Ca2+ signaling in mitochondria is important to tune mitochondrial function to a variety of extracellular stimuli. The main mechanism is Ca2+ entry in mitochondria via the Ca2+ uniporter followed by Ca2+ activation of three dehydrogenases in the mitochondrial matrix. This results in increases in mitochondrial NADH/NAD ratios and ATP levels and increased substrate uptake by mitochondria. We review evidence gathered more than 20 years ago and recent work indicating that substrate uptake, mitochondrial NADH/NAD ratios, and ATP levels may be also activated in response to cytosolic Ca2+ signals via a mechanism that does not require the entry of Ca2+ in mitochondria, a mechanism depending on the activity of Ca2+-dependent mitochondrial carriers (CaMC). CaMCs fall into two groups, the aspartate-glutamate carriers (AGC) and the ATP-Mg/Pi carriers, also named SCaMC (for short CaMC). The two mammalian AGCs, aralar and citrin, are members of the malate-aspartate NADH shuttle, and citrin, the liver AGC, is also a member of the urea cycle. Both types of CaMCs are activated by Ca2+ in the intermembrane space and function together with the Ca2+ uniporter in decoding the Ca2+ signal into a mitochondrial response.
Ca2+ signaling in mitochondria is important to regulate mitochondrial function in response to intra- and extracellular cues. The main mechanism whereby Ca2+ modulates mitochondrial function involves Ca2+ entry in mitochondria via the Ca2+ uniporter or rapid uptake mode (RAM) mechanisms (143, 145, 190) followed by the activation by Ca2+ of three dehydrogenases in the mitochondrial matrix (247). This causes an increase in mitochondrial NADH/NAD ratios which may result in increased energy available for mitochondrial functions. In addition, matrix Ca2+ has been suggested to activate ATP synthesis through a direct effect on F0F1-ATPase (31). Here, we review work dating more than 30 years ago and new evidence indicating that substrate uptake, mitochondrial NADH/NAD ratios, and ATP levels may be also activated in response to cytosolic Ca2+ signals via a mechanism that does not require the entry of Ca2+ in mitochondria, a mechanism depending on the activity of the Ca2+-binding mitochondrial carriers.
Mitochondrial carriers (MCs) are integral proteins of the mitochondrial inner membrane that function in the shuttling of metabolites, nucleotides, and cofactors between the cytosol and mitochondria (408), a subset of which are the Ca2+-binding MCs (CaMCs) (95). CaMCs fall into two groups: the aspartate/glutamate carriers (AGC) (92, 95, 195, 286) and the ATP-Mg/Pi carriers, also named SCaMC (for short CaMC) (94, 122). The two mammalian AGCs, aralar (AGC1) and citrin (AGC2), are members of the malate-aspartate NADH shuttle, and citrin, the liver AGC2, is also a member of the urea cycle (217). The transport reaction catalyzed by the AGCs is the only irreversible step in the malate-aspartate shuttle (217, 218) and a main point of regulation. The ATP-Mg/Pi carriers exchange cytosolic ATP-Mg for mitochondrial Pi and are involved in loading newly formed mitochondria with adenine nucleotides and modulating mitochondrial adenine nucleotide content in response to hormones (10, 13). The activity of both types of CaMCs is stimulated by Ca2+ acting on the external face of the inner mitochondrial membrane (286, 163, 276).
To perform a role in transducing Ca2+ signals to mitochondria, CaMCs must be poised for activation under normal cell conditions. The existing data indicate that the reaction catalyzed by these carriers is energetically favored under resting conditions but limited by the low activity of the carrier. Ca2+ activation removes this limitation, allowing the transport reaction to occur. Available data indicate that such processes occur at the expense of relatively small Ca2+ signals, with limited capacity to activate Ca2+ entry in mitochondria, and may be responsible for the energizing role of low levels of hormones and neurotransmitters. These carriers exist in organisms such as Saccharomyces cerevisiae that lack a Ca2+ uniporter. In particular, S. cerevisiae Sal1p appears to be the most basic CaMC, in fact the only mechanism conveying Ca2+ signals to mitochondria in this organism.
II. CALCIUM BINDING MITOCHONDRIAL CARRIER GENE AND PROTEINS
Metabolite transport in and out of mitochondria is essential for mitochondrial function. The MC family of proteins comprises a group of structurally related proteins that exist only in eukaryotes (see Refs. 96, 211, 283, 418 for recent reviews). MC members are integral proteins in the inner mitochondrial membrane, with a molecular mass of ∼35 kDa, six transmembrane spanning segments, and a tripartite structure (408) that catalyze the transport of metabolites, nucleotides and derivatives, and cofactors in mitochondria. The CaMC are a subset of MCs that have long hydrophilic amino-terminal extensions harboring EF-hand Ca2+-binding motifs that face the intermembrane space (92, 95) (Fig. 1).
The CaMC family as such was discovered when the genome sequences of S. cerevisiae (136) and Caenorhabditis elegans (78) became available. At that time and still now, the identity of the mitochondrial Ca2+ transporters responsible for Ca2+ signaling to mitochondria was unknown despite extensive efforts to characterize these proteins (342, 405; see also Refs. 63, 340 for review). The genome of S. cerevisiae was found to contain sequences for ∼35 members of the MC superfamily (117), and database analysis revealed that the genome of C. elegans also contained a number of similar sequences. A S. cerevisiae gene, YNL083w (364) (Table 1), and three sequences in C. elegans code for putative proteins containing sequence motifs corresponding to EF-hands (92), a property not described so far in members of the MC family. EF-hands (where EF stands for the E and F alpha helices of parvalbumin) are Ca2+-binding domains present in different Ca2+-binding proteins (269), and their presence in a putative MC protein opened up the possibility that the protein might be involved in the handling of Ca2+, or in Ca2+-regulated metabolite transport in mitochondria. The realization of the potential implications of this finding was the starting point for the cloning and characterization of all members of the CaMC subfamily. Soon after, it became clear that CaMCs are not involved in Ca2+ transport.
The CaMC family is made up of two types of carriers: the AGCs and the ATP-Mg/Pi carriers. The AGCs have long amino-terminal extensions, hence the name of aralar, coined by Satrústegui to designate the first of those carriers cloned by Araceli del Arco (Araceli, hiperlargo) (95). AGCs are represented by two isoforms in mammals, aralar (or aralar1) (93) encoded by SLC25A12, and citrin, named after type II citrullinemia, a disease caused by mutations in citrin (92, 195), encoded by SLC25A13 (195, 358). These isoforms are mainly expressed in brain, skeletal muscle, and heart (aralar, AGC1) and liver (citrin, AGC2).
The ATP-Mg/Pi carriers were first described as CaMCs with short amino-terminal extensions or SCaMCs and are represented by three main isoforms: SCaMC 1–3 (91, 94, 122) with a number of variants arising from alternative splicing. The identification of the transporter function of SCaMCs was carried out by Palmieri’s group (122) after reconstitution of the recombinant proteins into proteoliposomes. Throughout this manuscript the acronyms SCaMC 1–3 will be used to designate the particular isoforms of the ATP-Mg/Pi carrier, rather than the acronyms used previously (APC 1–3, from ATP-Mg/Pi carrier, Ref. 122), to avoid confusions with the well-known APC protein, adenomatous polyposis coli, a tumor suppressor multifunctional protein that is mutated in a majority of colon cancers (160).
The transporter function of the AGCs was reported long ago (27). These carriers are important for the urea cycle and malate-aspartate NADH shuttle and have been the subject of a number of studies including purification and functional reconstitution in proteoliposomes. However, Ca2+ regulation of the AGCs had escaped attention.
An ATP transport pathway different from the adenine nucleotide translocases was first described in fetal liver mitochondria where it was found to be important for the net uptake of adenine nucleotides immediately after birth (299, 300, 373). A few years later, a series of careful studies carried out by Aprille and co-workers (276) and Haynes and co-workers (163) resulted in the discovery that the ATP-Mg/Pi carrier was regulated by Ca2+ from the external side of the inner mitochondrial membrane (163, 276). In retrospect, this seminal finding set the stage for the identification of MCs endowed with Ca2+ regulation properties.
III. ASPARTATE-GLUTAMATE CARRIERS
A. Mammalian AGCs
Aralar/AGC1 and citrin/AGC2 are coded for SLC25a12 and SLC25a13, in human chromosomes 2q31 (80, 339) and 7q21.3 (358), respectively. The two genes have the same genomic structure, with 18 exons (358), also conserved in mice (180, 359).
1. Development and tissue distribution of aralar/AGC1 and citrin/AGC2
Aralar, citrin, or both isoforms are present in early preimplantation embryos, as lactate supports development from the two-cell stage onwards, through the operation of the malate-aspartate NADH shuttle (214). Thus the AGCs are present very early in mouse development. Aralar and citrin have an overlapping distribution in embryogenesis (93, 358). Both isoforms are expressed strongly in branchial arches, dermomyotome, limb, and tail buds at early embryonic stages. The characteristic strong expression of citrin in liver is acquired postnatally, while that of aralar in heart and skeletal muscle begins during embryonic development (93). In the adult mouse, aralar is highly expressed in almost every tissue including brain, skeletal muscle, and heart (95) as well as in cells from the immune and hematopoietic tissues both at embryonic stages (embryonic liver) and in adults (Table 2). Many tissues express both isoforms, except hepatocytes from adult mouse liver where citrin is the only detectable AGC (93).
Aralar and mitochondrial aspartate aminotransferase expression decrease, whereas that of citrin increases in heart from desmin-deficient mice (129), and the parallel decreases of the former have been taken as an indication of decreases in malate-aspartate NADH shuttle (MAS) activity in these animals (129). Aralar expression decreases both in mouse and human skeletal muscle after prolonged feeding on high-fat diet, in parallel with the decrease of many other transcripts for oxidative phosphorylation proteins (359).
2. Import of AGC precursor proteins to mitochondria
Like other MCs, AGC precursor proteins are synthesized in the cytosol and then imported to mitochondria. Targeting information in CaMCs is contained in their carboxy-terminal half, and all amino-truncated variants are targeted into mitochondrial membranes (91, 92, 94, 96, 286).
The mechanisms of import of MC precursor proteins and CaMCs have been discussed in detail in recent reviews (96, 196, 309). The insertion into the inner membrane of members of MC family is mediated by the TIM22 (translocase of the inner membrane) complex or carrier translocase (Fig. 2B). The TIM22 pathway contains components at the inner membrane as well as in the intermembrane space, referred to as the small Tim proteins, Tim8p, Tim9p, Tim10p, Tim12p, and Tim13p (198). These small Tims act as chaperone-like molecules to guide hydrophobic precursors across the aqueous intermembrane space, whereas the inner membrane complex mediates their insertion into the inner membrane (88, 89, 412). The small Tim proteins assemble into 70-kDa complexes, Tim8p with Tim13p and Tim9p with Tim10p in a 3:3 ratio (Fig. 2B) (88, 89, 197). These 70-kDa complexes have different substrate specificities, with the classical MC precursor proteins binding predominantly to the Tim9p-Tim10p complex (89, 289). In contrast, the Tim8p-Tim13p complex binds to the human and yeast AGC precursor proteins through their long hydrophilic amino-terminal domains (318). Pathway selection seems to be determined by the presence of the hydrophilic amino-terminal extension of CaMCs; thus the import of full-length aralar decreases 80% in mitochondria that lack the Tim8p-Tim13p complex, and shortening of the aralar amino-terminal domain results in preferential binding of aralar to the Tim9p-Tim10p complex (318). Mutations in the locus for human Tim8p (DDP1/TIMM8a locus) cause Mohr-Tranebjaerg syndrome (MTS/DFN-1, deafness/dystonia syndrome) (198, 317), and a lymphoblast cell line derived from an MTS patient shows decreased mitochondrial NADH levels due to defects in the MAS where AGC is the rate-limiting component (318). The Tim8p-Tim13p complex probably also participates in SCaMCs import as suggested by the differences in import efficiency detected for overexpressed full-length and amino-truncated versions of SCaMC3 protein (91).
3. Transport activity
The AGCs aralar and citrin catalyze the exchange of aspartate with glutamate plus a proton, as demonstrated in intact mitochondria (215, 216, 218) and in proteoliposomes reconstituted with the native purified protein (97, 98) or recombinant AGCs (286). Thus there is one positive charge that moves across the membrane together with the two amino acids, and the exchange is electrophoretic (114, 215–218, 352). This property of the AGC is rather uncommon among the MC family, only shared by the adenine nucleotide translocases and the uncoupling proteins. Because the aspartate-glutamate exchange reaction is strongly dependent on the membrane potential, the release of aspartate in exchange with glutamate is strongly favored (97, 98, 215, 216, 218, 286). Thus, in energized mitochondria, this carrier is essentially unidirectional. The affinity for glutamate is lower than that for aspartate (Km values of ∼0.2 and 0.05 mM, respectively), both for the native or recombinant AGCs (97, 99, 286), and these values do not change with the membrane potential (286). The effect of the membrane potential is a pronounced increase in transport activity that depends on an electrogenic modulation of the transport rate constant of the negatively charged carrier-aspartate complex, and not to a change in the substrate affinity of the carrier (286). The two AGCs also catalyze 1:1 electroneutral homoexchange reactions with aspartate, glutamate, and cysteine sulfinate and are strongly inhibited by pyridoxal phosphate (45, 101, 284, 286).
4. Functional roles of AGCs in urea cycle and the MAS
AGCs are required for the transfer of mitochondrial aspartate to the cytosol, an important step in urea synthesis (Fig. 3) and in gluconeogenesis from lactate and amino acids. Arginino-succinate synthetase (ASS), the cytosolic enzyme condensing citrulline with aspartate, uses preferentially but not exclusively, mitochondrial versus cytosolic aspartate (250, 319, 334, 413). For recent reviews on this subject, see References 333 and 334. The role of citrin in the urea cycle is discussed in section iiiA5.
AGCs are also needed for the transfer of mitochondrial aspartate to the cytosol in the MAS, described by Borst (48). The shuttle transfers reducing equivalents of NADH from cytosol to mitochondria and is required during glycolysis and lactate metabolism. Besides the cytosolic and mitochondrial isoforms of malate dehydrogenase (cMDH, mMDH) and aspartate amino transferases (cAAT and mAAT), the shuttle uses two MCs, the AGC and the malate/α-ketoglutarate carrier (OGC) (Fig. 4). The presence of the AGCs in this cycle provides a unidirectional substrate flow along the shuttle that functions in the direction of reducing equivalent transfer from cytosol to mitochondria. MAS is important in reducing equivalent transfer in many tissues including liver, particularly human liver, where other NADH shuttle systems such as the glycerol-3-phosphate shuttle (GPS) have low activity, ∼1/20 of that in rodents (330).
