The molecular identity of the mitochondrial Ca²⁺ sequestration system

Abstract

There is ample evidence to suggest that a dramatic decrease in mitochondrial Ca²⁺ retention may contribute to the cell death associated with stroke, excitotoxicity, ischemia and reperfusion, and neurodegenerative diseases. Mitochondria from all studied tissues can accumulate and store Ca²⁺, but the maximum Ca²⁺ storage capacity varies widely and exhibits striking tissue specificity. There is currently no explanation for this fact. Precipitation of Ca²⁺ and phosphate in the mitochondrial matrix has been suggested to be the major form of storage of accumulated Ca²⁺ in mitochondria. How this precipitate is formed is not known. The molecular identity of almost all proteins involved in Ca²⁺ transport, storage and formation of the permeability transition pore is also unknown. This review summarizes studies aimed at identifying these proteins, and describes the properties of a known mitochondrial protein that may be involved in Ca²⁺ transport and the structure of the permeability transition pore.

The standard model

The ability to accumulate, retain and release Ca²⁺ is a fundamental ubiquitous function of animal mitochondria. Extensive research during the last 50 years has resulted in a consensus model of mitochondrial Ca²⁺ handling that adequately accommodates most if not all experimental data, referred to here as ‘the standard model of mitochondrial Ca²⁺ handling’. This model is shown in Fig. 1 in a greatly simplified form, and assumes that mitochondria accumulate exogenous Ca²⁺ by means of an electrogenic carrier that facilitates Ca²⁺ transport across the inner mitochondrial membrane (IM) into the matrix. The transport is coupled to simultaneous accumulation of inorganic phosphate. Inside the matrix, accumulated Ca²⁺ and phosphate are stored in the form of osmotically inactive precipitates, and eventually are slowly released back into the cytosol with the assistance of Ca²⁺/nNa⁺ and/or Ca²⁺/2H⁺ exchangers (Fig. 1) that are also situated in the IM. When accumulated above a certain threshold, Ca²⁺ triggers opening of the so-called permeability transition pore (PTP). This may also be mediated by matrix proteins such as cyclophilin D (CypD). Opening of the PTP is thought to have a detrimental effect on mitochondria and cell well-being in general. The Ca²⁺ uniporter system and the PTP structure are thought to consist of proteins, but the molecular identities of these proteins are unknown. The only two Ca²⁺ transport-related proteins that have been identified are the Ca²⁺/nNa⁺ exchanger and Ca²⁺/2H⁺ exchanger: the gene for the CGP37157-sensitive mitochondrial Ca²⁺/nNa⁺ exchanger has recently been identified as NCLX (SLC24A6 family) [1], and that for the Ca²⁺/2H⁺ exchanger has been identified as Letm1 [2]. These two proteins will not be reviewed here; please see the review by Chinopoulos & Adam-Vizi [3] and that by Pivovarova & Andrews [4] in this issue for details. Nevertheless, extensive studies to identify proteins involved in Ca²⁺ transport, storage and the PTP have been performed over the last 50 years. This review is not concerned with kinetic, biophysical channel-related, bioenergetic or pathophysiological aspects of Ca²⁺ handling in mitochondria; numerous excellent reviews on these subjects can be found elsewhere. Here, the present review describes some of the most prominent and followed-up research efforts to identify the proteins involved in Ca²⁺ transport, storage and PTP, and presents a hypothesis on this subject that somewhat modifies the ‘standard model’.

Fig. 1.

Standard model of mitochondrial Ca²⁺ handling. Mitochondria accumulate exogenous Ca²⁺ by means of an electrogenic carrier (calcium uniporter, ‘U’) that facilitates Ca²⁺ transport across the inner mitochondrial membrane (IM) into the matrix. The transport is coupled to simultaneous accumulation of inorganic phosphate (not shown). Inside the matrix, accumulated Ca²⁺ and phosphate are stored in the form of osmotically inactive precipitates (‘precipitate’), and eventually slowly released back into the cytosol through a Ca²⁺/nNa⁺ (not shown) and / or a Ca²⁺/2H⁺ exchanger that is also located in the IM. The process of Ca²⁺ uptake is driven by the membrane potential; the process of Ca²⁺ release is driven by the pH gradient, in the case of the Ca²⁺/2H⁺ exchanger. Elevated intramitochondrial Ca²⁺ can stimulate the activities of enzymes of the tricarboxylic acid cycle (TCA), thereby boosting energy production in the mitochondria. When it accumulates above a certain threshold, Ca²⁺ triggers PTP opening, and this is also modulated by matrix-located protein cyclophilin D (CypD). E, exchanger; RC, respiratory chain.

