those observations imply that even low cytosolic Ca²⁺ elevations can be sufficient to activate MCUC activity, and it is interesting to note that particularly high Ca²⁺ binding affinity has been estimated for Arabidopsis MICU in vitro (K_d ~1 µM; Wagner et al., 2015a) as compared with mammalian MICU1 (16–21 µM and 4.4 µM; Wang et al., 2014; Waldeck-Weiermair et al., 2015), which may be linked to the presence of an additional EF-hand (see section on ‘MICU’ above). Even more, the electrochemical gradient across the IMM may drive the generation of matrix Ca²⁺ transients without the need for a cytosolic transient to occur in the first place (Fig. 4B). Activation of transport with adequately high affinity would then be sufficient to trigger uptake of Ca²⁺ from the cytosol/IMS as a ‘low-concentration Ca²⁺ store’. This would, however, require overcoming the MICU-based inhibition of uptake by increasing its Ca²⁺ binding affinity or by a non-Ca²⁺-binding mechanism. Alternatively, the activity of a hypothetical channel other than MCUC would be necessary. Without the need for a primary cytosolic Ca²⁺ transient, such a scenario predicts autonomous Ca²⁺ transients in individual mitochondria. Interestingly spontaneous fluctuations in the chemiosmotic gradient of single mitochondria have indeed been observed in plant and animal cells and linked to influx of Ca²⁺ (Duchen et al., 1998; Schwarzländer et al., 2012a; Hou et al., 2013), although the chemiosmotic fluctuations did not coincide with matrix Ca²⁺ transients in other cases (Santo-Domingo et al., 2013; Breckwoldt et al., 2014).

Understanding of mitochondrial Ca²⁺ control in vivo

To understand how a matrix Ca²⁺ transient is generated and tuned in a realistic physiological context, in vivo monitoring of Ca²⁺ dynamics is currently indispensable. Direct deductions are not without problems, however. Combination with genetic approaches, such as heterologous expression or removal of involved proteins, has already been intensely exploited for the functional analysis of MCUC components on matrix Ca²⁺ physiology in intact animal cells and tissues (Perocchi et al., 2010; Baughman et al., 2011; De Stefani et al., 2011; Plovanich et al., 2013; Raffaello et al., 2013; Sancak et al., 2013). Similarly, the assessment of Ca²⁺ dynamics in Arabidopsis mutants of MICU was the basis for deducing an inhibitory function of MICU in plants, based on increased steady-state concentrations and more rapid transients reaching higher peaks (Wagner et al., 2015a). In the current model, inhibition of uptake can be lifted by cytosolic/IMS Ca²⁺ binding to the regulatory EF-hands of MICU. This allows MICU to shape matrix Ca²⁺ dynamics by throttling influx, dependent on the properties of cytosolic/IMS Ca²⁺. Although in vivo sensing of subcellular Ca²⁺ dynamics in mutants can deliver new mechanistic insights, conclusions need to be drawn with caution. It is in fact likely that manipulation of expression of any component from the Ca²⁺ regulation machinery of the mitochondrion will have system effects and might also alter expression of the other regulatory components, as has been shown for MICU1 and MICU2 in mammals (Patron et al., 2014). Functional redundancy can provide a back up for the absence or inhibition of even those players that may be centrally important in the wild-type scenario. Even when Ca²⁺ dynamics are modified, as in the case of the Arabidopsis micu lines, it remains unclear to what extent the status of the Ca²⁺ handling system, being particularly dynamic and delicate, is comparable with the wild-type situation.

A combination with pharmacological approaches can circumvent acclimation, but may introduce off-target effects. Established inhibitors of mitochondrial Ca²⁺ uptake, such as ruthenium red and lanthanum, act with low specificity. Investigation of the structure–activity relationship of the known players and their comparison across systems, such as plants and animals, offers a handle for rational improvement and the development of novel, more specific regulators with promise for clinical use. Reports on the structures of a truncated MCU variant (Lee et al., 2015) and MICU1 (Wang et al., 2014) have provided first insights, but further improvements towards a high quality MCUC structure are urgently needed.

