Mitochondrial Ion Channels

Abstract

In work spanning more than a century, mitochondria have been recognized for their multifunctional roles in metabolism, energy transduction, ion transport, inheritance, signaling, and cell death. Foremost among these tasks is the continuous production of ATP through oxidative phosphorylation, which requires a large electrochemical driving force for protons across the mitochondrial inner membrane. This process requires a membrane with relatively low permeability to ions to minimize energy dissipation. However, a wealth of evidence now indicates that both selective and nonselective ion channels are present in the mitochondrial inner membrane, along with several known channels on the outer membrane. Some of these channels are active under physiological conditions, and others may be activated under pathophysiological conditions to act as the major determinants of cell life and death. This review summarizes research on mitochondrial ion channels and efforts to identify their molecular correlates. Except in a few cases, our understanding of the structure of mitochondrial ion channels is limited, indicating the need for focused discovery in this area.

1. INTRODUCTION

1.1. Historical Perspectives

Interest in mitochondria has risen and fallen over the past century in the quest to describe the fundamental processes of the cell. In many cases, the significance of claims about the role of mitochondria is realized only after a period of intense criticism, gradual accumulation of evidence, and later recognition that part or all of the original conjecture was correct. This began in 1890 with Altmann’s (1) description of mitochondria as the fundamental living elements of the protoplasm. He noted the similar morphology and staining characteristics of bacteria and the grains and filaments of cells (mitochondria), which he called bioblasts (bio, “life”; blasts, “germs”), and proposed that these bioblasts were the centers of biological function. Although he was roundly criticized at the time for his enthusiastic view of mitochondria, it seems that the next 100 years largely proved his point.

Novel functional roles for mitochondria often survive in modified form after the initial hypothesis has been overstated. The suggestion that mitochondria may be responsible for genetic inheritance came prematurely and was supplanted by recognition that the nucleus contained the bulk of the genetic material, only to be resurrected as our understanding of the mitochondrial genome and maternal inheritance emerged. Based on the observation that mitochondria can take up enormous amounts Ca²⁺, researchers initially proposed that mitochondria could be the main intracellular storage organelles involved in Ca²⁺ cycling. This suggestion was superceded by the elucidation of the specialized role of the endo-(or sarco)plasmic reticulum. However, recent experiments have rejuvenated interest in the role of mitochondria in fast Ca²⁺ responses (2).

Other roles for mitochondria have only grown in importance since they were originally suggested. Mitochondria’s participation in oxidation-reduction reactions and cellular respiration was proposed early in the twentieth century, and the mechanisms underlying mitochondrial bioenergetics have been elucidated through a steady series of discoveries and conceptual realignments, including the description of the Krebs cycle, the concept of oxidative phosphorylation, and the chemiosmotic hypothesis (3, 4).

The vectorial movement of ions is central to the chemiosmotic theory, and the four main postulates are as follows: (a) H⁺ translocation down its electrochemical gradient across the mitochondrial inner membrane is reversibly coupled to ATP phosphorylation through the ATP synthase (F₁F_O ATPase). (b) The flow of electrons down the respiratory chain is coupled to H⁺ pumping from the matrix to the intermembrane space to establish the protonmotive force. (c) Exchange-diffusion carrier proteins are present on the inner membrane to transport metabolites and selected inorganic ions into and out of the matrix. (d) The mitochondrial inner membrane is generally impermeable to ions other than H⁺.

The last point applies only to mitochondria with tight coupling between oxygen consumption and the phosphorylation of ADP. The extent of coupling varies with membrane leakiness. An increase in the ion permeability of the mitochondrial inner membrane can be induced artificially with protonophoric chemical uncouplers, by a change in metabolic demand [through the stimulation of proton flux through the ATP synthase by ADP and phosphate (P_i)], or in response to increased flux through mitochondrial ion channels (e.g., by Ca²⁺ influx via the uniporter). The energy dissipated by ion flux has a depolarizing influence on mitochondrial membrane potential (ΔΨ_m), which stimulates NADH oxidation, proton pumping, and respiration. An increase in NADH production is required to compensate for higher rates of respiration and to avoid a mismatch in energy supply and demand.

To understand mitochondrial ion movements and their consequences, mitochondria must be considered as a subsystem of the cell’s integrated processes, which include ion transport across the plasmalemma, intracellular Ca²⁺ cycling, and the mechanical or synthetic work of a given cell type. Only then can one appreciate the central role of mitochondria in maintaining the pseudohomeostasis known as normal cell physiology and understand how the failure of mitochondrial function can lead to cell death through a catastrophic necrotic event, autophagy, or the orchestrated process of apoptosis. Recognition of the role of mitochondria in cell life and death in the production of reactive oxygen species (ROS) and nitrogen species and in aging and disease has fueled a resurgence of mitochondrial interest in recent years. Mitochondrial ion channels, although still a nascent subject of investigation, appear to play a fundamental role in most of these functions.

1.2. Ion Transport Across the Inner Membrane: Background

Early observational studies of mitochondria in living cells suggested that mitochondria were sensitive to changes in osmolarity, suggestive of ion and water transport across the mitochondrial compartment. In 1915, Lewis & Lewis (5) reported that mitochondria in cultured embryonic chick cells could reversibly swell and contract in response to changes in the pH or osmotic pressure of the medium. These researchers also observed presciently that the morphology of mitochondria was extremely plastic (undergoing fission, fusion, elongation, etc.) and that mitochondrial structure degenerates in the presence of noxious agents, including hydrogen peroxide.

Other studies showed that a tissue could accumulate potassium against its concentration gradient in a metabolism-dependent manner, which supported the concept of active transport of ions. This in turn led investigators to study how isolated mitochondria actively accumulate ions. Lehninger (6) reviewed early work correlating mitochondrial electrolyte and water movement with oxidative phosphorylation and also discussed the behavior of mitochondria as “osmometers.” The technique of measuring light-scattering changes as an indicator of mitochondrial volume has since been extensively employed to assay the activity of mitochondrial ion channels.

The binding (or uptake) of Ca²⁺ to mitochondria was first recognized in the 1950s. Work by Chance & Saris reported that energized mitochondria took up Ca²⁺ to uncouple mitochondria transiently (reviewed in Reference 7). Mitochondria eventually irreversibly uncoupled and discharged Ca²⁺ when a certain threshold level of Ca²⁺ was exceeded [now understood as the mitochondrial permeability transition (MPT)]. The ability of mitochondria to take up large quantities of Ca²⁺ in a manner that depended on the presence of the respiratory substrates ATP, P_i, and Mg²⁺ was reported soon after. An elegant study by Chance (8) in 1965 demonstrated that Ca²⁺ (a) induces a decrease in pH in the suspending medium, (b) stimulates respiration, (c) causes a cycle of swelling and contraction of the mitochondria, (d) oxidizes NADH, and (e) alters the redox state of respiratory chain carriers with millisecond response times (t1/2 ∼ 70 msec). Additionally, Chance noted a crossover point in the redox chain between cytochromes b and c. This effect on the respiratory chain was relieved by the addition of 1.8 mM P_i, which also enhanced the maximal Ca²⁺-stimulated rates of respiration and proton accumulation. Small additions of Ca²⁺ yielded cycles of oxidation and reduction. However, when the total amount of Ca²⁺ added exceeded 300 µM, irreversible, large-amplitude swelling occurred and was accompanied by sustained oxidation of the pyridine nucleotide pool and cytochrome b (8). These early studies have inspired abundant interest in the role of mitochondrial Ca²⁺ uptake in cell physiology.

Although the relative impermeability of the mitochondrial inner membrane to ions is a basic tenet of the chemiosmotic hypothesis, Mitchell recognized that several modes of ion transport are present, and he referred to them as symporters, antiporters, and uniporters (discussed in Reference 4). Symporters cotransport multiple ions (or an ion and a metabolite) in the same direction across the membrane (e.g., the mitochondrial P_i/H⁺ carrier or the plasma membrane Na⁺/glucose transporter), often utilizing the asymmetric electrochemical ion gradient to drive the transport in a thermodynamically favorable direction. Antiporters exchange ions on different sides of the membrane and can be either electroneutral (e.g., the Na⁺/H⁺ antiporter of the mitochondrial or plasma membrane) or electrogenic (e.g., the Na⁺/Ca²⁺ exchanger of the plasma membrane). For electrogenic transporters, ion flux is driven by both the electrochemical gradients of the transported ions and the membrane potential. In a uniporter, ions flow electrophoretically down their electrochemical gradient, with transport rates in the range of 10⁴–10⁶ ions s⁻¹ (e.g., the ion channels of the plasma membrane, the mitochondrial Ca²⁺ uniporter, etc.). With advancements in the ability to record ion channel activity in membranes and bilayers, the concept of the uniporter has evolved into the study of mitochondrial ion channels.