Ca²⁺ uniporter system

Although the molecular identity of the mitochondrial calcium uniporter is still unknown, experimental data have suggested that it is a highly selective, inward-rectifying ion channel [5], a ‘gated’ pore containing a Ca²⁺ binding site on the cytosolic side of the inner mitochondrial membrane that activates Ca²⁺ transport [6,7]. It has been suggested that mitochondrial calcium uniporter contains at least two subunits, one of which is a dissociable intermembrane factor that is glycoprotein in nature, and that the mitochondrial calcium uniporter is regulated by association and dissociation of this factor, activated by calcium binding [8]. It has been shown that mitochondria depleted of endogenous Ca²⁺ exhibited low initial rate of energy-dependent Ca²⁺ uptake. Pre-incubation of de-energized mitochondria with added Ca²⁺ stimulated their energy-dependent Ca²⁺ uptake up to 10-fold, with strong cooperativity in the velocity–substrate curves for Ca²⁺-depleted mitochondria. To explain these and other kinetic peculiarities of Ca²⁺ transport, a model has been proposed in which the Ca²⁺-transporting system is present in a de-activated state in the absence of cytosolic Ca²⁺, and formation of the active Ca²⁺ uniporter is triggered by an increase in external Ca²⁺. The uniporter is formed by oligomerization of two or more protomers, resulting in formation of the ruthenium- and lanthanides-sensitive Ca²⁺-conducting gated channel [9].

As already mentioned, the molecular identity of the mitochondrial calcium uniporter remains unknown, despite considerable efforts by many prominent researchers. Since the pioneering reports of Sottocasa et al. [10] and Lehninger [11], numerous attempts have been made to isolate the calcium uniporter [12–23]. Various Ca²⁺ binding proteins and peptides have been isolated and characterized, all of which are able to bind Ca²⁺ in a ruthenium red- and La³⁺-inhibited fashion, and some of which are able to transport bound Ca²⁺ through artificial bilayer membranes. Reviewing all this literature is beyond the scope of the present review: Lars Ernster’s [23a] and Saris and Carafoli’s [24] recent review provide comprehensive literature surveys on the history of Ca²⁺ transport and attempts to isolate the Ca²⁺ uniporter. The present review covers only the most followed-up and detailed studies.

Calvectin

The earliest extensively studied preparations of mitochondrial Ca²⁺ binding protein(s) were isolated by Sottocasa et al. from intermembrane space of rat liver mitochondria [10] and ox liver mitochondria [14]. These preparations were capable of high-affinity Ca²⁺ binding that was inhibited by ruthenium red and La³⁺. These preparations showed a single band of approximately 30 kDa on PAGE, and contained sialic acid and neutral and amino sugars, typical of glycoproteins, a high content of dicarboxylic amino acids, and some bound Ca²⁺ and Mg²⁺. This preparation was capable of binding Ca²⁺ with high affinity (K_d of approximately 100 nm), and also contained a number of low-affinity Ca²⁺ binding sites. It was named ‘calvectin’ [25], and was suggested to represent the mitochondrial Ca²⁺ carrier or a major component thereof. Similarly isolated glycoprotein increased the conductance of artificial lipid bilayers in the presence of Ca²⁺, and the conductance was sensitive to ruthenium red [22], implying that it may be the Ca²⁺ uniporter or part thereof. Further studies revealed a set of unique features for this preparation. One of them was that the glycoprotein was found primarily in the inter-membrane space in both a free soluble form and also tightly bound to the inner membrane, but was absent in the matrix of mitochondria [26]. Binding to inner and outer membranes apparently required Mg²⁺ and/or Ca²⁺ [27]. Moreover, calvectin appeared to be able to move reversibly between mitochondrial compartments in the presence of Ca²⁺. The binding to the membrane could further be modulated by pyridine nucleotides, which also bind to calvectin; bound NAD+ decreased the association of calvectin with the membrane [28]. Mitochondria could be depleted of calvectin by treating them with uncoupling concentrations of pentachlorophenol in the presence of phosphate and acetate. This treatment affected the ability of mitochondria to release pre-loaded Ca²⁺ in response to the addition of pentachlorophenol, with an almost linear correlation between the amount of released glycoprotein and the rate of Ca²⁺ efflux [29]. Adding the glycoprotein back to mitoplasts (mitochondria stripped of their outer membrane) depleted of it by swelling in oxaloacetate/EDTA restored the Ca²⁺ uptake if Mg²⁺ was also included in the mixture [28]. Antibodies raised against calvectin were able to inhibit Ca²⁺ transport in mitoplasts, indicating that this glycoprotein is a required part of the mitochondrial Ca²⁺ transport machinery [30], (to note, a review by Saris and Carafoli mentions that “Saris found that the antiserum formed four precipitation bands in Ouchterlony immunodiffusion tests and did not inhibit Sr²⁺ uptake by the uniporter” [24]. We were not able to find another published record of that finding which is important because mitochondria are known to accumulate both Ca²⁺ and Sr²⁺ with about similar efficiency and ruthenium red sensitivity. Hence, this finding might imply that a conformation of the “uniporter” that transports Ca²⁺ is different from that transporting Sr²⁺). The authors suggested an interesting but rather simple ‘two-step’ model of calvectin involvement in Ca²⁺ transport: first, soluble calvectin in the intermembrane space binds Ca²⁺ and associates spontaneously with the inner membrane; second, it carries Ca²⁺ through the membrane and somehow returns back to the outer surface of the inner membrane [31]. Eventually, a single protein was purified from these crude preparations that migrated at approximately 14 kDa on SDS/PAGE and had a minimum molecular weight of 15 577 calculated on the basis of its amino acid composition. However, no sugars were found in this protein, although it had a high content of glutamic and aspartic acids. This protein also carried fewer low-affinity Ca²⁺ binding sites than the original ‘calvectin’ [23a].