Despite more and more structural, pharmacological, biochemical, physiological, and genetic data from the mitochondrial Ca²⁺ machineries of animals and plants, it is just emerging how matrix Ca²⁺ transients are shaped in vivo. Modelling approaches may offer an elegant way to make use of the existing information to start dissecting the choreography that occurs at the IMM. Such a strategy could generate testable hypotheses about the properties of the players involved and inform synthetic approaches to generate and manipulate subcellular Ca²⁺ signatures rationally, with the potential to re-wire intracellular Ca²⁺ signalling in a targeted manner.

Physiological relevance of mitochondrial Ca²⁺

Matrix Ca²⁺ tunes mitochondrial metabolism in mammals

In mammals, Ca²⁺ elevations in the mitochondrial matrix stimulate respiration and ATP synthesis to cover temporarily high energy needs of cells (Denton, 2009). Ca²⁺ overload, in contrast, can trigger cell death (Duchen, 2000). Increased biosynthesis rates of ATP rely on the activation of three mitochondrial dehydrogenases by Ca²⁺ (McCormack et al., 1990). Pyruvate dehydrogenase (PDH; Denton et al., 1972), NAD-isocitrate dehydrogenase (NAD-ICDH; Denton et al., 1978), and oxoglutarate dehydrogenase (OGDH; McCormack and Denton, 1979) are activated by physiologically relevant Ca²⁺ concentrations (100nM and 1 µM) in mitochondria isolated from mammalian tissues (Denton and McCormack, 1980; Denton et al., 1980). Ca²⁺ elevations in intact cells result in NAD(P) reduction (Duchen, 1992; Pralong et al., 1992), supporting a central role for Ca²⁺-dependent regulation of mitochondrial metabolism. In animals and plants, PDH activity is regulated through reversible phosphorylation (Holness and Sugden, 2003; Tovar-Méndez et al., 2003). The involved phosphatase of mammals, PDP1, is Ca²⁺-dependent, and an increase in free matrix Ca²⁺ switches PDH from an inactive to an active state, boosting the rate of oxidative phosphorylation. Knockout of the MCUC regulator MICU1 that results in an increased basal Ca²⁺ concentration in the matrix of cultured mammalian cells accordingly reduced PDH phosphorylation (Mallilankaraman et al., 2012b). Vice versa, lower levels of basal matrix Ca²⁺ in MCU^−/− mice increased PDH phosphorylation (Pan et al., 2013). In contrast, PDH phosphatase in plants is not activated by Ca²⁺in vitro or in intact mitochondria (Miernyk and Randall, 1987; Budde et al., 1988). Comparative studies further found that while the activity of the tricarboxylic acid (TCA) cycle enzymes NAD-ICDH and OGDH from various vertebrate sources (human heart, frog, and pigeon) is increased in the presence of Ca²⁺, the same does not hold true for the respective homologues from insect flight muscle, yeast, Escherichia coli, potato, and the spadix of Arum (McCormack and Denton, 1981; Nichols et al., 1994). Prediction of alternative physiological targets of Ca²⁺ in plant mitochondria is complicated by the fact that Ca²⁺ often exerts an indirect regulatory effect or the mechanism of Ca²⁺ regulation remains unknown, due to lack of obvious Ca²⁺-binding motifs and Ca²⁺-binding interactors. For instance, mammalian PDH is activated through Ca²⁺-controlled PDH phosphatase, while NAD-ICDH and OGDH do not contain any typical Ca²⁺-binding motifs and it remains unclear how their regulation by Ca²⁺ works mechanistically.

EF-hands make mitochondrial proteins candidates for Ca²⁺ regulation

Intracellular Ca²⁺ can be sensed by either Ca²⁺ sensor relays or sensor responders (Sanders et al., 2002). While sensor relays undergo a conformational rearrangement on Ca²⁺ binding that is passed on to a target protein, Ca²⁺ binding changes the function of a sensor responder directly. The EF-hand helix–loop–helix motif, which arranges four amino acid residues (X, Y, Z, and –Z in Fig. 5C, D) to co-ordinate Ca²⁺, is a typical feature of Ca²⁺ sensors in animals and plants. Yet, not every Ca²⁺-binding protein carries an EF-hand (e.g. annexins and proteins carrying a C2 domain) and not every EF-hand binds Ca²⁺ (e.g. Gelhaye et al., 2004). The human genome encodes at least 83 EF-hand proteins, and the Arabidopsis genome 250 (Day et al., 2002). For an appraisal of the role that Ca²⁺ plays in regulating mitochondrial function in mammals and plants, we queried the recently updated ‘MitoCarta’ list of human mitochondrial proteins (Pagliarini et al., 2008; Calvo et al., 2015) and an Arabidopsis mitochondrial proteome data set (Wagner et al., 2015a) for EF-hand motifs (ProSite pattern PS00018 and PS50222) using the ProSite algorithm (De Castro et al., 2006). Both data sets contain 10 EF-hand proteins each, associated with similar protein classes including proteins associated with Ca²⁺ transport, such as APC/AGC carrier proteins, LETM-like proteins, and MICU proteins (Hajnóczky et al., 2014) (Table 1; Fig. 5A, B).