The large electrical driving force for ion movement (∼180 mV) across the mitochondrial inner membrane strongly favors ion flow through any open ion channel, unless there is an equal and opposite chemical gradient for the ion (which is normally never the case for mitochondria). The theoretically enormous accumulation of cations in the mitochondrial matrix is limited by the restrictions imposed by the mitochondrial membrane–delimited compartment: Charge neutrality must be maintained, and changes in the osmolarity of the matrix have limits. For example, the entry of a cation through an open channel depolarizes ΔΨ_m and stimulates respiration and proton pumping. This increases the pH gradient component of the protonmotive force and also balances the positive charge entering with the ejection of a proton. If the cation accumulation in the matrix is accompanied by either the movement of a permeant weak acid (such as acetate) or the electroneutral uptake of P_i driven by the proton gradient, the pH change is prevented, and there is a net increase in matrix osmolarity and an increase in mitochondrial volume. Mitochondrial swelling can be counteracted by the concomitant stimulation of cation/H⁺ antiporters. A futile cycle of K⁺ influx [through a mitochondrial ATP–sensitive K⁺ channel (mitoK_ATP)] and efflux (via the K⁺/H⁺ antiporter) may regulate physiological mitochondrial volume (9) (see Section 3.4.1). Conversely, the efflux of anions through an inner membrane anion channel [e.g., inner membrane anion channel (IMAC); see Section 3.5.1], coupled with cation efflux, may mediate mitochondrial contraction (10).

If the mitochondrial permeability transition pore (PTP) is not activated, mitochondria can take up large amounts of Ca²⁺. This capacity as a Ca²⁺ sink is facilitated by the fact that P_i accumulates in parallel and can reversibly precipitate with Ca²⁺, effectively lowering the free Ca²⁺ level to the µM range and minimizing osmotic effects. Ca²⁺ uptake is indirectly coupled to proton fluxes through direct H⁺/Ca²⁺ exchange in some tissues, through the simultaneous uptake of P_i with H⁺, and through Na⁺/H⁺ exchange secondary to the action of the Na⁺/Ca²⁺ exchange Ca²⁺ efflux pathway. Mitochondrial Ca²⁺ overload can also result in rapid Ca²⁺ efflux from the matrix through PTP opening or PTP-independent pathways. In the case of PTP activation, matrix constituents with mass <∼1.5 kDa exit the mitochondrion, and the ion permeability barrier of the inner membrane is lost, effectively short-circuiting proton-coupled energy transduction and dissipating ΔΨ_m.

The other principal rapid cation uptake pathway that has been studied extensively is the K⁺ uniporter. K⁺ conductance can be substantial in energized mitochondria, but recent work has indicated that energy depletion (e.g., ischemia) or treatment with K⁺ channel openers activates fast K⁺ uptake. K⁺-selective ion channels may account for this response and are currently the subject of intense investigation because they confer protection against ischemia- or oxidative stress–mediated cell injury and apoptotic cell death (Section 3.4).

Researchers have also recognized for some time the transport of anions across the inner membrane in concert with cation movement. Mitchell & Moyle (11) reported that anions, including P_i, succinate, and malonate, accelerated the rate of decay of the pH gradient induced by a pulse of oxygen. This suggested the presence of anion transport systems coupled to proton movement, leading to the identification of the anion/metabolite-coupled cotransporter family. Inner membrane anion uniporters have been less well studied, but in the 1980s, an IMAC was postulated to account for anion-selective mitochondrial swelling responses (12). Subsequently, several anion channels were found in single-channel patch-clamp studies of mitoplasts (Section 3.5). Furthermore, investigators have identified certain mitochondrial membrane proteins that display anion channel activity under some conditions, including the mitochondrial uncoupling protein (UCP) (Section 3.1).

The best-studied mitochondrial ion channel in terms of structure and function is the voltage-dependent anion channel (VDAC), which is abundant in the mitochondrial outer membrane. VDAC is the primary route of entry and exit of metabolites and ions across the outer membrane and is a component of the mitochondrial contact site between the outer and inner membrane. Recent studies have focused on the regulation of outer membrane permeability as a physiological or pathophysiological control mechanism (see Section 2.1).

Table 1 summarizes known or postulated mitochondrial ion channels. The properties of mitochondrial ion channels are described below.

Table 1.

Mitochondrial ion channels. A summary of mitochondrial ion channel types identified either in isolated mitochondria, proteolipid bilayers, or patch-clamp experiments. Detected single-channel conductances have been tentatively assigned to a given class of channel, but these assignments have not been unequivocally proven. See text for details and abbreviations

Location	Type		Conductance (∼150 mM salt)	Modulators/inhibitors	Putative role(s)	Selected refs.
Outer membrane	VDAC (porin)		0.5–4 nS	Bax/Bak/Bcl-xL/Bcl-2, TOM20, Ca2+, pH, ΔV, NADH, VDAC modulator	Metabolite transport, cytochrome c release/apoptosis, PTP complex	16
	TOM40 (PSC)		0.5–1 nS	Signal peptides	Protein transport	54
	MAC (BH proteins)		2.5 nS	Bax/Bak	Cytochrome c release/apoptosis	37
	Miscellaneous		10–307 pS	ΔV(for > 100 pS)	-	31
Inner membrane	Ca2+ uniporter		6 pS	Divalents, nucleotides, RuRed, ryanodine	Ca2+ uptake	94
	PTP	MCC	0.03–1.5 nS	Ca2+, ΔV, signal peptides, CsA	Protein transport	174
		MMC	0.3–1.3 nS	CsA, pH, Ca2+, thiols, Bax, ANT inhibitors	Necrosis, apoptosis	178
	UCP		75 pS	Fatty acids	Thermogenesis	88
	KCa		295 pS	Ca2+, ΔV, ChTx, IbTx	Volume regulation	137, 138
	KATP		9.7 pS	ATP, GTP, palmitoyl CoA, Mg2+, Ca2+	Volume regulation, protection against apoptosis/ischemic injury	116, 119
	KV1.3		17 pS	Margatoxin	Cell death	147
	IMACs		45, 450 pS	ATP	Volume regulation (in yeast)	155
			15 pS(LCC)107 pS (centum pS)	Mg2+, pH, Pi, thiols, DIDS,cationic amphiphiles	Volume regulation	149, 150, 153

2. MITOCHONDRIAL ION CHANNELS OF THE OUTER MEMBRANE

2.1. The Voltage-Dependent Anion Channel

VDAC was originally purified and reconstituted in phospholipid bilayers in 1975 (13). A highly conserved protein with homology to bacterial porins, it forms an outer membrane pore with a diameter of 2.5–3 nm in the full conductance state (approximately 3 nS in 1 M NaCl). The channel is thought to be a polypeptide monomer that forms a beta barrel with 13 β-strands and one α-helix (14). It is partially anion selective in the open state, allowing the passage of metabolites such as ATP, ADP, and P_i, but it also permits the free diffusion of cations, including Ca²⁺, K⁺, and Na⁺ (14). Nonelectrolytes (<3 nm diameter; <6 kDa) can also pass through the open channel. When the channel is in the closed state, the pore is still conductive but constricts to ∼1.8 nm, at which point it favors cation permeability. The conductance to K⁺ decreases by ∼60%, and Ca²⁺ ions still permeate, but ATP flux is blocked, owing to a shift in the electrostatic profile and selectivity of the pore (15).

