The mitochondrion is at the core of cellular energy metabolism, being the site of most ATP generation. Calcium is a key regulator of mitochondrial function and acts at several levels within the organelle to stimulate ATP synthesis. However, the dysregulation of mitochondrial Ca2+ homeostasis is now recognized to play a key role in several pathologies. For example, mitochondrial matrix Ca2+ overload can lead to enhanced generation of reactive oxygen species, triggering of the permeability transition pore, and cytochrome c release, leading to apoptosis. Despite progress regarding the independent roles of both Ca2+ and mitochondrial dysfunction in disease, the molecular mechanisms by which Ca2+ can elicit mitochondrial dysfunction remain elusive. This review highlights the delicate balance between the positive and negative effects of Ca2+ and the signaling events that perturb this balance. Overall, a “two-hit” hypothesis is developed, in which Ca2+ plus another pathological stimulus can bring about mitochondrial dysfunction.
- reactive oxygen species
- free radicals
- permeability transition
Ca2+ AND MITOCHONDRIAL PHYSIOLOGY
Mitochondrial oxidative phosphorylation (ox-phos) is the major ATP synthetic pathway in eukaryotes. In this process, electrons liberated from reducing substrates are delivered to O2 via a chain of respiratory H+ pumps. These pumps (complexes I-IV) establish a H+ gradient across the inner mitochondrial membrane, and the electrochemical energy of this gradient is then used to drive ATP synthesis by complex V (ATP synthase) (136).
Chemically, the stepwise reduction of O2 (O2 → O2−· → H2O2 → OH· → H2O) proceeds via several reactive oxygen species (ROS). These ROS can damage cellular components such as proteins, lipids, and DNA (70), but recent evidence also highlights a specific role in redox cell signaling for mitochondrial ROS (55, 203). In the fine balancing act of aerobic metabolism, mitochondrial ox-phos accomplishes the reduction of O2 to H2O while maximizing ATP synthesis and maintaining ROS production to only the amounts required for microdomain cell signaling (19, 87).
In addition to ATP synthesis, mitochondria are the site of other important metabolic reactions, including steroid hormone and porphyrin synthesis, the urea cycle, lipid metabolism, and interconversion of amino acids (39, 141). Mitochondria also play central roles in xenobiotic metabolism, glucose sensing/insulin regulation (113), and cellular Ca2+ homeostasis (65, 66), which affects numerous other cell signaling pathways.
Despite these critical metabolic roles of mitochondria, classic “mitochondriology” was considered a mature field as recently as 1990. However, several important observations have fueled a renaissance in mitochondrial research, including 1) mitochondrial ROS are not just damaging by-products of respiration, but important for cell signaling (19, 23); 2) mitochondrial release of factors such as cytochrome c is an important step in programmed cell death (100, 110, 112); 3) nitric oxide (NO·) is an potent regulator of mitochondrial function (19, 23, 34); 4) mitochondrial morphology is far from static, with the organelles being subject to fission, fusion, and intracellular movement on a rapid timescale (95, 218); and 5) mitochondria actively orchestrate the spatiotemporal profiles of intracellular Ca2+, under both physiological and pathological conditions (65, 66). Together these observations suggest an extensive regulatory role for mitochondria in both normal and pathological cell function.
The interplay between the conventional and novel roles of mitochondria has received little consideration, and an examination of recent mitochondrial science reveals several incompatibilities with classic bioenergetic viewpoints. An example is the requirement of ATP for apoptosis (137). How does the cell maintain ATP synthesis in the face of mitochondrial disassembly that occurs during apoptosis? Another example, which is the focus of this review, is the role of Ca2+ in regulating organelle function and dysfunction. How can Ca2+, a physiological stimulus for ATP synthesis (5, 72, 118), become a pathological stimulus for ROS generation, cytochrome c release, and apoptosis? As will be discussed extensively, this apparent mitochondrial Ca2+ paradox revolves around a “two-hit” hypothesis (Fig. 1) in which a concurrent pathological stimulus can turn Ca2+ from a physiological to a pathological effector.
MITOCHONDRIAL Ca2+ UPTAKE AND RELEASE PATHWAYS
Most of the mitochondrial effects of Ca2+ require its entry across the double membrane into the matrix. Although the mitochondrial outer membrane was thought to be permeable to Ca2+, recent studies suggest that the outer membrane voltage-dependent anion channel (VDAC) is a ruthenium red (RuRed)-sensitive Ca2+ channel and thus serves to regulate Ca2+ entry to mitochondrial intermembrane space (60). Furthermore, transport across the inner membrane is highly regulated (for review see Ref. 65).
