The field of mitochondrial ion channels has recently seen substantial progress, including the molecular identification of some of the channels. An integrative approach using genetics, electrophysiology, pharmacology, and cell biology to clarify the roles of these channels has thus become possible. It is by now clear that many of these channels are important for energy supply by the mitochondria and have a major impact on the fate of the entire cell as well. The purpose of this review is to provide an up-to-date overview of the electrophysiological properties, molecular identity, and pathophysiological functions of the mitochondrial ion channels studied so far and to highlight possible therapeutic perspectives based on current information.
Mitochondria, the key bioenergetic organelles, consist of an outer membrane (OMM), an inner membrane (IMM), the intermembrane space between OMM and IMM, and the IMM-enclosed matrix. The protein complexes carrying out oxidative phosphorylation reside in the IMM, while the matrix is the site of metabolic processes such as the citric acid cycle and fatty acid beta-oxidation. Passive ion movement across membranes is driven by its electrochemical gradient (Δμion), which has two components: 1) the electrical one, i.e., the voltage difference between the two sides of the membrane (ΔΨm); and 2) the chemical one, i.e., the difference between the activities (often approximated as concentrations) of the ion on the two sides (ΔpIon) (for a discussion of these principles see, e.g., Refs. 80, 912). In energized mitochondria, the membrane potential difference (ΔΨm) created by the respiration-driven efflux of protons from the matrix is a major component of the Δμion across the IMM for any ion. Under normal circumstances its value is believed to be in the 160–200 mV range (negative on the matrix side). Whether a sizeable electrical field exists across the OMM (and, in case, its modulation) remains to be clarified. It has been argued that it might exist as a Donnan potential across the semipermeable membrane, reaching a value as high as ∼40 mV (1028).
In this review we discuss channel activities studied by electrophysiological techniques such as patch-clamping isolated mitochondria and mitoplasts devoid of the outer membrane or incorporating mitochondrial membrane vesicles or purified native or recombinant proteins into planar lipid bilayers. Thus the list includes not only bona fide ion channels, but also proteins with other functions (e.g., the ones involved in protein import) which conduct ions under certain conditions. Last, we briefly mention putative mitochondrial channels, i.e., proteins believed or shown to function as channels in other membranes and identified in mitochondria as well, whose channel activities have however not been proven directly in the IMM or OMM. We do not include discussion of proteins facilitating the transport of neutral species, such as aquaporin, and of carriers of the inner membrane for which no ion-conducting activity has been described.
II. ION CHANNELS AND PORES OF THE OUTER MITOCHONDRIAL MEMBRANE
A. Voltage-Dependent Anion Channel
Figure 1 illustrates the various ion channels and proteins able to conduct ions which reside in the OMM, and which include the mitochondrial porin, voltage-dependent anion channel (VDAC). VDAC1, one of the three isoforms of the most abundant protein of the OMM, was the first mitochondrial channel to be discovered (274, 1137) and has been the object of a large number of studies. Many excellent reviews are available (115, 150, 153–155, 180, 222, 272, 276–278, 313, 315–317, 544, 752, 956, 970, 1081, 1087, 1192, 1195, 1196, 1198, 1346), including comprehensive, detailed, and recent ones (273, 275, 1190, 1191, 1193) and a special issue of Biochimica Biophysica Acta dedicated to “VDAC structure, function and regulation of mitochondrial metabolism” [vol. 1818(6), June 2012]. We will refer to VDAC1 as “VDAC” unless we want to emphasize the isoform involved.
Some 30 years after the field was opened by the studies by Schein and Colombini, and much intervening work seeking to clarify its architecture, the structure of mammalian VDAC was reported in 2008 (122, 263, 549, 1045, 1335, 1336, 1430). The channel was confirmed to be a β-barrel, composed of 19 β-strands and an NH2-terminal α-helix adhering to the inner wall of the pore but possessing considerable freedom of motion (1354). The NH2-terminal segment is a crucial region, involved in many molecular interactions. The barrel measures 3.1–3.5 nm (it is slightly elliptical) in the plane of the membrane, and ∼4 nm in the perpendicular dimension. The path for solute passage measures ∼1.5 by 1 nm.
2. Electrophysiological properties
The channel properties of the purified and reconstituted protein have been thoroughly defined in studies using mostly the planar bilayer technique. In experiments of this type, VDAC reproducibly forms a large, voltage-dependent pore, whose characteristics, rather well conserved among species, are summarized in Table 1. A weak selectivity for anions is conferred to the fully open state by a higher density of positive versus negative charges on the inner walls (263, 1336). The channel can in any case conduct a substantial flow of cations, including Ca2+, for which it also possesses binding sites (318, 447, 575, 1044, 1192). Interestingly, the fully open channel reportedly conducts Ca2+ rather poorly (PCl/PCa ∼25 in experiments with a CaCl2 gradient), while partially closed states (known to be cation selective) are much better at it (PCl/PCa ∼1–4.5) with the somewhat paradoxical result that Ca2+ flux through a partially closed VDAC may actually be a fewfold higher than through the fully open pore (1282). Relatively small molecules such as metabolites and nucleotides can also permeate the channel. A binding site for nucleotides has been identified in the NH2-terminal segment (1423, 1424), but NMR data have not confirmed its existence (549). NAD(P)H sensitizes the channel to voltage, stimulating its closure, and reduces the permeability of the OMM for ADP and ATP (746, 747, 1456). A binding site involving strands 17 and 18 has been identified (549). VDAC may be capable of conducting macromolecules as well, provided they are in a suitably unfolded form: in a completely in vitro system, reconstituted mammalian VDAC was found to allow the translocation of DNA sequences across a planar membrane (1261). That VDAC may be accessible to polymeric oligonucleotides is confirmed by the open-channel block produced by phosphorothioate oligonucleotides, specific VDAC inhibitors (1237, 1283, 1284). Interestingly, VDAC has been proposed to mediate entry of tRNA into mitochondria in vivo as well (1111).
The dynamics of voltage-induced gating, variously envisioned in the past, cannot be deduced from the available static structure, but much evidence points to a role of the NH2-terminal sequence (see discussion in Ref. 1197). A recent paper reports that a double cysteine mutant VDAC in which a disulfide bond stabilizes the interaction between the NH2-terminal helix and β-strand 11 of the barrel wall still exhibits normal gating, suggesting that any conformational changes involved may not be major ones (1292) (this paper has generated a discussion: see Refs. 1292, 1298). Another recent study based on computational modeling (1434) returns to a long-standing proposal (820, 821), suggesting that gating may actually correspond to changes in the structure of the barrel (partial collapse) following removal of the stabilizing NH2-terminal α-helix. Data have been presented supporting a model in which the NH2-terminal can actually leave the pore lumen and participate in interactions with neighboring molecules (e.g., another copy VDAC) and the membrane (437, 489).
Curiously, a clear-cut identification of VDAC activity has not been reported in patch-clamp experiments purportedly recording from the OMM (651). While this unexpected absence is one of the elements suggesting that the seals might not have been established on that membrane, it is also compatible with the view that VDAC might actually be a strictly regulated pore, rather than a permanently open one. In line with this latter hypothesis, purified VDAC has been shown to sometimes adopt completely closed state(s) in bilayer experiments (117). In fact, VDAC, once seen as the inert hole of a sieve, have come to be considered, at the other extreme, a master gatekeeper regulating the flux of metabolites and ions between the mitochondria and the cytoplasm, with a crucial role in inter-organellar communication (cross-talk). The ruth is probably somewhere in between. Despite its abundance, VDAC may indeed pose a limit to the traffic of nutrients and metabolites across the OMM, since its downregulation by shRNA had a dramatic effect on cell growth (3, 682).
3. Regulation of VDAC
VDAC can be modulated in various ways (reviews in Refs. 305, 638). It can be phosphorylated at multiple positions (at least in vitro) by Ser-Thr kinases GSK3β (306, 828, 1180), PKA (108, 1180), PKCε; (94), Nek1 [never-in-mitosis A (Nima)-related kinase] (245, 246) and other kinases (see Refs. 305, 638), as well as by tyrosine kinases (1114). Phosphorylation is thought to regulate interaction with cytoskeletal components and is considered to be in most cases pro-apoptotic (638) except in the case of phosphorylation by Nek1 at Ser193, which has been shown to have the opposite effect (245, 246). At least four lysine residues can be N-acetylated as well (638), and VDAC also undergo O-linked β-N-acetylglucosamination (911) and S-nitrosylation (666, 667).
A list of factors potentially modulating VDAC (directly or indirectly) includes pyridine dinucleotides (549, 746, 747, 1456), glutamate (445, 446), ATP and ADP (1084, 1085, 1424), lipids including cholesterol, phosphatidylethanolamine, and cardiolipin (169, 369, 1081, 1083), possibly Ca2+ (114, contra: Ref. 1282). Proteins like VDAC itself, creatine kinase (4, 36, 76, 171, 195, 396, 768, 839, 906, 907, 909, 970, 1003, 1077, 1144, 1254, 1359), the ANT (24, 171, 1349), c-Raf kinase (739), endothelial nitric oxide synthase (eNOS; Ref. 30), tubulin (229, 490, 493, 1081, 1082, 1086), G-actin (1073, 1409), gelsolin (722), dynein light chain (1159), mitoHSP70 (1159), the TSPO (formerly mitoBzR) (849, 850, 1159), glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (1196, 1290), aldolase (1196), mutant (G37R) SOD1 (576), Bcl-xL (54, 549, 814), Bax (562), possibly t-Bid (1079, but see Ref. 1081), and COX-I (1074) are part of its interactome. These interactions constitute the basis of the involvement of VDAC in cell death and in cancer.
Tubulin has emerged as a modulator of possible physiological significance: its action as inducer of a reversible partial block, which limits ATP permeation, depends on the phosphorylation state of VDAC (1180) and may be relevant for a complete understanding of the action of microtubule-targeting chemotherapeutics (380, 1091). Block is strongly voltage dependent, requiring a polarity negative on the side of tubulin addition in BLM experiments. This poses again the question of whether a significant Δψ exists across the OMM. It has been proposed that interaction with tubulin may result in an increased sensitivity of the channel to Δψ, resulting in significant closure at potentials as low as 10 mV (1081, 1082). Microtubule-associated protein 2 (MAP-2) binds to VDAC (actually the dimer) in affinity chromatography experiments (770). Gelsolin, a Ca2+-dependent acting-regulating protein, has also been reported to be able to close VDAC, thereby exerting an anti-apoptotic effect (722).
4. VDAC and hexokinase
Overexpression of hexokinase-1 and -2 (HK1, HK2) and its association with VDAC are key features of glycolytic cancers (e.g., Ref. 1391). Hexokinase binding (for structural details on binding, see Refs. 4, 1191, 1077, 1199, 1401) to the conduit channeling ATP out of mitochondria provides a metabolic benefit to glycolytic cancer cells (Warburg effect; for reviews focussed on the Warburg effect, see Refs. 227, 839, 840, 841, 842, 976, 986–988; more general, see Refs. 324, 343) and it antagonizes cell death (76, 462, 809, 810, 1199, 1254) via inhibition of Bax-induced cytochrome c release (420, 971, 1360) and/or inhibition of the mitochondrial permeability transition (MPT) (256). Elimination of the NH2-terminal peptide of VDAC1, or introduction of a single mutation (K20S or G21A) antagonized both HK binding and HK-mediated protection against cell death (2, 63, 1199). Phosphorylation (on Thr51) by GSK3β prevents binding of hexokinase to VDAC (969; reviewed in Ref. 970). One of the ways the master pro-life kinase Akt is understood to produce its effects is by migrating to mitochondria, phosphorylating GSK3β (on Ser9), thereby inhibiting it and thus promoting HK-VDAC interaction (177, 969). This pro-life association of VDAC and HK appears furthermore to be dependent on cyclophilin D (801) and its acetylation status which is in turn modulated by Sirt3 (1203, 1348). Akt can also phosphorylate HK2 (on Thr473) and is thought to act as a regulator of metabolic fluxes (e.g., Ref. 1069). HK can be detached from mitochondria by G6P (76), by HK-dissociating peptides modeled on either HK (256, 435, 809, 876, 971, 1202) or VDAC1 (63) sequences and by small molecules such as clotrimazole (809, 971), bifonazole (998) (both fungicides, in the 10−5 M range), and methyl jasmonate (455). A soluble variant of the pilus protein FimA of pathogens such as E. coli K1, Salmonella, and Shigella can instead strengthen HK-VDAC interaction, antagonizing host cell apoptosis (1250). The mechanism by which HK detachment favors cell death is not yet clear and may comprise more than one factor. Proposals include disruption of aerobic glycolysis and of the energy balance of the cell, regulation of ROS production, altered interaction of Bcl-2 family proteins with mitochondria, facilitation of VDAC oligomer formation (reviewed in Refs. 1191, 1193), and MPTP opening (since death is influenced by expression of Cyp D and by cyclosporin A; Ref. 256) (see sect. IIIF). Regardless, the interaction of HK and VDAC may well represent an Achilles' heel of cancer (e.g., Ref. 421).
5. VDAC2 and VDAC3
VDAC2 and VDAC3 (858, 1040) are much less abundant than VDAC1, and less studied. Despite its comparatively low abundance, VDAC2 is the major isoform in bovine testis and spermatozoa, where it is located in part in the cellular membrane of the head and it undergoes tyrosine phosphorylation during capacitation (in human spermatozoa). The porin has been shown to be associated with the outer dense fiber of the sperm flagellum, a nonmembranous compartment (550). The axoneme and mitochondria of VDAC-less mice are altered (1116), and VDACs 2 and 3 are very important for male fertility (728). This suggests a function other than pore formation. VDAC3, purified and reconstituted after expression in yeast (it has never been obtained as a reasonably pure preparation from native mammalian tissue), gave rise to channel activity only with difficulty (1408), and whether this scant activity was really due to VDAC3 is doubtful (858). All three isoforms reportedly permeabilized liposomes with similar solute size exclusion thresholds (1408). VDAC2 forms pores resembling those of VDAC1 (854, 1408). VDAC2 has been reported to associate with ryanodine receptor 2 at SR/mitochondria junctions in the heart, and to be of major importance for the direct transfer of Ca2+ from the SR to mitochondria (865). VDAC1 has in turn been reported to interact with the inositol trisphosphate (IP3) receptor and to channel Ca2+ from the ER to mitochondria selectively, transferring an apoptotic signal (318, 495). The fact that VDAC2 deletion is embryonically lethal (while deletion of VDAC1 or VDAC3 is not, and does not seem to induce alterations of apoptosis) and its interaction with Bak (see sect. IIA6) single out VDAC2 for special consideration as a target for anti-cancer agents acting at the mitochondrial level, also called “mitocans” (1040, 1219).
6. VDAC, apoptosis, and cancer
The role of VDAC in cell death is multifaceted and complex (e.g., Refs. 847, 1193, 1190). Several proteins of the Bcl-2 family have been reported to interact with VDAC1 and/or VDAC2. Tsujimoto's group (reviewed in Ref. 1330) first reported that pro-apoptotic Bax and Bak could stimulate the efflux of cytochrome c from VDAC-containing liposomes via formation of a large pore comprising VDAC and Bax/Bak (1184, 1186; see also Refs. 6, 107, 1182). Pro-apoptotic BH3-only Bid (and t-Bid) and Bik were found not to interact with VDAC1, and did not affect its activity in liposomes (1187), whereas Bim, another BH3-only family member, was reported to activate VDAC directly (1248). The effective existence of a cooperative interaction between Bax and VDAC1 is, however, controversial. Studies in yeast used as a model system led to the conclusion that VDAC was not involved in cytochrome c release (1030), while Madesh and Hajnóczky (805) found evidence that VDAC alone was sufficient to allow cytochrome c efflux through membranes. Rostovtseva et al. (1079) could not detect any electrophysiological indication of an interaction between reconstituted mammalian VDAC and Bax, while t-Bid was observed to induce channel closure. VDAC1 and VDAC3 were found to be dispensable for the formation of the mitochondrial apoptosis-induced channel (MAC), proposed to correspond to the conduit for cytochrome c release (832), which requires instead the presence of Bax and/or Bak.