VDAC gating is highly voltage dependent, with peak currents near 0 mV and a dramatic reduction in channel open probability at positive or negative voltages in the range of 20–40 mV (15). This gating property may be physiologically important; recent evidence suggests that there is a significant membrane potential across the outer membrane (16). Although the large conductance of VDAC would tend to dissipate any membrane potential, the partial anion selectivity and the fact that fixed anionic charges on immobile macromolecules are asymmetrically disposed across the outer membrane would result in differential mobility of ions and the establishment of a Donnan potential. Assuming that protons rapidly equilibrate across the outer membrane, investigators hypothesized that if a pH gradient were detected, it could be justified only if a membrane potential existed. In fact, Porcelli et al. (17) detected a pH gradient by using pH-sensitive fluorescent proteins targeted to the intermembrane space. They reported a pH gradient of ∼0.7 between the cytosol (pH 7.59) and the intermembrane space (pH 6.88) and calculated that the outer membrane potential necessary to support such a gradient would have to be 43 mV. If this is true, then VDAC would tend to be in the closed (low anion permeability) state in intact cells, raising the question of whether VDAC regulates mitochondrial metabolism by limiting metabolite flux in vivo.

Notably, several cytosolic factors decrease the open probability of VDAC channels reconstituted in bilayers. For example, the reduced pyridine nucleotides NADH and Mg²⁺-NAD(P)H decrease ADP flux through VDAC, with K_is in the µM range (18). Moreover, endogenous inhibitors of VDAC, including an intermembrane space protein (19) and tbid (a cleaved form of Bid) (20), have been reported. Using a novel microfluidic and surface-plasmon resonance method, Roman and coworkers (21) recently surveyed a vast number of epitopes expressed from a liver cDNA library for these epitopes’ ability to bind to reconstituted VDAC. Roman et al. were able to identify (a) positive interactions with ∼40% of known VDAC binding partners and (b) 55 novel interactions, some of which inhibited VDAC in swelling assays.

VDAC is phosphorylated by protein kinase A (PKA) (22), protein kinase Cε (PKCε) (23), and tyrosine kinases (24). The phosphorylation of mitochondrial proteins by PKCε was correlated with the inhibition of PTP opening and may be related to protection against ischemic injury (23). VDAC is also a binding site for hexokinase II (HKII), and the dissociation of HKII from mitochondria has been correlated with induction of the MPT and apoptosis. The binding of HKII to VDAC appears to prevent VDAC closure (25), whereas hexokinase I may have the opposite effect (26).

Whether VDAC modulation of metabolite flux is a physiological control mechanism remains to be determined, and little is known about VDAC conductance in intact cells. However, recent reports indicate that a large outer membrane channel can be activated in intact neurons by a train of synaptic action potentials (27) or by a proteolytically cleaved form of Bcl-xL and that NADH or hypoxia can inhibit these channels (28). Additionally, ethanol metabolism, by shifting the cytosolic NADH/NAD⁺ redox potential to a more reduced state, may inhibit ADP flux across the outer membrane (29). Also, one study has reported that the molecular cut-off size and conductance of VDAC are Ca²⁺ dependent at submicromolar Ca²⁺ concentrations (30). Ca²⁺ pretreatment of partially permeabilized cells enhanced subsequent ATP uptake, implying VDAC-mediated regulation of metabolism. Notably, the large VDAC conductances seen in bilayers are not typically observed in patch-clamp experiments of intact mitochondria; rather, a variety of conductances in the range of 10–307 pS have been reported (31).

The implications of changes in outer membrane permeability on metabolism and cell death have been recognized in the context of various pathological states (29, 32). Apoptosis is initiated by the release of proapoptotic factors from the mitochondria, including cytochrome c, Smac/Diablo, HtrA2/Omi, AIF and Endo G, and others (reviewed in Reference 33). For some noxious stimuli [e.g., ceramide (34)], mitochondrial outer membrane permeability increases prior to, or in the absence of, an immediate change in inner membrane permeability (35, 36). At least three different mechanisms to explain the loss of intermembrane factors such as cytochrome c have been hypothesized: (a) physical rupture of the outer membrane as a result of mitochondrial swelling (usually linked to PTP opening), (b) a modification of VDAC structure, perhaps induced by proapoptotic proteins, such that the pore size increases enough to allow cytochrome c release (36), and (c) the formation of a new pore as a consequence of oligomerization and membrane insertion of proapoptotic proteins (37).

An oft-repeated model of the structure of the PTP (Section 3.6) includes, at a minimum, the adenine nucleotide translocase (ANT), VDAC, cyclophilin, and the F₁F_O ATPase (38). VDAC has been implicated both because it coimmunoprecipitates with the other proteins present at mitochondrial contact sites and because it is the main pathway for molecular transport across the outer membrane under normal conditions. Moreover, reconstitution of (at least) VDAC, ANT, and cyclophilin in phospholipid bilayers can form cyclosporin-sensitive channels (38). In addition, several common factors—for example, NADH and HKII binding—inhibit both the reconstituted VDAC channel and the PTP. The precise requirement for VDAC and the three-dimensional orientation of the components in the complex that form the pore are presently unknown, but possibly either a long pore spanning the two mitochondrial membranes forms from the association of VDAC with ANT or mitochondrial swelling simply ruptures the outer membrane (the role of VDAC is unclear in this case). In evaluating whether the current model of the PTP is correct (see Section 3.6), recent data demonstrating that the permeability transition can still be observed in ANT knockout mice (39) must be taken into account (40). Furthermore, large multiconductance channels in the inner membrane can still be observed in mitochondria devoid of VDAC (41), and Ca²⁺-induced MPTs with properties indistinguishable from controls occurred in mitochondria from VDAC1^−/− mice (42).

An ongoing debate is whether apoptosis is favored by VDAC opening or closure (14). Capano & Crompton (43) showed that the proapoptotic BH3-homology protein Bax coimmunoprecipitates with VDAC. In reconstitution studies in liposomes and phospholipid bilayers, Shimizu and coworkers (36) have reported that Bax can induce a novel high-conductance state of VDAC that permits cytochrome c to escape from the intermembrane space. Similarly, a recent report demonstrated that Bax and tBid increased the conductance of VDAC in planar bilayers and enhanced cytochrome c translocation from the cis to the trans chamber (44). Interestingly, VDAC phosphorylation by PKA inhibited the effects of Bax and tBid (45). Hexokinase II dissociation from VDAC and enhanced VDAC permeability as a trigger for cell death have been reviewed recently (32, 46).

In contrast, others have observed no effect of monomeric or oligomeric Bax on VDAC gating or conductance (20), although the antiapoptotic protein Bcl-x_L promoted VDAC opening and helped to maintain the rate of ATP/ADP exchange across the outer membrane (47). In addition, the proapoptotic tBid was found to induce VDAC closure (20), and other studies have also linked the induction of apoptosis with a decrease in outer membrane permeability to metabolites (48).

2.2. The Mitochondrial Apoptosis–Induced Channel

Although debate is ongoing about whether VDAC constitutes a route for cytochrome c release, another candidate has emerged as a possible outer membrane permeation pathway. A recent study found that a novel high-conductance channel was frequently observed in proteoliposomes prepared from mitochondrial outer membranes that were isolated from cells undergoing apoptosis induced by growth factor withdrawal (49). This mitochondrial apoptosis–induced channel (MAC) had a fully open conductance of ∼2.5 nS (∼5 nm diameter) and multiple substates (37). Unlike VDAC, it was partially cation selective (P_K+:P_Cl− = 3) and was voltage independent. Because both the antiapoptotic Bcl-2 (50) and proapoptotic Bax (51) proteins can form channels when reconstituted in lipid bilayers, their role as mediators/regulators of MAC was investigated. MAC was not detected in proteoliposomes from cells overexpressing the antiapoptotic Bcl-2 protein. In contrast, the expression of human Bax in a VDAC-less yeast strain resulted in the appearance of channels identical to MAC (49). Moreover, the increased probability of observing MAC correlated with the translocation of Bax to the outer membrane and cytochrome c loss from the intermembrane space, and interventions that prevented Bax activation (e.g., Bcl-2 overexpression) reduced MAC formation. In agreement with MAC constituting a cytochrome c efflux pathway, cytochrome c may modify the conductance of reconstituted MAC. Bax and Bak appear to be functionally redundant with respect to the formation of MAC because MAC can still be observed in Bax knockout cell lines but not in Bax + Bak double knockouts.