Figure 2 outlines the major mechanisms for mitochondrial Ca2+ transport, with Ca2+ uptake achieved primarily via the mitochondrial Ca2+ uniporter (MCU). Uptake is driven by the membrane potential (Δψm), and therefore the net movement of charge due to Ca2+ uptake consumes Δψm. A recent patch-clamp study suggests the that MCU is a highly selective (Kd < 2 nM) Ca2+ channel (99), but attempts to define its molecular nature have been largely unsuccessful. The channel is known mostly for its pharmacological sensitivity to RuRed (127), and a colorless component of RuRed (Ru360) is the active MCU-binding agent (156, 216). Saris et al. (165) identified a 40-kDa glycoprotein of the intermembrane space as an MCU regulatory component, although the transmembrane component of the MCU has been more difficult to isolate, with limited reports of such an entity (124). Interestingly, reverse MCU transport (Ca2+ export) was shown to be regulated by Ca2+ binding to the outer surface of the inner membrane (86) and was also linked to a soluble intermembrane space component.
Das et al. (41) showed that a complex of Ca2+-polyphosphate and β-hydroxybutanoate can form a Ca2+ channel indistinguishable from that in Escherichia coli, raising the possibility that the MCU (by virtue of mitochondrial/bacterial relationships) may be a nonproteinaceous entity. However, the second-order Ca2+ transport kinetics of the MCU suggest a more complex structure with separate activation and transport sites (169, 206). From a physiological perspective, a role was recently demonstrated for p38 MAP kinase in regulating RuRed-sensitive Ca2+ transport (126). Clearly, identification of the molecular nature of the MCU will aid greatly in understanding the physiological and pathological regulation of mitochondrial Ca2+ uptake.
Two additional mechanisms of Ca2+ entry into mitochondria have also been identified. The first, called “rapid-mode” uptake (RaM), occurs on a millisecond timescale and allows fast changes in mitochondrial matrix Ca2+ concentration ([Ca2+]m) to mirror changes in the cytosol ([Ca2+]c) (186). Second, we have found (11) that ryanodine receptor isoform (RyR)1 is localized to the inner membrane of mitochondria in excitable cells and have termed this channel “mRyR.” Kinetic analysis of the MCU predicts a tetrameric structure like RyR, which exists as a tetramer of ∼500-kDa subunits (17). Together, mRyR and RaM are thought to underlie the phenomenon of excitation-metabolism coupling, in which [Ca2+]c-induced contraction is matched by [Ca2+]m stimulation of ox-phos (see below).
A fast response of [Ca2+]m to [Ca2+]c requires rapid Ca2+ efflux from the mitochondrial matrix, and several mechanisms exist for this purpose (65). Primarily, Ca2+ efflux is achieved by exchange for Na+, which is in turn pumped out of the matrix in exchange for protons (Fig. 2). Thus both Ca2+ uptake and efflux from mitochondria consume Δψm and are therefore reliant on H+ pumping by the respiratory chain to maintain this driving force. In addition to these pathways of Ca2+ efflux, an additional mechanism exists in the form of the permeability transition (PT) pore (10). The PT pore is assembled from a group of preexisting proteins in the mitochondrial inner and outer membranes (38), with Ca2+ binding sites on the matrix side of the inner membrane believed to regulate pore activity. Normally, “flickering” of the PT pore between open and closed states serves to release Ca2+ from the matrix (84, 141, 205). However, prolonged PT pore opening due to [Ca2+]m overload can result in pathological consequences (38).
Ca2+ AS POSITIVE EFFECTOR OF MITOCHONDRIAL FUNCTION.
The primary role of mitochondrial Ca2+ is the stimulation of ox-phos (5, 40, 72, 118, 123). As shown in Fig. 3, this occurs at many levels, including allosteric activation of pyruvate dehydrogenase, isocitrate dehydrogenase, and α-ketoglutarate dehydrogenase (118), as well as stimulation of the ATP synthase (complex V) (40), α-glycerophosphate dehydrogenase (211), and the adenine nucleotide translocase (ANT) (123). Overall the effect of elevated [Ca2+]m is the coordinated upregulation of the entire ox-phos machinery, resulting in faster respiratory chain activity and higher ATP output. Thus mitochondrial ATP output can be changed to meet the cellular ATP demand. An example of this is β-adrenergic stimulation in cardiomyocytes signaling the demand for increased contractility. The concomitant upregulation of ox-phos via [Ca2+]m elevation provides the ATP needed for increased contractile force.
Many other mitochondrial functions are also regulated by Ca2+. For example, Ca2+ activation of N-acetylglutamine synthetase generates N-acetylglutamine (92), a potent allosteric activator of carbamoyl-phosphate synthetase, the rate-limiting enzyme in the urea cycle (119). In addition, Ca2+- and diacylglycerol-sensitive protein kinase (PKC) isoforms and calmodulin have been reported in mitochondria, although their precise targets within the organelle are less well understood (54, 160).