Tsujimoto's group was also the first to report that Bcl-2 and BclxL inhibited VDAC via their BH4 domain, and ascribed to this inhibition the protection from apoptotic death afforded by BclxL (1185). BclxL binding has been confirmed by structural studies, and localized to strands 17 and 18 of VDAC1 for BclxL (549, 814). The interaction results in a shift from the oligomeric towards a monomeric state of VDAC in micelles. Shoshan-Barmatz's group has defined domains involved in interaction with Bcl-2, which include the NH2-terminal segment, by mutagenesis studies (2, 56). Vander Heiden et al. (1343) found that BclxL favored the open conformation of the pore, but a recent study (54) has confirmed that Bcl-2 and BclxL cause a reduction of pore conductance in vitro. In any case, all authors concur in attributing an anti-apoptotic significance to this interaction. In fact, a peptide derived from the BH4 domain of BclxL was found to have anti-apoptotic action, reducing ischemia/reperfusion (I/R) injury in isolated rat hearts (944, 1247) and apoptosis in isolated human islets (659), while, conversely, expression of peptides copying VDAC1 sequences antagonized the protection offered by Bcl-2 (55) or BclxL (54) against staurosporine-induced apoptosis. The potential biomedical relevance of these studies is evident.
The formation of dimers and higher oligomers of VDAC1 has been reported 30 years ago (769, 819) and confirmed by several studies (2, 347, 438, 555, 702, 1013, 1197, 1335, 1435). These oligomers have been proposed to form the conduit for the efflux of cytochrome c from the mitochondrial intermembrane space of preapoptotic cells (2, 4, 1195, 1191, 1198, 1435). A relationship appears to exist between VDAC1 oligomerization and apoptosis (2, 637, 879, 1198): oligomerization is highly enhanced upon apoptosis induction as recently observed in living mammalian cells using bioluminescence resonance energy transfer (637). Lack of VDAC1 prevented cisplatin activation of Bax (1280) or selenite-induced cytochrome c release (1308). The formation of heterooligomeric complexes involving Bcl-2 family proteins has also been proposed (BclxL: 814) (Figure 2). Overexpression of VDAC1 has been found to be conducive to cell death (804, 1044, 1179). This may be rationalized in terms of an increased formation of pro-apoptotic oligomers, or of an increase in the fraction of channels not bound to HK, equivalent to a partial detachment of HK from its VDAC binding sites. In contrast with the view that BclxL inhibits VDAC thereby antagonizing apoptosis (see above), block of VDAC by the phosphorothioate oligonucleotide G3139 (1285) or by avicins (plant saponins with anticancer activity) (517) has been observed to be pro-apoptotic. VDAC2 has been heavily implicated in the regulation of Bak activity. Korsmeyer's group reported (250) that this isoform (but not VDAC1) inhibits Bak activation and apoptosis, as later confirmed by other groups (237, 318, 737, 1058). In lymphocytes, VDAC2 deletion is fatal, but the cells can be rescued by the simultaneous deletion of Bak (1058). In cells lacking VDAC2, Bak relocates from the OMM to the ER (1040). According to Hajnóczky and colleagues (1091) VDAC2 can however also have a pro-apoptotic role, mediating sensitivity to t-Bid. Tsujimoto's group (1412) has reported that VDAC2 was necessary for death induction by Bax in Bak-deficient cells.
Along with the evidence in favor of the participation of VDACs in the processes of cell death, results have been presented which argue against the whole notion. In their paper reporting the irrelevance of VDAC isoforms for the MPT (see below), Baines et al. (93) also reported that Bax-induced cytochrome c release from mitochondria isolated from WT or VDAC1-, VDAC3-, and VDAC1/VDAC3-null cells was the same. MEFs lacking either one or both of these isoforms showed no difference in their apoptotic response to Bax overexpression, staurosporine, or tumor necrosis factor (TNF)-α. However, cytochrome c release can be induced via different mechanisms (even in the absence of Bax and Bak), depending on the cell type, its apoptotic protein expression profile, and the stimuli used. Lack of VDAC2 is fatal (250), suggesting that this particular isoform is the one with the most important role in apoptosis, while VDAC3 seems to have no role in cell death (318). The involvement of VDAC (93, 669) and the ANT (669) in the MPTP has similarly been nearly ruled out on the basis of genetic deletion experiments. We may remark however that compensating changes tending to reestablish physiological homeostasis are notoriously difficult to assess.
One possibly confounding factor in evaluating the role of VDAC1 in cell life or death is its localization in cell compartments other than the OMM. VDAC (called porin 31HL) was discovered in the plasma membrane of human B lymphocytes (636, 1296, 1431). This surprising observation was confirmed by a series of careful studies employing a variety of techniques, including the development of monoclonal antibodies and electrophysiological recordings (81, 116, 152, 231, 270, 327, 582, 623, 674, 1033, 1160, 1235; reviews in Refs. 316, 1068, 1297), also in plants (1068).
VDAC1 was found to localize to subdomains: caveolae (90, 116, 544, 826, 1042) and postsynaptic density fraction (884). Expression in the PM appears to depend on the presence of a PM-targeting sequence introduced by alternative splicing in the NH2-terminal region, which directs it to the secretory pathway through the Golgi apparatus (89, 214). Alternative splicing of VDAC gene products is known, but its significance is, in general, obscure (858). PM VDAC1 may function in volume regulation and ATP release (938), and it has been reported to act as a redox enzyme (100, 736), although this activity is rather difficult to rationalize and could not be demonstrated for VDAC isolated from mitochondria (1197). It has been found to act as the receptor for plasminogen Kringle 5 (457, 458) and tissue-type plasminogen activator (459). Interestingly, the latter appears to be a substrate for VDAC's NADH-dependent reductase activity (459). VDAC1 has also been observed in the endo/sarcoplasmic reticulum (624, 758, 836, 1178, 1194, 1195, 1196, 1200).
B. Inward Rectifying Potassium Channel
An inwardly rectifying voltage-dependent potassium selective ion channel (Kir) has been observed in the outer mitochondrial membrane by patch-clamp experiments (398). The activity was regulated by cAMP and by osmolarity and blocked by cesium cations. However, inhibition by other classical potassium channel inhibitors, TEA+ and 4-aminopyridine, was not achieved. The lack of molecular identity and of a more detailed biophysical/pharmacological characterization of this channel prevents up to now the elucidation of its physiological role, if any. Furthermore, further studies will need to exclude that this activity is related to a cationic subconductance state of VDAC or is the result of contamination by ER via the so-called mitochondria-associated endoplasmatic reticulum membranes (MAMs) (532). It is of note in this respect that Kaasik and colleagues (726) have reported the functional expression in the ER of the small-conductance calcium-activated potassium channel (SKCa), required for calcium uptake into ER.
C. Protein Import Pores TOM40 and Sam50
The transport of proteins and genetic material across membranes is now known to involve “threading” of the polymeric chain through an at least partly proteinaceous pore in a number of systems. Examples include bacterial Sec (297, 798); type II (686), type III (217, 688), and type IV (1439) machineries; and the intracellular systems of the ER (1455), peroxisomes (852), and chloroplasts (34, 695, 1243). The complex major mitochondrial systems have been extensively studied and are now known in considerable biochemical detail (e.g., Refs. 133, 134, 234, 363, 373, 519, 520, 880, 881, 910, 1147). Whether the pores formed by this apparatus qualify as “channels” might be questioned, since their physiological function is not passive ion transport. We briefly cover the topic in this review because a considerable amount of excellent electrophysiological work has been carried out to characterize the protein-conducting pores, and because they might account for some of the “orphan” activities that are not ascribable at the moment to a given protein.
The mitochondrial protein import system relies on five pore-forming complexes. The electrophysiological description of mitochondrial import systems (520, 521) developed briskly in the 1990s. The earliest relevant work was done by the group of M. Thieffry and J.-P. Henry (540), who in 1988 reported a high-conductance activity of mitochondrial origin studied by the “tip-dip” technique (1294) and later also in planar bilayer experiments and by patch-clamping proteoliposomes (1295). Trypsin digestion, membrane fractionation, and selective digitonin solubilization were used to establish that the channel belonged to the outer membrane (259), but was not related to VDAC1, since it was present also in porin-less yeast mitochondria (394), and had different properties (717). The channel was observed to undergo fast open-channel block by peptides copying or resembling the addressing sequence of mitochondria-targeted proteins (395, 539, 540, 1295). This led to the nickname “peptide-sensitive channel,” or PSC. The behavior strongly suggested that the peptides were translocating through the channel, i.e., that this was a component of the protein import system (618, 1340). This was later confirmed by experiments with a chimeric fusion protein comprising parts of cytochrome b2 and dihydrofolate reductase, which blocked the channel when added on the cytoplasmic (NH2-terminal) side (548). The channel seemingly corresponds to TOM40, involved in translocation across and insertion into (518) the outer membrane. Isolated TOM40 and TOM40 complex reconstituted into planar bilayer membranes after solubilization or by fusion of proteoliposomes from membrane preparations exhibited channel activity very similar to that of the PSC (132, 521, 548, 718). Anti-TOM40 antibodies coimmunoprecipitated PSC activity. TOM40 is a rather well-behaved channel, as perhaps expected of a protein with a significant amount of β structure, often appearing in doublets or triplets exhibiting a degree of cooperativity (e.g., Ref. 132). This is coherent with the view that TOM40 exists as an oligomeric complex, and that two to three copies are likely to be present even in the isolated complex (14, 15, 132, 717, 1043). We note that an analogous activity has not been reported to have been observed in patch-clamp experiments considered to involve seals on the outer mitochondrial membrane of intact mitochondria. Single-channel conductance varies to some extent depending on the species of origin, and also exhibits intrinsic variability. Thus, for example, the yeast complex in tip-dip and patch-clamp (on proteoliposomes) experiments characteristically showed conductance steps of 320–330 pS (619, 1295) in 150 mM NaCl (550 pS in 150 mM KCl in planar bilayer experiments with recombinant TOM40; see Ref. 14), similarly to the Neurospora crassa pore (∼300 pS; Ref. 717), while the most prominent current variations observed with the mammalian counterpart corresponded to a conductance of ∼200–220 pS (1295). The single-channel current-voltage relationship is ohmic or slightly rectifying. In planar bilayer experiments with S. cerevisiae and N. crassa channels application of potentials higher than ∼70 mV of either polarity decreased the open probability (548, 718). This latter behavior resembles that described for VDAC and also for the mammalian MMC (e.g., Refs. 1267, 1273), although in both cases lower voltages are involved. The channel exhibits a spontaneous flickering that is eliminated by trypsinization of the cytoplasmic side (259), and possibly is due to the elimination of a 2.5-kDa NH2-terminal fragment (548). Fast “closures” are furthermore produced, both with the native and the trypsin-treated channel, by cationic peptides applied to the extramitochondrial (trypsin-sensitive) side, and this may be considered one of the defining features of the channel. Selectivity also varies somewhat depending on the species, but is always cationic [e.g., PK/PCl ∼8 for the S. cerevisiae channel (545); ∼3 for the N. crassa channel (717)]. The permeabilities of the various cations follow the order of their mobility in water, suggesting that the pore may indeed provide a wide permeation pathway. The diameter of the roughly cylindrical structure has been estimated at 2.0–2.2 nm by the polymer-exclusion method (548) and by EM imaging (14, 15, 717). Whether this diameter may be sufficient to admit a polypeptide is not certain, and there has been some speculation that the polypeptide permeation pathway may actually be formed by the assembly of a few monomers plus cofactors (14). The characteristics of mammalian TOM40, which have been much less studied, appear to be similar to those of the yeast counterpart (807, 1255).
The “sorting and assembly machinery” (SAM) and the TOB/SAM complex (363, 1364), centered on Tob55, are devoted to the incorporation of β-barrels (which transit first in the intermembrane space) into the OMM and participate in determining mitochondrial morphology (186, 685, 1440). While different supramolecular complexes involving also TOM components can assemble and operate (1300), the core SAM complex consists of a membrane-embedded component, SAM50, homologous to bacterial Omp85, and two smaller subunits, SAM35 and SAM37, the latter one not essential for cell viability (135, 374). SAM50 have conserved roles in VDAC biogenesis from yeast to mammalian systems (701). The SAM pore from yeast has been characterized after reconstitution in the planar bilayer, both as the recombinant purified SAM50 protein and as the purified complex (725). It is slightly cation-selective (PK/PCl ∼4). With the reconstituted complex the channel is observed as a flickering activity with a major conductance of 160 pS in symmetrical 250 mM KCl. Addition of the β-signal peptide (identified in the same piece of work) in this case reduced the gating frequency and induced a shift of the most prominent gating transitions to 320 pS, with occasional 640 pS (i.e., 4X) events. The effect of the targeting peptide was found to depend on the presence of SAM35, which appears to “present” the incoming protein to the pore-forming subunit inducing it to assume an active, dynamic state. The pore may actually be formed by an oligomeric assembly of SAM50 (itself a β-barrel) monomers, a model which can better account than a monomer-based one for the postulated sidewise release of the transiting protein into the membrane and for the flexibility required to accomodate partially folded substrates.
D. Pores Formed by Bcl-2 Family Members
1. Electrophysiology of Bcl-2 family proteins
A key step in apoptosis is the release of cytochrome c and other pro-apoptotic proteins from the intermembrane space of mitochondria. Various mechanistic models have been proposed for this process. One that enjoys considerable popularity envisions the formation of macropores in the OMM by oligomers or aggregates of the Bcl-2 family proteins Bax and/or Bak. Mitochondrial pro-apoptotic factors (cytochrome c, AIF, Smac/Diablo, some procaspases) would permeate the OMM via these pores. The origins of this model go back to the structural resemblance between the anti-apoptotic Bcl-2 family member BclxL and the membrane translocation (and pore-forming) domain of diphteria toxin and bacterial colicins (892, 1140). An analogous similarity was then noticed for Bax (1259) and various other family members (1010). The three-dimensional structures comprise two central hydrophobic alpha-helices surrounded by six or seven amphipathic alpha-helices. Accordingly, at least BclxL (110, 866), Bcl-2 (1141, 1145), Bax (43, 321 687, 767, 1145, 980, 1139, 1187, 1269), t-Bid (1139), and Bad (1024, 1025) have been observed to form channels in experimental membrane systems (planar bilayer or vesicles). In general, these are not well-behaved channels, and their properties are variable. Even within the same experiment, they can display a range of conductances. Bax channels can be voltage-dependent or not (767; Tombola, Zoratti, and Szabò, unpublished observations) and can be weakly or moderately selective either for cations or anions, at times switching selectivity during the recording (767; Tombola, Zoratti, and Szabò, unpublished observations). Schlesinger et al. (1145) found Bax channels to be slightly anion-selective, while Antonsson et al. (43), Pavlov et al. (980), and Dejean et al. (321) reported a weak preference for cations. Bcl-2 and BclxL appear to produce unselective or cation-selective pores, depending on pH. BclxL, Bcl-2, Bax, and Bid have been reported to be pH-sensitive, with acidic pH values favoring activity, although channels can often be observed also near neutrality (43, 866, 1139). The composition of the artificial membrane can also influence the channels' behavior (e.g., Ref. 1139).