The role of other outer membrane proteins in the formation of MAC and the precise time course of MAC activation in intact cells remain to be elucidated. With respect to the latter, Guihard et al. (52) detected MAC in mitochondria isolated from apoptotic rat liver only during the later stages of apoptosis, but Dejean et al. (53) argued that this was because of the particular apoptotic stimulus used (which activated the extrinsic pathway of apoptosis).

2.3. Translocase of the Outer Membrane

The insertion and translocation of mitochondrial preproteins encoded by the nucleus require specialized molecular machinery consisting of complexes of proteins on the outer and inner membranes. The constituent proteins of the translocase of the outer membrane of mitochondria (TOM) and its partner on the inner membrane, TIM, can form large conductance channels (54). Early electrophysiological studies of reconstituted mitochondrial membranes showed channels that were blocked by small basic peptides derived from mitochondrial presequences (e.g., the first 12 amino acids of cytochrome oxidase preprotein), and this peptide-sensitive channel (PSC) was proposed to be involved in mitochondrial protein import (55). Subsequent studies have shown that immunodepletion of the preparations with antibodies against TOM40p correlates with loss of channel activity, whereas control antibodies have no effect (56). Furthermore, antibodies against the carboxy terminus of TOM40p altered channel activity (56).

The properties of the channels formed by TIM or TOM proteins are almost identical (54). Both show a full conductance open state of 1 nS and a half-open state of 500 pS, and they have partial cation selectivity (P_K+:P_Cl⁻ = 5) (54). TOM channels show an asymmetric voltage dependence, with closure favored at positive voltages between +20 and +60 mV, but little inactivation at negative voltages. Estimates of the pore diameter based on conductance are in the range of 2.4–2.7 nm, whereas polymer exclusion methods give a molecular weight cutoff of ∼1000 and a dimension of ∼2 nm (54). The slightly larger estimates of pore size based on conductance are complicated by the finding that the channel may consist of a double-barrel structure in which there may be cooperative gating of the two pores.

2.4. Peripheral (Mitochondrial) Benzodiazepine Receptor

Mitochondria contain binding sites for benzodiazepine receptor (BR) ligands that are pharmacologically distinguishable from the central-type receptor. Because these sites were also found in tissues outside the nervous system, they were labeled peripheral benzodiazepine receptors (PBR) (57). The PBR was preferentially enriched in tissues with high rates of oxidative phosphorylation, and membrane subfractionation studies demonstrated that the receptor was mitochondrial (58). The influence of the mitochondrial benzodiazepine receptor (mBzR) on mitochondrial physiology is incompletely understood, but we know that it plays a vital role in cholesterol transport and steroid biogenesis, accounting for the high density of this receptor in the adrenal gland and testes (59). The mBzR resides in the outer membrane as a complex of proteins; the primary component is the 18-kDa isoquinoline carboxamide–binding protein (IBP). As its name implies, IBP has a high affinity for isoquinoline carboxamide ligands like PK11195 (60), and it is found in close association with VDAC, ANT, and other proteins of the mitochondrial contact site. The binding of isoquinoline carboxamides and benzodiazepines is mutually competitive for the mBzR, and it was originally proposed that the binding site for benzodiazepines, such as the prototypical Ro 5–4864 (4′ Cl-diazepam), included the IBP, VDAC, and ANT (61). However, studies of yeast transformed with the IBP, but lacking either the VDAC or ANT protein, show high-affinity binding sites for isoquinoline carboxamides and benzodiazepines that are no different from binding sites of native receptors (62). More recently other proteins have been reported to associate with the IBP; these include the PBR-associated protein 1 [PRAX-1 (63)], the PBR-associated protein 7 [PAP7, or PKA with regulatory subunit 1α (64)], and a 10-kDa protein [pk10 (65)]. Several endogenous ligands (endozepines) interact directly with the mBzR, including an 86-amino-acid polypeptide known as the diazepam binding inhibitor (DBI; homologous to liver acyl CoA–binding proteins) (66, 67) and its truncated derivative triakontatetraneuropeptide (TTN; DBI fragment 17–50) (68) as well as porphyrins (69), which bind to the receptor with nM affinities.

The first step of steroid synthesis and metabolism is the conversion of cholesterol to pregnenolone. This reaction is catalyzed by the cytochrome P-450 side-chain cleavage enzyme (P-450scc) located on the mitochondrial inner membrane. The overall reaction rate is limited by the transport of cholesterol across the outer membrane (70). PBR ligands facilitate this cholesterol transport (71) and enhance steroid production (72), whereas knocking out the IBP by homologous recombination in a cell line decreases steroid production (73). Molecular modeling indicated that the mBzR may form a barrel-like outer membrane cholesterol-permeable transport pathway composed of five membrane-inserted α-helices (74). Transmission electron and atomic force microscopic studies showed that anti-PBR antibodies are organized into complexes of 4–6 proteins (75).

If this structure forms a membrane-spanning pore in the outer membrane, it will need to accommodate the ∼6.4-Å diameter of cholesterol, so ions (with Pauling radii <2 Å) should also be permeable. It is well known that benzodiazepine binding to GABA_A receptors activates Cl⁻ conductance in the central nervous system. However, ion flux through the outer membrane via mBzR has not been demonstrated, although the close association between the IBP and VDAC begs the question. Several early studies showed that anions competitively inhibit ³[H]Ro 5–4864 binding in kidney membranes with a profile correlating perfectly with their permeability through Cl⁻ channels, and inhibitors of anion transport similarly inhibited Ro 5–4864, but not PK11195, binding (76). Other modulators of mBzR ligand binding include lipids (77), phospholipase A2 (78), König’s polyanion, and cyclosporin A (CsA) (79), but their effects on mBzR-associated mitochondrial ion flux have not been investigated. The relationship between mBzR ligands, mitochondrial inner membrane ion channels, and function is discussed below (see Section 3.5.1).

3. MITOCHONDRIAL ION CHANNELS OF THE INNER MEMBRANE

3.1. Proton Leak and Uncoupling Protein

Proton transport across the mitochondrial inner membrane is at the heart of chemiosmotic theory. However, in addition to the passage of protons down the electrochemical gradient via the mitochondrial ATP synthase, other routes of “proton leak” are present. Proton leak can contribute significantly to the control of respiration in mitochondria in state 4 (limited by the availability of ADP) and, to a lesser extent, in mitochondria in state 3 (in the presence of substrate and ADP) (80). The proton conductance of the inner membrane displays an ohmic voltage dependence at low to intermediate ΔΨ_m and a steep increase in conductance (nonohmic) at large ΔΨ_m (>200 mV). The mechanisms responsible for the background conductances are incompletely understood, but several possible explanations include “slips” in the redox-driven proton pump stoichiometries and leaks of protons across the lipid bilayer.

In the context of this review, the contribution of proton-permeable channels must be considered. The pioneering work of Nicholls showed that in brown fat mitochondria, which have high rates of respiration dissipated as heat during nonshivering thermogenesis, GDP and albumin (which binds fatty acids) could regulate the proton leak (81). Later, this proton leak was attributed to a 32-kDa UCP (82). Three isoforms (UCP1, UCP2, and UCP3) are expressed in a variety of tissues (83). Besides the role of UCP1 in heat production, the function of UCP in physiology remains obscure; however, recent interest has focused on the modulation of ROS production, particularly in the context of diseases such as diabetes and obesity. Mild uncoupling of mitochondria, perhaps by altering the rate of production of free radicals, may be protective (84). Alternatively, partial uncoupling by UCP2 and/or UCP3 may optimize the efficiency of energy metabolism (85). Theoretical work suggests that the optimal ATP flow is achieved when the conductance of oxidative phosphorylation is matched to the conductance of the workload, most efficiently at coupling ratios slightly less than one (86).

It is not clear whether the proton and Cl⁻ conductance (87) mediated by UCP involves a carrier or ion channel–type mechanism. A 75pS anion channel was recorded in giant liposomes reconstituted with UCP1 (88), and reconstitution of the individual transmembrane domains of human UCP2 showed that all six transmembrane peptides form helical conformations in lipid model membranes. Only the second transmembrane peptide exhibited voltage-dependent anion channel behavior (89). Other evidence indicates that the anion and proton conductances are separable, depending on the presence or absence of coenzyme Q (90). Alternatively, UCP-mediated uncoupling may involve fatty acid cycling accompanied by protonation/deprotonation to cause net H⁺ transport (85).