Overall, it appears that Ca2+ is a global positive effector of mitochondrial function, and thus any perturbation in mitochondrial or cytosolic Ca2+ homeostasis will have profound implications for cell function, for example, at the level of ATP synthesis. Also, it cannot be ignored that Ca2+, particularly at the high concentrations experienced in pathology, appears to have several negative effects on mitochondrial function, as discussed in the following sections.
Ca2+ OVERLOAD AND PERMEABILITY TRANSITION PORE
In contrast to the beneficial effects of Ca2+, the PT pore embodies the pathological effects of Ca2+ on mitochondria. The PT pore, as described nearly 25 years ago by Haworth and Hunter (81–83), is an assembly of preexisting proteins of the inner and outer mitochondrial membranes into a large conductance channel permeable to solutes of <1,500 Da. Debate still surrounds the composition of the PT pore, and although a full discussion is beyond the scope of this review, Fig. 4 shows the key components. Along with VDAC, ANT, and cyclophilin D (Cyp-D), several proteins are believed to regulate the pore (for review see Ref. 38).
The PT pore is triggered by high [Ca2+]m and other stimuli including oxidants and the depletion of adenine nucleotides. It is inhibited by acidic pH, antioxidants such as reduced glutathione (GSH), and cyclosporin A, which binds to Cyp-D, a matrix cis/trans-prolyl-isomerase. The ANT appears to be a key modulator, with ANT thiols proposed as a target for oxidative stress induction of the PT pore (69). Also, bongkrekic acid and attractyloside, two inhibitors of ANT, inhibit or activate PT pore opening, respectively. Recently, doubt has been cast on the role of ANT, because mitochondria isolated from double-knockout ANT1−/−, ANT2−/− mice appear capable of undergoing PT (101). However, the ANT is a mitochondrial carrier family protein, all of which share significant sequence homology, making it likely that other proteins in this family can substitute for the ANT in PT pore assembly.
The role of the PT pore in pathological cell injury and death has been cemented by the discovery that opening of this pore is mechanistically linked to cytochrome c release, a key event in apoptosis (112). Despite the recent popularity of this research topic, it is worth noting that Knyazeva et al. (100) discovered mitochondrial cytochrome c release in ischemic liver nearly 30 years ago! Several studies suggest a role for the PT pore in this process, including findings that 1) PT pore inhibitors (e.g., cyclosporin A) inhibit cytochrome c release and apoptosis (176, 222), 2) the Bcl family proteins have been shown to functionally interact with PT pore components such as VDAC (132, 177, 200), and 3) the loss of Δψm is a hallmark of apoptotic cell death and is thought to signal the recruitment of Bcl family proteins to the mitochondrion (44).
Despite strong evidence linking the PT pore, cytochrome c release, and apoptosis, the precise mechanism of cytochrome c release is still unknown and is likely to be dependent on cell type, apoptotic stimulus, and precise cellular conditions. Several non-PT pore-mediated mechanisms of cytochrome c release may exist, and it is important to emphasize that cytochrome c does not exit through the PT pore itself. Also, although in vitro PT pore opening results in mitochondrial swelling and outer membrane rupture (38, 143), this is unlikely to occur in vivo, because mitochondrial swelling is not typically observed in apoptosis (although it is in necrosis) (111). This is in agreement with our data from time-course experiments in isolated mitochondria showing that cytochrome c release is temporally unrelated to swelling (24). Overall, the PT pore can be considered an important signaling pathway leading to cytochrome c release, but its involvement in the physical mechanism of cytochrome c release is still debated.
Cytochrome c is highly positively charged and binds to negatively charged cardiolipin on the outside of the inner membrane. There are also binding sites on respiratory complexes III and IV, and it has been shown that cytochrome c release is a two-step process (143), involving release of the protein from its inner membrane binding sites followed by outer membrane translocation. In addition, PT pore opening appears to be accompanied by a burst of ROS (61, 62), and this phenomenon is proposed to be involved in the autoamplification phase of the pore (62, 102). Because ROS can cause oxidation of cardiolipin, changing its physical properties (204), this may also enhance cytochrome c release (62, 143). Therefore, it is possible that high [Ca2+] in the intermembrane space may enhance cytochrome c release by competing it off binding sites (89), through a mechanism involving ROS oxidation of cardiolipin. Within the overall context of Ca2+ as a mitochondrial pathological stimulus, we have shown (21) that PT pore triggering by ROS is potentiated by Ca2+. This is an example of the two-hit hypothesis (Fig. 1), in which the combination of Ca2+ plus a pathological stimulus such as ROS can elicit mitochondrial dysfunction.
In addition, recent evidence has suggested that cytochrome c can bind to the endoplasmic reticulum (ER) inositol 1,4,5-trisphosphate receptor (IP3R), rendering the channel insensitive to autoinhibition by high [Ca2+]c and resulting in enhanced ER Ca2+ release (14, 15). Thus Ca2+-induced mitochondrial cytochrome c release may propagate apoptotic signaling by promoting further Ca2+ overload. The close proximity between the ER and mitochondria (115) facilitates this cross-talk and is discussed in the next section.