Calciphorin

An integral low-molecular-weight membrane protein with the properties of a Ca²⁺ ionophore was isolated from calf heart inner mitochondrial membrane [15– 17,32,33]. It was characterized as a 3000 Da high-affinity calcium carrier and named ‘calciphorin’ [16]. In contrast to hydrophilic calvectin, the calciphorin was hydrophobic and lacked phospholipids, sugars and free fatty acids. Calciphorin was able to extract Ca²⁺ into an organic solvent phase and to transport Ca²⁺ through a bulk organic phase in the presence of a lipophilic anion (picrate), indicating the electrophoretic nature of the calciphorin–Ca²⁺ complex. The Ca²⁺ extraction was strongly inhibited by ruthenium red and lanthanum. The selectivity of ion extraction by calciphorin was Zn²⁺ > Ca²⁺, Sr²⁺ > Mn²⁺ > Na⁺ > K⁺ [32]. The Ca²⁺ binding site had a dissociation constant of 5.2 pm, with 1 mole Ca²⁺ bound per mole of calciphorin [32]. Calciphorin was shown to transport Ca²⁺ in a lipid bilayer membrane model such as reconstituted phospholipid vesicles. Furthermore, calciphorin-mediated Ca²⁺ transport across the vesicle membrane was toward the negatively charged side of the membrane. This calciphorin-mediated calcium transport in vesicles was also strongly inhibited by ruthenium red and La³⁺ [33].

The role of calciphorin as the Ca²⁺ ionophore was subsequently challenged by Sokolove and Brenza [34], who isolated a mixed protein–lipid fraction from rat liver mitochondria that had properties similar to those of calciphorin. They attributed all the Ca²⁺-binding and transporting ability of that fraction to its lipid components. In another study, these authors demonstrated that cardiolipin binds Ca²⁺ with high affinity (apparent K_d = 0.70 ± 0.17 µm) and can extract Ca²⁺ into a bulk organic phase. The interaction of cardioli-pin with Ca²⁺ was insensitive to Na⁺, but was inhibited by divalent cations (Mn²⁺ > Zn²⁺ > Mg²⁺). In addition, La³⁺ and ruthenium red were found to be strong inhibitors of Ca²⁺ binding by cardiolipin [35]. However, it should be noted that the isolation procedure used by Sokolove and Brenza was similar but not identical to that originally reported by Shamoo’s group who later successfully isolated ‘calciphorin’ from rat liver mitochondria [36]. Nevertheless, it is still not clear whether ‘liver calciphorin’ and ‘heart calciphorin’ are the same proteins, or indeed whether the procedure described by Jeng and Shamoo is reproducible. As can be seen from Table 1 in [36], the ‘calciphorin’ isolated from liver and two ‘calciphorin’ isolates from calf heart were quite different in terms of their estimated molecular mass and other parameters. To the best of our knowledge, there were no new reports on calciphorin after 1984.

Mironova’s glycoprotein and peptide

Mironova’s group worked on isolation and identification of Ca²⁺-transporting substances in mitochondria for almost two decades since approximately 1976, but most of the earlier results were published in hard-to-access Russian journals. The authors isolated a component capable of inducing selective Ca²⁺ transport in artificial bilayer lipid membranes from mitochondria and homogenates of various animal and human tissues. The Ca²⁺-transporting properties of this component were ascribed to the presence of a glycoprotein and a peptide. The 40 kDa glycoprotein and 2 kDa peptide from beef heart homogenate and mitochondria induced highly selective Ca²⁺ transport through bilayer lipid membranes. The glycoprotein contained 60–70% and 30–40% protein and carbohydrate, respectively. Sulfur-containing amino acids (1 mole per 1 mole of glycoprotein) and sialic acids (2 or 3 moles per 1 mole of glycoprotein) were also detected in the glycoprotein, and it was enriched in asparagine and glutamine [21], similar to calvectin. Lipids were not essential for the Ca²⁺-transporting activity of glycoprotein. Micromolar concentrations of the glycoprotein and the peptide were found to greatly increase the conductivity of bilayer lipid membranes. Ruthenium red abolished the glycoprotein- and peptide-induced Ca²⁺ transport in bilayer lipid membranes. A transmembrane Ca²⁺ gradient induced an electric potential difference whose magnitude was close to the theoretical value for optimum Ca²⁺ selectivity. The authors also identified thiol groups that were essential for Ca²⁺ transport in both the glycoprotein and the peptide. On the basis of these studies, the authors proposed a model in which the peptide is an active Ca²⁺-transporting portion of the glycoprotein, which lacks Ca²⁺-transporting activity when the peptide is detached. Ca²⁺ moves through special channels in the membrane formed by the peptide, and the glycoprotein, which has many Ca²⁺-binding centers, creates a high concentration of Ca²⁺ near the channel mouth. Functioning of the channels is controlled by thiol-disulfide transitions of sulfur-containing groups of the glycoprotein–peptide complex [21].