2. Channel formation and apoptosis
The physiological relevance of the results obtained with purified proteins in reconstituted systems is uncertain. In many studies the proteins used were actually constructs with partial deletions. For example, Basañez et al. (110) reported that anti-apoptotic full-length BclxL did produce pores upon reconstitution into lipid vesicles, but their size was insufficient to allow permeation by cytochrome c. Pores with a larger diameter, possibly involving lipidic components and capable of allowing cytochrome c release, were formed by the product of caspase-3 cleavage, or by the pro-apoptotic cleaved forms ΔN61-BclxL and ΔN76-Bclxl. Considering the variability displayed in each case, the channel activities of pro- and anti-apoptotic proteins in vitro are rather similar. Monomeric Bax or Bak are generally considered not to form pores, which require instead oligomerization induced by treatment of isolated recombinant Bax with detergents (but see Ref. 577) or the interaction with caspase 8-activated Bid (t-Bid) (44, 45, 329, 687, 860, 861, 1376), Bim (e.g., Ref. 432) or other BH3 domains (358). Saito et al. (1106) have concluded that while a Bax dimer can form a pore, a tetramer is needed for cytochrome c release (687). Prokaryotes provide many examples of aggregation of monomers to form a large pore, e.g., pneumolysin (444, 1303).
It may be surmised from all this that oligomerization most likely underlies pore formation by Bcl-2 and BclxL as well. The possible functions of pores formed by anti-apoptotic proteins have not been identified, insinuating the doubt that an ion channel-forming capacity of pro- and anti-apoptotic Bcl-2 family proteins may actually not be relevant for cytochrome c release (39, 738). In fact, we have found that a single-point mutated Bax form (BaxK128E) which has lost its pro-apoptotic character when expressed in cells is still capable of forming pores upon reconstitution in planar bilayers (1269). Overexpression of Bcl-2 did not lead to the detection of activity attributable to this protein in a patch-clamp investigation of isolated mitochondria (898). On the other hand, the direct addition of Bax, Bak, or Bid to isolated mitochondria has been reported to result in OMM permeabilization, while the mitochondria maintain transmembrane potential, oxidative function, and the ability to import proteins (328, 1187, 1357, 1376). Toroid-shaped structures formed by Bax in artificial membranes have been visualized directly (376). Membrane-inserting alpha helices 5 and 6 of Bax are necessary for the protein to have its pro-apoptotic function (535, 844), and peptides corresponding to this segment are sufficient to permeabilize membranes (423, 424), although whether they give rise to channel activity was not investigated.
In experiments by Martinou's group (1089), the pores formed by full-length Bax in the presence of t-Bid were deemed not large enough to allow the passage of cytochrome c, and interactions with other components of the OMM were proposed to be required. Among the proteins proposed to control OMM permeabilization by Bax and/or Bak are VDAC-2 and VDAC-1 (2, 93, 250, 562, 737, 1058, 1092, 1136, 1190, 1412). However, Baines et al. (93) have shown that all three mammalian VDAC isoforms are dispensable for cell death.
a) mac mitochondrial apoptotic channel. The groups of K. W. Kinnally and of E. A. Jonas have reported electrophysiological studies showing the formation of large pores presumably in the outer membrane of mitochondria of apoptotic cells (reviews in Refs. 322, 323, 611, 651, 830, 1103). Kinnally's group had previously studied a high-conductance, multilevel mitochondrial pore, but their first publication describing what was considered to be an OMM pore involved in cytochrome c release (dubbed “Mitochondrial Apoptotic Channel”, MAC) appeared in 2001 (980). The study utilized mitochondria isolated from hematopoietic FL5.12 cells. With these cells apoptosis can be induced by IL-3 depletion and involves Bax migration to mitochondria. Recordings from “intact” (unswollen) mitochondria, assumed to be from the OMM, did not reveal single-channel transitions, but the average patch conductance was reportedly higher in the case of organelles isolated from apoptotic cells (∼13 versus ∼6 nS). Single transitions of various sizes up to ∼2.5 nS could instead be observed in proteoliposomes formed with OMM preparations from apoptotic cells (which contained more Bax than nonapoptotic controls). The authors reported that the frequency of detection of MAC (in proteoliposomes) was about fourfold higher when the preparation came from cells deprived of IL-3. It could not be detected if the cells overexpressed Bcl-2 (interestingly, no novel channels attributable to Bcl-2 could be detected either), but it was found with normal frequency if the Bcl-2 expressed was a mutant, functionless Bcl-2 variant. MAC was characterized in this system as being voltage-independent, slightly cation-selective, and easily distinguishable from activity attributed to TOM40 or VDAC. Important similarities between the activity produced by reconstituted Bax and MAC were noted, the difference consisting in practice in the size of the most common events. In fact, an analogous activity could be observed in proteoliposomes obtained from a VDAC-less yeast strain expressing human Bax, strongly suggesting that Bax, but not VDAC, was present in MAC. Subsequent studies in the same proteoliposome system revised upward the upper limit of MAC conductance, produced evidence for an effect of cytochrome c and other bulky molecules compatible with permeation through the channel (488), and found an inhibitory effect of the amphiphilic cationic peptides propranolol, trifluoperazine, and dibucaine, while cyclosporin A had no effect (831). MAC was also characterized in some more detail by F. Vallette's group, who used a rat model of liver apoptosis and patch-clamping of proteoliposomes containing membrane fractions isolated from mitochondria having different properties reflecting the progress of the apoptotic process (484). MAC was found to be clearly distinguishable from VDAC and the MMC/MPTP (both also observed) on the basis of its biophysical properties and pharmacology. Its conductance was often close to 3 nS (150 mM KCl) with occasional higher values. Practically nonselective, it was found to close in response to low-to-moderate voltages of either polarity. Cytochrome c reduced its conductance to a low level in ∼50% of trials. In these experiments, MAC was found to appear late in the apoptotic process, in a mitochondrial fraction (LAM2) which had already lost 90% of its cytochrome c and contained Bax predominantly as an ∼80 kDa oligomeric complex (while Bax was present as lower oligomers or monomers at earlier stages of apoptosis). Kinnally's group also presented further immunological and pharmacological data confirming that Bax (and/or Bak) was at least a component of MAC (321) and in favor of its role in cytochrome c release from the intermembrane space and hence in apoptosis (992). The model arrived at envisions t-Bid-induced progressive MAC formation from Bax/Bak. While the involvement of other proteins is not ruled out, MAC forms independently of VDAC1 or VDAC3 (832). No conclusion has been reached concerning VDAC2 (see above for references) or the adenine nucleotide translocase(s), an IMM protein which has also been proposed as a partner of Bax (147, 197, 1446). Overall, these findings are in excellent agreement with the model, based on biochemical studies by several laboratories, envisioning the formation by Bax of wide cytochrome c-permeable openings in the OMM.
b) in-cell recordings of bcl-2 family pores. Jonas and co-workers were the first, and so far to our knowledge the only, researchers to achieve gigaohm seals and single-channel recording from intracellular membranes reached by perforating the outer membrane with a two-concentric-electrodes device (610) and used it to, record activity of both low- and high-conductance channels from squid presynapse intracellular membranes reputed to belong to mitochondria (605, 609). High-conductance activity was reportedly potentiated by electrical stimulation of the synapse; it was dependent on elevation of Ca2+ levels and was eliminated by pretreatment with FCCP, a ΔΨm dissipator (605, 612). Unitary conductance steps observed ranged up to 2.5 nS (605). Large channel activity was also potentiated by the presence of BclxL in the patch pipette (607, 609). In this case conductances mostly fell in the 100- to 760-pS interval (609). In the same experimental system, large multi-conductance (0.3–2.0 nS) pore activity was also readily observed by intracellular (mitochondrial) patch-clamp recording under hypoxia (608, 612). This development was inhibited by the cysteine protease inhibitor Z-VAD-fmk. Similar activity (0.3–3.8 nS) was elicited by a pro-apoptotic form of BclxL produced by deletion of NH2-terminal residues 2–76 (607, 612), and it was observed in mitochondria isolated from ischemic rat brain (187). Ischemic preconditioning was found to block index ischemia-induced BclxL cleavage (yielding ΔN61-BclxL) and large channel activity. The effect was mediated by Akt and Bad, and resulted in block of casp-3 activitation and prevention of apoptosis (877). These observations led to the conclusion that NH2-terminal-truncated form(s) of BclxL may, together with other proteins, generate large pores in the OMM of ischemic cells, accounting for the release of pro-apoptotic factors (187, 934). Note that this model envisions the involvement of the cleaved BclxL forms. Assuming cleavage by casp-3, which is known to occur (e.g., Ref. 415), runs into an apparent catch-22 difficulty: cleavage would be needed for the formation of a cytochrome c efflux pathway, necessary for caspase activation. Nonetheless, mice expressing a knocked-in caspase-resistant form of BclxL exhibited a reduction in the relevant mitochondrial activity and in the susceptibility to ischemic neuronal death (934). Thus the mechanism may serve in the acceleration/amplification relatively late phase of the apoptotic process (934) (delayed cell death; compare with the late appearence of MAC observed by Guihard, Ref. 484), or some other protease(s), possibly a zinc metalloprotease or a calpain, might intervene to generate pro-apoptotic forms as well.
Remarkably, in similar experiments application of Bax resulted in the observation of relatively small channels, similar to those observed with full-length BclxL, and only occasionally of giant conductances comparable to those elicited by ΔN-BclxL (606, 611). Expression of ΔN-BclxL was sufficient to induce death of neurons and of Bax- and Bak-less MEFs (934). Indeed, the authors point out that the process they investigated may not coincide with apoptosis as narrowly defined (934), making the point that in neurons, the cells they studied, Bax is downregulated, while BclxL is abundant (611).
Patch conductance in seals on isolated mitochondria was observed to be increased by Zn2+ (100 μM in bath). Of note, Zn2+ is an activator of the MMC/MPTP (593, 775, 1397; contra Ref. 330). Large-conductance channel activity was inhibited by the permeant Zn2+ chelator TPEN, but not by impermeant EDTA (187). This points to a process taking place in the matrix. The authors recognized this, but still assigned the activity to the OMM, hypothesizing that matrix Zn2+ might indeed activate the MPTP, resulting in the release of a messenger which in turn would activate the conductances they recorded in the OMM (187, 611). Mitochondria from ischemized brain had a higher content of chelatable Zn2+ and of ΔN-BclxL. Zn2+ is known to be involved in ischemic neuronal death (e.g., Ref. 219).
The same group also discovered an interaction of BclxL with the β subunit of the mitochondrial F0F1 ATPase, an interaction which appears to reduce proton leakage across the IMM (19, 249). This obviously demands that BclxL is present in the mitochondrial matrix, as indeed found by the authors. If BclxL is present at the level of both the OMM and matrix, it is reasonable to assume that it is also there at the IMM. These considerations, plus the characteristics of the pores observed (high-conductance, multi-level) the role of matrix Zn (see above), and the common involvement of the conductances under discussion and of the MMC/MPTP (e.g., Ref. 95) in ischemic death raise the question of the relationship between these two entities. May they coincide? Since the MPT involves the inner membrane, an obviously important point here is the identity of the membrane the seals were established on in Jonas' experiments. The authors adopted early on the view that it was the OMM, but the direct evidence provided is limited. The first argument, the effects of BclxL, a protein thought to be located exclusively at the OMM, was invalidated by the later report, by the same group, showing that BclxL was present also in other mitochondrial districts. A second criterion used was that the probability of observing ΔN-BclxL-elicited large channels was reduced when the intracellular patch pipette contained NADH (607). Millimolar NADH has been reported to reduce the permeability of the OMM to ADP (747) and the conductance of VDAC (1456). Jonas et al. (607) therefore surmised that VDAC may participate in the formation of the megachannels they observed, and performed a few experiments on mitochondria isolated from wild-type as well as from VDAC1-less yeast (607). With wild-type yeast mitochondria, an activity with conductances in the 200–400 pS range was attributed to VDAC (implying that the seal had been established on the OMM). Whether this activity was influenced by NADH was not reported. The presence of ΔN-BclxL (lacking amino acids 2–76) in the pipette resulted in a different pattern of activity, with larger (500–1,000 pS) and NADH-attenuated conductances. This behavior is in keeping with the already mentioned conclusion that N-cleaved pro-apoptotic forms of BclxL form large channels. VDAC1-less yeast mitochondria exhibited larger conductances (700–800 pS) that were unaffected by either NADH or ΔN-BclxL. We may note that NADH (used in the millimolar range) is a reducing agent which might, in general, counteract oxidative stress-induced phenomena. Another relevant point may be that in the 1997 methodological paper (610) the fluorescent dye present in the intracellular pipette appears to label entire, discrete subcellular compartments, suggesting a whole-organelle rather than an organelle-attached configuration. It thus seems to us that the identity of the patched membrane is a matter deserving further attention. In particular, nobody has tried yet to simulate the electrical properties of the electrical circuit representing a seal on the OMM in the presence of an intact IMM separated by only several nanometers of a conductive medium and joined to the OMM at contact sites. Be this as it may, the observations summarized above led Jonas (611) to propose that ischemic conditions lead to the appearence in the OMM of a pore composed by ΔN-BclxL, dephosphorylated Bad and VDAC.
III. ION CHANNELS OF THE INNER MITOCHONDRIAL MEMBRANE
A. Potassium-Selective Channels
Figure 3 gives an overview of all electrophysiological activities in the IMM.
1. ATP-dependent potassium channels
a) the plasma membrane katp channel. The mitochondrial KATP (mito KATP) channels, namely its very existence, its composition, properties, regulation, function(s) and physiological and medical relevance, constitute the most controversial topic in the field of mitochondrial channels. Part of the uncertainties fueling the debate derive from the complexities characterizing the much-studied plasma membrane (PM) counterpart, which it is therefore worthwhile to briefly recapitulate. Via binding of nucleotides at different sites, and other forms of regulation, PM KATP channels act as an important sensor of the cellular metabolic/energetic state, closing when energy (ATP) is abundant, and opening when the cell is “starved.” Given the nature of this review, we do not discuss here the varied functions of these channels. For more detailed information on all aspects of PM KATP lore, the reader is referred to the several reviews on the topic (11, 16, 66, 67, 212, 213, 403, 750, 889, 914, 1031, 1181, 1304, 1444).
I) Composition. The proteins giving rise to the ATP-sensitive channel activity observed in patch-clamp experiments (see below) were identified in the mid-1990s (12, 31, 32, 569, 570, 1109). KATP channels are heterooligomers composed by a tetramer of ∼40 kDa Kir potassium channel-forming monomers associated with four regulatory sulfonylurea receptor (SUR) ∼160 kDa subunits in an octameric complex (1204; reviews in Refs. 11, 84, 1173).
“Kir” stands for “K+-selective inward rectifier,” a family which differs from the general template of K+ channels in that each monomeric polypeptide comprises only two membrane-spanning α-helices (1, 46, 545, 913, 1098, 1366). The namesake inward rectification is due to voltage-dependent block by cytoplasmic Mg2+ or polyamines. These channels are classified into seven subfamilies: ATP-sensitive channels are formed by the members of the Kir6 group, which comprises Kir6.1 and Kir6.2. These “core” channels have an inhibitory ATP binding site (403), which can be modulated by phospholipids (1331, 1402).
Sulfonylurea receptors (SURs) (16, 211, 403, 750, 889, 1181) are members of the family of ATP binding cassette (ABC) transporters (552). ABC transporters have also been classified into seven subfamilies. The ones involved in the formation of KATP channels belong to the ABCC/MRP subfamily, which includes also, e.g., the cystic fibrosis transmembrane conductance regulator (CFTR). As far as is known, SURs do not act as “pumps,” and their function seems to be essentially that of regulating KATP channel activity.