MITOCHONDRIAL MORPHOLOGY, DYNAMICS, AND ER COORDINATION
Morphological alterations of mitochondria have been observed in multiple pathologies, implying a close relationship between mitochondrial structure and function. However, even in normal cells (95) mitochondria display dynamic tubular networks undergoing constant fission, fusion, and cytoskeletal trafficking (8, 9, 138). These networks appear to intertwine closely with tubules of the ER. Although the reason for dynamic mitochondrial behavior is not fully understood, it is believed to ensure appropriate mitochondrial distribution to provide ATP to localized cytosolic regions.
Mitochondrial fission requires the dynamin-like protein DLP1, a large GTPase that is transiently recruited to mitochondria by interactions with the outer membrane protein hFis1 (13, 142, 151, 179, 180, 217–219). Fusion is also mediated by GTPases; mitofusin (Mfn) is an outer membrane anchored GTPase and a homolog of the Drosophila “fuzzy onion” protein (31, 67, 85, 104, 158, 163, 164). Two mammalian Mfn isoforms (Mfn1 and Mfn2) have both redundant and distinct functions, forming homo- and heteromeric complexes (31). The coordinated fusion of inner and outer membranes is thought to occur when Fzo1p (a yeast homolog of fuzzy onion) forms a complex with Mgm1p, which is an inner membrane-associated dynamin family GTPase (174, 213).
Mitochondrial fission and fusion must occur at balanced levels to maintain normal morphology (173), and an imbalance results in excessive fragmentation or tubulation, with pathological consequences. Increased fission appears to be a prerequisite for cytochrome c release, because mitochondrial morphology in many apoptotic cells changes from tubular networks to a fragmented phenotype (95). For example, hFis1 overexpression causes mitochondrial fragmentation, cytochrome c release, and apoptosis (90, 217), whereas a dominant-negative DLP1 prevents cytochrome c release in staurosporine-induced apoptosis (57). Similarly, Mfn−/− mice die at midgestation (31), and the human homolog of Mgm1p has been identified as OPA1, which causes hereditary blindness when mutated (4, 46). Notably, DLP1 and Mfn2 have been shown to closely appose to proapoptotic Bax (94), and the development of mitochondrial fragmentation has been demonstrated in several apoptotic cell systems, indicating that the mitochondrial fission/fusion machinery directly or indirectly interacts with apoptotic components to regulate cell death (18).
Recent studies have suggested coordination between Ca2+ signaling of the ER and mitochondria, facilitated by strategic location of mitochondria at sites of ER Ca2+ release (115). The ER is the major intracellular Ca2+ store, and regulated Ca2+ release from this organelle is essential for cellular signaling. However, in pathological situations this may be detrimental to mitochondrial function, and Ca2+ released during the ER stress response may promote mitochondrial fragmentation and apoptosis. As noted above, it has also been reported that cytochrome c can bind to the IP3R, causing further ER Ca2+ release (14, 15). Because mitochondrial cytochrome c release during apoptosis is an “all-or-nothing” event occurring within a rapid time frame (116), it has been suggested that Ca2+ is the coordinating signal for cytochrome c release (15).
Several questions arise from these studies, such as the following: How can the cell avoid cytochrome c release during normal fission/fusion events? Is cytochrome c release in apoptosis merely an accidental by-product of mitochondrial fission (leakage?), or is fission a deliberate part of the mechanism of cytochrome c release? It is possible that when the distance between two fission sites is extremely close, the short lipid bilayer may not be able to provide sufficient curvature for membrane sealing, causing membrane rupture and leakage. In support of this notion, cytochrome c release on hFis1 overexpression is not inhibited by classic inhibitors of PT pore opening (90), suggesting that the PT pore is not involved in the mechanism of cytochrome c release under these conditions.
An additional link between ER Ca2+, mitochondria, and apoptosis is provided by studies showing that overexpression of antiapoptotic Bcl-2 lowers the steady-state ER [Ca2+] ([Ca2+]ER), leading to lower mitochondrial Ca2+ uptake on IP3R activation (56, 150). In addition, proapoptotic Bax−/−-Bak−/− fibroblasts also display lower [Ca2+]ER (172). These data suggest that increased ER Ca2+ release is a key event in apoptosis regulated by both pro- and antiapoptotic Bcl proteins. Furthermore, a direct role for ER Ca2+ in regulating mitochondrial fission has been proposed; ectopic expression of the p20 fragment of BAP31 (an ER protein that binds to Bcl-2 and Bcl-XL) causes mitochondrial translocation of DLP1, resulting in fragmentation and cytochrome c release (18). The signal for mitochondrial recruitment of DLP1 is likely to be elevated [Ca2+]m, because inhibition of mitochondrial Ca2+ uptake attenuates the fragmentation (18).