MAS is also the dominant NADH shuttle in heart (219, 221, 328, 331, 346–348, 424), and both isoforms aralar and citrin are expressed in heart, including human heart (380) with a preferential enrichment of aralar in atria (93). In the fetal and immediate perinatal period, the heart relies on glycolysis as a source of energy, with ∼40% of the ATP production derived from glycolysis, but within days oxidative phosphorylation increases to provide 93% of the ATP as fatty acid utilization increases (see Ref. 369 for review). In neonatal hearts, lactate oxidation accounts for the majority of oxygen consumption (127). MAS activity of neonatal porcine heart mitochondria is about three times higher than that of adult heart (346, 347), in which fatty acid metabolism bypasses the NADH shuttles. Strikingly, the activity of the heart mitochondrial Ca2+ uniporter undergoes a similar decrease along postnatal life in the rat (32). In contrast to the developmental changes in MAS activity, mRNA and protein levels of AGCs in porcine and mouse heart are higher in adult than in neonates (35, 347), although a decrease in steady-state aralar mRNA levels was observed from postnatal day 9 onward (35). Studies of the levels of all other MAS members indicated that those of OGC decrease postnatally in pig and rabbit (141, 347) and probably limit shuttle function in adult heart (347). An upregulation of MAS could be of potential interest in heart failure and myocardium hypertrophy, when substrate utilization shifts away from fatty acids toward glucose and lactate (328, 369).
Glucose metabolism drives the stimulation of insulin secretion in pancreatic β-cells through increased ATP levels that lead to closure of KATP channels (237, 238, 251). Lactate dehydrogenase activity is low in pancreatic β-cells (177, 178, 354), and NADH shuttle systems are important for glucose-induced activation of mitochondrial metabolism and insulin secretion (112, 135, 377). Among NADH shuttle systems, MAS is particularly important, as insulin secretion is normal in mice with no functional mitochondrial FAD-GPDH, the mitochondrial member of the glycerol-phosphate shuttle, whereas it is blocked when MAS activity is inhibited in these transgenic mice (119). Aralar is the only AGC present in islets and β-cells (93, 326), and its levels or activity appear to limit MAS function in these cells. Thus overexpression of aralar in INS-1E or pancreatic β-cell leads to increases in glucose-induced reduction of mitochondrial NAD(P)H, mitochondrial membrane potential, and ATP levels along with an increase in glucose-induced insulin secretion (326). Glucose utilization was not changed by aralar overexpression, but lactate formation was decreased, suggesting that MAS activity regulates the fraction of pyruvate being either oxidized in mitochondria or reduced to lactate (326). On the other hand, aralar overexpression did not modify insulin secretion when pyruvate, rather than glucose, was used as fuel, showing that it contributes primarily upstream of mitochondria to efficiently couple glycolysis to mitochondrial metabolism (326). MAS is also important in neurons, especially during lactate utilization, and this is discussed in section iiiA5b.
Interestingly, MAS is also required in early embryos. Studies on the preimplantation mouse embryo carried out more than 30 years ago have shown that the zygote has an absolute requirement for pyruvate to complete the first cleavage division, whereas lactate and glucose are only capable to support development from the two- and eight-cell stages, respectively (56, 57). Mouse zygotes can use both lactate and pyruvate (213), but only pyruvate supports the first cleavage. In zygotes, but not in blastocysts, lactate oxidation and NAD/NADH ratios decreased with increasing lactate concentration, suggesting that lactate utilization by lactate dehydrogenase was limited due to an inability to provide NAD for the reaction (213). Although zygotes express all four MAS enzymes, the inability to transfer reducing equivalents to mitochondria was found to be due to a lack of MAS activity, aspartate being the rate-limiting factor probably due to the high Km of the cytosolic aspartate aminotransferase (214). This suggests that AGC levels or activity may limit MAS function at this particular developmental stage. Thus addition of exogenous aspartate enabled mouse zygotes to utilize lactate and develop normally (214).
5. Mutations and polymorphisms
Much of the actual roles of aralar and citrin in humans and rodents have been unveiled through the discovery of human diseases involving the two carriers and with the development of mouse models carrying disruptions in the two genes.
A) CITRIN, AGC2, AND SLC25a13.
A deficiency in citrin, the AGC2 isoform specifically expressed in liver, kidney, and heart encoded by gene SLC25A13, causes type 2 citrullinemia (CTLN2), an adult-onset liver disease predominantly found in the Japanese population (195). Unlike classical citrullinemia in children, which is associated with a deficiency in ASS, in CTLN2 the ASS gene is normal and ASS deficiency is only found in liver, but not in other organs that express ASS. Moreover, the residual amount of enzyme present in the patients displays normal kinetic properties (333, 335). The disorder was first reported in a group of patients with high serum citrulline and ammonia levels and neurological symptoms that closely resemble those of hepatic encephalopathy. However, a different disease, neonatal intrahepatic cholestasis caused by citrin deficiency (NICCD), was found in a number of neonates with idiopathic neonatal hepatitis that carry the same mutations in citrin found in CTLN2 patients (280, 392). These neonates have multiple metabolic abnormalities including aminoacidemia, galactosemia, hypoproteinemia, hypoglycemia, and cholestasis. NICCD is usually not severe, and it can be managed by nutritional manipulation, but a few patients develop CTLN2 after 10 years or more (332).
In CTLN2 patients, mutations in citrin cause a complete lack of citrin protein (195). The generation of citrin-knockout mice has revealed many of the consequences of citrin deficiency (359). Liver mitochondria from citrin(−/−) mice had a drastic reduction of malate-aspartate NADH shuttle activity and aspartate efflux with no compensatory changes in the glycerophosphate NADH shuttle. Gluconeogenesis from lactate, but not from pyruvate, was severely impaired, in agreement with MAS requirement for gluconeogenesis from reduced substrates. Urea production from ammonia was severely reduced in liver from citrin(−/−) mice, and this was strongly alleviated in the presence of asparagine, indicating that aspartate was the major limiting factor. All of these metabolic characteristics are consistent with a primary failure in aspartate efflux from mitochondria and malate-aspartate NADH shuttle activity (359). However, despite the deficits in AGC-dependent pathways, citrin(−/−) mice failed to show any overt pathological consequences. In particular, these mice had normal serum ammonia and citrulline levels and responded to increases in protein load in their diet by increasing urea formation. This suggests that mitochondrial-derived aspartate is probably not essential for the ASS reaction in mice (265, 359).
The characteristic decline in hepatic ASS levels and activity of CTLN2 patients has been attributed to the defect in the transport of aspartate from mitochondria to the cytosol that could inhibit the ASS reaction and destabilize the liver enzyme (286). ASS protein has full activity in NICCD neonates (381) and declines gradually during the life of CTLN2 patients. Whether the shortage of aspartate by itself destabilizes ASS or other factors participate in this defect is still unknown, but this particular defect in hepatic ASS levels is also absent in the citrin(−/−) mouse (359). Species-specific differences in gene regulation and in alternative metabolic pathways have been proposed to explain the different phenotypes. In particular, the glycerol phosphate shuttle is ∼20-fold more active in rodent than in human liver (330), suggesting that this shuttle could compensate for the lack of MAS activity in mice. There is indirect evidence that, to compensate for the lack of MAS, CTLN2 patients rely heavily on an alternative NADH shuttle, the malate-citrate shuttle pathway, which is used for the transfer of acetyl CoA from mitochondria to the cytosol for fatty acid synthesis (332, 333). Fat accumulation that could be supported by increased malate-citrate shuttle activity has been observed in the livers of patients with citrin deficiencies (both NICCD and CTLN2) (175, 334, 381, 392). The defects in ureogenesis in the citrin(−/−) liver could be partially corrected by the administration of asparagine, which also normalized the high levels of citrulline and low levels of ornithine (265, 359). These findings suggest that the aspartate arising from asparagine hydrolysis could serve as a substrate for the ASS reaction. In addition, pyruvate administration corrected the defect in ureogenesis, by lowering the cytosolic NADH/NAD ratio (265).
B) ARALAR, AGC1, AND SLC25A12.
Aralar is an ubiquitous AGC isoform highly expressed in brain and skeletal muscle. The liver parenchyma seems to be the only adult tissue devoid of aralar, although it is present in Kupffer cells, and also in the embryonic liver (93), a major hematopoietic organ.
The existence of aralar-knockout mice (180) has revealed new functions of this carrier, particularly a role in the supply of brain N-acetyl aspartate (NAA) for myelin lipid synthesis (Fig. 5). Aralar−/− mouse embryos are produced in normal numbers, as expected from the overlap of aralar and citrin expression in fetal tissues (35, 93). However, soon after birth these mice develop severe growth defects, neuromuscular deficiencies, and a reduced myelination in the central nervous system (CNS), but not in peripheral nerves. Brain and muscle mitochondria from aralar-knockout mice have a complete loss of MAS activity, a drastic decrease in respiration on glutamate plus malate, with no compensatory increase in respiration on pyruvate plus malate, and a pronounced decrease in aspartate efflux from mitochondria (180). Aralar deficiency results in a large drop of NAA and aspartate in the brain and primary neuronal cultures derived from Aralar−/− mice, together with a drop in myelin lipids and similar reductions in the levels of the major proteins in myelin, myelin basic protein (MBP), and myelin-associated oligodendrocyte basic protein (MOBP).
NAA plays a role in myelin lipid formation through the supply of acetyl groups (71, 90, 249) (Fig. 5). NAA produced in neurons (398) undergoes transaxonal transfer to oligodendrocytes where it supplies acetyl groups for the synthesis of myelin lipids. The NAA cleavage enzyme aspartoacylase is restricted primarily to oligodendrocytes (43, 191). Mutations in aspartoacylase cause Canavan’s disease, characterized by a spongy degeneration of the CNS, increased levels of NAA in the brain and body fluids, and an extensive loss of myelin. Aspartoacylase-deficient mice, generated as model for Canavan’s disease (244), show decreased myelin lipid synthesis and a very drastic (∼80%) reduction in the brain acetate concentrations (235), showing that the lack of NAA results in a lack of acetyl groups for myelin lipid synthesis.
The fall in brain aspartate levels and the lack of aspartate production in brain mitochondria from Aralar−/− mice indicate that the major route of aspartate production in the CNS is mitochondrial and that it depends on aralar for aspartate efflux to the cytosol. Indeed, the CNS pathology in aralar-null mice appears at about the time where the blood-brain barrier develops, preventing the entry of aspartate and other amino acids into most regions of the adult brain except for selected circumventricular regions (302). Therefore, from the first week of life onward, brain cells have to rely on their own endogenous production of aspartate, which takes place mostly in neurons, the brain cells with highest aralar expression (303). NAA is derived from neuronal aspartate, and its synthesis is catalyzed by aspartate-N-acetyltransferase in brain mitochondria (236, 334) and microsomes, which have four to five times higher activity than mitochondria (231). Thus the fall in NAA in brain and in cultured neurons from Aralar−/− mice is clearly associated with the defect in aspartate production.
The important brain functions underscored from the study of aralar-knockout mice may be relevant for two human diseases of the nervous system. A strong linkage and association of the gene for aralar, SLC25A12, with autism has been reported (304, 353; but see Ref. 46). Two polymorphisms that corresponded to single nucleotide polymorphisms (SNPs) located in flanking intronic sequences of exons 4 and 16 of SLC25A12 were associated with the disease, although the functional relevance of these SNPs is yet unknown (304, 353). The expression of a number of genes is regulated by cis-acting elements, and inherited variation in gene expression may contribute to disease (52). An altered regulation of gene expression without variations in its coding sequence could affect cellular functions when the levels of the gene product are limiting, as appears to be the case for aralar, as an overexpression of this protein in β-cells increases MAS function (326).
Impaired aralar activity could also contribute to the pathology of Mohr-Tranebjaerg syndrome (MTS/DFN-1, deafness/dystonia syndrome) caused by mutations in the locus for human Tim8p (DDP1/TIMM8a locus) (198, 317; as discussed in sect. iiiA2). Together with Tim13, Tim8p forms a 70-kDa complex specifically utilized by AGCs (318), and possibly also the SCaMCs (89) on their way to the inner membrane, a complex with prominent expression in large neurons in the brain (318). The mutations in Tim8p in MTS/DFN-1 patients appear to result in impaired import of aralar, as a lymphoblast cell line derived from a MTS patient shows decreased mitochondrial NADH levels due to defects in MAS activity (318).
Thus genetic variations that change the expression of aralar, or an impaired import of aralar in mitochondria, could lead to changes in NADH shuttle function, respiration on glutamate and malate, aspartate synthesis by neuronal mitochondria, and NAA production, which could be related to these CNS diseases. A decrease in NAA in different brain regions of autistic patients has been reported (130, 281).
B. Nonmammalian AGCs
A glutamate-aspartate MC is present in all known eukaryotic genomes (96). In most higher eukaryotes AGCs are highly similar to the mammalian genes and have EF-hand Ca2+ binding motifs in equivalent positions. A prominent exception to this rule is S. cerevisiae, in which the long amino-terminal extension of the agc1p protein lacks Ca2+-binding motifs (69) (Table 1).
1. Transport activity
The carboxy-terminal domain of yeast agc1p, that encompasses the whole MC homologous region, has been expressed in bacteria and reconstituted in phospholipid vesicles. Its transport properties were found to be similar to those of the mammalian AGCs in terms of substrate specificity, affinity, inhibitor sensitivity, and voltage dependence but differed from mammalian AGCs in that yeast agcp is able to catalyze the uniport of aspartate, and to a greater extent of glutamate, in addition to the aspartate/glutamate exchange (69).
The carboxy-terminal domains of agc1p and mammalian AGCs are highly similar, suggesting that both arise from a common ancestor. This ancestor probably duplicated in mammals to give rise to the AGCs that catalyze a strict asp/glut exchange, and glutamate carriers (GCs) that catalyze the uniport of glutamate driven by the pH gradient (123), and evolved into agc1p in S. cerevisiae, that catalyzes both transport activities (69). The second of these activities, the uniport of glutamate, is thought to be required to provide mitochondrial glutamate for transamination reactions and as a carbon and nitrogen source for ornithine synthesis, as discussed below (69).
2. Role of yeast AGC in amino acid biosynthesis and in shuttling reducing equivalents to mitochondria
The activity of agc1p as a glutamate uniporter is probably responsible for changes in amino acid levels and labeling from [13C]acetate in yeast agc1p deletion mutants. These mutants had reduced valine, ornithine, and citrulline levels (69) that were explained as a consequence of reduced uptake of glutamate in mitochondria, with subsequent reduction in mitochondrial glutamate levels, since glutamate dehydrogenase is not mitochondrial but cytosolic in S. cerevisiae (169). Mitochondrial glutamate is required for transamination reactions within mitochondria (valine synthesis), and as carbon and nitrogen source for ornithine synthesis, all of which are reduced in agc1p mutants (69).