A decade later, the same group (in collaboration with Saris) generated polyclonal rabbit antibodies against a ‘Ca²⁺-binding mitochondrial glycoprotein’ (presumably the former glycoprotein). These antibodies were found to inhibit the uniporter-mediated transport of Ca²⁺ in mitoplasts prepared from rat liver mitochondria. Spermine, a modulator of the uniporter, decreased the inhibition [37]. The peptide was isolated and purified to homogeneity and shown to form a Ca²⁺-transporting channel in bilayer lipid membranes, requiring addition of the peptide from both sides of the membrane, [20]. This suggested that the channel is formed by two or more subunits, as in formation of the gramicidin D channel [38]. The authors also demonstrated that the Ca²⁺-binding 40 kDa glycoprotein previously reported as a precursor of the peptide may in fact be an irrelevant contaminant, as it was immunologically indistinguishable from beef plasma orosomucoid protein. However, antibody raised against the orosomucoid was not able to inhibit mitochondrial Ca²⁺ uptake [20], in contrast to the antibodies derived against mitochondrial glycoprotein in the previous study [37]. Nevertheless, the authors concluded that the presence of the 40 kDa glycoprotein in association with a channel-forming peptide [39] was due to co-purification.

Most recent ‘Ca²⁺ uniporter’ isolations

Chavez’s group isolated a semi-purified extract of proteins from rat kidney cortex mitochondria that conferred Ca²⁺-transporting capacity to energized cytochrome oxidase-containing proteoliposomes, and generated a mouse hyperimmune serum that inhibited Ca²⁺ transport in mitoplasts and proteoliposomes. The serum recognized three major proteins of 75, 70 and 20 kDa. The purified antibody recognizing the 20 kDa component inhibited Ca²⁺ transport by approximately 70% in mitoplasts, suggesting that this 20 kDa protein is a necessary component of the Ca²⁺ uniporter [23]. In a follow-up study, the same group isolated an 18 kDa protein that binds Ru360 (an inhibitor of Ca²⁺ uniporter) with high affinity, and proposed that it is part of the uniporter [40]. Most recently, these authors isolated a Ca²⁺-transporting protein fraction and separated it further by preparative electrofocusing. After incorporating the separated fractions into cytochrome oxidase containing proteoliposomes, they recovered two Ca²⁺-transporting activities, only one of which was inhibited by Ru360. On the basis of these results, the authors suggested that the Ca²⁺ uniporter is composed of at least two different subunits that become partially dissociated at low pH. The Ru360-resistant proteins are dissociated at low pH and represent the Ca²⁺ channel, whereas the subunit that binds to Ru360 remains linked to the channel at higher pH [41]. The same group also showed that glycosyl residues on the putative Ca²⁺ uniporter are not required for Ca²⁺ transport activity: deglycosylation of mitoplasts using glycosidase F removed the ruthenium red sensitivity of Ca²⁺ uptake but did not inhibit it [42].

It is very surprising that so much effort spanning several decades of research did not result in molecular identification of any of the isolated putative Ca²⁺-transporting proteins. Even the most recent studies by Zazueta et al. [40], performed when the majority of the new proteomics approaches, sequencing techniques and a wealth of genetic information were already available, did not identify the isolated proteins. Unfortunately, the chances of reproducing the older research and isolating the same proteins are low. Protein purification from mitochondrial membranes that carry hundreds of proteins is akin to magic: unless a spell is cast precisely (in this case a step-by-step isolation protocol listing all the reagents, procedures and conditions), the result could be just a sore throat. It may be more productive to adopt a targeted approach, selecting a few known mitochondrial proteins fitting the required ‘profile’ and using genetic approaches to prove their involvement in Ca²⁺ transport. What kind of ‘profile’ for a putative Ca²⁺ uniporter can be deduced from these older studies? The structure of this protein should accommodate the following features: the protein should be of moderate to low molecular mass, approximately 15–40 kDa, it should be capable of binding Ca²⁺, the binding should be inhibited by ruthenium red and other known Ca²⁺ uniporter inhibitors, it also should be able to bind to the inner mitochondrial membrane from at least the cytosolic side, preferably in the presence of Ca²⁺ (like calvectin), and, according to all the hypotheses reviewed above and a wealth of other known characteristics regarding Ca²⁺ transport, should be able to form a gated pore comprising several identical protomers or as a complex with other proteins. A known protein that mostly fits this profile is discussed below.