In summary, coordination between ER and mitochondria occurs both spatiotemporally and biochemically and has the potential to regulate mitochondrial morphology. Perturbation of the ER can lead to [Ca2+]m overload, fragmentation, and apoptosis. This is a further example of the two-hit hypothesis (Fig. 1), in which a normal mitochondrial process (fission/fusion) can have pathological consequences in the presence of high [Ca2+].
MITOCHONDRIAL ROS GENERATION
Mitochondria are a significant source of ROS, although the assumption that 1–2% of all O2 consumed ends up as ROS is likely an overestimate. Although an extensive review of mitochondrial ROS generation is beyond the scope of this review (see Refs. 88 and 202), an outline of the major sources within the organelle serves to highlight that this is both a physiological and a pathological phenomenon and will aid in the understanding of its regulation by Ca2+.
The primary ROS made by mitochondria is superoxide (O2−·), which is converted to H2O2 either by spontaneous dismutation or by the enzyme superoxide dismutase (SOD). H2O2 can be further transformed to OH· in the presence of metal ions by Fenton chemistry, although metal chaperone proteins in the mitochondrial matrix (37, 106) likely prevent this from occurring in the organelle.
The main source of O2−· in mitochondria is the ubisemiquinone radical intermediate (QH·), formed during the Q cycle at the QO site of complex III (128, 192, 201). Generation of ROS is accelerated by complex III inhibitors distal to this sitg (e.g., antimycin A), although evidence that inhibition further down the respiratory chain (e.g., at complex IV) can also elevate ROS is sparse (19). The majority of O2−· from QO is made facing the intermembrane space (192), leading to suggestions that O2−· may be released through VDAC into the cytosol (71). However, the intermembrane space contains both Cu/Zn-SOD (139) and ∼20 mM cytochrome c, which can be reduced by O2−· and pass the electrons on to complex IV (114). Indeed, exogenous O2−· is an excellent substrate for generation of Δψm and ATP synthesis (114), and mitochondria have been championed as a significant cellular ROS sink (63, 147, 224). Some HO2· is probably also made from QO at the intermembrane space, because the acidic dissociation constant (pKa) of O2−· is ∼4.8 (45). Uncharged HO2· can pass through membranes but is highly reactive and therefore likely initiates membrane lipid oxidation (45). Furthermore, the spontaneous dismutation of HO2· (to H2O2) is five to eight orders of magnitude faster than for O2−·. Under certain conditions, it has been shown that some O2−·/HO2· from QO may reach the matrix space (128, 192), but Mn-SOD in this compartment would rapidly convert it to H2O2.
Complex I is also a source of ROS, although the mechanism of generation is less clear than for complex III. Rotenone and other distal complex I inhibitors can cause O2−· generation facing the matrix side of the inner membrane (201, 192, 108), where Mn-SOD would convert it to H2O2. Recent reports suggest that glutathionylation of complex I (197) or phosphorylation by PKA (145, 153) can elevate ROS generation, but the physiological or pathological significance of this is unclear. In vitro, electrons entering at complex II (succinate dehydrogenase) can flow backward through complex I to make ROS (201, 202). In vivo this would be prevented by forward electron flow through complex I from NADH, except under pathological conditions in which NADH is depleted (see below).
Another important regulator of mitochondrial ROS is Δψm. The generation of ROS is exponentially dependent on Δψm (187), and both chemical uncouplers (e.g., 2,4-dinitrophenol) (140) and the novel uncoupling proteins (UCPs) (133) appear to decrease mitochondrial ROS generation in whole cells and organs, although in vitro experiments with isolated mitochondria have revealed opposing effects (27). Interestingly, as we originally hypothesized in 1998 (20), it has now been shown that ROS can stimulate mitochondrial uncoupling (51, 130) and that the processes of uncoupling and ROS generation exist in a feedback loop (20, 130, 125).
Our understanding of the roles of mitochondrially derived ROS has been transformed in recent years. Previously ROS were thought to be damaging by-products of respiration, responsible for oxidative damage and contributing to aging (for review see Ref. 75). However, recent evidence has shown that mitochondrially derived ROS are important for a multitude of cell signaling processes (23). Illustratively, cytosolic Cu/Zn-SOD−/− mice are viable, whereas mitochondrial Mn-SOD−/− mice are not (79). The fact that Cu/Zn-SOD cannot compensate for the loss of Mn-SOD (121) suggests a much more subtle, compartmentalized role for these enzymes than mere damage limitation and ROS detoxification. Several recent studies have shown that perturbations in mitochondrial ROS generation can affect diverse redox signaling pathways such as the cell cycle (167), cell proliferation (98), apoptosis (108), metalloproteinase function (155), oxygen sensing (30), protein kinases (154, 155), phosphatases (152), and transcription factors (78).