As a cautionary note, a leftward shift in the respiration-versus-ΔΨ_m curve (i.e., higher oxygen consumption at any given ΔΨ_m) is often referred to as an increase in “H⁺ leak.” However, the ion selectivity of the leak pathway is usually not examined, so a sizable increase in the conductance of an ion channel with an equilibrium potential different from ΔΨ_m (essentially any ion channel), be it selective or nonselective, would affect the curve similarly, by partially depolarizing ΔΨ_m. For example, mitochondria isolated from postischemic hearts have an increased proton (ion) leak rate that can be reversed by agents that block the PTP, a large nonselective channel (91). In mitochondria from hearts subjected to an ischemic preconditioning (IPC) protocol, GDP reversed the increase in proton leak, implicating UCP as a leak pathway potentially responsible for the cardioprotective effect of preconditioning. Unfortunately, there are no selective pharmacological tools available to block UCP in intact cells or tissues, so GDP and/or albumin sensitivity of isolated mitochondria is the only way to test for the contribution of UCP to a given process.

3.2. The Mitochondrial Ca²⁺ Uniporter

As mentioned above, mitochondrial Ca²⁺ uptake has been recognized for more than 50 years, yet the protein mediating this essential physiological transport process is still unidentified. The mitochondrial Ca²⁺ uniporter (MCU) is present at low density (0.001 nmol mg⁻¹ (92) on the mitochondrial inner membrane, but it has a high V_max of >1000 nmol min⁻¹ mg⁻¹, corresponding to a flux of ∼20,000 Ca²⁺ s⁻¹ (93), in the range of a fast gated pore.

Other evidence also supports a model of the Ca²⁺ uniporter as an ion channel. For example, the dependence of Ca²⁺ uptake on ΔΨ_m is consistent with electrodiffusion through a pore according to the Goldman constant field equation. Moreover, in a recent patch-clamp study of intact mitoplasts (94), an inwardly rectifying, highly Ca²⁺ selective (affinity 2 nM), voltage-dependent Ca²⁺ channel (MiCa) with properties that match the Ca²⁺ uniporter was recorded. Extramitochondrial concentrations of Ca²⁺ in the range of 20–100 µM yielded whole mitoplast currents of 20–30 pA, for a current density of ∼55 pA pF⁻¹ at −160 mV Remarkably, the Ca²⁺ flux did not saturate until [Ca²⁺] exceeded 100 mM, with a K_1/2 of 19 mM. Single-channel recordings in 105 mM Ca²⁺ showed a single-channel amplitude of 0.5–1 pA and a channel density of 10–40 channels µm⁻² (channel conductance ∼6 pS). MiCa had a high open probability at negative potentials and displayed rapid, but partial, Ca²⁺-independent inactivation. It was argued that the much higher fluxes (5 × 10⁶ Ca²⁺ s⁻¹) and K_1/2 recorded for MiCa in comparison to earlier studies of the Ca²⁺ uniporter were due to the inability to clamp the membrane potential in isolated mitochondria with such large currents flowing (94). Divalent ion selectivity for MiCa was also very similar to that reported for the MCU (Ca²⁺ ≈ Sr²⁺ ≫ Mn²⁺ ≈ Ba²⁺; blocked by Mg²⁺), and the channel was sensitive to ruthenium red and Ru360 at low nM concentrations.

Despite a number of attempts to identify the MCU protein over the years, its molecular structure is still unknown. A soluble mitochondrial Ca²⁺-binding glycoprotein, whose Ca²⁺ binding could be inhibited by La³⁺ or ruthenium red, was isolated in 1972 (95), and antibodies against it inhibited Ca²⁺ uptake (96). Later, Mironova and coworkers (97) also isolated a 40-kDa Ca²⁺-binding glycoprotein from beef heart mitochondria and found that a 2-kDa peptide fragment in the preparation could reconstitute a ruthenium red–sensitive Ca²⁺ selective conductance in black-lipid membranes. Antibodies against this peptide inhibited Ca²⁺ uniporter activity in mitoplasts (98). The 2-kDa protein was further purified and formed 20-pS ruthenium red–sensitive channels upon reconstitution (99).

In other studies, antibodies were raised against a mitochondrial Ca²⁺ transporting extract; an antibody that recognized a 20-kDa protein was most effective in inhibiting Ca²⁺ transport (100). An 18-kDa protein fraction with high specificity for ¹⁰³Ru360 was further purified, and antibodies against it inhibited Ca²⁺ uptake in cytochrome oxidase–containing vesicles (101). None of the putative Ca²⁺ transporter proteins have been sequenced thus far.

Gunter and coworkers (102) have reported a second rapid mode of mitochondrial Ca²⁺ uptake (RaM). In liver and heart mitochondria, RaM transports Ca²⁺ very rapidly at the onset of a Ca²⁺ pulse and inactivates during a pulse. As compared with liver mitochondria, heart RaM requires a longer time to reset between pulses when a series of pulses is applied. The rapid uptake is less sensitive to activation by spermine, ATP, and GTP but is inhibited by AMP (103). It is presently unknown whether RaM is mediated by a separate transport protein or if it is a second kinetic mode of the Ca²⁺ uniporter. The proteins involved in RaM-mediated Ca²⁺ uptake have not been identified.

3.3. Ryanodine Receptors

Mitochondrial Ca²⁺ uptake and endoplasmic reticular Ca²⁺ release are inhibited by common pharmacological inhibitors, including ruthenium red and ryanodine. Recent studies have reported that type I (skeletal type) ryanodine receptors (RyR1) are present on the cardiac mitochondrial inner membrane and may participate in either Ca²⁺ influx or efflux (104). Intriguingly, RyR1 immunoreactivity could not be detected in mitochondria from RyR1 knockout mice, providing strong support for the specificity of the antibodies used in these immunoprecipitation studies (104).

Any consideration of the ryanodine receptor as a viable candidate for the protein responsible for Ca²⁺ uniport activity in mitochondria must reconcile the discrepancy between the extremely high Ca²⁺ selectivity of the uniporter (94), which allows for mitochondrial Ca²⁺ uptake at submillimolar concentrations without permitting energy dissipation by monovalent ion flux, and the nonselectivity of ion permeation through ryanodine receptors. In addition, ion channels resembling RyR1 have not yet been reported in intact mitoplast patch-clamp recordings. Hence, further studies will be required to elucidate the role of RyR1 in mitochondrial Ca²⁺ handling.

Depending on the orientation of the Ca²⁺-triggering site of the RyR1, i.e., toward the matrix or toward the intermembrane space, RyR1 may be a candidate either for rapid uptake of Ca²⁺ by mitochondria or for Ca²⁺-triggered Ca²⁺ release from the mitochondrion through a non-PTP mechanism. Ruthenium red–sensitive Ca²⁺ efflux from Ca²⁺-loaded mitochondria, attributed to reversal of the Ca²⁺ uniporter, has been reported (105).

3.4. K⁺ Channels and Protection Against Ischemic Injury

More than 20 years ago, Lamping & Gross (106) demonstrated that pharmacological agents capable of opening K⁺ channels protect hearts against ischemia-reperfusion injury. K⁺ channel openers have come to be viewed as chemical preconditioners, i.e., compounds that can mimic the protective effects of brief cycles of ischemia and reperfusion. Moreover, the finding that K⁺ channel inhibitors such as glybenclamide and 5-hydroxydecanoate could block the protective effects of either IPC or K⁺ channel openers (107) suggested that K⁺ channels were a native trigger and/or effector of preconditioning. Although earlier studies naturally presumed that the target of the K⁺ channel openers was the sarcolemmal ATP–sensitive K⁺ channel (sarcK_ATP), the focus has shifted recently to the mitochondria as the primary target of these compounds.