The existence of a malate-aspartate NADH shuttle in S. cerevisiae has been questioned for a long time, because yeast mitochondria harbor external NADH dehydrogenases that oxidize NADH on the external face of the inner mitochondrial membrane (233) and lack complex I in the respiratory chain. However, all the components of the shuttle exist in yeast, particularly the MCs for α-ketoglutarate (α-KG) and malate exchange, Odc1 and Odc2, oxodicarboxylate carriers that can transport α-KG, α-ketoadipate, and malate (285, 386, 396), and the aspartate-glutamate exchanger agc1p. Growth on oleic acid produces large amounts of NADH in peroxisomes that require shuttle systems for reducing equivalent transfer to mitochondria. MAS is the shuttle system involved in the transfer of reducing equivalents from the cytosol to the mitochondria in this particular condition, as neither the odc1Δodc2Δ double mutants (386) or agc1p deletion mutants (69) grow on oleic acid. Because growth on ethanol, but not on acetate, is unimpaired (69), this defect is not due to defects in gluconeogenesis, but rather to defect in NADH shuttle function.
C. Ca2+ Regulation of the MAS
In this section we review evidence gathered many years ago suggesting that the MAS is regulated by Ca2+. The mechanism was long considered to rely on Ca2+ entry in mitochondria through the uniporter and subsequent activation of mitochondrial dehydrogenases, or on direct activation by calmodulin. After the molecular identification of the AGCs, it has been shown that the shuttle is regulated by Ca2+ from the external side of the inner mitochondrial membrane. However, the interactions between the Ca2+ uniporter-mediated Ca2+ signaling in mitochondria and the mechanism mediated by the AGCs are complex and may vary from tissue to tissue.
1. Ca2+ activation of MAS in liver
It has been known for a long time that α-adrenergic agents stimulate gluconeogenesis in hepatocytes. The mechanism is independent of cAMP and related to changes in cytosolic Ca2+ concentration ([Ca2+]i) (205, 372, 379). Increased gluconeogenesis is observed with reduced substrates, such as glycerol, xylitol, and sorbitol, that require the transfer of reducing equivalents to mitochondria, but is minimal or absent with oxidized substrates such as pyruvate or dihydroxyacetone phosphate (193, 226). The proposed sites of regulation are at the MAS or the GPS (194, 225, 379).
According to some groups, stimulation of rat hepatocytes or perfused liver with α-adrenergic agonists such as phenylephrine leads to sustained activations of MAS and GPS, or to a transient stimulation of MAS followed by a stable stimulation of GPS (194, 225, 372, 379). Increased MAS activity was typically inhibited by the transaminase inhibitor aminooxyacetate. The mechanisms responsible for increases in MAS activity were thought to involve either changes in intermediates affecting AGC activity or an increase in the mitochondrial membrane potential caused by a direct stimulation of respiration that would increase the driving force of the AGC and shuttle activity (225). In fact, α-adrenergic agonists transiently increased the cytosolic/mitochondrial aspartate concentration gradient and strongly reduced glutamate and especially α-ketoglutarate concentrations (226, 370, 372), a finding thought to reflect a rapid activation of α-KG dehydrogenase (α-KGDH) activity (379). Careful studies by LaNoue’s group (371) showed that the mitochondrial membrane potential does not change significantly in glucagon-stimulated hepatocytes, arguing in favor of changes in AGC activity at the expense of changes in intermediates. The fall in α-KG concentrations was regarded as the mechanism responsible for α-adrenergic agonist-dependent increase in liver AGC activity (370, 371). The effect of the decrease in α-KG levels was explained by the finding that α-KG was a strong competitive inhibitor of aspartate formation (competitive with respect to oxaloacetate) via glutamate transamination in isolated liver mitochondria (371).
Hepatocyte activation by Ca2+-mobilizing agents was thought to induce Ca2+ entry in mitochondria coupled to Ca2+ activation of α-KGDH activity. Glucagon, which increases intracellular cAMP levels, synergistically stimulates net inflow of Ca2+ to hepatocytes through receptor-activated Ca2+ channels (4, 121, 167) and net uptake of Ca2+ by hepatocyte mitochondria (4, 9, 121, 167, 264). As Ca2+ increases the affinity for α-KG of α-KGDH (246, 273), Ca2+ entry in mitochondria would result in a reduction of α-ketoglutarate levels, with subsequent activation of glutamate transamination and aspartate efflux from mitochondria, in spite of a concurrent decrease in oxaloacetate levels brought about by the increase in the NADH/NAD ratio (370). Studies of LaNoue’s group have shown that there is a preferential channeling of external glutamate to mitochondrial transaminase and to the active site of α-KGDH, suggesting microcompartmentation or channeling between α-KGDH and the mitochondrial transaminase (351, 361 but see Ref. 267), especially in liver. This could explain the high sensitivity of liver AGC activity to a decrease of mitochondrial α-KG levels caused by Ca2+-induced activation of α-KGDH. However, to the best of our knowledge, a demonstration of this mechanism with the use of Ca2+ uniporter blockers was never reported.
It should be noted that experiments carried out by Sugano’s group (155) in perfused rat liver indicated that MAS activation by α-adrenergic agonists and vasopressin is completely suppressed by calmodulin antagonists (155), whereas agonist-evoked Ca2+ signaling was not modified. This was interpreted as a direct effect of calmodulin on MAS activity. While the molecular basis for this effect is still unknown, it would be reasonable to propose a cytosolic site for calmodulin-dependent effects of Ca2+, in contrast to the interpretations based on a mitochondrial matrix Ca2+ effect on α-KGDH.
2. Ca2+ regulation of MAS in heart
Liver is not the only organ where Ca2+ activates MAS activity. Evidence for this was obtained in kidney (350) and heart (342). However, the activation by Ca2+ of MAS activity in heart has proven to be elusive. This is due to the fact that in the heart, Ca2+ has opposing actions on the two carriers of MAS, an activating role for the AGC, which is observed only under specific conditions, and an inhibitor role for the OGC. As a result, the overall activating role on the AGC or on MAS is not always apparent.
In studies with heart mitochondria, physiological increases in matrix Ca2+ resulted in a marked activation of α-KGDH and in a decreased accumulation of α-KG (409). However, Scaduto (342) found that under conditions of MAS operation (incubation with glutamate plus malate) the decrease in α-KG levels led to an inhibition rather than an increase of aspartate formation. Stimulation by Ca2+ of aspartate efflux in rat heart mitochondria was only observed after imposing a rise in mitochondrial NADH/NAD ratio through addition of pyruvate (together with malate and glutamate), to decrease OAA levels and basal aspartate efflux rate [reduced to ∼30% of control levels (342)]. Because α-KG is a competitive inhibitor of OAA transamination to aspartate, Ca2+-induced changes in α-KG levels controlled aspartate formation only under conditions where OAA levels were very low, as those provided by pyruvate. In the absence of pyruvate, the influence of Ca2+ on AGC activity was attributed to a balance between an inhibitory effect, caused by the increase in the NADH/NAD ratio and consequent reduction in OAA levels, and a stimulatory effect caused by a decrease in α-KG levels (342).
Regarding the OGC, in heart mitochondria there is a strict competition for α-KG between α-KGDH and this carrier (Fig. 4), by virtue of the similarities of their apparent Km values. An apparent Km of ∼1.5 mM for α-KG on the matrix side of the OGC has been reported (360), while the Km for α-KG in the case of α-KGDH is ∼0.67 mM (219) so that both steps will be very sensitive to the concentration of α-KG in the mitochondrial matrix. Consequently, activation or inhibition of OGC transport will lead to a reduction or stimulation in α-KGDH activity, respectively (219). Conversely, in perfused rat heart and isolated heart mitochondria, increased α-KGDH activity results in a decrease in α-KG efflux from mitochondria (278).
The relevance of the effects of Ca2+ on competition for α-KG has been underscored by a recent study in which heart metabolism was studied at both basal and high work loads (279). Lewandowski’s group (279) showed that high work loads, which lead to increases in mitochondrial Ca2+ (see sect. iii, C3b and E), resulted in a drastic reduction in OGC and MAS activity (to ∼20% of basal levels). Consistently, the decreased transfer of reducing equivalents to mitochondria was accompanied by increased tissue lactate levels (279).
On the basis of the above findings, it is interesting to note that under physiological conditions of high cytosolic and mitochondrial Ca2+ signals, MAS activity is activated in the liver but not in the heart. There are several reasons that could account for this difference. The competition for α-KG between α-KGDH and the OGC has been detected in heart but not in the liver. There is only one OGC gene (283), with no known variants, so that the difference between heart and liver is unlikely due to kinetic differences among isoforms. Tissue specificity could arise from differences between OGC levels in liver and heart; however, the activity and expression levels of OGC in heart are severalfold higher than in liver (171).
Another reason is related to microcompartmentation and substrate channeling, which may differ in heart and liver. In liver but not in heart mitochondria, flux along α-KGDH is higher when α-KG arises from glutamate by transamination rather than being provided directly as a substrate (219, 361). As indicated above, interaction between mitochondrial transaminase and α-KGDH with preferential α-KG channeling between the two enzymes explains why Ca2+ activates MAS through stimulation of α-KGDH in the liver. Experiments in intact mitochondria and in reconstituted systems suggest that α-KGDH can be bound to mitochondrial aspartate aminotransferase and that mitochondrial malate dehydrogenase can have a higher affinity for this binary complex than for any of the two constituents alone (120). It has been argued that this ternary complex could facilitate direct transfer of oxaloacetate to the transaminase, activation of malate dehydrogenase due to a decrease in the Km of the enzyme within the complex, and direct transfer of α-KG to α-KGDH (120). The formation or stability of this complex may be less important in heart than in liver mitochondria, and this may explain the lack of a clear activation of the AGC upon Ca2+ stimulation of α-KGDH in heart.
3. Ca2+ regulation of citrin and aralar
The mechanism described above for the activation of liver MAS by Ca2+, i.e., activation of α-KGDH by Ca2+ uptake on the Ca2+ uniporter (CU-α-KGDH), has long been the accepted explanation of MAS activation by Ca2+ mobilizing agents (370, 371). The discovery of the AGCs and the existence of a mechanism to activate AGC by Ca2+ acting on the external face of mitochondria (95, 286) therefore came as a surprise.
It is self-evident that if this Ca2+ activation mechanism of the AGC is physiologically relevant it would need to operate under conditions where the mitochondrial Ca2+ uniporter is not active. This question has been addressed recently in liver mitochondria that contain a single AGC isoform, citrin, and in brain and skeletal muscle mitochondria, that also have a single AGC isoform, aralar. Ca2+ activation of reconstituted MAS activity has been studied in isolated rat and mouse mitochondria incubated with ruthenium red (RR), under conditions preventing Ca2+ uptake in mitochondria (287).
Liver ACG coupled to MAS activity was found to be modestly (∼1.5-fold) activated by extramitochondrial Ca2+, with an S0.5 (free Ca2+ concentration for half-maximal activation) of ∼120 nM, close to the resting [Ca2+]i (79). On the other hand, activation by Ca2+ of brain and skeletal muscle MAS activity was more pronounced (∼3-fold) with an S0.5 of 324 ± 114 nM for brain mitochondria (287), with no changes in the Km for glutamate (79). These S0.5 values are substantially lower than the Km of the Ca2+ uniporter of mitochondria [∼10–20 μM, (143, 145, 190)], suggesting that Ca2+ activation of citrin- and aralar-coupled MAS activity may contribute to NADH reduction in mitochondria at Ca2+ concentrations where the Ca2+ uniporter is still inactive.
Unlike the CU-α-KGDH mechanism, in which the control on AGC activity is secondary to an increase in matrix aspartate level, i.e., through mass action ratio, Ca2+ activation of the AGC from the external face of the inner membrane leads to an increase in AGC flux and MAS activity at constant substrate concentrations, as expected for a classical regulatory reaction.
A) ROLE OF ARALAR AND THE MALATE-ASPARTATE NADH SHUTTLE IN TRANSDUCTION OF SMALL CA2+ SIGNALS TO MITOCHONDRIA.
The role of aralar-MAS pathway in transduction of small Ca2+ signals to mitochondria has been studied in neurons. Pardo et al. (287) have employed primary neuronal cultures from control and aralar-deficient mice and NAD(P)H imaging with two-photon excitation microscopy to study MAS activity. Lactate utilization involves NADH shuttle activity and is important in neurons, as part of the astrocyte-neuron lactate shuttle (292). Lactate is produced by astrocytes and taken up by neurons that use it as an oxidizable substrate, especially during periods of high activity (187, 293, 294). Addition of 20 mM lactate caused an increase in cytosolic and mitochondrial NAD(P)H and a substantial transfer of NAD(P)H to mitochondria in control, but not in aralar-deficient neurons (287), indicating that MAS is the main NADH shuttle in neurons. This was also confirmed by the marked increase of the lactate-to-pyruvate ratio observed in the culture medium of aralar-deficient neurons (Pardo and Satrústegui, unpublished observations).
To study the effect of small [Ca2+]i signals on MAS activity, neurons were incubated in Ca2+-free medium plus 100 μM EGTA and exposed to lactate together with 100 μM ATP or different Ca2+-mobilizing agonists (Fig. 6). ATP-induced [Ca2+]i transients were small, with departures of ≤100 nM from resting values that lasted 1 min at most and were not accompanied by any detectable increase in mitochondrial Ca2+ concentration ([Ca2+]mit) measured with rhod 2, in control or aralar-deficient neurons. However, this small Ca2+ signal resulted in a remarkable potentiation of lactate-dependent increase in mitochondrial NAD(P)H fluorescence in control neurons, which was notably absent in aralar-deficient neurons. Other Ca2+-mobilizing agonists, such as carbachol (50 μM) or thapsigargin (1 μM), also potentiated in a significant way the response to lactate in control neurons (287).
Since lactate-derived pyruvate contributes to NAD(P)H generation, Pardo et al. (287) also tested whether the effects of ATP involve Ca2+-stimulated pyruvate metabolism rather than Ca2+-stimulated MAS activity. Remarkably, the small [Ca2+]i signals triggered by ATP had no effect on the increase in mitochondrial NAD(P)H induced by pyruvate (Fig. 6) (287). Therefore, the results clearly showed that the neuronal aralar-MAS pathway is selectively activated by small Ca2+ signals that are below the threshold for Ca2+ uniporter activation.
Synaptosomal mitochondria start taking up Ca2+ at a global [Ca2+]i of ∼350–400 nM (241). This indicates that Ca2+ signaling through the aralar-MAS pathway in neuronal mitochondria, which displays an S0.5 of 324 nM, would be significant only below these Ca2+ concentrations, which is exactly what Pardo et al. found. On the other hand, the fact that mitochondria respond to [Ca2+]i signals with a substantial increase in NAD(P)H through a Ca2+ uniporter-independent pathway was new and surprising, especially because the magnitude of the response via the aralar-MAS pathway is very large in neurons, not far from that evoked by the activity of the Ca2+ uniporter-mitochondrial dehydrogenases pathway (see sect. iiiC3b). The availability of aralar-knockout animals will make it possible to assess the contribution of this pathway to Ca2+ signaling in different tissues. These studies may provide novel insights in Ca2+ signaling in mitochondria.