Each SUR monomer has three transmembrane domains (TMD0–2) and two cytoplasmic nucleotide binding folds (NBFs) that come together to generate two interacting sites that can bind MgATP/MgADP. Engagement of these sites promotes activity of the channel. The major SUR forms, produced from two genes with nearly 40 exons each (1181), are SUR1 (high affinity for sulfonylurea) and a couple of splicing isoforms, SUR2A and SUR2B (lower affinity). These are the proteins generally considered to partner with Kir6.x in the formation of PM KATP channels. However, several other “short” splice variants of SUR2, some with a compromised or missing COOH-terminal NBF, have been reported in the heart (403, 1181) and can form “unconventional” KATP channels with Kirs (1032). Several alternative splicing forms of SUR1, lacking transmembrane helices or an NBF domain, or even most of the protein (470, 510, 1110, 1146), have also been detected. While some are unproductive (470, 510), others can form channels by pairing up with Kirs (510, 1110, 1146, 1181). A further recent study provides evidence that these short forms are present in mitochondria and, therefore, may actually be a component of “mito KATP” (1421). Not much is currently known about the overall functional significance of these forms, but they may well contribute to the complexities of KATP regulation and pharmacology.
Even considering only the major forms of SURs, it is obvious that in principle a variety of complexes can assemble. At least in most cases, mercifully, octamers appear to comprise only one type of Kir and only one type of SUR (443) (but see below). This restriction leads to the expectation that six major types of Kir/SUR complexes may form, which have all been reported (see Table 2). For both Kir6.x and SURs, the expression of one or the other form is organ- and cell-type specific, but this segregation is not absolute, and the formation of heterocomplexes has been demonstrated in several studies (review in Ref. 1444). Different forms may coexist in the same cell (271). Molecular biology studies have provided convincing evidence that heteromers comprising combinations of different Kir6.x (e.g., Refs. 82, 85) or SURs (e.g., Refs. 251, 777, 1378) can form functional channels.
II) Biophysical properties of plasma membrane KATP channels. The biophysical properties of the native PM KATP channels need to be kept in mind when considering the questions associated with mito KATP. Given our purposes here, we will focus on the conductance, which obviously depends on ionic conditions.
Kir6.2-based channels are more commonly expressed, and the cardiac type (Kir6.2-SUR2A) is often referred to as representative of KATP channels. They are most readily observed in the absence or at lowered levels of ATP, e.g., in excised membrane patches or in the membrane of cells treated with mitochondrial poisons (e.g., KCN). However, Kir6.1-based (KNDP) channels need Mg2+ and nucleotides for activity. KATP channels are strongly K+-selective channels (current reversal potential close to theoretical). For the cardiac type (less so for the Kir6.1-based channels), the single-channel current-voltage relationship shows inward rectification, pronounced in cell-attached experiments or when activity is recorded in the presence of low-millimolar range Mg2+ on the cytoplasmic side. This is an example of “unidirectional” fast channel block, which results in the reduction of the average amplitude of the current conducted by the open channel. Physiological blockers, acting from the cytoplasmic side, have little effect on the conductance in the voltage range negative to the K+ reversal potential. The slope of the current-voltage relationship in this latter range is referred to here as the conductance of the channel. Discrete subconductance levels can be observed (86, 628), and cooperativity upon clusterization has been proposed (86).
The expression of components in suitable model cells has allowed researchers to precisely define a set of intrinsic properties of the various KATP channels. Conductance values of defined-composition (cloned) channels are as summarized in Table 3.
In these heterologous expression systems, the values depend essentially on the Kir isoform expressed: in symmetrical 140 mM KCl Kir6.1 complexes exhibit a conductance of ∼33–35 pS, versus 67–80 pS for Kir6.2. The conductance of Kir6.1/6.2 concatemers depends not only on the proportions of the two monomers within the tetramer, but also on their relative position (86). In native cells, the picture is less well defined. In some cases also smaller conductances than those reported in Table 3 can be observed (1037). Considering the values recorded in at least approximately symmetrical high K+, in ventricular cardiomyocytes and skeletal muscle myocytes the channels do exhibit the conductance expected of Kir6.2-SUR2A complexes (with values reaching however up to ∼100 pS). In pancreatic cells or insulin-producing cultured cells, thought to express Kir6.2-SUR1 channels, conductances seem to cluster around 50–65 pS, and even lower values are observed in some cases in recordings from neuronal membranes (e.g., Ref. 993). Smooth muscle cells, believed to express predominantly SUR2B, present a range of conductance values, including conductances close to the values expected for Kir6.1-based channels as well as for Kir6.2-based ones (see Table 3 and Ref. 1037). This is in agreement with expression studies. Some of the lowest conductances reported in the literature were obtained with cells of this type (e.g., ∼13 pS in symmetrical 140 mM KCl in rat mesenteric artery vascular smooth muscle cells; Ref. 1286). Overall, conductances ranging roughly from 10 to 100 pS have been attributed to PM KATP channels.
In several cases, mitochondrial KATP channels have been studied after reconstitution in planar lipid bilayers. The technique has also been used in a few studies of the plasma membrane channel. Parent and Coronado (967) recorded it upon fusion of vesicles of rabbit skeletal muscle t-tubule membrane. In 260/60 mM KCl (cis/trans; vesicles added in cis) they observed a conductance of 67 ± 2 pS in the absence of Mg2+. The channels apparently tended to associate in pairs. Kovacs and Nelson (694) incorporated ATP-sensitive channels from canine aorta smooth muscle. The paper does not report a conductance value, which can be estimated from the linear portion (at cis-positive potentials) of the rectifying plot at ∼80 pS with 120/5 mM (cis/trans) KCl. With similar ionic conditions, Mayorga-Wark et al. (845) measured a conductance of 180 ± 6 pS (no Mg2+) for ATP-modulated K+ channels from basolateral membranes of mudpuppy (a salamander) enterocytes. Oosawa (945) used a preparation of membranes from insulin-secreting HIT T15 cells. In this study an ohmic conductance of ∼53 pS was measured in symmetrical 140 mM KCl.
In single-channel recordings, the KATP channels exhibit typical “bursting” behavior, with at least two time constants needed to describe the distribution of both open and closed times (e.g., Refs. 628, 1231, 1232, 1326).
III) Regulation of PM KATP channels. The defining features of the KATP channels are generally considered to be their regulatory and pharmacological properties, preminently regulation by ATP (and also by GTP, UTP; e.g., Refs. 121, 400, 628, 914, 1031, 1332). The “core” Kir channels have an inhibitory ATP binding site (e.g., Ref. 82), which in the case of Kir6.2 accounts for inhibition in the 10–100 μM range. In experiments with excised membrane patches, ATP concentrations in the double- or triple-digit micromolar range are often enough to cause a drastic decrease of the channel open probability (Po) (while there is no effect on its conductance; e.g., Refs. 628, 918). Free ADP is also inhibitory, although with a ∼10-fold lower Ki than ATP (627, 1231). From the point of view of channel behavior, the effect of ATP is to reduce the burst duration and the number of openings per burst and to increase the frequency and duration of interburst closures (e.g., Refs. 628, 1231), as well as to stabilize a long-lived closed state (1231, 1232). High Hill coefficients for inhibition point to cooperativity by multiple receptor sites (628). In fact, all four Kirs need to be in the ATP-free permissive state for the channel to be open. Inhibition does not involve phosphorylation, since it is brought about also by nonhydrolyzable analogs (401, 628, 1231). The effect is very rapid and reversible so that KATP may be regarded as a ligand-gated channel (121). Importantly, the Kir subunit also binds to phosphatidylinositol 4,5-bisphosphate (PIP2) (and other lipids, such as PIP1 and, importantly, long-chain acyl CoAs; e.g., Ref. 734). This interaction is activating, i.e., in the absence of ATP PIP2-Kir interactions tend to keep the channel open (e.g., Ref. 602). These features account for the increased activity of Kir6.2-based channels when membrane patches are excised and ATP is washed away, and also for the subsequent channel rundown as PIP2 is lost (403). This interaction obviously implies regulation of KATP activity by the cellular signaling pathways which use phosphatidylinositolphosphates as messengers.
While binding of ATP to the Kir subunit inhibits, binding of Mg-adenine nucleotides to the SUR subunits activates the channel, contrasting the inhibitory effect on Kir (e.g., Ref. 627). MgADP can also reverse channel rundown, at least partially. Affinities differ among the various subunits. In the case of the Kir6.1-SUR2B subtype, the dependence of the open probability on ATP is bell-shaped. The channels strictly require nucleoside di- or triphosphate binding to SUR to open, and will become inhibited at higher concentrations via the Kir6.1 site. Contrary to the “cardiac type,” these channels (the “KNDP-type”) do not therefore appear upon membrane patch excision, and need the presence of, typically, MgADP to be observed. ATP/ADP levels probably contribute, along with PIPs, to setting the levels of KATP activity, even though ATP is not expected to fall below millimolar levels in living cells. There is, however, no scarcity of other regulatory biochemistry and interactions.
The modulation of PM KATP by pH is complex, with early studies leading to seemingly contradictory conclusions. In studies involving expression of recombinant rodent Kir6.x and hamster SUR1 in Xenopus oocytes, Xu and colleagues established that high levels of CO2 (hypercapnia) and, equivalently, intracellular acidification to pH 6.8–6.6 activates the channels via a conserved histidine of the Kir subunit (1369, 1395, 1396, 1404, 1405). Extracellular acidification had only a modest effect. The Kir6.2-SUR2A channel of ventricular heart cells was found to exhibit a bell-shaped pH dependence, with acidification increasing its Po in the 7.6–6.0 range and decreasing it below pH 6. Recombinant Kir6.2-SUR1 expressed in HEK cells has furthermore been reported to be directly activated (Po increase) by alkaline intracellular pH, with a peak at pH 8.0 (822).
A further layer of complexity is provided by phosphorylations. The channels are in fact targets for a number kinases, including PKC, PKA, and metabolism-sensing AMPK. The resulting effects depend on channel composition and on the site of phosphorylation (403; see also Ref. 75). Thus phosphorylation by PKC can be inhibitory in smooth muscle and insulin-secreting cells and activating in cardiomyocytes. PKA activates KATP in vascular smooth muscle cells. At least both Kirs, hSUR1 and SUR2B, can be phosphorylated by this kinase (403). AMPK activates cardiac-type channels (1249) but inhibits those of insulin-secreting cells (238). Activation of KATP may also take place downstream of PKG (and thus downstream of NO) (e.g., Refs. 635, 1104): it involves phosphorylation (not necessarily of the KATP itself) and it may also involve ROS, specifically H2O2 (235, 236). H2O2 is in fact believed to act as an activator of KATP (72). NO can also act by direct nitrosylation (635).
IV) Pharmacology of PM KATP channels. Given the complexities of their biophysical properties and modulation, researchers have resorted to pharmacological tools in their efforts to characterize and distinguish KATP channels and their effects, in particular as pertains to the mitochondrial one(s). Needless to say, this approach is not without its own difficulties. Like most ABC “pumps,” SURs can interact with a number of exogenous ligands, and this results in a rich pharmacology (889, 1031, 1174). Some of these ligands can interact also with other cousin ABC transporters (e.g., glibenclamide with CFTR, MRP1, MDR1) as well as other proteins. This promiscuity is a pharmacological headache and contributes to the disputes surrounding mito KATP.
Sulfonylureas (e.g., tolbutamide, glibenclamide) and meglitinides (e.g., repaglinide, nateglinide), used to manage diabetes mellitus, inhibit β-cell KATP by binding to SUR with affinities in the micromolar range. A binding site on Kir6.2 has a much lower affinity. Currently binding to SURs is envisioned as involving two sites: one (A) which is specific of SUR1, and one which is common to all three major SURs (B) (1174). A first subset of drugs (including tolbutamide, nateglinide) binds specifically to site A, a second (including glibenclamide/glyburide) to both sites. Full inhibition can only be obtained in the presence of nucleotides interacting with the Kir site. Sulfonylureas thus seem to act essentially by preventing the stimulatory effect of MgADP at SURs. For our purposes it is important to mention 5-hydroxydecanoate, often considered to be a specific inhibitor of mito KATP (see below). Li et al. (763) have recently reported that this compound can inhibit cardiac PM KATP activated by oleoyl-CoA in excised membrane patches (provided some ATP is present). In intact myocytes however, it failed to reverse KATP activation induced by metabolic inhibition or by rilmakalim, a KATP opener.
Mirror-like, KATP channel openers, a group of chemically diverse molecules (e.g., cromakalim, nicorandil, pinacidil), act similarly to MgADP, antagonizing inhibition by MgATP at Kir (64, 121, 888, 889). A common activating site is involved, comprising amino acids in the last transmembrane helix of SUR (888). Only one activator of SUR1-containing channels is known, namely, diazoxide (and its derivatives). Although known to activate also SUR2B-made KATP, this compound was initially thought not to activate SUR2A-comprising channels. This notion led researchers to attribute its cardioprotective effect in ischemia to mito KATP. However, D'hahan et al. (332) have reported that it can actually activate also cardiac-type (Kir6.2-SUR2A) channels, provided intracellular ADP is present at concentrations >10 μM. Diazoxide and ADP were in fact found to act cooperatively in inducing channel opening (843). Diazoxide can therefore activate all known types of KATP channels, while the other activators act on SUR2-based ones. Generalizing, SUR2B has an approximately fourfold higher affinity for openers than SUR2A (888). Thus pharmacological tools are in principle available to distinguish between SUR1- and SUR2-containing channels. Opener binding is greatly facilitated by the presence of Mg2+ and hydrolyzable ATP (889).
b) mitochondrial atp-dependent potassium channels. An IMM ATP-sensitive K+ conductance has been the object of much interest and of ongoing debate due to its putative involvement in “ischemic preconditioning.” More than 20 years have elapsed since the first reports of an ATP-sensitive mitochondrial inner membrane K+ channel (571, 976), but its very existence, let alone its molecular composition and its functions, are a matter of dispute (e.g., Refs. 58, 59, 138, 167, 210, 304, 428, 498, 514, 928–930, 1275–1277). The majority of investigators are convinced that such a conductance is present (while some still maintain it does not), but its identity remains a question mark. Pharmacological data are inconclusive, and the presence in mitochondria of proteins corresponding to PM KATP components has not been unequivocally established by biochemical or immunological methods.
I) Electrophysiology: not quite enough? The logically foremost hypothesis is that the mito KATP channel(s) may represent a mitochondrial population of whatever KATP is present in the plasma membrane, e.g., Kir6.2-SUR2A in the case of ventricular myocytes. Electrophysiology in general is a convenient fingerprinting tool which ought to allow a definite answer and provide hints as to the possible alternative identiti(es) of the channels. Surprisingly few patch-clamp studies have been devoted to the study of mito KATP (Table 4). The first was the seminal study by Inoue et al. (571). This work used rat liver mitoplasts obtained by digitonin treatment followed by hypotonic swelling and Ca2+-induced fusion. Mitochondrial preparations normally contain ER and PM contaminants, and thus the mitochondrial origin of the channels cannot be taken for granted. In excised inside-out patches the ATP-inhibited K+-selective channel had a conductance of ∼10 pS in 100 (pipette) vs. 33 (bath) mM KCl, lower than the conductances reported for PM KATP variants under comparable ionic conditions (see above). Patching in the mitoplast-attached mode mitoplasts obtained from human Jurkat lymphocytes, Dahlem et al. (296) observed an outwardly rectifying ATP-modulated channel with a conductance of 15 pS at matrix-negative and of 82 pS at matrix-positive potentials in 150 mM KCl (bath and pipette). The conductance of this channel may be considered to be in line with expectations based on PM KATP studies, although the voltage dependence is inverse.
We note that small-conductance Ca2+-activated K+ channels [SK1–3; K(Ca)2.1–3; many variants reported, heterocomplexes can be formed] (reviewed in Refs. 7, 985, 1239), have conductances compatible with the lower values reported in these studies, are not voltage-dependent but exhibit inward rectification due to open channel block, and are activated by Ca2+ in the submicromolar range, i.e., at contamination levels. They are furthermore activated by some compounds having a distinct, although partial, resemblance to diazoxide, pinacidil and other activators of KATP, suggesting the possibility of a pharmacological crossover (see Figure 4). Sensitivity to apamin is characteristic. SK1–3 are strongly expressed in the nervous system, but they are present in many other cell types, including atrial and ventricular cardiomyocytes (e.g., Ref. 1333), smooth muscle (e.g., Ref. 1067) and liver (e.g., Ref. 279).