Ca2+ AND MITOCHONDRIAL ROS
At the heart of understanding how Ca2+ can be both a physiological and a pathological effector of mitochondrial function is the issue of how Ca2+ modulates mitochondrial ROS generation. In this section, the theoretical and experimental considerations underlying this “mitochondrial Ca2+ paradox” are discussed. Because the QH· intermediate in the Q cycle is a significant source of ROS (vide supra), two related parameters that can regulate ROS generation are 1) the effective concentration of QH·, which is increased when the distal respiratory chain is inhibited, and 2) the frequency of QH· occurrence, which is increased when the respiratory chain turns over more quickly. Thus both stimulation and inhibition of mitochondria can result in enhanced ROS generation.
Figure 5 shows the theoretical mechanisms by which Ca2+ can enhance ROS generation. Stimulation of the TCA cycle and ox-phos by Ca2+ would enhance ROS output by making the whole mitochondrion work faster and consume more O2. Indeed, mitochondrial ROS generation correlates well with metabolic rate (148, 184), suggesting that a faster metabolism simply results in more respiratory chain electron leakage. In addition, Ca2+ stimulation of nitric oxide synthase (NOS) (3) generates NO·, which inhibits complex IV (34), and this would enhance ROS generation at QO. Indeed, we hypothesized (19) that a major physiological role for NO· is the regulation of mitochondrial ROS output. Thus mitochondria can act as a “redox signaling box,” converting an NO· signal into an ROS signal (19). This signaling axis operates within a physiological window of NO· concentrations, and disruption of this axis by pathological levels of NO· is detrimental to mitochondrial ATP synthesis and cell function (178). Furthermore, NO·, in conjunction with high [Ca2+]m, can inhibit mitochondrial complex I (91), and this is another example of the two-hit hypothesis (Fig. 1) because NO· and Ca2+ are not detrimental to complex I function. Although other inhibitors of complex I are known to enhance its ROS generation (108, 197), it is not known whether this is the case for NO· + Ca2+. Furthermore, Ca2+ can enhance cytochrome c dislocation from the mitochondrial inner membrane, either by competing for cardiolipin binding sites or by inducing the PT pore, and this results in an effective block of the respiratory chain at complex III, which would enhance ROS generation (62, 143). Another possibility is that Ca2+ can perturb mitochondrial antioxidant status. We have observed (unpublished data) that mitochondrial GSH is released very early in Ca2+-induced PT pore opening, suggesting that Ca2+-exposed mitochondria may generate more ROS because of diminished GSH levels.
Despite the theoretical events depicted in Fig. 5, experimental observations on Ca2+ and mitochondrial ROS are diverse. Overall it appears that Ca2+ diminishes ROS from complexes I and III under normal conditions but enhances ROS when the complexes are inhibited. Differences in the literature seem to stem from nuances of the specific respiratory chain inhibitors used, the presence of uncouplers, the source of the mitochondria, and the respiratory state (i.e., state 4 vs. state 3).
Addition of Ca2+ to rat heart mitochondria with a complex III inhibitor (antimycin A) caused a sharp increase in ROS generation, similar to that observed with FCCP (27). Because Ca2+ uptake causes mild uncoupling of mitochondria (Δψm dissipation), some of its effects on ROS generation may be due to this uncoupling effect. The mechanism by which uncouplers enhance ROS generation is unclear (201) but may involve perturbation of the pH gradient (ΔpH) across the mitochondrial membrane, which would affect the topology of HO2·/O2−· generation (45).
In contrast, addition of Ca2+ to brain mitochondria in the presence of antimycin A did not stimulate ROS generation from complex III, but Ca2+ did elicit a complex I ROS generation when rotenone was present (185). Notably, this effect was reversible by addition of EGTA, suggesting that the PT pore (irreversible cytochrome c loss) is not involved and hinting at a reversible Ca2+ binding site that may modulate ROS generation. This is supported by data showing that in liver mitochondria with antimycin A the ability of Ca2+ to stimulate ROS generation is inhibited by the local anesthetic dibucaine and by Mg2+ (102). The authors attribute this effect to the ability of these species to displace Ca2+ from its membrane protein binding sites (102).
A further study on Ca2+ and brain mitochondria showed that in the presence of rotenone plus complex I substrates, Ca2+ decreased ROS generation (188). This was not due to uncoupling, because Ca2+-induced loss of Δψm lasted only a few seconds whereas ROS generation was depressed for several minutes. With oligomycin present, Ca2+ caused large-scale uncoupling that lasted for several minutes. Notably, this was reversed by EGTA, again suggesting a reversible Ca2+ binding site that regulates ROS (188).
Overall these experiments suggest that in a tissue-specific manner, Ca2+ diminishes ROS from both complexes I and III under normal conditions and enhances ROS when these complexes are inhibited. The exact mechanism of Ca2+-induced ROS generation is unclear, although it may involve changes in the three-dimensional conformation of the respiratory complexes. Indeed, Ca2+ is reported to alter the spectrum of cytochromes a/a3 in isolated complex IV (212), and we have shown (21) that Ca2+ exposes novel mitochondrial targets for nitration by ONOO−, consistent with protein conformational changes. Ultimately, differences in the subunit composition of the respiratory complexes between tissues may underlie the tissue specificity of Ca2+ effects on ROS generation.