Fig. 5.

Mitochondrial EF-hand proteins in humans and Arabidopsis as targets of Ca²⁺ regulation. (A) Submitochondrial localization of selected EF-hand-containing proteins detected in human and Arabidopsis data sets of mitochondrial proteins (Table 1). Topology of EF-hands in Arabidopsis transmembrane proteins is inferred from their mammalian homologues. (C) Structure model of the canonical EF-hand motif. Amino acid positions X, Y, Z, and –Z are responsible for Ca²⁺ co-ordination. (D) Conservation of positions X, Y, Z, and –Z in human mtGPDH, Arabidopsis GDH2, and their EF-hand-lacking homologues. Purple background indicates compatibility of the amino acid with EF-hand function according to ProSite. Numbers in grey indicate the total sequence similarity between related proteins. (This figure is available in colour at JXB online.)

Table 1.

Mitochondrial EF-hand proteins in animals and plants

Proteins in the human MitoCarta and an Arabidopsis mitochondrial proteome data set (Wagner et al., 2015a) that possess Ca²⁺-binding EF-hands according to ProSite. Protein IDs refer to UniProt entries (human) and AGI codes (Arabidopsis). References refer to protein and/or EF-hand localization studies. Due to a lack of data, localization of EF hands is not further specified for Arabidopsis proteins. In the MitoCarta, additional proteins (NEFA (NUCB2), RCN2 (ERC55), and FKBP10) were found, but are not shown since they probably represent false positives and localize to the Golgi apparatus or ER instead (Patterson et al., 2000; Weis et al., 1994; Nesselhut et al., 2001).

Human MitoCarta
Process	Name	ID	Location (protein/EF hand)
Transport-related	APC1, SCAMC1, SLC25A24	Q6NUK1	IMM/IMS ( Nosek et al., 1990; Del Arco and Satrústegui, 2004)
	APC2, SCAMC3, SLC25A23	Q9BV35
	APC3, SCAMC2, SLC25A25	Q6KCM7
	AGC1, SLC25A12	O75746	IMM/IMS ( Palmieri et al., 2001)
	AGC2, SLC25A13	Q9UJS0
	LETM1	O95202	IMM/matrix ( Nowikovsky et al., 2012)
	MICU1	Q9BPX6	IMS/IMS ( Csordás et al., 2013; Sancak et al., 2013; Petrungaro et al., 2015)
	MICU2	Q8IYU8	IMS/IMS ( Sancak et al., 2013; Patron et al., 2014; Petrungaro et al., 2015)
Dehydrogenase	mtGPD2	P43304	IMM/IMS ( Klingenberg, 1970; MacDonald and Brown, 1996)
Other	MIRO1, RHOT1	Q8IXI2	OMM/cytosol ( Fransson et al., 2006)
	MIRO2, RHOT2	Q8IXI1
	NDUFAB1, SDAP	O14561	Associated with complex I and/or matrix-localized/? ( Runswick et al., 1991; Cronan et al., 2005)
	EFHD1, mitocalcin	Q9BUP0	IMM/? ( Tominaga et al., 2006)
	GRP75, HSPA9, PBP74, mortalin	P38646	Matrix and OMM/? ( Dahlseid et al., 1994; Szabadkai et al., 2006)
Arabidopsis mitochondrial proteome data set
Process	Name	ID	Location (protein)
Transport-related	APC1	AT5G61810	IMM ( Stael et al., 2011)
	APC3	AT5G07320	IMM ( Stael et al., 2011)
	LETM1	AT3G59820	IMM ( Zhang et al., 2012)
	LETM2	AT1G65540	IMM ( Zhang et al., 2012)
	MICU	AT4G32060	IMS ( Wagner et al., 2015a)
Dehydrogenase	GDH2	AT5G07440	Matrix ( Ito et al., 2006)
	NDB1	AT4G28220	IMM ( Elhafez et al., 2006)
	NDB2	AT4G05020	IMM ( Elhafez et al., 2006)
Other	MIRO1	AT5G27540	OMM ( Duncan et al., 2011)
	Calmodulin	AT1G66410 AT2G27030 AT2G41110 AT3G43810 AT3G56800 AT5G21274 AT5G37780	?