The role of the citrin-MAS pathway in transduction of Ca2+ signals to liver mitochondria has not been investigated so far, but based on the very low S0.5 for Ca2+ of citrin (79), it may be relevant at small departures from the basal Ca2+ concentration, which are below the range of the Ca2+ uniporter. Larger Ca2+ signal increases would be relayed to the mitochondrial matrix, resulting in Ca2+ activation of α-KGDH, decrease in α-KG, and subsequent stimulation of aspartate efflux from mitochondria (370, 371) and MAS activity. Indirect evidence that this may be so is examined in section iiiE.
B) EFFECTS OF LARGE CA2+ SIGNALS ON THE ARALAR-MAS PATHWAY.
Pardo et al. (287) studied the increase in mitochondrial NAD(P)H in neurons in response to large [Ca2+]i signals brought about by depolarization-dependent activation of voltage-operated Ca2+ channels (VOCC) in the presence of millimolar external Ca2+ concentrations. In addition to Ca2+-mediated MAS activation, which was abolished in aralar-deficient neurons, it was expected that VOCC-dependent Ca2+ entry would result in Ca2+ uptake by mitochondria and subsequent dehydrogenase activation (110) in both control and aralar-deficient neurons.
There was no difference between the increases in [Ca2+]i and [Ca2+]mit obtained after lactate plus KCl addition in control and aralar-deficient neurons, in agreement with the lack of effect of aralar overexpression on Ca2+ homeostatic mechanisms reported earlier (222). The increase in the mit/cyt NAD(P)H fluorescence ratio obtained in the presence of KCl was significantly larger than that obtained by lactate addition itself, but in this case, there were no differences between control and aralar-deficient neurons (287). As the increase in mitochondrial NAD(P)H obtained with high K+ was expected to result from the additive effects of Ca2+ activation on MAS activity and Ca2+ activation of mitochondrial dehydrogenases, the lack of difference between wild-type and aralar-deficient neurons indicated that MAS activity contributes very little to the increase in mitochondrial NAD(P)H under high-K+ depolarization, suggesting that MAS activity may be inhibited under conditions allowing mitochondrial Ca2+ uptake.
Studies of MAS activity in isolated mitochondria in the presence or absence of 200 nM RR showed that Ca2+ activation of MAS was diminished if no RR was present (Fig. 7). This concentration of RR inhibits mitochondrial Ca2+ uptake completely (Fig. 7). Thus, in neurons, Ca2+ activation of MAS is blunted when Ca2+ is allowed to enter the mitochondria (287). This is possibly due to the effect of mitochondrial Ca2+ on the affinity for α-KG of α-KGDH and the competition for α-KG between α-KGDH and the OGC (Fig. 4), as described earlier for the heart MAS activity (see sect. iiiC2). Another factor that may blunt MAS activation when all mitochondrial dehydrogenases are activated by Ca2+ is the thermodynamic constraint of proton gradient formation. If the proton electrochemical gradient obtained under those conditions is close to maximal and cannot be increased, this would limit the entry of reducing equivalents into mitochondria through MAS. This situation has been described in β-cells incubated with different fuel secretagogues that bypass glucose metabolism, each of them setting a maximal hyperpolarization of the mitochondrial membrane that correlated with insulin secretion (7).
The results obtained in neurons suggest that, as in heart, large Ca2+ signals generate an increase in NAD(P)H arising mainly from the activation of tricarboxylic acid cycle dehydrogenases rather than from MAS activity. The inhibition of MAS probably disappears when the activation by Ca2+ of mitochondrial dehydrogenases comes to an end, and α-KG levels and mitochondrial proton electrochemical gradient return to control values. Thus it is possible that after the decay of [Ca2+]mit transients, MAS activity could prolong the increase in mitochondrial NAD(P)H induced by high [Ca2+]i signals, and thus contribute to ATP synthesis and recovery of the resting state, conditions that rely on neuronal lactate utilization (293).
D. Other Mechanisms of Ca2+ Regulation of NAD(P)H Transfer to Mitochondria Independent of Ca2+ Entry
1. Mitochondrial glycerol-3-phosphate dehydrogenases
The mitochondrial glycerol-3-phosphate dehydrogenases (FAD-GPDH), together with cytoplasmic NAD-linked GPDH, catalyze the reactions of the GPS. FAD-GPDH is present at high levels in tissues capable of rapid oxidative metabolism of glucose and is thought to be preferentially used over other shuttle systems when glucose is used as the carbon source, since glycerol-3-phosphate arises directly from the glycolytic substrate dihydroxyacetone phosphate. FAD-GPDH enzymes from insect flight muscle (158), lung (125, 126), liver (414), brown adipose tissue and brain (36), and beta-cell mitochondria (329) are stimulated by Ca2+ (Fig. 8). Ca2+ activation in all cases results in a decrease in the Km for glycerol-3-phosphate with no changes in the apparent Vmax. Half-maximal activation occurs at ∼100 nM free Ca2+, i.e., at around resting Ca2+ concentration (174, 329), and similar values were obtained for glycerol-3-phosphate-induced ATP production (174) or respiration (75) in mitochondria from pancreatic β-islets. Inhibition of mitochondrial Ca2+ uptake with RR did not interfere with the stimulation in blowfly flight muscle mitochondria (66), indicating that Ca2+ exerts its effects on the extramitochondrial side of the inner mitochondrial membrane. Since the apparent Km of liver FAD-GPDH, 4–5 mM (414), is far higher than the concentration of glycerol-3-phosphate in liver of fasted rats, ∼0.1 mM (308), the Ca2+-dependent increase in affinity will result in an increased glycerol-3-phosphate oxidation in the liver.
Activation by Ca2+ of FAD-GPDH is hardly modified in the presence of calmodulin and calmodulin inhibitors (234, 414), suggesting that the enzyme reacts directly with Ca2+. In fact, the protein sequence of FAD-GPDH from mammalian sources (58, 234) shows two EF-hand motifs close to the carboxy terminus of the enzyme, a region facing the intermembrane space. These putative Ca2+ binding domains are absent in the enzyme from S. cerevisiae (273, 321), which is not known to be Ca2+ sensitive. EF-hand motifs are also absent in the FAD-GPDH proteins from Candida albicans, Aspergillus nidulans, and Neurospora crassa databases. The mammalian enzyme shows Ca2+ binding in 45Ca2+ overlay experiments, suggesting that the EF-hand motifs bind Ca2+ and are responsible for Ca2+ activation (234). A mutation in the Ca2+-binding domain associated with a reduced Ca2+ activation was found in a family of diabetic subjects (277). These results suggest but do not prove that the carboxy-terminal Ca2+-binding domain of FAD-GPDH is responsible for Ca2+ activation. The sequence of the Ca2+-regulated blowfly FAD-GPDH (158) is still unknown, but it is interesting to note that a Drosophila FAD-GPDH sequence with 62% identity with respect to that of the human enzyme (NP_611063, CG8256-PC, isoform C) also harbors two EF-hand motifs in conserved positions.
Pancreatic β-cells have a highly active glycerol-3-phosphate shuttle. NADH shuttles are important in glucose-stimulated insulin secretion, as glucose metabolism in β-cells occurs essentially through aerobic metabolism, and lactate dehydrogenase levels are extremely low (354). By using aminooxyacetate to inhibit malate-aspartate NADH shuttle function in mice that also have targeted disruption in FAD-GPDH, glucose-induced insulin secretion was abrogated (119). Whereas the metabolic role of the enzyme is clearly established, the actual role of Ca2+ regulation of FAD-GPDH is not yet known.
In β-cells, stimulation of insulin secretion by glucose involves a rise in [Ca2+]i. This rise requires mitochondrial metabolism to increase the ATP/ADP ratio and results in the closure of ATP-sensitive K+ channels, membrane depolarization, and influx of Ca2+ through voltage-sensitive channels (for review, see Ref. 189). The resulting increase in cytosolic Ca2+ displays oscillations that correlate with the glucose-evoked rhythmical plasma membrane potential activity (338), and with fluctuations in the ATP/ADP ratio, islet respiration, and insulin secretion (189).
Ca2+ entry mitochondria via the Ca2+ uniporter are critical to maintain cytosolic Ca2+ oscillations in β-cells (189), and although GPS could also participate in the maintenance of the oscillatory behavior, studies with mice deficient in FAD-GPDH have shown that this is not the case. In these studies, the oscillatory [Ca2+]i response and the fluctuations in plasma membrane potential with bursts of action potentials in response to glucose were unchanged by FAD-GPDH deficiency. Moreover, an imposed and controlled rise in [Ca2+]i in β-cells similarly increased NAD(P)H fluorescence in control and GPDH-deficient islets (307). These results have led to the conclusion that a possible cyclic activation of the enzyme by Ca2+ is not required for the oscillatory behavior (307). However, because MAS and GPS shuttles play partly redundant roles, it is still possible that a cyclic, Ca2+-dependent activation of GPDH could be unmasked under conditions where MAS is inactive. Alternatively, it has been argued that the high affinity for Ca2+ of GPDH causes a constitutive activation of GPS (307).
2. External NAD(P)H dehydrogenases in fungi and plants
Mitochondria from fungi and plants contain non-proton-pumping NAD(P)H dehydrogenases as alternative systems to feed electrons from NAD(P)H to the respiratory chain in a rotenone-insensitive manner. These enzymes are the only system to feed electrons from NAD(P)H to ubiquinone in those organisms, such as S. cerevisiae, that lack complex I. Unlike complex I, non-proton-pumping NAD(P)H dehydrogenases are single polypeptide chains that oxidize NAD(P)H in the cytosol (external enzymes) or mitochondrial matrix (internal enzymes), and pass the electrons to FMN or FAD in the mitochondrial membrane (67).
Plants contain four alternative NADH dehydrogenases, and S. cerevisiae has one internal and two external enzymes, which oxidize matrix or extramitochondrial NADH, respectively. The filamentous fungus Neurospora crassa has complex I and three or four alternative NAD(P)H dehydrogenases, a situation similar to that of plants (67, 257). Two of these dehydrogenases are external, NDE1 and NDE2, and NDE1 was shown to be Ca2+ dependent (251). NDE1 is anchored to the inner mitochondrial membrane by an amino-terminal domain, and its catalytic site faces the intermembrane space (Fig. 8). It contains an insertion that harbors an EF-hand motif (252). NDE1 is thought to be mainly involved in the oxidation of cytosolic NADPH and is absolutely dependent on Ca2+ at physiological pH (251). Ca2+-dependent NAD(P)H dehydrogenases exist in the genome of filamentous fungi Aspergillus nidulans (EAL87637 and EAA62080), and Magnaporte grisea (EAA50381), and even in Yarrowia lipolytica (XP_503592), an alkane-utilizing yeast that shares several properties with filamentous fungi, but not in the budding yeast S. cerevisiae or Schizosaccharomyces pombe.
Non-proton-pumping, Ca2+-dependent NAD(P)H dehydrogenases are also found in plant mitochondria (257, 258). The plant enzymes are similar to fungal external and internal NAD(P)H dehydrogenases and fall in the nda, ndb, and ndc families (253, 254, 256, 258). The ndb family is characterized by the presence of an insertion carrying an EF-hand motif roughly in the same position of N. crassa NDE1, and potato ndb1 has been shown to correspond to an external Ca2+-sensitive NADPH dehydrogenase (253). The external NADPH dehydrogenases of plant mitochondria are activated by cytosolic Ca2+ with an S0.5 of ∼0.3 μM or higher (257, 271, 306).
The kinetics of Ca2+ activation of NADPH oxidation in N. crassa mitochondria have not been directly assessed, but it is possible that these external, Ca2+-dependent NAD(P)H dehydrogenases provide a pathway to oxidize cytosolic NADPH under conditions where cytosolic Ca2+ levels increase. In the case of the plant enzymes, it is believed that external NADPH dehydrogenases will be inactive under normal resting conditions when the cytosolic Ca2+ concentration is ∼100 nM. However, when the plant is stressed, an increase in [Ca2+]i to 1–2 μM contributes to its activation (257). Since a high-reduction level of the respiratory chain leads to reactive oxygen species (ROS) production, it has been proposed that activation of external NADPH dehydrogenase may help minimize ROS production under stress conditions by keeping the cytosolic NADP pool relatively oxidized (257).
E. Interplay Between AGCs, GPS, and the Ca2+ Uniporter in Regulation of Mitochondrial Metabolism in Response to Cytosolic Ca2+ Signals
Ca2+ signals in excitable and nonexcitable tissues are made up of single or periodic fluctuations in cytosolic Ca2+ concentrations. These are driven by electrical activation or by stimulation by hormones or paracrine agents that evoke Ca2+ release from intracellular stores mainly through the operation of inositol trisphosphate (IP3) and ryanodine (Ry) receptors (see for reviews, see Refs. 40–42, 314).
Local Ca2+ signals are produced spontaneously or by low levels of a locally applied stimulus and involve either IP3Rs or RyRs together with different types of plasma membrane channels. The elementary Ca2+ signal events generated this way are thought to reflect the coordinate opening of clusters of channels and have been termed puffs, sparks, or QEDs depending on whether the channels are IP3Rs, RyRs, or voltage-operated channels, respectively (40, 42). These local signals are very brief Ca2+ pulses that do not extend spatially to the rest of the cell (42).
At higher level of stimulation, the Ca2+ transients last longer and the resulting signal usually spreads out as a Ca2+ wave to reach targets that are distributed throughout the cell. For the waves to occur, most of the IP3Rs or RyRs must respond to each other through the process of Ca2+-induced Ca2+ release (40, 42).
Hormones and agonists acting through receptors coupled to phospholipase C (PLC) generate periodic Ca2+ spikes or oscillations whose amplitude is not affected by the agonist dose, which rather modulates the spiking frequency (314, 322, 421). This phenomenon described as frequency-modulated Ca2+ signaling (384) was first found in nonexcitable cells, particularly in hepatocytes (322, 421), exocrine pancreas (385), smooth muscle, and egg cells from animals and plants, where fertilization elicits single or repetitive Ca2+ transients in response to fertilizing sperm (40, 113). Once initiated at a particular subcellular locus, Ca2+ waves propagate in the cell at a fixed velocity, regardless of the agonist dose. Because Ca2+ has a limited diffusion coefficient in the cytosol, particularly in cells with high levels of Ca2+-binding proteins, it is thought that Ca2+ waves may provide a mechanism to deliver localized Ca2+ release to distal parts of the cell (42, 384).
Ca2+ waves arising from any of these mechanisms can reach the whole cell and can even spread from cell to cell through intercellular waves. However, in common with the local Ca2+ signals, every single wave or oscillation is a brief pulse of Ca2+, arising from the balance of reactions that introduce Ca2+ in the cytosol and those that remove it. It has been suggested that many Ca2+-sensitive processes can be regulated more accurately by low-frequency Ca2+ oscillations than by an equivalent time-averaged Ca2+ signal, which would yield a small elevation of Ca2+ above basal levels.