Plasma membrane K_ATP channels have been extensively studied at the molecular level, and their physiological roles in insulin release and in the modulation of vascular tone are indisputable. However, the presence of high densities of sarcK_ATP in muscle cells has not been adequately explained. The recent availability of gene knockout mouse models in which components of the sarcK_ATP channel have been ablated provides a new opportunity to answer this question. In the mouse, the primary physiological role of sarcK_ATP apparently is to help the animal cope with metabolic stress. Mice that lack the pore-forming subunit of the cardiac sarcK_ATP (Kir6.2) have a severely compromised ability to tolerate ischemia—even short periods of ischemia lead to rapid ischemic contracture of the heart (108, 109). Similarly, when the K_ATP channel is pharmacologically inhibited in mice with HMR1098, a sarcK_ATP blocker, ischemic dysfunction is accentuated. Neither ischemic nor chemical preconditioning can mitigate this effect. Function is also compromised with exercise, a more physiological form of metabolic stress (110). These studies indicate that the mouse is highly dependent on sarcK_ATP channels for survival under conditions of high energy demand.

In contrast, in larger animal species (e.g., rabbits) (111) and in humans (112), sarcK_ATP appears to play a minor role in protecting the heart during ischemia. Selective pharmacological inhibition of sarcK_ATP has little or no effect on infarct size after ischemia and reperfusion or on the cardiac preconditioning response (113). Rather, sarcK_ATP contributes to postischemic electrical dysfunction by increasing the dispersion of repolarization and the heterogeneity of electrical excitability (114, 115). Moreover, the action potential–shortening effects of sarcK_ATP activation during ischemia are not correlated with the extent of protection afforded by K⁺ channel openers (107). Thus, other targets of these compounds, including the mitochondria, have been investigated in the context of protection against metabolic stress.

3.4.1. Mitochondrial K_ATP

In 1991, Inoue et al. (116) reported the presence of ATP-sensitive K⁺ channels in the liver mitochondrial inner membrane, using the direct mitoplast patch-clamp method. The mitochondrial K_ATP channel (mitoK_ATP) had properties similar to the K_ATP channel observed in the sarcolemma of cardiac cells, albeit with a lower conductance. Thus, a link was established between the effects of K⁺ channel openers on mitochondrial function and a specific ion channel target. The effects of K⁺ channel openers on mitoK_ATP have been correlated with protection against ischemic injury in intact hearts (117) and in isolated myocytes (118). The general hypothesis that an increase in mitochondrial inner membrane permeability to K⁺ improves cellular tolerance to ischemia-reperfusion injury has found widespread support in various tissues, including the liver, gut, brain, kidney, and heart.

Although studies of mitoK_ATP in mitoplasts by patch-clamp have been scarce (116, 119), functional evidence supporting mitoK_ATP has been obtained by several methods, including (a) electrophysiological recordings of channels in proteoliposomes reconstituted with mitochondrial membrane proteins, (b) measurements of K⁺ uptake into mitochondria or reconstituted liposomes, (c) mitochondrial swelling assays, (d) changes in mitochondrial redox potential, and (e) alterations in mitochondrial energetic parameters such as respiration and ΔΨ_m. Several groups have also reported the reconstitution of diazoxide-activated, 5-hydroxydecanoate-inhibited ion channels in planar lipid bilayers (for recent reviews of the evidence supporting a specific mitoK_ATP channel in the inner membrane, see References 120–122).

As for most of the mitochondrial channels, the lack of a specific molecular entity associated with mitoK_ATP has largely restricted putative identification to the use of pharmacological agents with known properties. A strong caveat is that some of these agents have substantial nonspecific effects on mitochondria that can alter the response to ischemia-reperfusion injury. A variety of K⁺ channel openers activate mitoK_ATP; these include diazoxide, nicorandil, BMS191095, cromakalim, levcromakalim, EMD60480, EMD57970, pinacidil, RP66471, minoxidil sulfate, and KRN2391. Only the first three show significant selectivity toward the mitochondrial versus the sarcolemmal isoform of the K_ATP channel in cardiac myocytes. Another drawback is that there is only one widely available K⁺ inhibitor that selectively inhibits mitoK_ATP without blocking sarcK_ATP (123), 5-hydroxydecanoate. The classical K_ATP channel inhibitor, glybenclamide, is a sulfonylurea that blocks both the sarcK_ATP and mitoK_ATP isoforms, whereas HMR1098 is usually selective for the sarcolemmal channel. In general, the pharmacological profile of the protective effect is consistent with mitoK_ATP, rather than sarcK_ATP, as the prime target (111).

Recent reports suggest that a number of other compounds—including sildenafil (124), levosimedan (125), YM934 (126), and MCC-134 (127)—modulate mitoK_ATP. MCC-134 could inhibit mitoK_ATP while activating sarcK_ATP. Importantly, MCC-134 prevented diazoxide-mediated protection against simulated ischemia, again supporting the argument that mitoK_ATP and not the sarcK_ATP channels mediated protection (127).

Another emerging area of interest is how signal transduction pathways either activate or are activated by the mitoK_ATP channel. Signaling pathways linked to phosphoinositide hydrolysis, protein kinase C (PKC) activation, or tyrosine kinases are mediators and/or effectors of cellular protection (128). In many cases, the downstream effects of receptor activation can be blocked not only by inhibitors of the kinases but by inhibition of the mitoK_ATP channel (129). This begs the question of whether the channel lies upstream or downstream of the posttranslational modifications mediated by either PKC or other kinases. One plausible link between the activation of mitoK_ATP and signaling would be a change in redox-sensitive pathways as a result of an increase in mitochondrially derived ROS (130). This could be due to an increase in respiratory rate (and consequent leak of electrons to superoxide) induced by the opening of the K⁺ channel. A common effector, glycogen synthase kinase 3β (GSK-3β), has been proposed as the integrator of various preconditioning stimuli, including the actions of K⁺ channel openers (131). Activation of GSK-3β blunts the effects of laser-induced oxidative stress on PTP activation in isolated cardiomyocytes. The mitoK_ATP channel is likely to be both a target and an effector in these pathways. For example, nitric oxide donors (132) and PKC activators (123) can enhance the activation of mitoK_ATP in isolated cardiac cells. In some studies, the mitoK_ATP inhibitor 5-hydroxydecanoate not only prevented IPC when applied during the preconditioning phase but also blocked the protection against infarction when given before the long index ischemia: this was the case for both early and delayed preconditioning protocols (133). More recently, the nitric oxide/cyclic GMP/G kinase signal cascade has been implicated in the activation of mitoK_ATP (134–136).

POSSIBLE STRUCTURE OF mitoK_ATP.

Because the pharmacology of mitoK_ATP parallels certain combinations of known plasma membrane sulfonylurea receptor (SUR) and inward rectifier K⁺ channel (Kir) subunits (191, 192), these proteins have been obvious molecular candidates for the mitochondrial channel. Antibodies against either SUR or Kir6.x (193–200) have been reported to give a strong immunoreactive band in purified mitochondrial membranes. However, there are several criteria to meet when considering these results: (a) The mitochondrial inner membrane preparation must be shown to be completely free of other contaminating components, (b) the subunit of interest is co-enriched with mitochondrial markers, and (c) independent methods must confirm that the protein identified by the antibody is actually the original target (i.e., proteomic analysis should return an SUR or a Kir sequence in the preparation). Although some studies have satisfied the first two criteria, the last has not been achieved, leaving open the question of whether or not mitoK_ATP is a homolog of surface membrane K_ATP channels.

3.4.2. Mitochondrial K_Ca

The second selective K⁺ channel to be identified in the mitochondrial inner membrane by direct patch-clamp of mitoplasts was the Ca²⁺-activated K⁺ channel (mitoK_Ca), detected in mitochondria from human glioma cells in 1999 (137) and in cardiac mitochondria by our laboratory in 2002 (138). In mitoplast patch-clamp experiments, this channel had a conductance of ∼300 pS and was inhibited by the K⁺ channel toxins charybdotoxin or iberiotoxin at nM concentrations, thus resembling the properties of Ca²⁺-activated K⁺ channels found in the plasma membranes of some cells. In partially permeabilized adult cardiomyocytes, these toxins also blunted K⁺ uptake into mitochondria (138). A K_Ca opener, NS-1619, accelerated mitochondrial K⁺ uptake and decreased infarct size in rabbit hearts subjected to 30 min of global ischemia and 2 h of reperfusion (138). We suggested that mitoK_Ca may be activated under pathophysiological conditions that increase mitochondrial Ca²⁺ uptake as a safeguard against excessive mitochondrial Ca²⁺ accumulation and may also play a physiological role to fine-tune mitochondrial volume and/or Ca²⁺ accumulation under conditions of increased cardiac workload. Ca²⁺ activation of this channel would be expected to cause a partial depolarization of ΔΨ_m, which would decrease the driving force for Ca²⁺ entry under conditions of positive inotropic stimulation or ischemia.