EF-hands appear in functionally related dehydrogenase systems in animals and plants

Not common to both sets are several dehydrogenases (Table 1; Fig. 5A, B): human mitochondrial glycerol-3-phosphate (G-3-P) dehydrogenase (mtGPDH/GDP2) resides on the outer surface of the IMM where it acts as part of the ‘G-3-P shuttle’, which consumes cytosolic NADH to generate G-3-P from dihydroxyacetone phosphate, subsequently re-oxidized by mtGDPH at the IMM, transferring electrons to the mitochondrial ubiquinone pool. The G-3-P shuttle has been thoroughly characterized in animals and yeast (Larsson et al., 1998; Rigoulet et al., 2004; Mráček et al., 2013). Homologues of GPDH have also been described in plants (Shen et al., 2003, 2006), but the EF-hand for direct Ca²⁺ binding and activation of mammalian mtGPDH (Hansford and Chappell, 1967; Klingenberg, 1970; MacDonald and Brown, 1996) is absent in related proteins from plants, yeast, and fungi (Brown et al., 1994; Satrustegui et al., 2007). Plants possess a particularly large diversity of mitochondrial dehydrogenases (Schertl and Braun, 2014), including additional dehydrogenases to mediate oxidation of cytosolic NAD(P)H through mitochondrial electron transport (Rasmusson et al., 2008). NDB-type NAD(P)H dehydrogenases in plants are also located at the outer surface of the IMM (Douce et al., 1973; Luethy et al., 1995; Rasmusson et al., 1999; Elhafez et al., 2006) and also contain a conserved EF-hand (Table 1; Fig. 5B; Michalecka et al., 2003). Arabidopsis NDB1 and NDB2, which were both identified as EF-hand proteins in the Arabidopsis proteome (Table 1), are specific for NADPH (NDB1) and NADH (NDB2), respectively, and the activities of both are controlled by Ca²⁺ (Geisler et al., 2007). The activating effect of Ca²⁺ on NDB2 additionally depends on cytosolic pH (Hao et al., 2015), putting this protein at the interface between cytosolic and mitochondrial metabolism, Ca²⁺ signalling, and redox regulation (Wallström et al., 2014).

EF-hands are acquired and lost during the course of evolution

Glutamate dehydrogenase 2 (GDH2) was identified as a plant mitochondrial EF-hand protein (Table 1; Fig. 5B). Arabidopsis possesses three NAD(H)-dependent GDHs (Fontaine et al., 2012), with GDH2 being the only one to carry an EF-hand (Fig. 5C, D). GDHs from multiple plant species have been shown to be activated by Ca²⁺ (Garland and Dennis, 1977; Kindt et al., 1980; Yamaya et al., 1984; Das et al., 1989; Itagaki et al., 1990; Turano et al., 1997). In plants and animals, GDHs reversibly convert glutamate to the TCA cycle intermediate 2-oxoglutarate and connect nitrogen and carbon metabolism. In plants, Ca²⁺ seems mostly to activate the amination reaction (Garland and Dennis, 1977; Turano et al., 1997; Yamaya et al., 1984), but it is an open question whether Ca²⁺ activates GDH in planta, since high micromolar Ca²⁺ concentrations were required for maximal activation of GDH in vitro (Turano et al., 1997), while matrix Ca²⁺ transients peak in the low micromolar range (Zottini and Zannoni, 1993; Logan and Knight, 2003; Wagner et al., 2015a). Two human homologues of plant GDH, GLUD1 and GLUD2, localize predominantly to the mitochondrial matrix (Mastorodemos et al., 2009) but lack EF-hands (Fig. 5B–D).