Electrophysiological data concerning mito KATP are provided by a number of papers using the BLM technique with purified proteins or membrane fractions (99, 136, 137, 139, 594, 870, 871, 903, 976, 1443). The first reconstitution of a mitochondrial protein increasing the K+ conductivity of a bilayer may have been the one by Mironova et al. (869), published in Russian. Subsequent developments of the research by this group are summarized below (see sect. IIIA1bIII) in the context of the discussion on the molecular identity of mito KATP.
Another set of studies employing BLM was carried out by the group of Adam Szewczyk. These authors fused BH submitochondrial particles to study a well-behaved slightly rectifying cation-selective channel with a conductance of ∼100 pS (cis-negative range) in 150/150 or 50/150 (cis/trans) mM KCl (136, 137, 139). The activity could be inhibited by ATP/Mg2+ (0.5 mM) added on the side (cis) of the membrane opposite to that of SMP addition (trans). ATP or Mg2+ alone in cis did not have this effect. Neither did ATP/Mg2+ added in trans, or the nonhydrolyzable ATP analog AMP-PNP/Mg2+ added on either side. Inhibition by cis ATP/Mg2+ was reversed by diazoxide and BMS-191095 (see below) as well as by cis-side addition of GDP. Mito KATP had been previously found to be activated by guanine nucleotides in experiments monitoring K+ transport-induced swelling (977). 5-Hydroxydecanoate (5-HD), glibenclamide, and quinine also inhibited, while HMR-1098, considered to be a specific PM KATP inhibitor (see below), did not. Mg2+ (1 mM) added in trans caused a marked decrease of the single-channel conductance measured in the cis-negative voltage range, coherent with an open channel block. The channel was modulated by trans (i.e., presumably matrix-side) pH, with a transition to basic conditions (pH 7.2 to >8.2) increasing both the mean Po and the single-channel conductance. Decreasing the trans (“matrix”) pH to 6.2 apparently reduced the Po, but unfortunately this effect was not studied in detail (pH 6.2 may or may not fall in the acidic inhibitory range of PM KATP values; a pH-dependence curve may contribute to establishing the identity of mito KATP). A shift of the cis-side pH to 8.2 had little effect, while a decrease from 7.2 to 6.2 resulted in a significant decrease of the Po.
Jiang and colleagues (594, 903) incorporated an IMM fraction from human ventricular myocytes into BLMs and observed activity with stepwise conductance variations up to ∼ 120 pS in symmetrical ∼ 150 mM KCl. The different conductance levels were attributed to cooperative behavior or subconductance states. This activity was inhibited by ATP and 5-HD, and activated by GTP and diazoxide and by PMA, a PKC activator [all added to the same (cis) chamber as the SMPs]. It was furthermore activated by isofluorane, putatively accounting for the preconditioning effect of this anesthetic (595, 903). Western blotting indicated the presence of PKC-δ (but not -ε) in the preparation. An anti-Kir6.2 antibody detected a 56-kDa protein in mitochondria. The apparent size reminds one of the analogous observation by Mironova et al. (871).
K+-selective conductances of various sizes (predominantly 56 pS in 150 mM KCl) were also observed by Zhang et al. (1443) upon reconstitution of SMPs obtained from beef heart (BH). In this work MgATP had no significant effect when added on the same side of the BLM as the SMPs (cis, in this case), but inhibition was observed upon addition to the opposite side. The effect was reversed by GTP (trans). This behavior is analogous to that observed by Bednarczyk and colleagues (137) and by Mironova et al. (871). 5-HD on the same side of SMPs and glibenclamide inhibited (while HMR-1098 did not), and diazoxide reactivated the activity. Superoxide provided by the xanthine/xanthine oxidase system (in cis) produced a rapid activation of the channels attributable to the oxidation of thiol groups.
Despite differences and uncertainties, the channels described by these groups may be considered to resemble each other and to probably correspond to the same molecular species.
II) Pharmacology and pathophysiological role. An important role has been assigned to mito KATP, especially in the heart, in protection from I/R injury, preconditioning (384, 895), and postconditioning (634). In ischemic preconditioning (IP) (900, 663, 664), the application of brief period(s) of reversible ischemia results in myocardial “stunning” (i.e., the temporary loss of the ability to contract) and “adaptive” biochemical changes, including increased phosphocreatine content. Upon subsequent prolonged ischemia, ATP decline and lactate production occur at a slower rate than in untreated tissue so that myocyte death and irreversible damage take longer to develop (e.g., Refs. 588, 589, 641, 900, 901). Elaborate signaling pathways have been identified as mediating acute ischemic preconditioning (for reviews on signaling, see Refs. 216, 527, 528, 664, 781, 1416). The belief that a mito KATP acts as a crucial element has probably influenced their tracing and the interpretation of some of the observations. One can distinguish between a so-called “trigger phase,” in which cells are primed for resistance to the final, severe (referred to as “index”) ischemic episode, and a “mediator phase,” i.e., the actual deployment of protective biochemical instruments when damage is being wrought, i.e., at reperfusion after index ischemia. Cell death associated with I/R damage is thought to be due largely to the onset of the mitochondrial permeability transition, and protective action is widely thought to consist in repressive measures preventing or limiting this phenomenon (e.g., Refs. 526, 766). This latter stage is discussed in section IIIF, but we may anticipate here that a major end-effector may be inhibition (by phosphorylation) of GSK-3β, whose activity is understood to facilitate onset of the MPT.
The biochemistry of ischemic preconditioning has been intensively studied. A sequence of brief ischemic episodes has been reported to result in changes in the antioxidant enzyme system (303) and cell membrane phospholipids (613). Ischemia causes the release of G protein-coupled receptor (GPCR)-coupled effectors like adenosine (269, 481, 1341), bradykinin (which induces production of NO and ROS; e.g., Ref. 941), and opioids (1153; see also Refs. 982, 983). Activation of kinases, including PKCε, p38 MAPK, PKG, and tyrosine kinase(s) (91), is part of downstream signaling. As already mentioned, KATP channels are targets for various kinases, including PKC, PKA, and PKG, and are activated by reactive oxygen species (ROS) (H2O2) (1268). ROS generation is widely considered to be a sine-qua-non of preconditioning (e.g., Refs. 996, 997, 996). KATP channels, and specifically mito KATP channels, along with mitoBKCa, have been assigned a central role in their generation (339).
Indeed, pharmacological evidence soon identified KATP activation as a major element in the mechanism of preconditioning, since inhibitors abolished the effect (471, 477, 478, 479, 481, 776, 791, 1299, 1307). Conversely, KATP openers mimicked preconditioning, reducing the size of the infarcted area, and KATP inhibitors again abolished this effect (e.g., Refs. 478, 480, 848, 1122). Logically, sarcolemmal KATP was initially considered to be involved (e.g., Refs. 974, 1321). However, chemical preconditioning was induced, among other KATP openers, by diazoxide, which appeared not to be able to activate cardiac-type (SUR2A-containing) channels (386, 569, 574, 940, 1176). Mito KATP had meanwhile been discovered, and in 1996 Garlid and co-workers (429, 430) reported, on the basis of swelling experiments with isolated mitochondria and reconstituted proteoliposomes, that it was vastly more sensitive to diazoxide than the sarcolemmal channel (implying, incidentally, that they cannot be the same complex, or at least that they partecipate in different regulatory interactions). Diazoxide, considered as a specific mito KATP opener, was reported to protect mitochondrial functionality during ischemia (e.g., Ref. 578). Like nicorandil (1128), another KATP activator, it induced the oxidation of mitochondrial ETC flavoproteins while leaving sarcoplasmic KATP unaffected (778, 779). Mitochondrial flavoproteins are components of the respiratory chain which become more oxidized when the mitochondrial ΔμH decreases and respiration is therefore stimulated, which is what is expected to happen if the K+ conductance of the IMM increases, but also upon protonophoric “uncoupling.” In several subsequent papers flavoprotein oxidation was used as a gauge of mito KATP opening (e.g., Refs. 668, 1437). Garlid's group (579) also showed that ATP-sensitive K+ flux in mitochondria could be inhibited (K1/2 ∼50 μM) by 5-HD, a known inhibitor of ventricular myocyte sarcolemma KATP (∼10–100 μM; Ref. 922) also known to counteract the protective effects of ischemic preconditioning (e.g., Refs. 70, 848). Following a later report that 500 μM 5-HD did not inhibit Kir6.2-SUR2A channels expressed in HEK293 (559), the drug came to be considered and used as a selective inhibitor of mito KATP (e.g., Refs. 668, 1129). Inhibition of sarcoplasmic KATP by 5-HD has recently been reported to take place in membrane patches, requiring the presence of some ATP, but not in whole myocytes (763). In fact, 5-HD is but one member of the family of fatty acids, which generally inhibit KATP (368, 428, 642). On the contrary, acyl-CoA esters are activating (467, 734, 823). Adding to the complexity, 5-HD can be converted to its acyl-CoA derivative (and further metabolized) (515).
Other purportedly selective agents have been reported. The effects of the most commonly used ones on PM KATP, mito KATP, and preconditioning are summarized in Table 5.
If a compound, say diazoxide, is considered a specific modulator of mito KATP, its effectiveness in cardioprotection supports a role of mitochondrial channels in the phenomenon, and, by implication, their existence. The bulk of the evidence supporting the involvement of mito KATP in protection against I/R damage is pharmacological. Unfortunately, the lack of activity of the drugs on sarcoplasmic KATP cannot always be taken for granted, and other cellular targets possibly accounting for their effects have been identified. This is particularly true of the most important of these agents, diazoxide (268). For example, the report that PKC is needed for diazoxide to activate mito KATP (1371) suggests that the effect of the drug may involve more than a one-to-one interaction with a single protein. Most, if not all, of the pharmacological agents reportedly activating mito KATP can behave as membrane-permeable weak acid/base pairs, and thus as uncoupling, ΔμH-dissipating agents (see Figure 4) (553) (including glibenclamide: Ref. 1274; diazoxide: Refs. 469, 697; contra for diazoxide: Ref. 196). As mentioned, an uncoupling effect is expected to lead to an increase of the respiratory rate and to the oxidation of flavin-based prostetic groups in the enzymes of the respiratory chain, as verified (e.g., Ref. 1124).
To assign a change of flavin fluorescence observed upon addition of a mito KATP activator to channel opening one needs therefore to show its reversibility by mito KATP inhibitors, and/or its modulation by signaling pathways expected to impact on KATP, as indeed has been done in many studies (e.g., Refs. 646, 668, 778, 968).
These controls weaken the contention that preconditioning by diazoxide and other agents may derive from an unspecific uncoupling resulting in the production of ROS. ROS can indeed be produced by protonophore-treated cells (e.g., Ref. 1125). ROS production upon treatment with mito KATP openers such as diazoxide (e.g., Refs. 365, 538, 708) and P-1075 (942) has also been reported, and it has been described to be inhibited by KATP blockers such as glibenclamide and 5-HD (e.g., Refs. 230, 815, 942, 996, 997, 1007). 5-HD has also been found to block ROS production induced by agents presumably acting only indirectly on mito KATP, such as 6-hydroxydopamine (1071), NMDA (409), acetylcholine (1418). These and other observations have led to a model in which ROS are considered to be the effectors of protection downstream of mito KATP (see below; e.g., Refs. 230, 302, 705, 950), although the matter is somewhat controversial (699, 815).
On the opposite front, at least one group has reported that in experiments with permeabilized cells diazoxide induced an increase in the rate of oxygen consumption which was not antagonized by 5-HD and did not depend on the concentration of K+ in the perfusion medium (677). The same group reported on the other hand that the effects of levosimendan were potassium dependent (678, 679). Others have reported that diazoxide does not induce an increase in mitochondrial ROS production (356, 384) or that it antagonizes generation of superoxide (356, 382, 1268), or that it may have either a pro-oxidant or an anti-oxidant effect, depending on the metabolic state and the transmembrane potential of the mitochondria (357, 1062). Superoxide production by the respiratory chain can take place mainly at two sites, namely, complexes I and III (e.g., Refs. 699, 896) but complex II also contributes (for references see, e.g., Ref. 745). In the case of the former, superoxide formation depends on the redox state of a flavine mononucleotide, linked in turn to the poise of the NAD/NADH couple: an increased supply of reducing equivalents (e.g., because of reverse electron flow from site II, or increased ΔμH) is expected to increase superoxide formation, while uncoupling (depolarization) is expected to decrease ROS production at this (matrix-facing) site (e.g., Ref. 1175). Thus activation of mito KATP would be expected to reduce superoxide production by this source (e.g., Ref. 392). At coenzyme Q:cytochrome c oxidoreductase (the bc1 complex, i.e., complex III of the respiratory chain), superoxide formation is believed to be accounted for by single-electron reduction of oxygen by semiubiquinone in the “o” center. This is likely to be the process involved in ROS production by uncoupler-treated mitochondria, and it would be expected to be stimulated by opening of mito KATP and consequent depolarization. Several studies have reached the conclusion that activation of mitochondrial K+ channels reduces or prevents the generation of ROS (for references, see Ref. 1268). Garlid's group also identified complex I as the source of superoxide (35).
“Mild uncoupling” with protonophores has in any case been proposed to result in ROS-mediated cardioprotection by itself (196, 293, 878). Uncoupling by the classical protonophore dinitrophenol (DNP) results in depletion of ATP and, in the presence of KATP activators which modify the nucleotide dependence of the channels, may cause activation of PM KATP (1124). The effect may however be due instead to a direct interaction of DNP with the channels (22).
Why would an increase of the IMM permeability to K+ cause ROS generation? One answer is formulated in terms of variation of Δψm (see above). Another invokes matrix alkalinization due to increased proton pumping to compensate K+ entry (35, 281, 295). One reason this conclusion was reached was that H2O2 production (followed using fluorescent probes) correlated with matrix pH (also monitored with a probe), but not with loss of ΔΨm induced by CCCP, a classical protonophore and uncoupler. Other groups (including ourselves: Ref. 1125) have however reported that ROS production is indeed stimulated by protonophores (such as FCCP or DNP), which are not suspected to induce matrix alkalinization.
An important role in signal transduction to mitochondria in the context of preconditioning is assigned also to the mitochondrial (183, 185, 868) population of gap-junction protein connexin 43 (reviews in Refs. 1096, 1154, 1156, 1157). IPC did not result in cardioprotection in connexin (Cx) 43-deficient mice (538) or in knock-in mice in which Cx43 had been substituted by Cx32 (1072). Cardioprotection is associated with increased phosphorylation of Cx43 at S262 and S368 by PKC (PKCε; Refs. 345, 346, 1155, 1234). The GSK3β signal (phosphorylation/inhibition of GSK3β is protective, because it leads to upregulation of mito KATP) has also been proposed to involve Cx43 (1293). On the other hand, a recent study has placed Cx43 upstream, as a cofactor in signal transduction from the δ opioid receptor, and more precisely in the activation of IB PI3K (573). Nicorandil may actually exert its mito KATP-opening effect at least partly via the NO-PKG pathway (719). ROS proceed to activate PKC (e.g., Ref. 684).
Finally, pharmacological evidence has also been mustered in support of the view that plasma membrane, not mitochondrial, ATP-sensitive channels are responsible for preconditioning. Suzuki and co-workers (1257, 1258) came to this conclusion in the case of diazoxide-induced preconditioning, considering HMR1098 to be a sarc KATP channel blocker, and 5-hydroxydecanoate a mito KATP channel blocker (see also Ref. 1249).
In summary, in our opinion the exact physiological role of mito KATP and the mechanism of its involvement in cardioprotection will be proven only following its definitive molecular identification.