Storage of Ca²⁺ in mitochondria

Net Ca²⁺ uptake into mitochondria requires co-transport of an IM-permeable anion such as acetate or phosphate. In the latter case, the accumulated Ca²⁺ forms a precipitate in the matrix of mitochondria in an apparently spontaneous process. The precipitate can store large amounts of Ca²⁺ and is readily observed in isolated mitochondria by electron microscopy [4,43]. The precipitates appear in the form of large (50– 100 nm diameter) electron-dense granules with a hollow electron transparent core, and are always found in immediate proximity to the IM [43]. Formation of the Ca²⁺ and phosphate precipitates is thought to be the major mechanism of Ca²⁺ storage in mitochondria [4,43,44]. It has been suggested that a protein or other matrix constituents may serve as a nucleation center facilitating formation of the Ca²⁺ precipitates [43]. Indeed, the presence of a protein may explain the always amorphous nature of Ca²⁺-phosphate precipitates, which is somewhat puzzling because hydroxyapatite [Ca₅(PO4)₃(OH)], the most commonly found composition of mitochondrial Ca²⁺ and Pi precipitates, is crystalline. In blood, where high levels of Ca²⁺ and Pi are standard, a protein called ‘fetuin’ had been shown to inhibit the precipitation of hydroxyapatite from supersaturated solutions of calcium and phosphate [45]. Perhaps a similar protein serves the same function in the mitochondrial matrix. The presence of substantial amounts of the Ca²⁺-binding proteins mitocalcin [46], calbindin-28k and calbindin-30k (calretinin) in a particulate fraction of rat brain [47] and in brain mitochondria [48,49] has been demonstrated previously, and annexin I [50] and annexin VI [51] were found in liver mitochondria. At least some of these proteins (annexin VI) serve as nucleation factors in vitro [52]. However, the contribution of these proteins to mitochondrial Ca²⁺ storage has not been examined. Although Ca²⁺-binding matrix-located proteins are the main candidates for the role of nucleation factors, non-protein factors such as mitochondrial DNA cannot be ruled out, as the Ca²⁺-binding ability of DNA is well known [53].

The Ca²⁺ and phosphate granules have been isolated from Ca²⁺-loaded rat liver mitochondria and their composition assessed [54]. The granules contained significant amount of carbon and nitrogen, indicating the presence of protein(s). However, the protein was not isolated or identified because the focus of that study was identifying the molecular form of the Ca²⁺ and phosphate precipitate. In addition, protein identification techniques were much more time-consuming and costly in 1967 when the study was performed than they are now. The Ca²⁺-phosphate precipitates are discussed in more detail elsewhere [3].

Proteins implicated in PTP formation

One of the most dramatic manifestations of abnormal Ca²⁺ homeostasis in mitochondria is the opening of a large channel called the ‘permeability transition pore’ (PTP) in the inner membrane that renders them incapable of energy production and can result in cell death by either apoptosis or necrosis. The functional and physiological aspects of PTP and mitochondrial Ca²⁺ transport have been reviewed extensively [55–61].

After a certain tissue-dependent threshold for the accumulated Ca²⁺ is reached, the permeability of IM to solutes abruptly increases due to opening of the PTP, a protein-mediated pore in the IM. It has been suggested that opening of the PTP is probably triggered by the increase in the free matrix Ca²⁺ concentration [55–61], although the evidence for this is ambiguous. The free matrix Ca²⁺ does increase somewhat upon loading mitochondria with Ca²⁺, but does not exhibit any abrupt changes immediately before PTP opening [44]. On the other hand, it increases significantly upon loading of brain mitochondria with Ca²⁺ if opening of the PTP is inhibited [44]. Therefore, it is not clear whether opening of the PTP is triggered by the free matrix Ca²⁺ or bound matrix Ca²⁺, or both together. In either case, factors affecting formation of the Ca²⁺ and phosphate precipitates are expected to influence the Ca²⁺ threshold for PTP activation. Changes in the precipitation characteristics are expected to have a strong effect on the overall Ca²⁺ retention and storage in mitochondria.

The molecular identity of the protein(s) that actually form the PTP channel remains a mystery. Past studies identified several proteins involved in the PTP formation or modulation, such as the voltage-dependent anion channel, the adenine nucleotide translocator (ANT), and, more recently, the mitochondrial phosphate transporter PIC [59], although none of these proteins are currently thought to directly form the PTP channel [59,60]. Until recently, mitochondrial ANT was viewed as the most likely PTP-forming protein [57,59,60]. ANT may also interact with another matrix protein, cyclophilin D (CypD) [62]. The latter is a target of cyclosporin A, a peptide inhibitor of PTP. Although the role of CypD in modulating the Ca²⁺ threshold for PTP activation had recently been strongly confirmed [63–66], the role of ANT in PTP formation was strongly challenged [67]. The PTP in mitochondria isolated from CypD-ablated mice is insensitive to cyclosporin A and exhibits a much higher Ca²⁺ threshold [63–66], whereas the PTP is activated by Ca²⁺ accumulation in ANT-deficient liver mitochondria isolated from mice that were genetically ablated of ANT in the liver [67].

Experimental evidence supporting a role for the voltage-dependent anion channel in PTP formation has been discussed in detail [60,61]. However, the recent finding that genetic ablation of any of the three mammalian voltage-dependent anion channel isoforms or all of them together does not affect Ca²⁺-induced PTP opening strongly suggests that voltage-dependent anion channels are not involved in PTP channel formation [68]. While the molecular identity of the PTP channel-forming protein remains unknown, a role for the mitochondrial phosphate transporter PIC in PTP formation cannot be ruled out [59].