Spatial contiguity between the site where the Ca2+ signal is generated and downstream effectors is critical for rapid responses to global Ca2+ signals and especially to respond to localized Ca2+ signals. This is especially important for mitochondria, where the main mechanisms whereby Ca2+ signals are transmitted are the Ca2+ uniporter and the RAM (for rapid uptake mode) (61, 144, 365). The Ca2+ uniporter has a low affinity for Ca2+, with a K0.5 of 5–10 μM (145, 190), and is allosterically activated by Ca2+ (206–208), while the RAM has a slightly higher affinity (145). The fact that there is a substantial mitochondrial Ca2+ uptake despite modest changes in peak [Ca2+]i is explained by the strategic location of mitochondria close to endoplasmic reticulum (ER) Ca2+ release channels, or plasma membrane Ca2+ channels, and therefore to microdomains of very high [Ca2+]i which escape detection by present imaging methods (44, 124, 133, 260, 311). These microdomains of high [Ca2+]i may be detected by an aequorin chimera targeted to the mitochondrial intermembrane space, where it is exposed to local high [Ca2+]i (295). The access of IP3-linked Ca2+ spikes, but not global Ca2+ signals, to the intermembrane space is limited by the permeability of outer mitochondrial membrane (305, 83). This barrier may be overcome by increasing voltage-dependent anion channels or tBid levels (83, 305), resulting in increased delivery of IP3-linked Ca2+ spikes and oscillations to mitochondria. However, it has been conclusively demonstrated that mitochondria from rat luteal cells, insulin-producing cell lines, rat adrenal glomerulosa cells, and osteosarcoma cells take up Ca2+ at low submicromolar [Ca2+]i, suggesting that under appropriate conditions the Ca2+ uniporter may operate at modest departures from resting [Ca2+]i (296, 367, 375). Recent findings indicate that the capacity for Ca2+ uptake in mitochondria is also influenced by mitochondrial morphology (with lower capacity when the mitochondrial network is fragmented) and kinase activity (261; reviewed in Ref. 44). Once the cytosolic Ca2+ has returned to resting levels, Ca2+ is also pumped out of mitochondria to the cytosol on the Na+/Ca2+ exchanger or the more sluggish H+/Ca2+ transporter (143, 145). Efflux through a possible low-conductance mode of the permeability transition pore (PTP) has also been suggested by some studies (172, 173) but not by others (376). Thus mitochondria participate in Ca2+ buffering through multiple mechanisms (see Refs. 310, 312 for review).
The mitochondrial matrix Ca2+ spikes that result from the coordinated function of uptake and efflux pathways may be the same all through the cell, but heterogeneous Ca2+ signals in the mitochondrial population may occur as a result of heterogeneity of the driving force for Ca2+ uptake, of the spacing between mitochondria and Ca2+ release sites, and of the cytosolic Ca2+ signal delivered in each site (148). This heterogeneity has been directly visualized in fluorescence imaging studies at the level of individual mitochondria, in response to global Ca2+ signals in pancreatic acinar cells (288) and primary cultured aortic myocytes (259) or to local sparks in H9c2 cardiac cells (282). The term Ca2+ marks has been coined to design miniature increases in mitochondrial Ca2+ in response to these elementary release events (148, 282). The mean duration of Ca2+ marks in H9c2 cardiac cells was longer than that of the Ca2+ sparks that evoked them, in agreement with the slow Ca2+ extrusion through the Na+/Ca2+ exchanger (282). These studies have led to the conclusion that not all Ca2+ sparks in cardiac cells (282) nor Ca2+ puffs in HeLa cells (77) elicit a response in mitochondria.
This is the scenario in which the role of AGCs as Ca2+ signaling mechanisms becomes apparent. The AGCs together with FAD-GPDH and the Ca2+ uniporter make up a mitochondrial decoding toolkit whereby each Ca2+ signal translates into a physiological response of mitochondria: an increase in substrate supply (NADH or FADH2) to the electron transport chain. The decoding devices are all different in their Ca2+ sensitivity. The AGCs and FAD-GPDH are activated by Ca2+ at concentrations below those where the Ca2+ uniporter operates. For unitary sparks, puffs, QEDs, or other brief and highly localized Ca2+ signals, AGCs and FAD-GPDH could respond both earlier than the Ca2+ uniporter and after the Ca2+ mark has ended, and even under conditions where the Ca2+ uniporter does not respond. In this way, AGCs may prime mitochondria to respond to Ca2+ entry by prolonging mitochondrial energization beyond the duration of the individual Ca2+ mark through an increase of the reduction state of mitochondrial NADH and a decrease of that of cytosolic NADH. This allows for a localized increase in glycolysis and pyruvate supply to mitochondria needed to respond to the Ca2+ activation of mitochondrial dehydrogenases.
In the case of global Ca2+ signals of low intensity, the present data suggest that AGCs and FAD-GPDH will operate as only Ca2+ signaling devices in mitochondria at the very start and at the end of a single transient, or at the front and tail of every Ca2+ wave, and may cooperate with the Ca2+ uniporter-mitochondrial DH (CU-mitDH) mechanism to elicit a full mitochondrial response to Ca2+ along the duration of the Ca2+ transient or wave. This is the situation in liver in response to α-adrenergic agonists and glucagon in which citrin, the Ca2+ uniporter, and the liver ATP-Mg/Pi carrier (see sect. ivB), are all activated, resulting in increases in MAS, gluconeogenesis, and urea synthesis. Whether the Ca2+ activation of citrin or aralar actually persists throughout the duration of the Ca2+ signal is unknown, but at least in the case of the aralar-MAS pathway in heart and neurons there is evidence that this may not be so. In these two cases, Ca2+ activation of the aralar-MAS pathway is blunted when Ca2+ activation of the mitochondrial dehydrogenases occurs so that the emerging scenario is one in which Ca2+ signaling switches from aralar-MAS to uniporter-mitochondrial DH and back to aralar-MAS as the global Ca2+ signal is initiated, reaches a maximum, and ends. The recruitment of the GPS which is not known to be inhibited at high Ca2+ concentrations could compensate for the inhibition of MAS at high Ca2+ concentrations. The combined or sequential activation of the AGC-MAS and CU-MitDH pathways may explain observations in hippocampal neurons (185, 186), hepatocytes (149, 301, 313), β-cells and adrenal glomerulosa cells (301), and pancreatic acinar cells (407) in which the responses to single or low-frequency cytosolic Ca2+ spikes evoke transient increases in mitochondrial NAD(P)H that outlast the duration of the mitochondrial and cytosolic Ca2+ spikes. The longer duration of the NADH transients compared with the duration of the Ca2+ transients is thought to have the advantage of prolonging enhanced energy production for some time after the termination of the Ca2+ signal so that all of the energy demands for the restoration of the steady state disturbed by the Ca2+ signal may be met (407). A differential recruitment of the AGC-MAS pathway by specific agonists (for example, by stimulating glycolytic flow) may explain findings in rat luteal cells where agonists such as PGF2α and ATP (acting at metabotropic P2 receptors) induce higher and longer lasting increases in mitochondrial NADPH than others, in spite of similar transmission of the Ca2+ signal to mitochondria (291). In pancreatic β-cells (1) and RBL-2H3 mast cells (84) undergoing sequential cytosolic Ca2+ transients, mitochondria respond to a second cytosolic Ca2+ transient with a larger increase in mitochondrial Ca2+ than the first. Whether mitochondrial priming through the AGC-MAS pathway could be involved in the facilitation effect of the first Ca2+ transient is an open question.
The role of the AGG-MAS pathway in response to high-intensity global signals is still unknown. The initial response to high-frequency stimulation in the CA1 area of the hippocampus is a global NAD(P)H dip caused by rapid stimulation of NADH oxidation by the respiratory chain in the postsynaptic neuron (187, 356) as also found in other neuronal types (110, 162). This initial response is attributed to the feedback regulation of oxidative phosphorylation in response to the fall of the ATP/ADP ratio and the increase in cytosolic and mitochondrial Ca2+, and modeling evidence indicates that it is due to increased uptake of extracellular lactate by neurons (25). Subsequently, this is followed by an increase in lactate production in astrocytes, which is accompanied by an overshoot of glial NADH levels which fuels a gradual recovery of neuronal NADH levels as a result of the activation of tricarboxylic acid cycle dehydrogenases (187). Both phases of the neuronal NAD(P)H response require the MAS-aralar pathway to utilize lactate, and the global sustained increase in NAD(P)H that follows the initial dip is inhibited in the presence of inhibitors of the neuronal monocarboxylate transporter MCT2 (131). However, most of the sustained increase in NADPH upon high-frequency stimulation probably occurs at high Ca2+ levels, when the CU-mitDH pathway is highly activated and where Ca2+ activation of the aralar-MAS pathway is blunted (287). Thus the actual role of the Ca2+ activation of the aralar-MAS pathway in this situation is unknown and awaits investigation.
Studies with mitochondrial and cytosolic expressed aequorin in heart cells have shown that overall myocardial work correlates with increases or decreases in beat-to-beat frequency and/or amplitude of cytosolic and mitochondrial Ca2+ transients (315), and with sustained changes in mitochondrial Ca2+ levels between oscillations, i.e., “diastolic” increases or decreases in mitochondrial Ca2+ (51, 315). Mitochondrial matrix Ca2+ tracks the [Ca2+]i that drives the increase in the work load (31, 315) and may activate Ca2+-regulated mitochondrial dehydrogenases resulting in relatively slower tonic increases in NADH (51). In intact heart muscle, an increase in work load causes an initial decrease in mitochondrial NADH levels, followed by a slow recovery towards control values (51). This recovery of NADH levels is Ca2+ dependent (50). As mitochondrial Ca2+ beat-to-beat spikes might not return to basal levels at high work loads (315), Ca2+ activation of mitochondrial dehydrogenases persists between spikes and has been considered responsible for the recovery of NADH levels (50, 51). If glucose or lactate is used as fuel, any increase in mitochondrial activity in cardiac cells involves the AGC (aralar or citrin)-MAS pathway. However, the persistent presence of Ca2+ in mitochondria and of Ca2+ activation of mitochondrial dehydrogenases may impose a constraint on the operation of the aralar-MAS pathway eventually leading to its inhibition at high work loads (279).
IV. ATP-MG/Pi CARRIERS, SCAMC GENES, AND PROTEINS
A. Mammalian SCaMCs
The ATP-Mg/Pi carriers were first identified as CaMCs with short amino-terminal extensions, or short CaMCs (SCaMCs), and are represented by three main isoforms, SCaMC 1–3 encoded, respectively, by SLC25A24, SLC25A25, and SLC25A23 (33, 91, 94, 122), with a number of variants arising from alternative splicing. Like human AGC genes, SCaMCs share exon-intron structure. SCaMC-1 maps at chromosome 1p13.3 and SCaMC-2 and SCaMC-3 genes at positions 9q34.13 and 19p13.3, respectively (33, 94).
1. Extended diversity of mammalian SCaMCs
Mammalian SCaMCs are one of most complex subsets of MCs. All code for highly conserved proteins (70–80% identity), of ∼500 amino acids, with an amino-terminal extension harboring EF-hand Ca2+-binding motifs homologous to calmodulin (94). Mammalian orthologs for each SCaMC isoform are also highly conserved (96).
Human SCaMCs present amino- and carboxy-terminal variants generated by alternative splicing (33, 91, 94, 122) (Fig. 9). Two SCaMC-1 amino-terminal variants, of 477 and 458 amino acids, are generated from an alternative first exon (94, 122). The 477-amino acid-long SCaMC-1 variant is ortholog to Efinal, a protein from rabbit small intestine reported to be present both in peroxisomes and mitochondria (413). A more complex pattern of amino-end variants are found for human SCaMC-2 (94). Three SCaMC-2 products have been characterized SCaMC-2a, -2b, and -2c of 469, 503, and 488 amino acids, respectively, with amino-terminal ends encoded by three different exons 1 (Fig. 9). As a result, SCaMC-2a and SCaMC-2c lack the first EF-hand motif, but SCaMC-2b has all four (Fig. 9). Other SCaMC-2 variants, SCaMC-2d, arise from the use of an alternative 5′-splicing donor site in the most proximal exon 1, which causes a frame shift. This transcript is still able to generate two shortened proteins of 366 and 344 amino acids using internal ATGs. Both SCaMC-2d proteins lack EF-domains 1, 2, and 3 but maintain the whole carboxy-terminal end (Fig. 9) (94). SCaMC-2a, -2b, and -2c specific sequences have also been detected in other mammalian species, and a rat SCaMC-2a ortholog with predominant expression in liver and skeletal muscle has been identified (243).
The main product of SCaMC-3 is SCaMC-3a, of 468 amino acids in length and four EF-hand motifs (91, 94) (Fig. 9). Three carboxy-end variants, SCaMC-3b, -3b′, and -3c/3d, of 422, 456, and 435 amino acids, respectively, have been described (33, 91). These shortened proteins are encoded by alternatively spliced transcripts that bear 3′-end sequences derived from repetitive transposable elements, Alu and a MaLR retrotransposon, instead of sequences derived from the last SCaMC-3 exon (33, 91). All 3′-spliced sequences introduce premature stop signals into the SCaMC-3 transcripts. These carboxy-end variants lack the last transmembrane domain (TM6), but that does not change their location in mitochondria (33, 91).
The available data about SCaMCs expression and tissue distribution are far from being conclusive (33, 91, 94, 122, 243, 413). Most SCaMC isoforms show overlapping expression patterns with no clear tissue specificity. SCaMC-1, -2, and -3 transcripts are found in several tissues such as skeletal muscle, brain, kidney, and pancreas and also in ESTs sequences derived from eye (417). Some isoforms appear to have a prominent expression in specific tissues. Thus a high level of SCaMC-1 expression is found in small intestine from rabbit (413) and a high SCaMC-2 expression is detected in skeletal muscle and liver (94, 243). In pancreatic rat AR2J cells, SCaMC-2a has been found to be upregulated by dexamethasone (243).
On the other hand, human SCaMC-1 and SCaMC-2 amino-variants, SCaMC-1a, SCaMC-2b, and SCaMC-2c (Fig. 9), show a more restricted distribution than SCaMC-1 and SCaMC-2a, the variants of each gene most widely expressed. Thus SCaMC-1a variant is exclusively expressed in high levels in testis (122). A brain-specific pattern is observed for SCaMC-2c isoform (94). SCaMC-2b, the only SCaMC-2 protein containing four EF-hand domains, is detected at low levels in kidney, brain, and lung.
2. Targeting of ATP-Mg/Pi precursor proteins to mitochondria
The mechanisms of mitochondrial import of CaMC precursor proteins have been discussed in section iiiA2 (Fig. 2). These same mechanisms, notably the use of small Tim proteins Tim8p-Tim13p to escort the precursor proteins along the intermembrane space (Fig. 2), apply to ATP-Mg/Pi carriers coded by the SCaMC genes.
Some of the SCaMCs lack one or several EF-hand domains and even most of the amino-terminal extension, while others lack one transmembrane domain (91, 94) (Fig. 9). The lack of transmembrane domains does not alter the import of these SCaMC variants (91), but it is still unknown whether the actual transporter function of the carrier may be affected, as the current model for a functional carrier based on the structure of the ADP/ATP translocase predicts that all transmembrane helices and the short surface helices on the internal face of the inner membrane participate in the transporter function (290, 291). On the other hand, the total lack of the amino-terminal extension appears to redirect SCaMC precursor proteins to the escort proteins used by most MCs, the Tim9p-Tim10p proteins (91) (Fig. 2).