III) Molecular identity based on biochemical/genetic evidence. All considered, the biophysical characteristics of at least part of the conductances observed in BLM and patch-clamp studies are in fair agreement with what may be expected of a Kir6.2-based channel. However, other hypotheses concerning the molecular identity of mito KATP have been put forward. In experiments employing purified proteins, Mironova and co-workers (869, 871) described a channel with a conductance in the order of 24 pS in 0.1 M KCl, formed by a protein with an apparent mass of ∼55 kDa obtained via a detergent-free ethanol/water extraction procedure from beef heart mitochondria. In 1999 this protein was reported to produce activity compatible with the presence of clusters of interacting K+-selective and redox-sensitive channels which were inhibited by millimolar ATP added on the opposite side of that of protein addition (871). A preparation based on Triton X-100 solubilization of rat liver and beef heart mitochondria and containing essentially a protein migrating in gels with an apparent molecular mass of 54 kDa, produced, upon incorporation into BLM, channels with a chord conductance of ∼30 pS at 85 mV in 1 M KCl (976). When reconstituted into liposomes, this preparation allowed transmembrane K+ fluxes to take place. This activity was inhibited by the combination of Mg2+ and ATP (neither of the two alone had a detectable effect). While the reported affinities for inhibiting nucleotides are in the range required for inhibition of KATP channels in the absence of Mg2+ (see above), the requirement for both ATP and a divalent cation seems to be at variance from that reported for classical PM KATP channels. K+ transport in the liposome assay was, remarkably, inhibited by glibenclamide with Ki ∼50–70 nM, even though only one band was seen in the preparation. In 2004 Mironova et al. (870) tentatively identified the ∼55 kDa protein band (which in 2D gel electrophoresis split into at least five bands; apparently the same band obtained with the Triton-based procedure, see below) as corresponding to the mitoKir component of the putative mitoKir-SUR complex, despite the difference in the apparent molecular mass (Kirs have molecular mass of ∼40 kDa). The current-voltage curve showed rectification (although divalent cations were not present). Single-channel chord conductance values in symmetrical 100 mM KCl ranged from 10 to 100 pS, a variability interpreted as evidence of cluster behavior. According to this report the K+ transport activity conferred by the protein was insensitive to 5-HD, glibenclamide, cromakalim and diazoxide, but it was inhibited by TPP+ with K1/2 ∼50 nM. ATP inhibited with a K1/2 of 500–900 μM (depending on the type of assay and on [Mg2+]). The gel electrophoretic analysis of the ethanol-extracted preparations used in the 1992 and 2004 papers shows only one band, but Triton solubilization-based preparations from rat brain and liver mitochondria were shown to contain also another protein with an apparent molecular mass of 63 kDa (99, 870). This was proposed to correspond to a SUR subunit, accounting for the properties of the reconstituted channel. In studies of K+ flux upon reconstitution in liposomes the transport mediated by this preparation was inhibited by ATP, 5-HD, and glyburide, and stimulated by diazoxide and cromakalim in the concentration range expected for a Kir-SUR channel.
The 63-kDa protein, whose sequence was not determined, may correspond to a “short” SUR splicing isoform (see sect. IIIA1aI). The idea that alternative forms may be involved in the formation of mito KATP has experimental support. Kir6.2 and SUR2A or short forms of the latter have been identified in ventricular myocyte (1211; Kir6.1 as well) and brain (729) mitochondria by immunofluorescence, immunogold labeling, and Western blotting. In an informative study, Ye et al. (1421) identified 28-, 55- and 68-kDa splicing short forms of SUR2A (in addition to the long form) and a 55-kDa form of SUR2B in gradient-purified mitochondria from mouse heart. Mitochondria from mouse brain contained the same species except for the 68-kDa form. Coexpression of the 55-kDa form of SUR2B and Kir6.2 in an heterologous system produced an ATP-sensitive channel with an ∼50% lower conductance than that of the Kir6.2-SUR2A cardiac-type KATP channel. This activity was glibenclamide-insensitive and could not be reinstated by diazoxide or pinacidil after inhibition by ATP, suggesting that at least this particular SUR short form may be pharmacologically inert. Interestingly, a 28-kDa protein binding glibenclamide with low affinity has been found in beef heart mitochondria (BHM) (1274, 1278). Short forms of SUR have been found in mitochondria also by other studies (729, 1032, 1211, 1279).
O'Rourke, Marbàn and collaborators have compared the pharmacological profiles of cardiac mito KATP and of molecularly defined PM KATP expressed in HEK cells, using patch-clamp to assess PM KATP activity and flavoprotein oxidation as an index of mito KATP (559, 777). The uncoupling effect of glibenclamide was taken into consideration. These investigators concluded (at the time): “Our results demonstrate that mito KATP channels closely resemble Kir6.1/SUR1s KATP channels in their pharmacological profiles.” A different nature of the SUR unit might explain some of the peculiarities of mito KATP modulation by ATP (see above). A recent study reports detecting Kir6.1 and SUR1 in hepatic mitochondria (905). It is however difficult to be sure that the activities and/or immunoresponsive material described do not originate from contaminating PM proteins or mitochondria-associated membranes (MAMs). Ardehali et al. (57), for example, did not detect Kir6.1 in their preparations of rat liver mitochondria (RLM). Concerns about the specificity of antibodies also linger. This objection has been forcefully made by Marbàn's group, who found that two commercial antibodies thought to be specific for Kir6.1 recognized instead completely unrelated mitochondrial proteins (413). In addition to the studies already mentioned (597, 729, 1211), immunoanalytical approaches have been used by Kuniyasu et al. (715), who found no evidence of Kir or SUR subunits in the mitochondrial fraction obtained from rat hearts (despite the fact that all were apparently present in the microsomal fraction). The adenoviral vector-mediated expression of Kir6.1 or Kir6.2 or dominant-negative constructs derived from them in rabbit ventricular myocytes did not affect mito KATP activity as assayed by flavoprotein fluorescence; immunohistochemistry versus Kir6.1 did not provide evidence for a localization of this subunit in mitochondria (1169).
Kir6.2 has been targeted to the mitochondria of HEK293 and HL-1 cells (780). Its expression in the organelles conferred resistance to hypoxic stress, a cell culture-level counterpart of ischemia-reperfusion. Mice KO for either Kir6.1 (863) or Kir6.2 (862, 1172, 1258), as well as mice overexpressing Kir6.2 in part of the brain (543) have been generated and used in a number of studies (supporting, inter alia, the involvement of Kir6.2 in protection from I/R damage: Refs. 543, 1107, 1252, 1253), but surprisingly so far to our knowledge they have not been exploited to investigate the nature of mito KATP.
SUR1- (1189) and SUR2-deleted (264) mice are also available. Mito KATP studies have apparently not been performed with the SUR1 KO mouse. Despite resulting in various cardiovascular problems (9), SUR2 disruption turned out to protect the heart against I/R damage compared with WT controls (1240), and ischemic preconditioning did not provide further protection (1421). As mentioned above, in these animals short-form variants of SUR2 continue to be expressed and were localized in mitochondria (1032, 1421). The short (55 kDa) form can cooperate with Kir6.x subunits to produce a KATP channel with somewhat altered properties, e.g., insensitivity to glibenclamide (1032). Protection has been associated with a “bioenergetic phenotype” comprising increased IMM permeability to K+, reduced ΔΨm, increased resistance to Ca2+ overload, increased basal rate of ROS generation (9). ΔΨm and K+ uptake (swelling) were not altered by diazoxide, suggesting that sensitivity to the opener may indeed be associated with the long form of SUR2. This is relevant because of the possibility that the effects of diazoxide may be unrelated to the presence of a KATP in mitochondria. Besides the possibility that they may “simply” derive from uncoupling (see above; Ref. 553), it has been suggested that they may have to do with inhibition of the succinate dehydrogenase complex (516, 867, 1135) (pinacidil, on the other hand, inhibits NADH oxidation; Ref. 516).
Recent studies in C. elegans have cast further doubts on the existence of Kir-SUR mito KATP. C. elegans exhibits “ischemic preconditioning,” and swelling assays on purified mitochondria show a behavior compatible with the presence of mito KATP channels with a pharmacological profile analogous to that of mammalian mito KATP (1386). Deletion of any or all three Kir-encoding genes in C. elegans did not grossly affect the preconditioning protection phenotype. Furthermore, mutant mitochondria (including triple-KO) exhibited near-normal mito KATP channel activity, as assayed by the Tl+ transport technique (1390). Tl+ transport was activated by diazoxide and atpenin A5, and activation could be reversed by 5-HD and glyburide (1387). These results obviously suggest that Kir subunits may actually not be involved in forming mito KATP.
Finally, a complex comprising SDH, the ANT, mitochondrial ATPase and mitochondrial ATP-binding cassette protein 1 (mABC1) was found to exhibit mito KATP-like properties upon reconstitution into proteoliposomes and planar lipid bilayers (57, 58; see also Ref. 485). K+ transport in proteoliposomes was enhanced by diazoxide, pinacidil, 3-nitropropionic acid (3-NPA), and malonate (see also Ref. 1385) (inhibitors of SDH) and antagonized by 5-HD, glibenclamide, and ATP. The channels observed displayed a predominant conductance of “< 100 pS in 100 mM K+.” Selectivity for K+ was however low. The data presented do not allow a meaningful comparison with the properties of the mito KATP channels observed in other planar bilayer experiments. Po was increased by diazoxide and 3-NPA. ATP, 5-HD, and glybenclamide decreased it, while atractyloside and HMR-1098 did not have a significant effect. Those who refute the identification of mito KATP with this complex point out that the concentrations required to activate mito KATP are much lower than those required to inhibit SDH and suggest that SDH may act as a regulator of mito KATP and as a mediator of mito KATP opening (428, 1384, 1388). The proposal has been made that mito KATP opener-induced ion fluxes may be mediated by the ANT (680). The existence of a supramolecular protein complex sensitive to mito KATP-targeting pharmacological agents is supported by results obtained with a mouse expressing a mutated mitofusin-2 (MFN2) gene (MFN2 is an OMM protein involved in mitochondrial fusion). In this animal complexes II and V are functionally impaired, but the defect can be rescued by 5-HD, while diazoxide-induced symptoms in controls resembled those associated with mutant MFN2 (485).
In constrast to all above hypotheses, impressive evidence has been presented that the ion-conducting portion of mito KATP may actually be formed by a mitochondrial population of the inward-rectifying K+ channel ROMK2 (Kir1.1b), a short form of ROMK (412). ROMK is ATP-sensitive and shares other characteristics with KATP channels, including Mg2+ block and the ability to associate with ABC proteins like the CFTR and SUR2B (see Ref. 412 for references). Its single-channel conductance at negative voltages is ∼32–43 pS (cell attached; 110 mM KCl in pipette) (e.g., Refs. 261, 1060). Conductance values reported for mito KATP tend to be either above or below this range, and as stated also by Foster and coworkers (412), more work is required to confirm the identification. ROMK-KO mice (790, 792) and rats (1451) have been obtained, although they are short-lived, and fairly selective inhibitors of the channel are being developed (174, 1287).
2. Big-conductance calcium-activated potassium channel
The large-conductance calcium- and voltage-activated K+ channel BKCa (also named MaxiK, Slo1, KCa1.1, and KCNMA1) is ubiquitously expressed at the plasma membrane of excitable and nonexcitable cells including sensory and epithelial cells and smooth muscle. BKCa is important in smooth muscle contraction but also for cytoprotection during I/R, hypertension, and cancer cell metastasis (292, 371, 388, 1105, 1222). The predicted molecular mass for a single subunit is 125 kDa; however, the channel has different forms due to alternative splicing. The single-channel conductance of BKCa ranges from 100 to 300 pS in 150 mM KCl (735). Either Ca2+ binding to its cytoplasmic (or matrix, for mitochondrial channels) domain or membrane depolarization induces its opening (598, 908, 1432), thereby coupling membrane potential and intracellular calcium concentration (557, 735). Calcium decreases the energy required to open the channel, causing a leftward shift in the Po/voltage relationship. Voltage dependence is conferred by the four TM segments (VSD: voltage sensor domain) preceding the pore-forming TM segments, as in the case of Kv channels. Activity is further regulated by beta subunits (β1-β4) having two transmembrane domains and a molecular mass around 26 kDa (1088). Importantly, leucine reach repeat-containing LRRC-subunits with a single transmembrane domain also regulate activity and have been shown to shift the voltage required for channel activation by −140 mV in case of the PM BKCa (1414). Both native and cloned BKCa channels are activated in electrophysiological experiments by calcium in the micromolar range and are stimulated by 12,14-dichlorodehydroabietic acid (diCl-DHAA) (1108); the benzimidazolones NS004, NS1619, and NS11021 (54, 149); and by the indole carboxylate compound CGS7181 (560). Only the last compound shows however high potency and selectivity towards BKCa, acting below 0.1 μM concentration when applied to inside-out patches (560). BKCa α-knockout mice are viable and are characterized by a phenotype of bladder hyperactivity, urinary incontinence, cerebellar ataxia and Purkinje cell malfunction, and a shortened lifespan (855, 1132).
a) mitochondrial bkca activity and pharmacology. The presence and activity of BKCa has been revealed also in intracellular membranes, including nuclear membrane, ER, Golgi, and mitochondria (1212). MitoBKCa (929, 1276) has been observed by direct patch-clamping of mitoplasts of mammalian cells as well as in planar lipid bilayer experiments. The channel has been identified in mitochondria of two glioma cell lines, in astrocytes as well as in ventricular cells, skeletal muscle and brain and in endothelial cells, with conductance values ranging from 276 to 307 pS in 150 mM symmetrical potassium solution (see Table 6). MitoBKCa was shown to be activated by calcium, diCl-DHAA (1108), NS1619 (1214), 17β-estradiol (937), and hypoxia (252) and was found to be blocked by specific inhibitors charybdotoxin (483, 1215), iberiotoxin (252, 253), and paxillin (536, 537). Table 6 reports the main properties of the mitoBKCa observed in different systems. In summary, modulators of PM BKCa act also on the mitochondrial form of this channel, but no drug acting exclusively on the mitochondrial channel is available.