An alternative model of PTP implicates no specific proteins in the role of the PTP channel [69]. According to this model, the pore is formed by aggregation of some misfolded integral membrane proteins; transport through these proteins is normally be blocked by cyclophilin D or other chaperones but Ca²⁺ accumulation and or oxidative stress increase the number of misfolded proteins. When the number of protein clusters exceeds the number of chaperones available to block transport, opening of ‘unregulated pores’ that are no longer sensitive to PTP inhibitors such as cyclosporin A would occur [69]. Although interesting, this model fails to account for approximately half of the known PTP features, such as its fast reversibility by Ca²⁺ chelation, its sensitivity to regulation by matrix pH, transmembrane voltage, fixed pore size, etc. [60].

Overall, literature analysis [55–61] allow us to formulate a minimum set of requirements to be fulfilled by a plausible candidate for the role of PTP channel-forming protein. First, it has to be able to bind to the IM. Although it has always been presumed that the PTP-forming protein is an integral protein embedded in the IM, there is no evidence to support this presumption. The PTP-forming protein does not have to be located in the IM before it forms a channel; it may simply bind to the IM and move into the IM upon activation. There are numerous examples of proteins moving between various cellular compartments and membranes upon activation. Second, it has to be able to form a large (approximately 2–3 nm diameter) transmembrane channel to allow the passage of charged and uncharged solutes up to 1500 Da. Third, it has to be able to form the channel in a fully reversible, fast and Ca²⁺-dependent fashion, as the full reversibility of PTP opening upon Ca²⁺ chelation and its fast transition between an open and a closed state are well known [55]. Finally, the molecular structure of this putative protein should ideally feature Ca²⁺-binding sites, reactive thiol groups to facilitate channel formation upon oxidation, and conformationally critical β-sheets, as suggested by the effect of cyclophilin D, which is a peptidyl-prolyl-cis/trans isomerase.

The above features are a minimum set of features that, if present in a single protein, would strongly implicate it as a plausible candidate for the role of PTP channel-forming protein. Other PTP features such as regulation by adenine nucleotides and effectors of ANT may be due to other proteins that interact with this putative PTP channel and modulate its activity.

gC1qR

Although there may be a number of unknown mitochondrial proteins that fulfil these requirements, at least one ubiquitous and evolutionary conserved eukaryotic protein, gC1qR, meets these requirements in full. As mentioned earlier, the molecular identity of the protein(s) that actually form the PTP channel remains the most intriguing question. On the basis of structural and other information, we hypothesize that the gC1qR protein, also known as p32, gC1QR/33, splicing factor 2 (SF2) and hyoluronan-binding protein 1 (HABP1), is the most plausible candidate for the role of PTP channel-forming protein. gC1qR is a 23.8. kDa multifunctional cellular protein (although it migrates at 33 kDa in SDS/PAGE, probably due to glycosylation and strong charges [70]) that was originally isolated and characterized as a plasma membrane protein with high affinity for the globular ‘heads’ of the complement component C1q, but was actually just one of its diverse binding partners. gC1qR is synthesized with an N-terminal mitochondrial targeting sequence that is cleaved after import into mitochondria. The matrix location of this protein is firmly established [71–73]; however, it is also found in other cellular compartments [74]. In humans, it is encoded by the C1qBP gene [75]. The function of this protein in mitochondria is not known. gC1qR is an evolutionarily conserved eukaryotic protein. Homologous genes have been identified in a number of eukaryotic species, ranging from fungi to mammals [74], compatible with its potential role in PTP as the latter is also ubiquitous among species. Its mitochondrial location and unique structural features make gC1qR protein a highly plausible candidate for the role of PTP channel, as discussed below.