Together, Ca2+-mediated events may have additive effects on ROS generation. For example, Ca2+ stimulation of the TCA cycle would increase electron flux into the proximal respiratory chain while Ca2+-induced cytochrome c release would simultaneously inhibit the distal respiratory chain. Although thus far our two-hit model (Fig. 1) has concentrated on the convergence of Ca2+ with other pathological stimuli, it should be recognized that Ca2+ itself is a pathological stimulus and in some cases Ca2+ is both the first and second “hit,” i.e., Ca2+ can modulate its own pathological effects.
Ca2+ AND PT PORE OPENING—A CAUSE OR EFFECT OF ROS GENERATION?
Apoptosis is reportedly accompanied by a burst of ROS (29, 61, 62, 102), although the recent discovery that the ROS fluorescent probe dichlorofluorescin (DCF) can be oxidized directly by cytochrome c raises the possibility of experimental artifacts due to apoptotic cytochrome c release (25). Studies on isolated mitochondria with other ROS probes have provided more information on this scenario, showing that a complex interplay exists between PT pore opening and ROS generation in response to Ca2+ (29, 61, 102, 204). What is still unclear is whether ROS generation is merely a consequence of PT pore opening and cytochrome c release or is an integral part of the signaling machinery of PT pore opening. Although evidence abounds that exogenously added ROS can trigger the PT pore (21, 29, 144), it is unclear whether endogenously generated ROS can perform this function. In support of a role for endogenous ROS in the PT pore, antioxidants and thiol reagents such as DTT are able to inhibit Ca2+-induced PT pore opening (38). Furthermore, Castilho et al. (29) report that Ca2+-induced PT pore opening is sensitive to the concentration of O2 (the substrate for O2−· generation) and is inhibited by catalase, suggesting a role for H2O2 generation.
Theoretically, PT pore-mediated cytochrome c release would enhance ROS generation by inhibiting complex III, but experimental evidence for this is limited. A key issue is how much cytochrome c is required to sustain respiration. Studies on isolated complex IV have shown that very low levels of cytochrome c (1 μM) can sustain activity (7), suggesting that with millimolar levels of cytochrome c in the intermembrane space (39, 171), significant loss would not result in respiratory inhibition. Indeed, it appears that mitochondrial electron transport and ATP synthesis are maintained during apoptosis (208), presumably to allow execution of the apoptotic program, which requires ATP (137). However, it is worth noting that cytochrome c is not the only factor lost from mitochondria on PT pore opening. Nearly 100 proteins are lost from the intermembrane space (146), as well as GSH and other matrix solutes. Any of these components (especially GSH) could be responsible for enhanced ROS generation during apoptosis. Another idea developed within this context is that cytochrome c is an antioxidant, because it is an effective scavenger of O2−· and is readily diffusible and recycled in the cell (147). This raises the possibility that cytochrome c release during Ca2+-/ROS-mediated mitochondrial dysfunction may have evolved as a protective strategy to scavenge extramitochondrial ROS before they reach the organelle.
Overall, the interplay among Ca2+, ROS, and ATP in both the life and the death of the cell is extremely complex. This interplay is illustrated in Fig. 6, showing that the roles of each player in this “love-hate triangle” are different depending on the physiological/pathological status of the cell.
CLOSING THE TRIANGLE—ATP/ROS/Ca2+
Thus far we have concentrated on the effects of Ca2+ on mitochondrial ROS generation and ATP, but it is important to briefly mention reverse signaling, i.e., the effects of ROS and ATP on Ca2+ homeostasis. A key example is the RyR, which has a critical thiol residue that can be modified by ROS (53) or reactive nitrogen species (RNS) (193, 214), thereby activating the channel and enhancing sarcoplasmic reticulum (SR) Ca2+ release. Thus mitochondrial ROS can theoretically modulate extramitochondrial Ca2+ pools. Consistent with these biochemical findings, it is reported that both SR and mitochondrial Ca2+ accumulation are inhibited by ROS (77, 93). Notably, at very high Ca2+ levels only SR Ca2+ uptake was inhibited, suggesting that in the face of a Ca2+ overload mitochondria continue to accumulate Ca2+, which may contribute to downstream pathology (93). Notably, the inhibition of mitochondrial Ca2+ uptake by ROS was different depending on the respiratory substrate and was almost absent when mitochondria were incubated with a broad mix of substrates supporting complexes I, II, and IV. This suggests that oxidative damage to the respiratory chain may depolarize the mitochondrial inner membrane, impairing Ca2+ uptake, and is consistent with our observations showing that ONOO− can uncouple mitochondria (22). Such uncoupling prevents Ca2+ accumulation by diminishing the magnitude of Δψm. The effects of NO· on mitochondrial Ca2+ have also been examined by several authors including ourselves (6, 24, 157); we (24) reported that NO· inhibition of mitochondrial respiration inhibits Ca2+ uptake by the organelle. It is tempting to speculate that under certain conditions Ca2+ activation of NOS, making NO·, may inhibit respiration as a protective mechanism to prevent mitochondrial Ca2+ accumulation.