Evidence for the presence of BKCa channel protein in mitochondria has also been provided by Western blotting, electron microscopy, and immunofluorescence microscopy (353, 633, 1020, 1215). Douglas and colleagues (353) provided biochemical evidence for the presence of a 125-kDa protein recognized by anti-BKCa antibody in mitochondria of several brain regions including Purkinje cells and cerebellum granule cell neurons. O'Rourke's group (1407) has reported the recognition of a 55 kDa protein in cardiac mitochondria and of an 80 kDa protein in liver mitochondria. The reason for the detection of proteins with lower than the predicted molecular was not clarified in this latter work. A 120-kDa band reacting with an anti-BKCa antibody was present in mitochondria from cardiomyocytes, while it was absent in liver organelles (1183). Finally, a protein with the expected molecular of 110 kDa was immunoprecipitated using an anti-BKCa antibody from mitochondrial preparations made from cochlea and brain (633). In summary, both electrophysiological and biochemical evidence locate a population of BKCa channels to mitochondria. It is not known whether the regulatory mechanisms mentioned in the case of PM-located BKCa operate also in mitochondria, where the channel is thought to be activated primarily by calcium, but the mitoBKCa channel β4 subunit protein (26 kDa) was found in mitochondria, e.g., in brain (1020). Src kinase as well as LRR proteins are also present in this organelle (565, 1113) and might a priori modulate the gating properties of the mitochondrial channel. Interestingly, a charybdotoxin-insensitive BKCa channel activity has recently been described in a whole brain preparation by using the basolateral membrane technique (385). The authors hypothesized that lack of charybdotoxin block might be due to regulation by β4 subunits. Alternatively, this behavior might be expected assuming, e.g., that the observed activity was due to heterooligomers between KCa1.1 and other BKCa family α subunits (e.g., Slo2, see below), given that heterooligomerization might lead to channel activity with intermediate pharmacological properties (604).
b) molecular identity. Mito BKCa has been proposed to be the same species as PM-located BKCa, and indeed pharmacological evidence currently favors KCa1.1/Slo1 as the mitochondrial channel. The molecular mass of the bands detected on Western blots by anti-BKCa antibodies suggests that the mitochondrial form is the same as the PM-located form or possibly corresponds to alternatively spliced BKCa. Interestingly, a BKCa isoform called glioma BKCa (gBKCa) was found to be expressed in different types of human cancer cells, including glioma, breast, colon, pancreas, stomach, duodenum, lung, and liver cancer. gBKCa is characterized by a 63-amino acid-long insert, against which a specific antibody was recently generated which is not able to recognize the unmodified BKCa channel protein. Both anti-BKCa and anti-gBKCa revealed localization of the respective isoforms in the PM, mitochondria, Golgi, and ER, indicating that distribution of gBKCa and BKCa within the cells is identical (434). This finding suggests that BKCa and gBKCa undergo the same modifications leading to organellar targeting within the cell. A splice variant, BKCa-DEC harboring a specific amino acid sequence at the COOH terminus of the protein (VEDEC) (643), was identified as a mitochondrial candidate (633). In another study BKCa located to the plasmamembrane while a splice variant (different from BKCa-DEC) was clearly targeted to mitochondria (1317). Both studies however addressed the question of targeting using a heterologous system where possible mistargeting due to overexpression was not ruled out. A recent study highlighted that ∼20% of the proteins that interact with BKCa, as identified by proteomics, are mitochondria related (633), including cytochrome c, ATP synthase, and chaperon HSP60. Whether these interactions regulate channel activity is still an open question. Recently the participation of the Slo1 (KCa1.1) channel in the formation of mitoBKCa has been questioned by using a genetic mouse model lacking the Slo1 gene: an identical mitoBKCa activity and pharmacological profile was found in both Slo1−/− and wild-type mice, suggesting that Slo1 does not give origin to mitoBKCa (1389). It is to mention that calcium did not activate mitoBKCa in wild-type mouse mitochondria in these experiments based on the use of Tl+-sensitive fluorophore loaded into the mitochondrial matrix, raising the question of whether mitoBKCa activity was measured. The authors proposed that the widely expressed Slo2 genes are crucial for the formation of mitoBKCa. The large-conductance K+ channel gene family indeed includes Slo2, encoded by two genes in mice [Kcnt1 (Slo2.2/Slack) and Kcnt2 (Slo2.1/Slick)]. In C. elegans, both Slo1 and a single gene coding for Slo2 exist; therefore, the authors used knockout worms to rule out a role for Slo1 and to propose instead that Slo2 contributes to the calcium-sensitive, charybdotoxin- and iberiotoxin-insensitive mitoBKCa in C. elegans. Given that Slo2, in contrast to Slo1, does not harbor the S0 segment (i.e., it has only 6 TM segments), application of a specific antibody against the S0 segment might a priori discriminate between Slo1 and Slo2 as mitoBKCa forming components in mammalian mitoBKCa.
c) physiological function. Information on the physiological role of mitoBKCa has been obtained prevalently by pharmacological means and only recently using a knock-out mice model. MitoBKCa has been proposed to be activated under pathophysiological conditions that increase mitochondrial Ca2+ uptake. Opening of mitoBKCa protects against damage to the heart and other organs caused by ischemia and reperfusion (for reviews, see e.g., Refs. 928, 929, 930). Preconditioning hearts with small molecule BKCa openers or β-estradiol resulted in decreased cardiomyocyte death due to ischemic insult and in reduced myocardial infarction (148, 149, 189, 937, 1241, 1370, 1407). For example, activation of the mitoBKCa channel by NS1619 (1183) or by NS11021 has been shown to be effective in protecting the normoxic hearts against ischemia (53). Hypoxia-activated mitoBKCa channels also protect cardiomyocytes isolated from chronically hypoxic rats (252), which are resistant to injury induced by acute oxygen deprivation (189). The involvement of mitoBKCa was further indicated by the finding that cardioprotective effects could be antagonized by paxillin, a BKCa inhibitor acting also on mitoBKCa.
The protective effect of BKCa openers has been attributed to a partial depolarization of the IMM reducing the driving force for calcium influx and thus preventing excessive calcium accumulation in the matrix (631, 1108). Opening of the calcium-inducible MPTP can thus be prevented (253). In isolated mitochondria from heart and brain, as well as in isolated heart, activation of mitoBKCa with NS1619 and CGS7184 was shown to reduce ROS production (537, 712, 1231). MitoBKCa may also fine-tune mitochondrial volume: its activation by application of NS11021 leads to charybdotoxin-sensitive K+ influx and swelling in the presence of permeable anions. Improved mitochondrial respiratory control might also explain the cardioprotective effect of mitoBKCa (53). Importantly, a functional coupling has been recently described between mitoBKCa and the respiratory chain complexes (142).
Small molecule mitoBKCa inhibitors used in the above studies to define the physiological role of mitoBKCa might however have also additional, mitoBKCa-unrelated effects. For example, NS1619 at high concentrations induces matrix K+ and H+ influx through leak, a nonspecific transport mechanism (20). Another study reported that NS1619 is an effective inducer of immediate neuronal preconditioning in cultured cortical neurons, but its neuroprotective effect is independent of the activation of BKCa channels (431). BKCa channel opener CGS7184 affects intracellular calcium homeostasis by interacting with the sarcoplasmic reticulum RYR2 channels (1394). BMS-204352, another BKCa channel activator, inhibits L-type calcium channels in ventricular myocytes with a KD of 6 μM (1223), while paxillin inhibits BKCa and various isoforms of SERCA, with IC50 values of 5–50 μM (178). Furthermore, NS11021, at concentrations that confer cardioprotective effects (149), induces a charybdotoxin-insensitive respiratory control decrease and a drop in membrane potential of nearly 30 mV even in the absence of K+ (53). Thus accumulating evidence suggests that depending on the concentrations used drugs acting on BKCa might have off-target effects, suggesting caution when interpreting results obtained using these drugs. In fact, for example, a recent study showed that isoflurane-mediated cardioprotection was abolished by paxillin in both wild-type and Slo1−/− hearts lacking the KCa1.1 α subunit (1389), suggesting a BKCa-independent and thus nonspecific action of paxillin. The same study also challenged the notion of a role of KCa1.1 in protection against ischemia/reperfusion injury, proposing instead an involvement of Slo2 (see above).
With regard to a possible role of mitoBKCa in the regulation of apoptosis, patch-clamp experiments using recombinant Bax showed an inhibition of mitoBKCa, which might contribute to opening of the MPTP which takes place during cell death (253). Opening of BKCa in isolated brain mitochondria was shown to inhibit ROS production by respiratory chain complex I and was hypothesized to be beneficial for neuronal survival (712). Paxillin but not iberiotoxin accentuated TRAIL-induced apoptosis in glioma cells, but the effect of paxillin was not due to its action on BKCa channels (632).
3. Intermediate-conductance calcium-activated potassium channel (IKCa): electrophysiology and pharmacology
KCa3.1, also called IK1 or SK4 or Gardos channel, is an intermediate-conductance potassium channel expressed in various tissues (epithelial and endothelial tissues, immune system, sensory neurons and microglia but not in excitable tissues), is composed of six transmembrane domains (S1-S6), with the pore loop located between S5 and S6, as well as cytoplasmic NH2 and COOH termini. It displays a non-ohmic conductance ranging between 10 and 80 pS in 150 mM potassium (105). Its activation by intracellular calcium has been reported to occur at EC50 values in the range of 95 to 350 nM (depending on experimental conditions) and depends on the constitutive binding of calmodulin to the calmodulin binding domain (CMBD) region at the COOH terminus of the channel. PM-located IKCa (KCa3.1) has been implicated in numerous physiological processes, including cell proliferation and differentiation by regulation of membrane potential and calcium signaling in erythrocytes, activated T and B lymphocytes, macrophages, microglia, vascular endothelium, epithelia, and fibroblasts. Despite the wide expression and important physiological functions of IKCa, the absence of this channel protein in transgenic knockout mice does not result in severe physiological changes, possibly because of developmental compensation (1399). Nevertheless, IKCa is currently being proposed as a target in autoimmune diseases (730), cardiovascular diseases (1400), treatment of hypertension (665), cancer therapeutics (e.g., Refs. 451, 864), and Alzheimer's disease (806). TRAM-34 (1-[(2-chlorophenyl) diphenylmethyl]-1H-pyrazole) (IC50 20 nM), cyclohexadiene 4 (IC50 1.5 nM) and clotrimazole are among the most potent and used inhibitors of IKCa, while DCEBIO (EC50 1 μM) and NS309 (EC50 30 nM) are known activators (1399).
Recently, this channel has been recorded from the inner mitochondrial membrane of human colon carcinoma cells (311), of HeLa cells (human cervix adenocarcinoma cells), and of mouse embryonic fibroblasts (1126). The channel protein was detected in purified mitochondrial fractions also by Western blot, revealing that the molecular weights of the IKs in the PM and IMM are the same (311, 1126). Mito IKCa was inhibited by clotrimazole and TRAM-34. The biophysical and pharmacological properties of the mito IKCa and of the PM IKCa channels of the same cells were indistinguishable, with mito IKCa showing conductance ranging from 10 to 90 pS in 150 mM KCl, activation by submicromolar Ca2+ and inhibition by TRAM-34 in the 10−8 M range (1126). The physiological role of mito IKCa has not been studied in detail so far, but a role similar to that of other mitochondrial potassium channels may be envisioned, i.e., a contribution to the regulation of membrane potential, volume, and ROS production. TRAM-34 induced a hyperpolarization of the mitochondrial membrane (as expected if a positive charge-carrying influx is inhibited), confirming that functional IKCa is expressed in the IMM. Given the proposed role for other mitochondrial K+ channels in cell death (see below), the membrane-permeant TRAM-34 was used in an attempt to induce cell death, and the effect of recombinant Bax on channel activity was studied. In contrast to BKCa (KCa1.1) and Kv1.3, IKCa was not inhibited by Bax (1126). Interestingly, while TRAM-34 used alone at 2–80 μM concentration did not induce apoptosis (1036, 1126), in combination with the death receptor ligand TRAIL it synergistically increased the sensitivity to TRAIL of melanoma cells (1036). Whether this effect is to be ascribed to the PM or the mitochondrial population of IKCa (or both) is unclear. Given that both TRAM-34 and TRAIL are characterized by relatively good safety profiles, coadministration of the two drugs might be exploited for melanoma treatment. TRAM-34 and clotrimazole have also been shown to decrease the viability of epidermoid cancer cells when applied together with cisplatin, apparently via an action on the PM-located IKCa, involved in the apoptotic volume decrease (748).
4. Small-conductance calcium-activated potassium channel (SKCa)
A recent report presents biochemical, microscopic, functional, and electrophysiological evidence of their presence in the IMM of guinea pig heart cells (1242). When the authors incorporated a purified mitochondrial membrane fraction into planar lipid bilayers, they observed Ca2+-modulated, apamin-inhibited activity. The conductance steps observed with symmetrical 200 mM KCl ranged from 180 pS at 1 μM Ca2+ (Po = 0.5) up to 730 pS with 100 μM Ca2+ (Po = 1). These values are unexpectedly high for SKCa, as the authors recognize. Dolga et al. (348) studied by patch clamp of HT-22 neuronal mitoplasts an activity ascribed to SKCa on the basis of its pharmacological properties. The study highlighted a role of the channel activity in protection against glutamate-induced oxytosis (348). These papers support the notion that a mitochondrial population of these channels may exist and might account for some of the electrophysiological observations ascribed to mito KATP.
5. Voltage-gated mitochondrial potassium channel Kv1.3
a) pm Kv1.3. Voltage-gated potassium channels (Kv) comprise a large family of channels that are expressed in both excitable and nonexcitable cells (reviews in Refs. 60, 492, 590, 803, 935, 1428). They control the resting plasma membrane potential and the frequency of action potentials in excitable cells, while in nonexcitable tissues, such as pancreatic islets, the immune system, and epithelial cells, they are involved in feedback regulation of the plasma membrane potential, and thus in processes ranging from secretion to cell proliferation. Downstream events are the result of a signaling cascade that most often involves modulation of voltage-dependent Ca2+ channels and variations of cellular Ca2+ levels. Each Kv gene encodes a protein subunit, four of which may form either homotetramers or heterotetramers within the same family (Kv1-Kv12) (492). Functional diversity of Kv activities is enhanced by factors including heterotetramerization and the association of accessory proteins, such as the β-subunits which can modulate gating properties and assist multimerization (1026, 1318). Alternative splicing and posttranslational modifications also contribute to variety of activity, and various mechanisms have been proposed to regulate protein expression itself.
Kv channels share with other potassium channels the general structure of the selectivity filter, in which coordination of a transiting K+ by four carbonyl groups belonging to the four polypeptide chains stabilizes the ion, allowing it to shed solvation water and to overcome the membrane lipophilicity barrier (354, 597, 598, 782, 788, 803, 886, 1339, 1422, 1450). Voltage dependence is mediated by a voltage-sensing region bearing a net positive charge (597, 1289, 1314). A characteristic of Kv channels is the presence of a vestibule which helps to gather and concentrate cations for eventual transport through the pore, due to the presence of a set of acidic amino acids which impart an overall negative charge to the vestibule walls. These conserved residues can interact with a strategic basic (positive) amino acid present in the peptide toxins of some venoms, such as ChTx, MgTx, IbTx, ShK, which can therefore “plug” and block the channel (1051, 1429). Other toxins approach instead the channel via the membrane lipid bilayer and interact with the voltage sensor (e.g., Ref. 803).
Kv members are known to be located in the plasmamembrane, but Kv1.3 has been localized to mitochondria as well (see below). This channel, first discovered in lymphocytes (216), is expressed also at least in the CNS, macrophages, kidney, testis, adipose tissue, osteoclasts, liver, and skeletal muscle (492). PM Kv1.3 has been implicated in the regulation of cell proliferation (218), apoptosis (for review, see Refs. 1260, 1270) and volume (731, 732), and in neurotransmitter release by excitable cells (e.g., Ref. 1201). Furthermore, Kv1.3 seems to participate in the pathways regulating energy homeostasis and body weight by a still poorly clarified mechanism (1406). One intriguing possibility, still to be explored, is that the mitochondrial Kv1.3 rather than the PM-located channel may be involved in energy homeostasis.