3. Transport activity of mammalian ATP-Mg/Pi carriers
A) NET INFLUX AND EFFLUX OF ADENINE NUCLEOTIDES IN MITOCHONDRIA.
The exchange of cytosolic ADP against mitochondrial ATP is essential to maintain oxidative phosphorylation and is carried out by two or more adenine nucleotide translocases present in all eukaryotes (96, 105, 283). The translocase can also work in the “reverse” mode, i.e., exchanging cytosolic ATP against mitochondrial ADP. In the reverse mode, the exchange of ATP4− to ADP3− is electrogenic, and it is thought to generate the mitochondrial membrane potential in yeast growing in anaerobiosis and in mitochondrial DNA (mtDNA) depleted cells (8, 60, 76, 223, 230). An upregulation of a chloride channel may also be involved in the generation of mitochondrial membrane potential in rho° cells (20). The reverse mode of the translocase can proceed as long as ATP is hydrolyzed in mitochondria by the F1F0-ATPase. Because proton pumping, which requires mtDNA encoded subunits of F1F0-ATPase, is not needed, this process operates in mtDNA depleted cells (8). The reverse mode of the adenine nucleotide translocase is also required during heart ischemia, to maintain a partially polarized mitochondrial membrane potential generated by proton pumping through the F1F0-ATPase at the expense of ATP hydrolysis (2, 111, 325).
It should be emphasized that the one-to-one exchange reaction catalyzed by the ANT does not result in net increase or decrease in adenine nucleotide (AdN) content in mitochondria. Although the translocase can also exchange AdNs against pyrophosphate, the affinity of the translocase for pyrophosphate is low compared with that of AdNs (200), and it is therefore doubtful that this exchange operates in vivo, at least under resting conditions, i.e., with normal pyrophosphate and AdN levels (13, 247). Thus it is generally believed that the ATP-Mg/Pi carrier found more than 20 years ago and only recently found to be encoded by the newly discovered SCaMC 1–3 genes is the only transporter responsible for net changes in AdN levels.
B) TRANSPORT ACTIVITY OF THE ATP-Mg/Pi CARRIER.
In the late 1970s, Sutton and Pollak (374) found that during rat neonatal development the AdN content of hepatic mitochondria increased more than twofold within the first 2 h after birth (374). The mechanism responsible for this rapid accumulation was found to be a saturable, atractyloside-resistant, and mersalyl-sensitive ATP transport (14, 26, 299, 300). ATP was the preferred substrate, and transport was inhibited by dATP, ADP, and AMP, but not by cAMP, adenosine 5′-[βγ-imido]triphosphate (p[NH]ppA), GTP, or CTP. ATP transport occurred in mitochondria incubated without added substrates but was inhibited by CCCP and KCN, indicating that the proton-motive force was required (300).
Studies carried out by Aprille (10, 13) established that the carrier present in liver mitochondria catalyzes an exchange of ATP-Mg for Pi across the inner mitochondrial membrane in either direction. To study the activity of the carrier, the AdN content of liver mitochondria is first decreased by in vitro incubation with pyrophosphate that enters mitochondria on the adenine nucleotide translocase in exchange with ATP or ADP, causing a net loss of AdN. Alternatively, matrix AdN can be depleted in preparation for studies of transport by incubating with Pi which catalyzes net flux over the ATP-Mg/Pi carrier itself. The matrix AdN content can then be restored by incubation of liver mitochondria with ATP, Mg2+, and phosphate (21). At equilibrium, the relationship between the concentration gradients of ATP-Mg and Pi across the inner mitochondrial membrane implies that, like Pi, ATP-Mg is taken up and retained in mitochondria against a concentration gradient relative to the cytosol. The pH gradient is the energy source for Pi accumulation, via the phosphate carrier, and this in turn drives ATP-Mg accumulation (via the ATP-Mg/Pi carrier), just as it also drives the transport of malate and other dicarboxylates through the dicarboxylate carrier. As a consequence, ATP-Mg uptake or efflux depends on the pH gradient and is abolished with nigericin and uncouplers while it is largely unaffected by dissipation of the membrane potential (15).
Net uptake or loss of AdNs occurs when flux in one direction exceeds that in the opposite direction. For example, if mitochondrial ATPase in inhibited by oligomycin, or mitochondrial ATP synthesis is low due to the lack of oxidizable substrates, then ATP-Mg uptake will take place. On the other hand, if cytosolic ATP levels are low, ATP-Mg efflux will prevail (13). According to Aprille (13), the activity of the ATP-Mg/Pi carrier increases or decreases the matrix AdN pool but does not substantially change the mitochondrial ATP to ADP ratio, because the carrier is much slower than the ATPase and ANT reactions, which exert control over that ratio (13). However, other authors have found that ATP-Mg accumulation increased the ATP/ADP ratio in the liver mitochondrial matrix (343).
The carrier catalyzes an electroneutral exchange of divalent phosphate (HPO) against divalent ATP-Mg (ATP-Mg2−) or divalent protonated ADP (HADP2−) (182, 275). The apparent Km for ATP-Mg in the presence of 2 mM Pi is ∼2.6 mM, and HADP2− competes poorly with ATP-Mg so that in vivo ATP-Mg and Pi probably are the physiologically relevant transported species (10, 108, 275). Remarkably, ATP-Mg/Pi carriers reconstituted into liposomes had the same substrate specificity (122). The transport rate of the ATP-Mg/Pi carrier is relatively slow (1–2 nmol · min−1 · mg protein−1) compared with that of the phosphate carrier or adenine nucleotide translocase (∼150–200 nmol · min−1 · mg−1; Ref. 217), a fact that was initially attributed either to the low abundance of the carrier or to a slow rate constant (10). In fact, the Vmax values of two of these carriers expressed in E. coli and reconstituted in liposomes (280 and 44 μmol · min−1 · mg protein−1) for SCaMC-1/SLC25A24 and SCaMC-3/SLC25A23, respectively (122), are much lower than that of the phosphate carrier (0.61 mmol · min−1 · mg protein−1) (142, 420).
ATP-Mg/Pi transport activity is inhibited by mersalyl and is insensitive to carboxyatractyloside (CAT) (10), a result also obtained with the carrier proteins reconstituted in liposomes. In addition, these studies revealed that pyridoxal-5-phosphate and bongkrekic acid considerably inhibited one of the mammalian isoforms, SCaMC-3/SLC25A23 (122).
C) Ca2+ REGULATION OF ATP-Mg/Pi TRANSPORT IN MAMMALIAN MITOCHONDRIA.
Haynes et al. (163) were the first to report that net AdN transport required micromolar Ca2+. In fact, net nucleotide uptake or efflux through the ATP-Mg/Pi carrier is completely blocked in the presence of excess EGTA and increases when free Ca2+ rises from 1 to 4 μM. This results in a lower apparent Km for ATP (4.4 and 2.4 mM at 1 and 2 μM Ca2+, respectively), with little changes in Vmax (276). Importantly, increased activity did not require Ca2+ entry in mitochondria, as it was unaffected by RR (276). Thus the ATP-Mg/Pi carrier was the first known MC with an absolute requirement for extramitochondrial Ca2+, a property which is not observed when the proteins are reconstituted in liposomes (122). Calmodulin antagonists inhibit the activation by Ca2+, suggesting that Ca2+ may stimulate ATP-Mg/Pi carrier activity via a specific calmodulin-related Ca2+ binding site (276). All SCaMCs have a set of EF-hand motifs highly similar to calmodulin (91, 94) that are thought to confer Ca2+ regulation to ATP-Mg/Pi activity, raising the possibility that calmodulin antagonists might directly interfere with the SCaMCs.
A related transport activity, the Ca2+-activated dCTP (and other dNTP) transporter, has been described in mitochondria from human acute lymphocytic leukemia cells (56), following the demonstration that a mechanism for mitochondrial dNTP uptake existed in isolated mitochondria that is essential for DNA synthesis (73, 118). The protein coded by the gene SLC25A19 has been shown to catalyze the transport of dNTP when reconstituted into proteoliposomes (105) and termed deoxynucleotide carrier (DNC). Mutations in SLC25A19 cause congenital microcephaly (323), and DNC is involved in the transport of ddNT and antiretroviral reverse transcriptase inhibitors (227). However, the DNC protein has no amino-terminal extension or Ca2+-binding domains, has a much lower affinity for dNTPs than the Ca2+-activated dCTP carrier (212), and overexpression or downregulation of DNC in RKO cells did not affect Ca2+-dCTP transport in mitochondria (212). Thus the molecular identity of the Ca2+-activated dCTP carrier is still unknown.
B. Hormonal and Developmental Changes in AdN Content in Liver Mitochondria
1. Hormonal regulation of liver mitochondria AdN content
Mitochondria isolated from hepatocytes stimulated with glucagon and Ca2+ mobilizing agents such as vasopressin, thapsigargin, or Ca2+ ionophores have increased AdN content (17, 59, 152, 156, 357, 362, 389), which is not due to an increase in mitochondrial volume (108). Glucagon-induced Ca2+ signals are transient and much smaller than those elicited by α-adrenergic agonists (phenylephrine) or vasopressin (82, 121, 368), but by increasing the intracellular concentration of cAMP and protein kinase A activity, glucagon and Ca2+ mobilizing agents synergistically stimulate the net inflow of Ca2+ to hepatocytes through receptor-activated Ca2+ channels (9, 168). Studies on the effect of mitochondrial inhibitors and RR on the effect of glucagon, forskolin, or dibutyryl cAMP on thapsigargin-stimulated Ca2+ inflow suggested that protein kinase A regulates the fate of Ca2+ entering the cell through store-activated Ca2+ channels by directing some of this Ca2+ to mitochondria (121). In fact, the increase in mitochondrial AdN content is not observed if hepatocytes are pretreated with EGTA, and this effect has been attributed to the strict Ca2+ requirement for the activation of the ATP-Mg/Pi carrier (163, 276). An alternative view held by Halestrap and co-workers (247) was that the increase in mitochondrial AdN content could also be caused by an exchange of cytosolic AdN with mitochondrial pyrophosphate on the ANT, as pyrophosphate accumulates in liver mitochondria after hormonal treatment as a consequence of Ca2+ inhibition of mitochondrial pyrophosphatase.
The fact that hormones can trigger a Ca2+-dependent activation of the ATP-Mg/Pi carrier in normal resting cells has been taken as an indication that the matrix/cytosol ATP-Mg gradient is lower than the Pi gradient in resting cells and ATP-Mg uptake is favored over release (108). However, the ATP-Mg/Pi carrier stays inactive unless triggered by an appropriate Ca2+ signal (108). By changing the matrix AdN content, the ATP-Mg/Pi carrier may regulate mitochondrial metabolic pathways that depend on AdNs either as substrates or cofactors, such as gluconeogenesis from lactate, urea synthesis, synthesis of mitochondrial proteins, RNA and DNA, and import of mitochondrial precursor proteins, among other processes. The first reactions in gluconeogenesis and urea synthesis are ATP-dependent carboxylations. In urea synthesis, CO2 is fixed into carbamoyl phosphate, which is formed by carbamoyl phosphate synthetase I, an ATP-dependent reaction. Citrulline arises from carbamoyl phosphate and ornithine, and citrulline synthesis in mitochondria isolated from hepatocytes treated with glucagon and α-adrenergic agonists is strongly activated (137, 156, 389). There is a correlation between AdN pool size and citrulline synthesis in mitochondria that suggests that the AdN pool size may control citrulline synthesis (137), although this has been questioned (156). Because the direction of net AdN movement occurs in both directions depending on the relative ATP/Pi concentration on each side, it has been suggested that in situations where energy status is low (i.e., hypoxia), Ca2+ activation of the carrier results in AdN loss to the cytoplasm. The result, at least in liver, would be downregulation of metabolic pathways that are energy consumers but are not necessary for survival of the liver cell itself (i.e., gluconeogenesis and urea synthesis). When favorable energy conditions (hypoxia to normoxia) are restored, mitochondria would take up AdN and intermediary metabolic pathways would be active again (13).
Interestingly, the emerging picture is one in which glucagon and α-adrenergic agents activate the two CaMCs existing in hepatocytes, causing both an increase in AdN content through Ca2+ activation of ATP-Mg/Pi carrier (which results in increased gluconeogenesis through an increase in pyruvate carboxylase and ureogenesis through an increase in citrulline synthesis) and increase in AGC activity, also required to activate MAS, gluconeogenesis, and ureogenesis (see sect. iiiA4).
2. Role of liver ATP-Mg/Pi transport in early postnatal development
In the newborn rat liver, initial AdN content in mitochondria is very low, but soon after birth the cytosolic ATP/ADP ratio and AdN content of mitochondria increase rapidly (14, 373, 374), more than twofold within the first 2 h after birth (373). A similar increase was found in neonatal rabbit liver (53, 327). Total AdN content of the liver increases more modestly, by ∼25% (83), so that the increase in the mitochondrial matrix AdN results in a decrease in AdN in the cytosolic compartment of ∼10–13% (see Ref. 12 for review).
The postnatal increase in mitochondrial AdN occurs shortly after catecholamine and glucagon are secreted at the time of parturition, while insulin levels decrease (85, 104, 134, 209). This leads to activation of glucagon and β2-adrenergic receptors in fetal liver (324), to increases in cAMP, and to increased rate of glycogenolysis and glycolysis-derived ATP production, resulting in an increased cytosolic ATP/ADP ratio (374). Activation of α-adrenergic receptors and increases in cAMP levels also lead to increases in free Ca2+ concentration in fetal hepatocyte cultures and to a stimulation of mitochondrial functions including the increase in mitochondrial AdN content (132). The availability of oxygen appears to be another factor controlling AdN uptake in postnatal mitochondria. At birth, as lungs inflate, there is a change in fetal liver circulation that results in sudden perfusion of the liver with well-oxygenated blood (12). This appears to be a permissive condition for mitochondrial uptake of AdN, since animals subjected to prolonged postnatal hypoxia fail to increase mitochondrial AdN content and related functions (12, 16, 54, 397).
The rapid postnatal accumulation of AdN in mitochondria was found to be due to a saturable, atractyloside-resistant, and mersalyl-sensitive ATP transport (11, 26, 299, 300, 373), carried out by the ATP-Mg/Pi carrier (reviewed in Ref. 13), which is thought to represent one of the targets of glucagon and catecholamines in fetal liver mitochondria. In preterm neonates that have lower ATP levels than term fetuses, the increase in liver mitochondrial AdN content is delayed by 2 h and is preceded by an equally delayed increase in cytosolic ATP (402), an increase required to drive the ATP-Mg/Pi carrier towards mitochondrial ATP accumulation (13). Control strength studies with appropriate inhibitors led to the finding that the early postnatal AdN accumulation in mitochondria results in an increase in the respiratory control index (298) mainly as a consequence of an increase in state 3 respiration (373, 400, 401) and also from structural changes in the inner mitochondrial membrane that result in decreases in mitochondrial volume (400) and permeability to protons (28, 298, 399, 401). The increase in state 3 respiration resulted from an increase in ANT activity (13, 14, 327) and/or F1-ATPase activity (13, 28, 401), which have been attributed both to a simple mass action ratio effect of the increase in matrix AdN content and to rapid de novo synthesis of mitochondrial proteins, particularly F0F1-ATP synthase (3, 248, 400), a process dependent on an increased stability and translational efficiency of mitochondrial transcripts (179, 232), but also requiring adequate ATP levels for chaperonin-assisted folding of the mitochondrial proteins.