Miro GTPases, detected in both data sets (Table 1; Fig. 5), are EF-hand proteins that decorate the OMM and mediate mitochondrial motility and morphology in animals and plants (Boldogh and Pon, 2007; Yamaoka and Leaver, 2008; Yamaoka and Hara-Nishimura, 2014).

A CaM protein was among the mitochondrial proteome from Arabidopsis (Table 1; Fig. 5B). CaM proteins are exceptionally highly conserved Ca²⁺ sensor relays, of which seven genes in Arabidopsis encode four protein isoforms that are considered to be genuine CaMs due to their high similarity to vertebrate CaMs (McCormack and Braam, 2003). These isoforms differ by a maximum of four amino acids and are thus indistinguishable in our proteomic data set. In contrast to animals, plants also possess CaM-like proteins (CMLs) that harbour 2–6 EF-hands and share at least 15% sequence identity with CaMs without having other identifiable functional domains (McCormack and Braam, 2003). Applying these criteria, 50 genes were predicted to code for Arabidopsis CMLs and they are involved in various processes covering growth and development, abiotic stress response, and pathogen defence (Perochon et al., 2011; Bender and Snedden, 2013). Although CaMs and CMLs are considered mostly nucleo-cytoplasmic, individual isoforms, as well as matching CaM-binding proteins (Bussemer et al., 2009), have been found within other cell compartments including the mitochondrion (Yamaguchi et al., 2005; Chigri et al., 2012).

Three additional EF-hand proteins, NDUFAB1/SDAP, GRP75/HSPA9/PBP74, and EFHD1, were found in the human data set without EF-hand-containing counterparts in Arabidopsis (Table 1). Briefly, they act as an acyl carrier protein (NDUFAB1/SDAP), at the physical interface between mitochondria and the ER (GRP75/HSPA9/PBP74), and in apoptosis and differentiation of mammalian neuronal and muscle precursor cells (EFHD1). The importance of Ca²⁺ binding in these processes is unclear.

The simple comparison of EF-hand proteins in protein data sets of human and Arabidopsis mitochondria results in a remarkably coherent picture. Although each of the the 10 proteins found is likely to represent only a subset of the full Ca²⁺-related inventory, there are clear parallels between the respective protein functions. Mitochondrial dehydrogenases appear to be able to obtain and lose EF-hand motifs in a modular manner in the course of evolution (Fig. 5D). This may correlate with the particular lifestyles and environments of plants versus mammals. On the functional level, similarities between the different dehydrogenases are striking, however, suggesting that a link between mitochondrial Ca²⁺ and respiratory redox metabolism is conserved between plants and animals. Notably, the EF-hands of most identified proteins are not exposed to the mitochondrial matrix (Fig. 5A, B). This does not mean, however, that regulation of these proteins is restricted to cytosolic Ca²⁺. Local IMS Ca²⁺ levels are likely to be dependent on the local import and export dynamics, giving rise to microdomains, which further emphasizes the complexity of the mitochondrion as a cellular integration platform of Ca²⁺ regulation.

Phenotypes of animals and plants with defective mitochondrial Ca²⁺ regulation

The sophistication of the regulation of mitochondrial Ca²⁺ at the molecular and cell physiological level, as well as the existence of several Ca²⁺-regulated mitochondrial proteins predicts that dysfunction gives rise to severe defects at the organismal level. It came as a surprise, therefore, when an initial report described the generation of viable and healthy MCU knockout mice, with only mild alterations at the whole-organism level (Pan et al., 2013). Their mitochondria lacked any capacity for rapid Ca²⁺ uptake and their cytosolic Ca²⁺ signatures were altered, indicating dysfunctional Ca²⁺ handling at the cellular level. Those results have been heavily debated, since viable mice could only be obtained in a mixed genetic background (Pendin et al., 2014). Several recent in vivo studies in mammals suggest that alterations in mitochondrial Ca²⁺ dynamics by interference with MCU function are indeed linked to various pathologies. Post-natal manipulation of MCU levels in mice demonstrated the contribution of MCUC to the regulation of skeletal muscle tropism. MCU overexpression or down-regulation caused muscular hypertrophy and atrophy, respectively, probably independently of metabolic alterations, but associated with Ca²⁺-dependent mitochondria-to-nucleus signalling (Mammucari et al., 2015). Finally, in mice with myocardial MCU inhibition by transgenic expression of a dominant-negative MCU, a strong correlation between MCU function, oxidative phosphorylation, and correct pacemaker cell function was found (Wu et al., 2015). In zebrafish (Prudent et al., 2013) and Trypanosome brucei (Huang et al., 2013), genetic manipulation of MCU also resulted in major developmental and energetic defects. Homozygous human patients carrying a loss-of-function mutation of MICU1 suffer from myopathy, cognitive impairment, and extrapyramidal movement disorder (Logan et al., 2014). At the cellular level, they show increased agonist-induced mitochondrial Ca²⁺ uptake at low cytosolic Ca²⁺ concentrations and decreased cytosolic Ca²⁺ transient amplitudes. However, at least under resting conditions, the fibroblasts from affected individuals do not display any severe metabolic defects. Instead, chronic elevation of the mitochondrial matrix Ca²⁺ load seems to lead to mitochondrial stress, resulting in fragmentation of the mitochondrial network.