Structural features of gC1qR as related to PTP and Ca²⁺ uniporter

There are striking structural features of this protein that make it perfect for the roles of PTP channel and calvectin-like ‘Ca²⁺ uniporter’. The putative role of gC1qR in PTP formation was suggested previously [76], but this hypothesis did not attract much interest, mostly due to then dominant view that the PTP is formed by ANT, which has now been disproved [67]. gC1qR is a doughnut-shaped trimer with an outer diameter of approximately 7.5 nm, a mean inner diameter of approximately 2 nm, and a thickness of approximately 3 nm. Each monomer consists of seven consecutive β-strands forming a highly twisted antiparallel β-sheet. The channel wall is formed by the β-sheets from all three subunits. This makes gC1qR a potential target for cyclophilin D, a well-known modulator of PTP sensitivity to Ca²⁺ [55,58–60,63–66]. The latter belongs to a class of enzymes called peptidyl-prolyl-cis /trans-isomerases that act upon prolines in β-sheets, resulting in a conformation change in the target protein. The gC1qR is a very acidic protein with a highly asymmetric negative charge distribution on the protein surface. One side of the doughnut and the inside portion of the channel possess a high number of negatively charged residues; the opposite side is much less negatively charged. These features permit the gC1qR trimer to interact with other charged surfaces such as phospholipid membranes or other proteins, and the ability of gC1qR trimer to bind to the plasma membrane surface is well documented [74]. Moreover, these interactions are inherently sensitive to modulation by divalent cations such as Mg²⁺ and Ca²⁺, which can bind to gC1qR and compensate its surface charges. As gC1qR is acidic, its interactions with other proteins may be weakened by increasing the acidity of the medium. It is well known that PTP is inhibited at low pH, probably due to release of cyclophilin D from its putative binding site on the PTP [77]. Thus, it is not unreasonable to suggest that gC1qR has a putative binding site for cyclophilin D. Each monomer of gC1qR has one cysteine at residue 186 (Cys186). This residue does not form inter-chain disulfide bonds between the monomers of a single gC1qR trimer [74]. However, under oxidative conditions, it forms a disulfide bond between monomers of different gC1qR trimers, thereby forming a hexameric structure consisting of two trimers. This complex has a much higher hydrodynamic radius and altered ligand-binding properties [78]. Lastly, a very important feature of the gC1qR trimer is that its inner channel is very large (approximately 2 nm diameter in the compact trimeric crystal), allowing easy passage of solutes with molecular mass 0.4–3 kDa [76], irrespective of their nature and electric charge, which is compatible with the necessary PTP channel properties. Moreover, the primary sequence of gC1qR predicts three putative N-linked glycosylation sites, and the protein was indeed found to be strongly glycosylated [70]. Considering that gC1qR has to be present at the cytosolic side of the inner mitochondrial membrane (see below) for its many activities to be possible, the presence of glycosyl residues should render it a target for ruthenium red binding, as it would be expected form a putative Ca²⁺ uniporter.

Interaction of gC1qR with pro-apototic and other Ca²⁺-dependent cell signaling cascades

Both the mitochondrial PTP and Ca²⁺ uniporter are implicated in so many pathological and physiological scenarios that it would be virtually impossible for the proteins involved in these systems to avoid multiple interactions with other signaling and regulatory proteins. It has been shown that gC1qR is a partner of the pro-apoptotic protein Hrk [79], a mammalian BH3-only protein. Multiple lines of experimental evidence verified a specific interaction and co-localization of Hrk and gC1qR, both of which depended on the presence of the highly conserved C-terminal region of gC1qR. Hrk-induced apoptosis was suppressed by expression of gC1qR mutants lacking the N-terminal mitochondrial signal sequence (gC1qR_74–282) or the conserved C-terminal region (gC1qR_1–221), which inhibit competitive binding of Hrk to the gC1qR protein and disrupt the channel function of gC1qR, respectively [79]. Another recently discovered pro-apototic protein, smARF, also binds to mitochondrial gC1qR. This protein is known to induce autophagic cell death. gC1qR physically interacts with both human and murine smARF, and co-localizes with them to the mitochondria. Remarkably, knock-down of gC1qR levels significantly reduced the steady-state levels of smARF by increasing its turnover. As a consequence, the ability of ectopically expressed smARF to induce autophagy was significantly reduced. gC1qR stabilizes the smARF [80]. Mitochondrial gC1qR has also been shown to be a substrate for ERK and an integral part of the MAP kinase cascade [81]. It is also a general protein kinase C (PKC)-binding protein [70]; it binds to and regulates the activity of PKC isoforms PKC-α, PKC-ζ, PKC-δ, and PKC-µ (the latter being constitutively associated with gC1qR at mitochondrial membranes) without being their substrate [82]. Several lines of evidence suggest that mitochondrial PKC may directly regulate PTP status, at least in heart [83], and the involvement of PTP in cell apoptosis is suggested in so many publications that it is difficult to provide a key reference. The most recent data strongly linking gC1qR to Ca²⁺-related mitochondrial dysfunction and apoptosis was obtained by Chowdhury et al. [84], who demonstrated that constitutively expressing gC1qR in a normal murine fibroblast cell line induced growth perturbation, swelling and derangements of cristae in cell mitochondria, release of cytochrome c and formation of apoptosome complexes. They also showed that mitochondrial dysfunction was due to a gradual increase in ROS generation in cells over-expressing gC1qR. Together with ROS generation, they found an increased Ca²⁺ influx in mitochondria, resulting in a decreased membrane potential and severe inhibition of the respiratory chain complex I [84].

How could gC1qR participate in both the PTP channel formation and in Ca²⁺ uniport?