Subsequent reports have confirmed that K_Ca channel openers protect hearts against ischemic injury (139, 140). Moreover, similar to the effect of mitoK_ATP activation, mitoK_Ca opening has been implicated in early and delayed preconditioning (141) and may participate in the cardioprotection triggered by ischemia or receptor activation (142, 143). NS-1619-mediated preconditioning can be prevented by blocking K_Ca channels with paxilline or by scavenging ROS during the exposure to the opener (140). The latter effect is reminiscent of the role of mitoK_ATP in preconditioning and implies that mitoK_Ca may be both a trigger and an effector of protection. A recent report (144) showed modulation of mitoK_Ca by PKA pathway but not by the PKC pathway, in contradistinction to the regulation of mitoK_ATP . Although both mitochondrial K⁺ channels appear to have similar effects on mitochondrial function, each has a distinct and nonoverlapping pharmacology (138, 144). This strongly supports the idea that enhanced mitochondrial K⁺ uptake is the common factor associated with resistance to cell injury.

Antibodies against the BK type K_Ca channel cross-react with purified cardiac mitochondrial membranes (138), and a recent immuno-electron microscopy study revealed BK immunoreactivity in brain mitochondria (145). In addition, the β subunit of the BK channel is present in the mitochondria (146). However, it is still not clear whether the mitochondrial protein is identical, or just antigenically similar, to the surface membrane channel.

3.4.3. Kv1.3

Kv1.3, a Shaker-type K⁺ channel, was reported to be present in the mitochondrial inner membrane of Jurkat T lymphocytes (147). In symmetric (134 mM) KCl, channels with a conductance of ∼17 pS were detected in a mitoplast preparation and inhibited by margatoxin, a Kv1.3-selective toxin. The mitochondrial channels had properties that were identical to those of plasma membrane Kv1.3 channels, which are also present in this cell line. The channels were not present in mitoplasts from Kv1.3-null cells (CTLL-2) and were restored when Kv1.3 was overexpressed in a mouse cytotoxic T lymphocyte cell line (CTLL-2). Interpretation of the results is complicated by the concomitant expression of the Kv1.3 channel in the plasma membrane, possibly contributing surface membrane contamination to the preparation. However, the authors provide several lines of evidence supporting a mitochondrial inner membrane localization, including observations of the small conductance channel in patches that also contained the 107-pS channel (see Section 3.5.1) and the PTP (see Section 3.6) and the finding that margatoxin can hyperpolarize ΔΨ_m.

3.5. Anion Channels

3.5.1. IMAC

Mitochondrial swelling induced by cation and anion movements into the matrix compartment has been extensively employed to define ion permeabilities across the mitochondrial inner membrane (4). Following early studies of mitochondrial anion flux from the laboratories of Azzone, Selwyn, and Brierly (reviewed in Reference 12), a series of papers published in the 1980s provided experimental evidence that an inner membrane anion channel (IMAC) was active under certain conditions (e.g., Mg²⁺ depletion, alkalinization) (10, 12, 148). IMAC is permeable to a variety of inorganic (e.g., SCN⁻ > NO₃⁻> Cl⁻ > P_i) and organic (e.g., oxaloacetate²⁻ > citrate³⁻ > malate²⁻ > ATP⁴⁻) anions (10). It is inhibited by protons and Mg²⁺ and by many different cationic amphiphiles, including amiodarone (IC₅₀ ∼ 1 µM), amitriptyline (IC₅₀ ∼ 10 µM), dibucaine (IC₅₀ ∼ 20 µM), propranolol (IC₅₀ ∼ 25 µM), and Ro 5–4864 (4′-chlorodiazepam; IC₅₀ ∼ 34 µM), among others (10). IMAC is modulated by thiol crosslinkers, including mersalyl and N-ethylmaleamide (10). IMAC activity is thought to be kept in check by endogenous inhibitors and by its strong voltage dependence under physiological conditions. However, the inhibition of IMAC by protons and Mg²⁺ is very temperature sensitive, suggesting that IMAC is poised to respond to small changes in pH or Mg²⁺ under normal conditions (148), perhaps as a mechanism for modulating mitochondrial volume.

Direct single-channel patch-clamp methods revealed a number of partially anion-selective conductances in the inner membrane; the most prominent is the so-called centum pS (or mCtS) channel, a strongly voltage-dependent outwardly rectifying current (149–151). In the first patch-clamp study of the mitochondrial inner membrane, a 107-pS channel was recorded with a P_Cl⁻:P_K+ permeability ratio of 4.5; this channel comprises the main background conductance of the inner membrane (149). In this initial study, no pH dependence was observed, so the authors concluded that the 107 pS was not likely to be the single-channel equivalent of IMAC. However, subsequent investigations revealed properties of the ∼100-pS channel in brown fat mitochondria that closely match the IMAC of swelling assays, including inhibition by propranolol, dihydropyridines, and the nucleotide analog cibacron Blue (10, 150). More recently, Schonfeld et al. (152) showed that IMAC was activated by long-chain fatty acids in both intact rat liver mitochondria and in patch-clamp studies of the centum pS channel. Alternatively, a 15-pS anion-selective channel displaying many of the same properties as IMAC (i.e., both are activated by matrix alkalization and blocked by amiodarone, propranolol, and tributyltin) has been described (153).

Of particular interest is the inhibition of IMAC by benzodiazepines (10, 154), considering that the mBzR (see Section 2.4) is present on mitochondrial membranes (albeit on the mitochondrial outer membrane). In studies comparing the effects of Ro 5–4894 and PK11195 on the mCtS and the large multiple-conductance channel (MCC) in mitoplasts, both channel types were blocked by mBzR ligands at concentrations of less than 1 µM (151), whereas central benzodiazepine ligands had no effect at concentrations up to 20 µM. The MCC was also sensitive to inhibition by CsA, but the mCtS channel was not. Another feature discriminating the two channel types was their sensitivity to protoporphyrin IX (PPIX), which is a known ligand of the mBzR (69). The MCC was inhibited by PPIX at low concentrations (IC₅₀ ∼ 24 nM) but was activated by PPIX at high concentrations (EC₅₀ = 244 nM). In contrast, the mCtS was inhibited by PPIX at all concentrations (IC₅₀ ∼ 35 nM) (151). It is still unclear how the modulation of a presumed outer membrane receptor, mBzR, regulates inner membrane conductances such as the MCC and the mCtS. Researchers have reported other anion-selective channels, including ATP-sensitive anion channels found in yeast inner membranes (155). However, nothing is known about their physiological role and distribution in other species.

3.5.2. IMAC and mitochondrial criticality

IMAC and the mBzR play a role in postischemic electrical and contractile dysfunction in the heart (115). Inhibitors of IMAC and/or the mBzR (e.g., 4′-chlorodiazepam or PK11195) prevent or reverse oscillatory (PTP-independent) mitochondrial depolarizations induced by substrate deprivation (156) or oxidative stress (157) in adult cardiomyocytes. The underlying mechanism of the mitochondrial oscillator, explored in both experimental and computational studies (157–159), incorporates the concept of mitochondrial ROS–induced ROS release [originally coined to describe direct laser-induced PTP-mediated depolarization of ΔΨ_m (160)], which triggers IMAC activation in a positive feedback loop (158). Remarkably, the local depolarization of just a few mitochondria in a cardiac cell can lead to cell-wide self-sustaining oscillations of ΔΨ_m (157). Hence, the term mitochondrial criticality was forwarded to refer to the cellular conditions leading up to a breakpoint between stable and unstable ΔΨ_m in the mitochondrial network (159). The approach to the critical state depends on whether mitochondrial ROS production by the respiratory chain exceeds a threshold level in a significant fraction of mitochondria (∼60%) in the network. At the critical point, the weakly coupled fluctuations of individual mitochondria transition into an emergent spatiotemporal pattern of synchronized limit-cycle oscillations. This mitochondrial network phenomenon is tightly coupled to the cardiomyocyte’s electrophysiological response through energy-sensitive sarcK_ATP channels on the surface membrane, leading to scaling of the organelle-level dysfunction to the whole organ during ischemia and reperfusion (115).