The increase in mitochondrial AdN content produced in the postnatal liver could stimulate pyruvate carboxylase and citrulline formation in response to glucagon, as discussed in section ivB1. Pyruvate carboxylase in mitochondria and phosphoenolpyruvate carboxykinase are rate limiting for gluconeogenesis immediately after birth (54, 86, 87), and pyruvate carboxylation rates increase in concert with the increase in AdN content (18). Increasing AdN content of newborn mitochondria in vitro mimicked the developmental changes in pyruvate carboxylase activity, indicating that pyruvate carboxylation is one of the consequences of postnatal activation of the mitochondrial ATP-Mg/Pi carrier (12, 18, 53). Citrulline synthesis is another candidate for regulation by the postnatal increase in mitochondrial AdN content, as it is low in newborn mitochondria and increases along with the AdN content (12). Intramitochondrial protein synthesis, which increases severalfold along the first postnatal hours, is also regulated by the mitochondrial AdN pool, and it greatly increases when 0-h newborn mitochondria are preincubated to accumulate AdN (183).
In sum, cAMP and Ca2+ converge to stimulate the ATP-Mg/Pi carrier of newborn liver mitochondria, through an increase in cytosolic ATP (mainly through cAMP-mediated stimulation of glucogenolysis and glycolytic ATP production) and an increase in cytosolic free Ca2+, which is required to trigger transport activity of the carrier. The consequences of the postnatal increase in liver mitochondria AdN content may vary in different species. In the rat, the mitochondrial AdN pool is very small at birth, and its rapid postnatal increase correlates with an increase in ANT activity and state 3 respiration (13). On the other hand, in the rabbit the mitochondrial AdN content at birth is larger and its rapid increase is not followed by changes in state 3 respiration (327). Therefore, the consequences of the small mitochondria AdN pool at birth depend on its size with respect to a certain threshold that may vary depending on the particular mitochondrial function.
3. ATP-Mg/Pi transport in extrahepatic tissues
A mechanism for net AdN transport similar to the one described in liver mitochondria has also been observed in rat kidney (146, 165). In kidney mitochondria, net AdN uptake and efflux takes place through two different mechanisms, a CAT-resistant and Ca2+-regulated pathway similar to the ATP-Mg/Pi carrier from liver mitochondria, and a CAT-sensitive pathway (inhibited at 5 μM CAT) that is resistant to EGTA (146). ATP efflux or uptake through these pathways was inhibited (25–50%) by high (10 μM) cyclosporin A concentrations (146, 165) but not by 1 μM cyclosporin A (146), a concentration that is adequate to fully desensitize the permeability transition pore in liver mitochondria to cyclophilin D (34). From its CAT sensitivity, it has been suggested that the mechanism responsible for this second pathway could be an isoform of the ANT different from that of liver (146).
Net changes in AdN content occur in heart mitochondria, but the mechanisms seem to be different from those in liver mitochondria. Net decreases in AdN content have been observed in isolated heart mitochondria exposed to high Pi concentrations, but the transport mechanisms are to a large extent CAT sensitive and do not match that of the liver mitochondria ATP-Mg/Pi carrier. AdN efflux has been shown to proceed through two pathways: a fast, Pi-dependent pathway, which is blocked by CAT, and a slow and CAT-resistant pathway (23, 415), to be discussed in section ivB3a. Unlike liver mitochondria, rat heart mitochondria were unable to accumulate AdN when incubated with 2 mM ATP (24). On the other hand, the postnatal increase in mitochondrial AdN pool size that has served as metabolic signature of the ATP-Mg/Pi carrier activity is a rapid process only in liver, but it is much slower in heart (349). Therefore, the identity of the possible carriers involved in AdN loading of mitochondria in extrahepatic organs is still unknown.
The SCaMC proteins are obvious candidates for such a role. Compared with other MCs, SCaMC-1 to -3 have an unusually large number of variants. These variants probably differ in their Ca2+ binding properties, as Ca2+ binding motifs are lost in some of them (Fig. 9) (91, 94), and perhaps also in catalytic activity, since some SCaMC-3 lack one transmembrane domain (91) (see Fig. 9). Fiermonte et al. (122) studied the catalytic activity of the proteoliposome-reconstituted carboxy-terminal regions of SCaMC-3 and SCaMC-1 and found that SCaMC-3 activity is about eight times lower than that of SCaMC-1, whereas the affinities for substrates were the same (122). SCaMC-1 and -3 were equally resistant to CAT (10 μM), but another translocase inhibitor, bongkrekic acid (10 μM), considerably inhibited SCaMC-3 but not SCaMC-1. This opens up the possibility that the different SCaMC isoforms and variants may have on the one hand differences in Ca2+ sensitivity, and on the other, partial sensitivity towards the classical translocase inhibitors. The particular set of SCaMC genes and variants in kidney and heart may be responsible for the changes in AdN content observed in isolated kidney or heart mitochondria.
As indicated in section ivB2, in intact cells the matrix/cytosol ATP-Mg gradient is lower than the Pi gradient, and the ATP-Mg/Pi carrier remains inactive in the resting state, unless stimulated by an appropriate Ca2+ signal (108). The activated carrier would direct ATP-Mg into mitochondria, and the resulting increase in mitochondrial AdN should increase ANT and F1-ATPase activity through mass action ratio, resulting in an increase in mitochondrial ATP synthesis (10). This should actually result in a Ca2+-dependent increase in mitochondrial ATP synthesis.
Increases in mitochondrial ATP synthesis that depend on Ca2+ signals have been reported in heart at different work loads (382, 383; see Ref. 31 for review). 31P-NMR studies have shown that these increases occur without the variations in ADP, ATP, Pi, or phosphocreatine concentrations that normally control oxidative phosphorylation in skeletal muscle and have been attributed to metabolic consequences of increase in mitochondrial Ca2+ (31). In intact myocardial cells, mitochondrial Ca2+ tracks the rapid changes in cytosolic Ca2+ levels occurring in each cycle of contraction (70, 176), and increased frequency or amplitude of cytosolic Ca2+ spikes is mirrored by similar changes in mitochondrial Ca2+ spikes (315, 395), and by elevations in interspike Ca2+ levels (diastolic increase in mitochondrial Ca2+) (103, 315). The mechanisms proposed for mitochondrial Ca2+ signaling in heart are a Ca2+-induced increased activity of mitochondrial dehydrogenases (but see Ref. 410), and activation of F0F1-ATPase (31, 161, 345, 382, 383, but see Ref. 159 for review), and ANT (262) by matrix Ca2+. It has been proposed that F1-ATPase is activated by Ca2+-induced release of a Ca2+-binding inhibitor protein, CaBI (422, 423), and Ca2+ binding to the β-subunit (170), or e-subunit of F1-ATPase (19). Experiments in which RR inhibited the activation of F1-ATPases at high work loads implicated matrix Ca2+ as the primary effector (161, 383). However, the effects of RR in intact cells are not specific for the Ca2+ uniporter (110), and Robert et al. (315) found RR or R360 to be totally inefficient in inhibiting the Ca2+ uniporter in intact neonatal cardiomyocytes. Thus a possible upregulation of heart ATP-Mg/Pi carriers by cytosolic Ca2+ emerges as an alternative possibility contributing to increased ATP synthesis with increasing work loads.
Early studies with mitochondria isolated from ischemic organs showed a diminished content of AdN in mitochondria from heart (22, 220), liver (270, 411), and kidney (164, 165) associated with a decrease in oxidative phosphorylation typically found in mitochondria isolated from ischemic tissues. After 20 min of ischemia, the levels of heart mitochondrial AdN decreased to ∼20% of control values, with a rapid loss between 10 and 20 min (22). State 3 respiration was also inhibited, and titration with CAT of state 3 respiration in mitochondria from control and ischemic heart suggested that the decrease in state 3 and uncoupled respiration caused by ischemia cannot be attributed to a loss of ANT activity but rather to the decrease of intramitochondrial AdN levels (22).
As Pi can induce an efflux of AdN from heart mitochondria, and Pi doubles within 1 min after the onset of heart ischemia (210, 272, 337), a number of studies by Asimakis and co-workers (23, 337, 355) have addressed the involvement of the ATP-Mg/Pi carrier in loss of mitochondrial AdN. The loss of AdN in rat heart mitochondria has been suggested to occur either by enzymatic degradation in the mitochondrial matrix or by movement of AMP, ADP, or ATP across the inner mitochondrial membrane. Since 5′-nucleotidase, which is responsible for the dephosphorylation of AMP and IMP to adenosine and inosine, is not found in the mitochondrial matrix (192) and in vitro studies of AdN efflux from isolated heart mitochondria incubated with Pi showed a correlation between the loss of mitochondrial AdN and the appearance of extramitochondrial AdN (23), it was concluded that the most probable mechanism was AdN efflux (337). AdN efflux has been shown to proceed through two pathways: a fast, Pi-dependent pathway, which is blocked by CAT, and a slow Pi-independent and CAT-resistant pathway (337). The Pi-dependent pathway has many of the characteristics of the liver ATP-Mg/Pi carrier, including the requirement for Mg2+ (23, 355), but appeared to be a different mechanism, as it was inhibited by atractyloside.
The extent of AdN loss from mitochondria and its implications for a decrease in state 3 respiration in ischemic heart are somewhat controversial. Dransfield and Aprille (107) did not find any decrease in the mitochondrial AdN pool in hepatocytes incubated under chemical hypoxia, while the mitochondrial AdN shifted to AMP (107). Not all studies show a clear-cut decrease in the activity of the respiratory chain during ischemia, and in some studies state 3 respiration was found to increase after 30 min of ischemia (49, 228; see Ref. 363 for review). These variations are perhaps associated with the preparation of heart mitochondria, which may not reflect that of the original tissue, particularly in substrate and ion content, especially in ischemic heart, that have more fragile mitochondria than those of normal heart (181). In addition, the procedures for isolation of heart mitochondria yield largely subsarcolemmal mitochondria rather than interfibrillar mitochondria, which provide most of the energy for the contractile apparatus (see Ref. 363 for review).
Although a role for ATP-Mg/Pi carriers in efflux of mitochondrial AdN to the cytosol during ischemia is uncertain, these carriers could be involved in the events triggered during reperfusion, particularly the opening of the permeability transition pore (see Refs. 37, 38, 81, 102, 153, 154, 274 for reviews). This pore, which allows the passage of solutes smaller than 1.5 kDa across the inner mitochondrial membrane, opens under the influence of Ca2+, lack of AdNs, and oxidative stress and is inhibited by cyclosporin A (CSA), an inhibitor of cyclophylin D (CypD), a peptidyl-prolyl cis-trans isomerase (PPIase) in the mitochondrial matrix (5, 378). The pore has been shown to open during reperfusion, but not ischemia (102, 153), possibly as a consequence of the increased ROS formation that accompanies reperfusion, and the critical role of mitochondrial-generated reactive oxygen species in PTP opening (201).
As the total level of mitochondrial AdN regulates PTP opening, Hagen et al. (147) have proposed that the ATP-Mg/Pi carrier could modulate the probability of PTP opening by changing mitochondrial AdN levels. Pi and Ca2+ increase and ATP decreases during ischemia, and under these conditions, the Ca2+-dependent efflux of mitochondrial ATP-Mg in exchange for cytosolic Pi on the ATP-Mg/Pi carrier was found to increase the susceptibility of PTP opening by high Ca2+, atractyloside, and prooxidants in liver mitochondria (147). However, whether this mechanism operates in vivo and its relevance to heart ischemia-reperfusion is still an open question.
The role of CypD in PTP opening has been proven conclusively after the generation of mice deficient in CypD (30, 34, 268, 344). Mitochondria from these mice require much higher Ca2+ loads to induce PTP opening (30, 34, 268, 344). Cell death in heart and brain as a consequence of ischemia-reperfusion was greatly reduced by CypD deficiency, indicating that PTP opening, which recruits CypD to the pore, is involved specifically in this type of necrotic death (30, 268, 344; reviewed in Refs. 39, 140, 150).
An involvement of ANT in PTP formation was initially suggested from the pore properties: sensitivity to the ANT substrates ATP and ADP (as closers of the pore) and inhibitors (CAT that locks ANT in a “c” conformation and favors pore opening and bongkrekic acid that promotes the “m” conformation of ANT and favors pore closure; see Ref. 426 for review). A favored model envisions that the ANT turns into a nonselective channel, an idea supported by the finding that a number of MCs, such as the ANT, PiC, and the AGC can behave as relatively unspecific uniporters by treatment with SH reagents (99, 100, 204). However, recent work by Kokoszka et al. (199) has shown that mitochondria from livers of ANT-knockout mice in which the two ANT genes expressed in liver, ANT1 and 2, were genetically inactivated, still have PTP activity. From this it was concluded that PTP can form in the absence of ANT, although regulation of PTP activity by atractyloside and ADP is lost and its sensitivity to Ca2+ is also decreased. Thus it appears that ANTs are nonessential components of the PTP but can regulate PTP activity in response to Ca2+ and ANT agonists (199). This opens the possibility that other MCs related to these ANT isoforms could take over part of the role, both in normal liver function and in PTP formation (151). Indeed, a most surprising finding was that these ANT-knockout mice survived normally for more than a year (199). Since mice do not have ANT3 (199), this could include the newly described ANT4 (106, 316), although this isoform is not normally expressed in mouse liver (316).
The ATP-Mg/Pi carriers are the closest homologs of ANT and are widely expressed in every tissue. In S. cerevisiae, Sal1p, the yeast ATP-Mg/Pi carrier, can partially take over the role of ANT, and vice versa (68, 74), and this opens up the possibility that this can also occur in mammalian tissues. It shares with ANT the sensitivity to AdNs, but not atractyloside sensitivity, at least not for isoforms SCaMC-1 and SCaMC-3 (122). However, one of the ATP-Mg/Pi isoforms, SCaMC-3, is sensitive to bongkrekic acid (122) so that the similarity among the ANT and ATP-Mg/Pi carriers could be even larger. As with all other MCs, all three SCaMC isoforms have prolines at conserved positions in the boundary of the TM domains facing the matrix (91, 94), that could interact with the mitochondrial PPIase CypD. In fact, ATP-Mg loading and efflux from kidney mitochondria through the ATP-Mg/Pi carrier have been shown to be inhibited by CSA, albeit at relatively high concentrations (146, 166). Being strictly Mg2+ dependent, the ATP-Mg/Pi carriers might also confer Mg2+ sensitivity to the PTP, a known property of the pore (37). PTP opening is triggered by Ca2+, and there is mounting evidence that this involves matrix Ca2+ rather than Ca2+ in the intermembrane space, the compartment sensed by the ATP-Mg/Pi carrier. However, under certain conditions, PTP opening has revealed an extramitochondrial site for Ca2+ regulation. Induction of PTP by thiol cross-linkers such as phenylarsine oxide does not require the presence of Ca2+ in the mitochondrial matrix and is activated by extramitochondrial Ca2+ (200, 201, 224a).