On the basis of the structural features of gC1qR, a novel mechanism of PTP formation and Ca²⁺ transport can be proposed that involves the same protein in both systems (Fig. 2). According to this model, gC1qR ‘protomers’ are present in the intermembrane space, the inner membrane and the matrix of mitochondria, but the ‘Ca²⁺ uniporter’, consisting in this case of IM-and perhaps matrix-located gC1qR protomers, is not assembled to its fully active form. When a threshold of Ca²⁺ is reached, a few protomers of gC1qR migrate from the intermembrane space to the IM and bind to the protomer located in the IM. This creates a fully functional Ca²⁺ uniporter. The binding does not occur in the presence of ruthenium red, which blocks it by interacting with the glycosyl residues of IM-embedded protomers. Prolonged accumulation of Ca²⁺ results in its concentration in the matrix and the intermembrane space exceeding another Ca²⁺ threshold, triggering the formation of a larger, multi-component PTP channel (Fig. 2). Both the Ca²⁺ transport system and the PTP have to be pre-assembled for their full activity, and our hypothesis is that they are two stages of the same process of Ca²⁺-dependent assembly of gC1qR protomers, perhaps in co-operation with some other regulatory proteins. The PTP channel in the IM may be formed by a gC1qR multimer comprising several (e.g. three in Fig. 2) identical gC1qR trimers stacked onto each other. Formation of the multimer means that the structure acquires sufficient hydrophobicity to move into the IM and form a transmembrane channel. Formation of this structure is caused by Ca²⁺ accumulation in the matrix, but not necessarily an increase in free Ca²⁺ concentration. For example, gC1qR trimers, which are inherently capable of binding Me²⁺ ions, may initially be sequestered by another Me²⁺-binding protein in the matrix as a gC1qR*Mg²⁺ complex, and the Me₂⁺-binding protein may also be capable of serving as a nucleation factor catalyzing the formation of Ca²⁺-phosphate precipitates. When the latter accumulate, they displace gC1qR from the complex, thereby priming it to PTP. The next step may be a change in gC1qR conformation that would increase the probability of its interaction with another gC1qR trimer to form a hexamer. This conformational change may be facilitated by binding to cyclophilin D. The last step is a further Ca²⁺-dependent change in conformation of this newly formed hexameric gC1qR–cyclophilin D complex to allow it to translocate into the IM and form the PTP. In this model, chelating free Ca²⁺ by EGTA would force the structure to leave the IM, suppressing channel formation, but will not reverse the entire process because it could not quickly remove the Ca²⁺-phosphate precipitates that have already formed. Therefore, the pore-forming complex will remain primed and ready for repeated cycles of PTP opening and closing. On the other hand, Cys186 mediated formation of disulfide bridges between the two gC1qR trimers would render the channel structure permanent and insensitive to modulation by Ca²⁺ or cyclophilin D, thereby producing an ‘unregulated’ pore. This PTP model can easily accommodate most if not all known data on PTP activation and regulation, including its sensitivity to a wide variety of chemically dissimilar compounds and even to the changes in the conformation of major IM proteins that are capable of modifying the surface charge of the IM, such as ANT.

Fig. 2.

Proposed model of CA²⁺ uniporter and PTP assembly. The ‘protomers’ (flat disks) of a putative protein forming the ‘Ca²⁺ uniporter’ (‘U’) and PTP (e.g. gC1qR as discussed in the text) are present in the intermembrane space, the inner membrane and the matrix of mitochondria, but the ‘Ca²⁺ uniporter’ consisting of IM- and matrix-located protomers, is not assembled to its fully active form. When a threshold amount of Ca²⁺ is reached, a few protomers migrate from the intermembrane space to the IM and bind to the protomer located in the IM. This creates a fully functional ‘Ca²⁺ uniporter’. Such binding does not occur in the presence of ruthenium red, which blocks it by interacting with the glycosyl residues of IM-embedded protomers. Accumulated Ca²⁺ and phosphate bind to an unidentified ‘nucleation factor’ (‘n.f.’) that prevents the formation of crystalline Ca²⁺-phosphate precipitates. Upon prolonged accumulation of Ca²⁺, the storage capacity of this ‘nucleation factor’ is exceeded, and the Ca²⁺ concentration in the matrix and the intermembrane space rises above the threshold for PTP assembly, thereby triggering formation of a larger multi-component PTP channel (see text for further details).

Concluding remarks

Obviously, this model is highly speculative as there are no data that directly support its key features. However, there are three predictions about this mechanism that can be verified experimentally. First, gC1qR has to physically move into the IM to form a PTP, i.e., upon accumulation of Ca²⁺ and phosphate, the distribution of free gC1qR between the mitochondrial compartments should change dramatically towards the IM, preceding opening of the PTP. Second, knocking out the gC1qR protein should prevent Ca²⁺-induced PTP formation or at least severely increase its Ca²⁺ threshold. Third, antibodies against gC1qR should inhibit Ca²⁺ uptake in mitoplasts. We are currently trying to verify these predictions experimentally.

Abbreviations

ANT

adenine nucleotide translocase

CGP37157

7-chloro-5-(2-chlorophenyl)-1,5-dihydro-4,1-benzothiazepin-2(3H)-one

CypD

cyclophilin D

EKR

extracellular-signal-regulated kinase

Hrk

a product of harakiri gene

IM

inner mitochondrial membrane

PTP

permeability transition pore

Ru360

C₂H₂₆Cl₃N₈O₅Ru₂

smARF

“short mitochondrial ARF”, a short isoform of p19ARF protein