There are currently no corresponding phenotypic observations for plants lacking MCU. A complete loss-of-function line of Arabidopsis may require knockout of all six MCU homologues (see section on ‘MCU’ above; Fig. 2B). However, with a better understanding of the individual plant MCU homologues, their subcellular localization. and their differential expression (Stael et al., 2012; Meng et al., 2015), less complex genetic lines may allow the first focused analyses over the next years. In Arabidopsis lines lacking MICU (see section on ‘MICU’ above), pronounced changes in matrix Ca²⁺ dynamics in root tips correlate with changes in mitochondrial ultrastructure and adjustments in the respiratory machinery, while no gross developmental phenotype was found (Wagner et al., 2015a). In contrast, Arabidopsis lines lacking mitochondrial GLR3.5 show abnormal mitochondrial ultrastructure and an early senescence phenotype without strong alterations of matrix Ca²⁺ dynamics, making it difficult to draw direct mechanistic links at present (see section on ‘GLR3.5’ above). The general incoherence of the current picture in both animals and plants may be partly explained by compensatory mechanisms operating through the subcellular physiological network (Schwarzländer and Finkemeier, 2013). Alternatively, acclimation between the cellular and whole-organism scale may be responsible for the observed robustness (see ‘Understanding of mitochondrial Ca²⁺ control in vivo’ above). Both levels are currently insufficiently understood for complex biological systems, including animals and plants. Contrary to common argument, there is no sound mechanistic basis to justify correlations between the importance of most cell physiological players or processes and gross developmental phenotypes that may be induced by their impairment on the organismal level.

Conclusions and future perspectives

While several key players of the regulation of mitochondrial Ca²⁺ have been identified in animal systems, the situation remains less clear for plants. Comparative approaches with concerted research efforts in both systems side-by-side represent a promising strategy if we are to understand the complexity of the machineries involved and to distil the minimal set of components and regulatory mechanisms required. The use of genetically encoded Ca²⁺ sensors combined with genetic manipulation of the organism has proven fruitful to dissect the individual contributions of the different players in live cells and whole organisms. Yet, the generation of genetic models to study mitochondrial Ca²⁺ in animals under in vivo conditions remains technically challenging. Arabidopsis T-DNA insertion collections offer a strong advantage, which is partly offset by genetic diversification with multiple family members, however, as is the case for MCU. Both constraints may be overcome by emerging techniques, such as genome editing by CRISPR–Cas9, which are usable in a wide range of organisms and potentially allow multigene targeting. Pleiotropic and compensatory effects during development may be counteracted by inducible loss- or gain-of-function approaches. Active interaction across the animal and the plant disciplines appears to hold particular promise to unravel further the fundamental roles that mitochondrial Ca²⁺ signalling plays in vivo.

Acknowledgements

We thank Marco Zancani (Università degli Studi di Udine) for critical reading of the manuscript. MS thanks the Deutsche Forschungsgemeinschaft (DFG) for support through the Emmy-Noether programme (SCHW1719/1-1), the Research Training Group 2064, and a grant (SCHW1719/5-1) as part of the package PAK918, and IS thanks the Ministero dell’Istruzione, dell’Università e della Ricerca for funding through the PRIN project (2010CSJX4F), to the Human Frontiers Science Program (HFSP 0052) and to the Italian Association for Cancer Research (IG 11814).

Author notes

Editor: Markus Teige, University of Vienna