The PM-located Kv1.3 channel displays an activation threshold between −50 mV and −60 mV, a single-channel conductance of 24 pS in T lymphocytes (955) and is activated by depolarization (962). The discovery of the physiological roles of K+ channels first in T-cell activation and later in other cell types was made possible mainly by the use of specific scorpion (ChTx, MgTx) and sea anemone (Shk) toxins (see above). In addition, potent small molecule inhibitors of Kv1.3 have been identified, including Psora-4 (IC50 3 nM) (1347), PAP-1 (IC50 2 nM) (1148) and clofazimine (IC50 300 nM) (1059). The clarification of its pathophysiological roles and the availability of potent inhibitors have made this channel a promising pharmacological target (157) for various diseases, including several autoimmune diseases (255, 1398), diabetes (262), and cancer (e.g., Refs. 56, 300, 389, 1244).
b) mitochondrial localization of Kv1.3. Interestingly and unexpectedly, functionally active Kv channels have been identified in the IMM. The first was Kv1.3, the main voltage-gated homotetrameric channel expressed in lymphocytes. Localization of Kv1.3 to the IMM in lymphocytes and subsequently in other cell types has been demonstrated by multiple techniques: patch clamp (see below), immunogold transmission electron microscopy in lymphocytes (1265), macrophages (1352) and postsynaptic medial nucleus of the trapezoid body (MNTB) neurons (433), immunofluorescence in gerbil hippocampus (140) and Western blot in the above cells as well as in PC-3 prostate cancer, MCF-7 breast adenocarcinoma (487), SAOS-2 osteosarcoma, B16F10 melanoma cell lines (741) and in J774 macrophages (744). In all these cases the channel protein was revealed in genetically nonmanipulated cell lines or tissues, either healthy or cancerous. The mitochondrial localization was further proven by biochemistry using purified mitochondria and different markers to assess contamination by ER and/or PM. Given that Kv1.3 as well as mitochondrial markers become enriched in the purified organellar fraction while the intensity of ER/PM markers decreases (487), it can be concluded that the protein is indeed present in the IMM. Different antibodies raised against distinct regions of Kv1.3 were used in the above studies revealing a 65-kDa band, suggesting that the primary sequence of the mito Kv1.3 is highly homologous (if not the same) to that of the PM Kv1.3. However, the targeting mechanism is still unclear: mitochondria-specific targeting due to possible alternative splicing or, more likely, posttranslational modifications are possibilities for consideration. It is interesting to note that Kv1.5 also has a dual localization to PM and mitochondria at least in macrophages (744) and another Kv channel, Kv10.1, has recently been described as functional channel in the nuclear membrane beside being active in the PM (247).
c) biophysical and pharmacological properties of mito Kv1.3. Functional expression of Kv1.3 in mitochondria even at the highly negative resting potential (approximately −180 mV) is indicated by hyperpolarization upon incubation of isolated organelles with specific Kv1.3 inhibitors, namely, margatoxin and ShK. Under physiological conditions, Kv channels are expected to allow an electrophoretic inward flux according to the electrochemical gradient for K+ (e.g., Ref. 167). Entry of potassium is compensated by the respiratory chain-driven efflux of protons and by activity of the electroneutral K+/H+ antiporter, to avoid volume changes and depolarization. Patch clamping of the IMM of lymphocyte mitochondria (1265) and of gerbil hippocampal cells (140) confirmed that active Kv1.3 channels are present in the IMM of various cells. Stable transfection of CTLL-2 lymphocytes with the Kv1.3-coding gene gave rise to electrophysiological activity in both the PM and the IMM, with characteristics of the PM Kv1.3 channel (1265). This finding further suggests that the mitochondria-located channel is a product of the Kv1.3 gene. In fact, channel activity recorded in the range between −60 to +60 mV displayed some of the known characteristics of Kv1.3 in lymphocytes: a slope conductance of 25 pS in 150 mM KCl, potassium selectivity, a slight rectification, and inhibition by specific inhibitors margatoxin and Psora-4 (1264, 1265). The activity attributed to mito Kv1.3 in gerbil hippocampal mitochondria was sensitive to 10 nM MgTx but was not inhibited by agitoxin, a blocker of Kv1.x channels (140). The channel was voltage dependent in the −50/+50 mV voltage range and showed a conductance of 109 pS in 150 mM KCl medium. Heterotetramerization of Kv1.3 with other Kv1.x channels in hippocampal mitochondria and/or possible association with regulatory β subunit might account for the biophysical/pharmacological properties of this channel being different from those found for Kv1.3; however, further work is required to identify subunits contributing to the observed activity.
d) mito kv1.3 and apoptosis. Mito Kv1.3 is expected to participate in regulation of mitochondrial membrane potential, volume, and ROS production, similarly to other K+ channels found in the IMM. The role of mito Kv1.3 under physiological conditions has not been elucidated in detail; however, a crucial role for this channel in apoptosis became evident first in lymphocytes and later in other systems as well. The mechanistic features of apoptosis are broadly understood, and it is widely accepted that mitochondria have a central role since the release of cytochrome c (and other proteins) contributes in a fundamental way to the activation of caspases, which marks the point of no return. Expression of a mitochondria-targeted Kv1.3 construct was sufficient to sensitize apoptosis-resistant CTLL-2 T lymphocytes, which lack Kv channels, to several pro-apoptotic stimuli. Mito Kv1.3 has been identified as a target of Bax, a pro-apoptotic Bcl-2 family protein, and physical interaction between the two proteins in apoptotic cells has been demonstrated (1264, 1269). Incubating Kv1.3-positive isolated mitochondria with Bax triggered apoptotic events including membrane potential changes (hyperpolarization followed by depolarization due to the opening of MPTP), ROS production, and cytochrome c release, whereas Kv1.3-deficient mitochondria were resistant. Highly conserved Bax lysine 128 protrudes into the intermembrane space (38) and mimics the already-mentioned crucial lysine in Kv1.3-blocking peptide toxins. Mutation of Bax at K128 (BaxK128E) abrogated its effects on Kv1.3 and mitochondria, indicating a toxinlike action of Bax on Kv1.3 to trigger at least some of the mitochondrial changes typical of apoptosis. It should be noted that inhibition of the channel is not equivalent, either logically or experimentally, to its absence. Evidence for the physiological relevance of the interaction of mito Kv1.3 with Bax via lysine 128 has been obtained also in a cellular context, using Bax/Bak less double knockout mouse embryonic fibroblasts (1269). Mutant BaxK128E was unable to mediate cell death in DKO MEFs challenged with various apoptotic stimuli. These findings are not in contradiction with the view that Bax contributes to cytochrome c release, but indicate that a K+ channel-dependent event is important for the release of cytochrome c and that expression of mito Kv1.3 sensitizes cells to apoptosis. As mentioned, TRAM-34, a blocker of mito IKCa has also been shown to induce an early mitochondrial hyperpolarization (1036). Blocking the IMM channel(s) would not per se be expected to result in permeabilization of the OMM to proteins. This might however be obtained with the intermediacy of ROS, which may prompt Bax migration to mitochondria (13, 298, 916, 1004) and/or the onset of the MPTP, which provides a mechanism alternative to Bax oligomerization for cytochrome c efflux (464, 948, 949, 1047). In accordance with our findings, platelets, where Kv1.3 is the exclusive Kv channel expressed, were resistant to apoptosis in Kv1.3 knockout mice (846).
Recently we reported that Psora-4, PAP-1, and clofazimine, three distinct membrane-permeant inhibitors of Kv1.3, induce death by directly targeting the mitochondrial channel in multiple human and mouse cancer cell lines, while membrane-impermeant, selective, and high-affinity Kv1.3 inhibitors ShK or margatoxin did not induce apoptosis, further proving the crucial role of mito Kv1.3 for this process (741). Importantly, the membrane-permeant drugs killed cells also in the absence of Bax and Bak, a result in agreement with the current mechanistic model for mitoKv1.3 action (Figure 5). Genetic deficiency or siRNA-mediated downregulation of Kv1.3 abrogated the effects of the drugs, proving specificity of their action via Kv1.3. The importance of our findings was validated in vivo: intraperitoneal injection of clofazimine reduced tumor size by 90% in an orthotopic melanoma B16F10 mouse model, while no adverse effects were observed in several healthy tissues. Similar results were obtained for primary human cancer cells from patients with chronic lymphocytic leukemia (B-CLL) which express higher level of functional Kv1.3 than B cells from healthy subjects (743). Importantly, only pathological ex vivo cells of patients underwent apoptosis when treated with any one of the three drugs, while residual healthy T cells were resistant. Such selective action of these drugs on tumor cells might be ascribed to a higher expression of Kv1.3 in cancer lines/tissues with respect to healthy ones; however, other factors are also expected to contribute. In fact, we observed that a synergistic action of channel expression and of an altered redox state accounts for apoptotic sensitivity of the pathological cells (743). The fact that clofazimine is already used in the clinic for the treatment of, e.g., leprosis (1059) and shows an excellent safety profile highlights the chance of exploiting mito Kv1.3 targeting for therapy. A positive correlation between expression of Kv1.3 in different cancer cell lines and the sensitivity of these cells to clofazimine as death-inducing agent has been observed (742).
In summary, inhibition of mito Kv1.3 results in the generation of ROS and facilitation of cytochrome c release. It may therefore be expected that inhibition of other Kv family channels present in the IMM of other cell types would promote apoptosis in a similar manner. It is to note that indeed the action of Bax is not restricted to Kv1.3: Kv1.1, mito Kv1.5, and mito KCa1.1 can also interact with Bax (253, 744). Indeed, in macrophages, which express both Kv1.3 and Kv1.5 in their mitochondria, downregulation of both channels is required to prevent staurosporine-induced apoptosis (741). Thus mitochondrial voltage-gated K+ channels beside those of the plasmamembrane are emerging as therapeutic targets for oncological diseases (740).
6. TASK-3 two-pore potassium channel: role in apoptosis
Recently TASK-3 (TWIK-related Acid-Sensitive K+ channel-3; KCNK9), a two-pore potassium channel known to reside in the plasma membrane, was identified in mitochondria of melanoma and keratinocyte cells by immunochemical and molecular biology methods (1099). A strong intracellular TASK-3 positivity was found also in healthy intestinal epithelial cells (693). PM TASK-3 is characterized by a small, 18-pS conductance in 140 mM symmetrical potassium at hyperpolarizing voltages, by an inward rectification at single-channel level, by high open-channel noise, and by burst-like behavior with the open probability increasing at depolarizing potential (68). In another study, in the presence of 2 mM Mg2+ on the extracellular side, the single-channel unitary conductances were 27 and 17 pS at −60 and +60 mV, respectively (648). The channel is maximally active at pH 7.4 and is inhibited with a pK value of 6.7 (375). Zinc at 100 μM concentration selectively blocks TASK-3 over TASK-1 in physiological conditions but is less effective at acidic pH and high extracellular potassium (265). Furthermore, TASK-3 is sensitive to other divalent cations (902) and to ruthenium red with an IC50 of 0.7 μM (294). PM-located TASK-3 has an important role in the regulation of apoptosis and in tumorigenesis (973, 990). A TASK-3 knockdown melanoma cell line displayed altered size, DNA content, and morphology. Furthermore, reduced expression of TASK-3 resulted in compromised mitochondrial function, suggesting that this two-pore potassium channel is functionally present in the mitochondria of the WM35 melanoma cells, and its function is essential for the survival of these cells (692). Interestingly, external ROS-induced mitochondrial ROS production was significantly lower in the knock-down cells with respect to control ones. Whether this alteration of ROS production is linked to resistance to apoptosis of cancer cells, as well as the contribution of the putative mitochondrial TASK-3 versus the PM-located channel in determinaing resistance to apoptosis still remains to be understood.
The mitochondrial localization of functional TASK-3 has been recently demonstrated by the group of Adam Szewczyk, using immunofluorescence and patch clamp on mitoplasts obtained from a human keratinocyte HaCaT cell line (1306). The channel conductance was 83 pS at positive voltages and 12 pS at negative voltages in symmetric 150 mM KCl and activity was voltage-dependent with the probability of channel opening (Po) increasing at positive potentials. Lidocaine (1 mM) and acidic pH (6.2), known to modulate PM TASK-3, completely blocked channel activity. These characteristics of mitoTASK-3 are compatible with those of PM TASK-3.
7. pH-sensitive potassium channel
Recent single-channel studies (626) on the IMM of mitochondria from embryonic rat hippocampus revealed the presence of a previously undescribed outwardly rectifying, voltage-dependent potassium channel displaying a conductance of 68 pS at positive voltages and of 10 pS at negative voltages in symmetric 150 mM KCl. The channel was insensitive to activators and inhibitors of other known mitochondrial potassium channels, including paxillin and those affecting two-pore channels and inward rectifying channels (lidocaine and tertiapin, respectively), but was regulated by pH: a decrease of the bath solution pH from 7.2 to 6.2 resulted in complete block of the channel. The lack of sensitivity to known potassium channel blockers did not allow to hypothesize a molecular candidate. The physiological function and tissue distribution of this channel are also unclear at the moment.
B. Calcium Channels
1. Mitochondrial calcium uniporter: molecular identity, electrophysiology, and physiology
Matrix-negative voltage-driven Ca2+ uptake across the IMM is carried by the mitochondrial Ca2+ uniporter (MCU). Furthermore, a non-MCU Ca2+ channel and a mitochondrial ryanodine receptors (mito RyR) have been also proposed to contribute to this process (for recent reviews, see Refs. 1021, 1066). As to calcium efflux, Na+-dependent (mito NCX) and Na+-independent Ca2+ efflux systems have been identified along with the mitochondrial permeability transition pore, which allows exit of various ions and metabolites (see sect. IIIF) (e.g., Refs. 290, 355). The long-studied calcium uniporter characterized in bioenergetic studies (1021, 1038) was shown to correspond to a highly calcium-selective ion channel by the seminal work of Kirichok et al. (658). Mitoplasts from COS-7 cells, obtained by osmotic swelling, were patch-clamped in the whole-mitoplast configuration, revealing an inwardly rectifying current (IMiCa) with an amplitude depending on Ca2+ concentration and a density of 55 ± 19 pA/pF at −160 mV with 100 μM [Ca2+]c. The current exhibited a high half-saturation at 20 mM Ca2+ and lacked typical Ca2+-dependent inactivation. The authors determined also the single-channel conductance, ∼6.5 pS in 105 mM Ca2+, leading to an estimate of 10–40 MCU/μm2. Interestingly, the open probability declined rapidly with depolarization, suggesting a feedback control on calcium uptake into mitochondria. Proof in favor of the identification of IMiCa being transported by the MCU was obtained by its inhibition with 200 nM ruthenium red or even lower concentrations of ruthenium 360, two classical blockers of the MCU. Furthermore, IMiCa displayed a selectivity for divalent cations similar to that of the uniporter and was sensitive to cytoplasmic Mg2+ (658). In a recent study the whole-mitoplast MiCa current was found to be different in mitochondria isolated from different types of tissues, but the mechanism underlying these differences was not investigated (397). A priori, posttranslational modifications and/or association with regulatory subunits might account for differences in activity (changes in protein expression level, which was not determined, cannot be excluded either).
In another study, three distinct activities with different conductances (11, 23, and 80 pS in 105 mM Ca2+) and different sensitivity to ruthenium red have been recorded in mitoplasts from endothelial and HeLa cells (587, 188). Recording of the mitochondrial megachannel, expected to be activated at the high calcium concentrations used (see below), was avoided in these studies either by the use of MMC inhibitor cyclosporin A (Bodnarenko; Ref. 188), or, apparently, by preparation of mitoplasts with the French press instead of hypoosmotic shock (397). Co-existence of multiple pathways for mitochondrial calcium uptake and/or the presence of tissue-specific channel activities is suggested by these results (see also Ref. 556). The characterization of MCU as a calcium-selective channel as well as the elaboration of MitoCarta, a compendium of mitochondrial proteins (953), allowed the long-sought molecular identification of the mitochondrial Ca2+ “uniporter” (120, 319). In this case the protein seems to be a specifically mitochondrial one with no predicted localization elsewhere within the cell. Silencing of the gene encoding for a coiled-coiled protein Ccdc109a resulted in a reduced calcium uptake, while its overexpression enhanced calcium entry into the matrix. Importantly, recombinant proteins obtained either by expression and purification from E. coli or by an in vitro transcription/translation system gave rise in both cases to the activity of a small, 6–7 pS (in 105 mM calcium) calcium-permeable channel when studied in BLM experiments (allowing a lower signal to noise resolution with respect to path clamp) (319), with characteristics compatible with those described for MCU activity studied in mitoplasts by direct patch clamping (658). Application of the in vitro expression system allowed us to avoid even trace contamination by other membrane proteins, which might have eventually led to MCU-independent channel activity. Further proof in favor of the identification of Ccdc109a as MCU was obtained by inhibition of the current with ruthenium red and gadolinium and by loss of channel activity and reduced calcium uptake upon mutation of two negative charges (D260 and E263) in the putative, highly conserved pore loop region (120, 319). Mutation of S259 to alanine resulted in loss of sensitivity to ruthenium red in calcium uptake assays (120). Nonetheless, it must be m