In conclusion, although largely speculative, the notion that the ATP-Mg/Pi carriers may be involved in the formation of the permeability transition pore deserves further work and consideration.
C. Nonmammalian ATP-Mg/Pi Carriers
ATP-Mg/Pi carriers are widely conserved, with orthologs in most of eukaryotic organisms (94). Yeast contains a single isoform, whereas higher eukaryotes have two (C. elegans or chicken) or three isoforms (Danio rerio, Xenopus, or Arabidopsis) with high similarity among each other (96). Except for the S. pombe ortholog, all proteins display EF-hand motifs at equivalent positions (94). It has been suggested that these isoforms arose by a gene fusion process between a primitive ANT gene and calmodulin-like sequences (96).
1. Transport activity of yeast Sal1p
Sal1p is the product of baker’s yeast gene YNL083w, renamed SAL1 (for suppressor of aac2 lethality) by Chen (74). YNL083w was identified during the sequencing of the genome of S. cerevisiae, and it was found to be nonessential (364, 427). The sequences of human SCaMCs and Sal1p are very similar, with 34–37% identity at the amino acid level along the 300-amino acid-long carboxy terminus (68, 94).
Like its human counterparts, Sal1p is localized to mitochondria (68). The transport activity of Sal1p was studied in mitochondria obtained from wild-type and Sal1p-deficient yeast (68) and in yeast deficient in all three ANTs, aac1,2,3Δ (109), in which the main mechanism for ATP uptake in mitochondria is Sal1p. Freshly isolated wild-type yeast mitochondria incubated in the presence of CAT increased their total AdN content by about threefold, after 5- to 10-min incubation in the presence of 4 mM ATP, and this was reduced by 40–50% in Sal1p-deficient mitochondria (68). ATP uptake on Sal1p was Mg2+ dependent, and ATP efflux was stimulated by external Pi (68), as reported for mammalian ATP-Mg/Pi carriers.
2. Ca2+ regulation of ATP-Mg/Pi transport in yeast mitochondria
ATP-Mg uptake on Sal1p was stimulated by extramitochondrial Ca2+, with an S0.5 of ∼30 μM (68), a value somewhat higher than that reported for the rat liver ATP-Mg/Pi carrier (163, 276). This is perhaps not too surprising as Sal1p lacks two (out of four) canonical EF-hand motifs present in most human SCaMCs (94). EF-hands 1 and 4 have two amino acids that do not satisfy the coordination bond of canonical EF-hands. EF-hands 2 and 3 are canonical, and Chen (74) showed that they are required for Ca2+ binding and other Sal1p functions. Thus the lower affinity for Ca2+ of Sal1p with respect to the liver ATP-Mg/Pi carrier may be explained on this basis.
3. Physiological role of yeast Sal1p
As there is a large number of mitochondrial functions that require ATP in the mitochondrial matrix, it was surprising that a triple deletant of all three ANTs (aac1,2,3Δ) was still able to grow on glucose (109). This finding led to the proposal of the existence of another transporter that could supply mitochondria with AdNs in these conditions (109). Wild-type and aac1,2,3Δ yeast express Sal1p during growth on glucose, and Sal1p expression was found to be required for growth in the absence of all three translocases (68, 74). Indeed, any combination of aac2Δ and sal1pΔ is lethal. These findings support the idea that Sal1p is the ATP carrier that compensates for the lack of ANTs, particularly Aac2p, during growth on glucose.
Glucose sensing in yeast is required to tune growth and metabolic activity to amount and type of sugars available. Glucose uses a signal transduction pathway similar to that employed by hormones in higher eukaryotes, in which both cAMP and Ca2+ are early second messengers (139, 320). Glucose binds to the G protein-coupled receptor Gpr1p (224) inducing a transient elevation of cytosolic Ca2+ through a pathway involving PLC stimulation (6, 387) and IP3 production (388). Although glucose-induced Ca2+ transient involves mainly Ca2+ influx from the external medium through the Cch1p/Mid1p plasma membrane Ca2+ channel (391), Ca2+ entry was not required to generate glucose-induced physiological changes as none of them was blocked in the presence of EGTA (138). In the face of these contrasting facts, the role of Ca2+ signaling in glucose sensing and the relevant targets, if any, of the glucose-induced Ca2+ transient have remained largely unknown.
Cavero et al. (68) tested the hypothesis that Sal1p may be a target of the glucose-induced Ca2+ transient that would be revealed in the absence of ANTs. By studying the kinetics of glucose-induced bud formation, it was found that in cells lacking all three translocases, the presence of EGTA during preexposure to glucose resulted in a total block of the effect of glucose. This effect was not observed for wild-type or Sal1p-deficient yeast. In fact, the kinetics of bud emergence in aac1,2,3Δ yeast that rely on Sal1p for ATP loading of mitochondria was the same as that obtained in the absence of glucose (68). These results show that Ca2+ binding to Sal1p is a component of the glucose sensing system that becomes essential in yeast lacking ANTs (Fig. 10). This is also supported by the results of Chen (74), who showed that mutations in the EF-hand motifs of Sal1p that eliminated Ca2+ binding also abrogated the ability of the protein to suppress the lethality of aac2Δ sal1Δ double mutants, suggesting that Ca2+ binding by the protein is critical for its function during growth on glucose (74). This vital and Ca2+-dependent function of Sal1p is ATP entry in yeast mitochondria, a function that can be taken over by ANT, possibly by exchanging cytosolic ATP for mitochondrial pyrophosphate in a Ca2+-independent way.
Baker and Schatz (30) proposed that the only mitochondrial proteins essential for viability in S. cerevisiae are those involved in protein import. Import and assembly of mitochondrial proteins requires matrix ATP hydrolysis at various steps including mHsp70 at TIM23 to translocate preproteins to the mitochondrial matrix (196), and for chaperonin-assisted folding of mitochondrial proteins (240). This requirement probably explains the vital role of Sal1p in the absence of ANTs.
V. EVOLUTION OF MECHANISMS FOR CALCIUM SIGNALING IN MITOCHONDRIA
Admittedly, the Ca2+ uniporter pathway is by and large the most potent system for Ca2+ signaling in mitochondria. Mitochondria from vertebrate tissues have the capacity for respiration-linked Ca2+ uptake (65), a capacity also present in insect flight muscle mitochondria (62, 66, 157, 419) and trypanosomatids (263, 266). However, this capacity does not appear to be universal. Ca2+ transport varies between species in plant mitochondria and between tissues within the same species (128). Mitochondria isolated from hypocotyls of plants such as coffee, soybean, and corn, but not from potato tubers and white cabbage leaves, have the ability to transport Ca2+ (128, 242). Neither energy-linked Ca2+ transport, high-affinity Ca2+ binding, nor Ca2+-stimulated respiration occurs in mitochondria isolated from S. cerevisiae or Torulopsis utilis (Candida utilis) (64, 65). The absence of a Ca2+ uniporter in mitochondria from S. cerevisiae has also been confirmed in later studies (184). Because mitochondria from N. crassa showed a slow respiration-dependent Ca2+ accumulation, a limited capacity to accumulate Ca2+, and lacked high-affinity Ca2+ binding sites, it is uncertain whether they have a real Ca2+ uniporter (65). Since the molecular identity of the Ca2+ uniporter or the RAM (365) are still unknown, studies demonstrating that such transport system is actually present or absent in Neurospora mitochondria could provide helpful clues.
A separate set of mechanisms for Ca2+ signaling in mitochondria relies on Ca2+ signals that reach the external face of the inner mitochondrial membrane. Three of these are functionally related, in that all three supply extra reducing equivalents to mitochondria, the Ca2+-sensitive external NAD(P)H dehydrogenases from fungi and plants (which transfer electrons to complex III), FAD-GPDH (which also transfers electrons to complex III), and the AGC-MAS pathway which supplies electrons to complex I.
The first two pathways are energetically equivalent, and either one of them appears to be present in eukaryotes. Thus most fungi and plants have the external Ca2+-dependent NAD(P)H DH system, but not the Ca2+-regulated FAD-GPDH system, whereas animals have the Ca2+-regulated FAD-GPDH system but lack Ca2+-regulated external NAD(P)H DHs. An interesting exception to this rule is the yeast S. cerevisiae. Mitochondria from this organism do not have a Ca2+-regulated external NAD(P)H DH or a Ca2+-regulated FAD-GPDH and thus are completely devoid of a mechanism whereby cytosolic Ca2+ can regulate the transfer of reducing equivalents from cytosolic NAD(P)H to complex III. The AGC-MAS pathway is energetically superior to the other two, in that the reducing power is transferred to NADH that can be oxidized in complex I. The Ca2+-regulated AGC-MAS system is present in all known plant and animal genomes, with the perhaps not surprising exception of S. cerevisiae, where agc1p is not regulated by Ca2+.
The singularity of these two exceptions is probably related to the uniqueness and extreme simplicity of the Ca2+ signaling mechanisms in Saccharomyces. Indeed, the Ca2+-signaling apparatus of S. cerevisiae is much simpler than that of animal and plant cells and is involved in the regulation of the cell cycle, mating, sensing of glucose, resistance to salt stress, and cell survival. In filamentous fungi with more complex growth patterns, Ca2+ is involved in other processes such as secretion, cytoskeletal organization, hyphal tip growth, branching, sporulation, and circadian clocks (47, 425). Thus filamentous fungi have a substantive “toolkit” of Ca2+-signaling proteins, far more complex than that present in budding yeast (47, 425).
An additional mechanism whereby cytosolic Ca2+ signals to mitochondria from the external face of the inner mitochondrial membrane is the Ca2+-regulated ATP-Mg/Pi carrier. The functional outcome of this regulation is quite different; it is not related to reducing equivalent transfer to mitochondria, but rather to loading or unloading of ATP in mitochondria, the consequences of which are still to be established. Interestingly, this mechanism for Ca2+ signaling is most probably the most basic of them, including the Ca2+ uniporter, as it is the only one present in the very small repertoire of Ca2+ signaling devices of S. cerevisiae.
The yeast Ca2+-regulated ATP-Mg/Pi carrier becomes essential to transport ATP in mitochondria when the ANTs are absent (68). Interestingly, transporters related to the ATP-Mg/Pi carrier and involved in a similar function are present in highly degenerate mitochondrion-related organelles such as hydrogenosomes (229) and mitosomes (239, 394). These organelles, known as hydrogenosomes or mitosomes depending on whether or not they generate molecular hydrogen as the end product, are present in a large range of microbial eukaryotes that lack mitochondria and have probably evolved from a common mitochondrial endosymbiont (404). Hydrogenosomes and mitosomes appear to derive from mitochondria that lost their respiratory function as a result of movement into anaerobic habitats. Hydrogenosomes use a fermentative pathway for pyruvate metabolism with substrate level phosphorylation to synthesize ATP (115, 116). There are two types of ADP/ATP transporters in hydrogenosomes. Those from chytridiomycete fungi, such as Neocallimastix, are true ADP/ATP carriers and are inhibited by CAT and bongkrekic acid (403, 406). Those of the anaerobic flagellates Trichomonas vaginalis and Trichomonas gallinae, the Hmp31 proteins (116), are closer to SCaMCs, except for the lack of amino-terminal extension (33–34% identity and 53–56% similarity to the MC homology region of SCaMC-1–3) and are not inhibited by bongkrekic acid (390). Mitosomes do not synthesize ATP (404), and the ADP/ATP transporter of mitosomes from the human parasitic protozoa Entamoeba histolytica (72) is related to the ATP-Mg/Pi carrier and Hmp31 proteins. It is the only member of the MC family present in this organism, which otherwise lacks mitochondria. This carrier has 25% identity and 44% similarity with Sal1p, and 26–28% identity and 47–49% similarity with SCaMC 1–3; it catalyzes an electroneutral transport of ADP and ATP not inhibited by CAT or bongkrekic acid (72) and is involved in the uptake of ATP in mitosomes, where ATP is needed for the chaperonin-assisted folding of nuclear-coded mitosomal proteins (72, 393).
In sum, the ATP-Mg/Pi carriers appear to be present not only in mitochondria, where they are Ca2+ regulated, but also in most types of mitochondrion-related remnant organelles, where they are not regulated by Ca2+.
There are many mechanisms for Ca2+ signaling in mitochondria independent of Ca2+ entry. Some of them have been known for a long time, particularly FAD-GPDH, the external Ca2+-dependent NADPH dehydrogenases, and the liver ATP-Mg/Pi carrier, while others, the AGCs, are relatively new. However, their physiological role as mechanisms for Ca2+ signaling in mitochondria has remained in the dark, probably because the Ca2+ uniporter (CU) has been considered the one and only mechanism responsible for that role. Indeed, the consequences of their activation are either increases in mitochondrial redox potential (NADH or flavines), or ATP levels, such as those expected to occur upon Ca2+ entry in mitochondria. It is thus not surprising that the existence and importance of some of them in relation to the CU has only become evident when data from genetically modified organisms has emerged. This is the case for two CaMCs: neuronal AGC1, aralar, which through its participation in the malate-aspartate NADH shuttle plays a prominent role in transducing small Ca2+ signals to neuronal mitochondria, and that of the yeast ATP-Mg/Pi carrier Sal1p and its role in sensing glucose-induced Ca2+ signals. A more complete understanding of the interplay between these different mechanisms requires the molecular identification of the CU, the development of more specific tools for the analysis of transporter function, and genetically engineered organisms, and this should establish the actual role of each of these CaMC in Ca2+ signaling in mitochondria.
We are thankful for financial support received from the Ministerio de Educación y Ciencia, Fondo de Investigaciones Sanitarias del Ministerio de Sanidad (FIS), Comunidad de Madrid and Química Farmaceútica Bayer, and the institutional support of the Ramón Areces Foundation to the Centro de Biología Molecular Severo Ochoa.
We are indebted to Dr. Kathryn F. LaNoue, Dr. José M. Cuezva, and Dr. June Aprille for critical reading of the manuscript and for many helpful suggestions; to Laura Contreras, Javier Traba, Patricia Mármol, Dr. Santiago Cavero, and Dr. Milagros Ramos for their excellent work in quest for the role of CaMCs; and to José Belio for assistance with the illustrations.
Address for reprint requests and other correspondence: J. Satrústegui. Dept. Biología Molecular, Centro de Biología Molecular Severo Ochoa, UAM-CSIC, Campus Cantoblanco, 28049 Madrid, Spain (e-mail:)
- Copyright © 2007 the American Physiological Society