3.5.3. CLC, mCLIC

Given the lack of information on the molecular structure of the anion-permeable mitochondrial channels described above, it is worthwhile to examine potential candidate channels known to be localized to mitochondrial membranes. Two classes of chloride channels, the ClC type and the CLIC type, have been reported to be present in mitochondria. The voltage-dependent chloride channel (ClC) family comprises a large class of structurally related membrane proteins with putative chloride channel activities. Although some members of the ClC family (ClC-3, ClC-4, ClC-5) may be targeted to intracellular locations, there is no evidence for mitochondrial localization. However, ClC-Nt, an anion channel isolated from tobacco plants, is enriched in mitochondrial membranes and has been proposed as a candidate for IMAC (161).

CLICs (chloride intracellular channels) represent a new class of intracellular anion channels that have been identified by their homology to the p64 protein. A mitochondrial homolog, mtCLIC, has been identified from differential display analysis of differentiating mouse keratinocytes from p53^+/+ and p53^−/− mice. MtCLIC colocalized with cytochrome oxidase in keratinocyte mitochondria but also was detected in the cytoplasmic compartment. This p53-regulated putative channel has been associated with apoptosis (162) and may exist as either a soluble or transmembrane form that may translocate to the nucleus in response to cell stress (163).

3.6. Permeability Transition Pore

Permeabilization of the mitochondrial inner membrane, first detected as a large-amplitude swelling response in response to a variety of effectors (e.g., thyroxine, Ca²⁺, P_i), has been known for more than 50 years (8, 164). In the 1970s, the MPT was characterized in greater detail (165). It was proposed to involve the activation of a PTP (reviewed in References 40, 166, and 167). The finding that the immunosuppressant drug CsA could inhibit the MPT provided a crucial tool for subsequent investigation (168, 169) and spurred the identification of its mitochondrial binding protein, cyclophilin D (170, 171). Definitive evidence that the PTP was truly an ion-permeable channel was obtained upon the application of the technique of patch-clamping mitoplasts and the detection of MCC (153, 172–175), or mitochondrial megachannels (176–179), which displayed conductances up to ∼1.3 nS that were inhibited by CsA. The PTP allows the passage of ions and metabolites up to ∼1500 molecular weight, for an apparent pore diameter of ∼3 nm. In addition to the most commonly employed trigger, Ca²⁺ overload of the mitochondria, PTP opening is promoted by ΔΨ_m depolarization, P_i, ROS, and thiol modification, and it can be inhibited by adenine nucleotides, Mg²⁺, or matrix protons. Evidence that the ANT is a component of the PTP or even the pore itself was provided by the findings that bongkrekic acid, an inhibitor that stabilizes the ANT in the “m” conformation, inhibits the PTP and that atractylosides, which stabilize ANT in the “c” conformation, activate the PTP (180). Moreover, purified ANT preparations formed large Ca²⁺-activated ion channels upon reconstitution in lipid bilayers (181). Similar channels of 600-pS conductance were obtained when Neurospora crassa ANT was expressed in Escherichia coli and reconstituted, a preparation that is devoid of possible contamination with other mitochondrial membrane components found in heart mitochondrial preparations (182). These channels were inhibited by ADP and bongkrekate, but not by carboxyatractylate, and were modulated by cyclophilin, whereas CsA abolished the cyclophilin effect.

Although the structure of the PTP is widely portrayed as consisting of a multiprotein complex prominently featuring the ANT, VDAC, cyclophilin, the F₁F_O ATPase, hexokinase, and other modulatory proteins, recent observations have challenged this model. The two proteins central to prior models of PTP structure, ANT (39) and cyclophilin D (183–186), have been ablated in recent gene knockout mouse studies. The results support a modulatory (i.e., involving an increase in the threshold for Ca²⁺ activation) rather than an obligatory role for these proteins in the MPT. As discussed above (Section 2.1), data from mitoplast patch-clamp studies and gene-targeted mice also indicate that VDAC is not an obligatory member of the PTP protein complex.

Regardless of the structural details of the PTP, multiple lines of evidence support the idea that in tissues such as the heart, PTP opening occurs only during reperfusion after ischemia (115, 187) and is a major checkpoint on the route toward cell injury and death, although classic stimuli can readily induce apoptotic cell death in MPT-resistant mouse strains (186). However, mitochondrial depolarization in response to stress is not always the result of PTP opening (157). Thus, multiple assays must be used to determine whether a MPT has occurred; these include sensitivity to PTP inhibitors (e.g., CsA or sanglifehrin) and the direct demonstration that small molecules can permeate the mitochondrial inner membrane.

3.7. Connexin 43

Rodriguez-Sinovas et al. (188) recently demonstrated that connexin 43 (Cx43), the protein responsible for forming gap-junctional channels between cells, is present in mitochondrial membrane fractions and is increased in mitochondria from hearts subjected to an IPC protocol. Moreover, the protection afforded by IPC in wild-type mice is absent in heterozygous Cx43-deficient (Cx43^+/−) mice (189). Diazoxide-mediated protection is also lost in Cx43^+/− mice. This effect was tentatively attributed to a decrease in cardiomyocyte ROS production during chemical preconditioning, as measured using a probe that loads (depending on ΔΨ_m) into the mitochondrial matrix, where it is then oxidized (190). Cx43 was specifically enriched in mitochondrial inner membrane fractions (188), and coimmunoprecipitation studies demonstrated an interaction of Cx43 with Tom20, part of the protein import machinery (see Section 2.5). However, it is presently unclear whether Cx43 forms a channel in the inner membrane. Several factors make it unlikely that it forms a pore: (a) Cx43 hemichannels generally do not conduct until the connexon is formed when the hemichannels meet at the junctional membrane, (b) there is no evidence that mitochondrial respiration or ΔΨ_m is influenced by the mitochondrial Cx43 content, and (c) the Cx43 appeared to be in the phosphorylated form, which can induce channel closure. However, the activity of Cx43 may be altered by ischemia and/or reperfusion. Inhibiting the Hsp-90-dependent protein import pathway with geldanamycin prevents the IPC-mediated Cx43 translocation to the mitochondria. But the role of mitochondrial Cx43 in cardioprotection is unclear because geldamycin did not alter the protective effect (188). Furthermore, although geldanamycin abolished diazoxide-mediated protection, there was no correlation between this protection and Cx43 levels (188), nor was there a correlation between the extent of injury and the level of Cx43 in the absence of the chemical preconditioning stimulus.

4. CONCLUDING REMARKS

Ion channels can transport millions of ions per second, and the electrochemical driving forces for ion movement across the inner membrane are enormous. Thus, the knowledge gap regarding mitochondrial ion channel structure is perhaps understandable—these proteins must be present in extremely low abundance or have a very low open probability to maintain the low permeability to ions required to exploit the protonmotive force for ATP generation. Nevertheless, although mitochondrial ion channels may be highly controlled and open only briefly, the significance of their effects cannot be overstated. Mitochondrial ion channels are crucial to the mechanism of energy supply and demand matching and are the decisive factor in determining whether a cell lives or dies.

Mitochondrial ion channels for Ca²⁺, K⁺, or anions have been functionally and pharmacologically characterized at many levels, spanning from the single channel to the intact cell and to whole-organ function. The challenge ahead lies in defining the molecular structures responsible for forming the ion-selective mitochondrial pores. Achieving this goal will undoubtedly spur the development of novel and specific therapeutic agents targeted to the mitochondria. As the organelles responsible for integrating and responding to environmental challenges, mitochondria are the hub of all cellular functions and play a central role as determinants of cell life and death in a variety of pathologies, including acute coronary syndrome, neurodegeneration, cancer, and aging. Perhaps the key to managing these health problems will come from the next phase in the history of mitochondria to arise from mitochondrial ion channel discovery.