A further level of cross-talk is suggested by Zoccarato et al. (224), who report that mitochondrial scavenging of external H2O2 is inhibited by Ca2+. They attribute this to a Ca2+-induced inactivation of glutathione reductase (GR) and glutathione peroxidase (GPX). The authors speculate that succinate-driven reverse electron flow through complex I, generating NADH (the substrate for GR and GPX), may underlie this phenomenon (224).
Finally, the importance of ATP in Ca2+ homeostasis cannot be understated, because the maintenance of ion gradients in all cell types depends on ATP (26). Therefore, any disturbance in ATP levels will have a large impact on SR, ER, and plasma membrane (PM) Ca2+ pumps. Overall, as highlighted in Fig. 6, it is clear that ATP, ROS, and Ca2+ exist in a triangular network, with each having the ability to control the others. The following sections discuss two pathological examples of this triangle in action.
EXAMPLES OF PATHOLOGICAL [Ca2+]M OVERLOAD—CARDIAC ISCHEMIA-REPERFUSION INJURY
Overload of [Ca2+]m is a key event in cardiac ischemia-reperfusion (I/R) injury (189, 190), with the end point being opening of the PT pore, cytochrome c release, and apoptosis/necrosis (38, 68). Figure 7 outlines the events preceding [Ca2+]m overload. Under normal conditions, the PM Na+/H+ exchanger (NHE) equilibrates internal and external Na+ and H+ and the PM Na+/Ca2+ exchanger keeps [Ca2+]c low. During ischemia, lactic acidosis forces the NHE to import external Na+, resulting in cytosolic Na+ overload (96). Subsequently, the PM Na+/Ca2+ exchanger is forced into reverse mode to dispose of excess Na+, resulting in [Ca2+]c overload. This Ca2+ is then taken up by mitochondria, resulting in [Ca2+]m overload and all its consequences as described above (for reviews see Refs. 38 and 68).
Several levels of communication among Ca2+, ROS, and ATP exist in this system, and at the center are the MAP kinases (ERK, JNK, p38). Mitochondrial ROS can activate MAP kinases (107, 154), and all three MAP kinases are activated during ischemia and, to a greater extent, on reperfusion (131, 221). Recently it was reported that mitochondria are the sole source of ROS for p38 activation in cardiomyocyte ischemia (103). Notably, ERK1/2 activation leads to phosphorylation and activation of NHE-1 (159, 183), and this may contribute to a feed-forward activation loop (Ca2+ → ROS → ERK → more Na+ → more Ca2+), enhancing [Ca2+]m overload in I/R injury (Fig. 7B). Additionally, ERK signaling is critically dependent on mitochondrially generated ATP, which is unavailable during ischemia (1). This raises the important point that ATP is a substrate for all signaling kinases and therefore mitochondrial ATP synthesis has the potential to impact on cell signaling at multiple levels, not just MAP kinases.
In support of a role for MAP kinases in the cross-talk between ROS and Ca2+, JNK has been shown to be elevated within mitochondria during I/R (77), raising the possibility of direct phosphorylation of mitochondrial respiratory complexes, with subsequent effects on ROS generation. In addition, p38 can inhibit the mitochondrial RuRed-sensitive Ca2+ uniporter (126), suggesting a protective role for this MAP kinase. However, in neurons, p38 can mediate the translocation of proapoptotic BH3-containing proteins to mitochondria (59), suggesting opposing roles for p38 in different cell types.
Mitochondrial Ca2+ overload does not have to result in PT pore opening and cytochrome c release to be detrimental to heart function. For example, the inhibition of complex I that occurs in I/R injury is dependent on [Ca2+]m, because it is prevented by RuRed (73, 74). Several other mitochondrial enzymes are inhibited in response to I/R or ROS exposure, including aconitase, α-ketoglutarate dehydrogenase, and complexes III and IV (105, 161). Because the heart is almost entirely dependent on mitochondrially generated ATP for its contractile energy, a defect at any level of the mitochondrial ox-phos machinery can have profound implications for contractile function on reperfusion.
Another important mitochondrial consequence of I/R injury is an increase in the H+ leak (uncoupling) of the inner membrane (Fig. 7B; Ref. ⇓). This may involve the generation of RNS such as ONOO−, which we demonstrated (22) can increase H+ leak. In addition, O2−· and lipid oxidation products can activate mitochondrial UCPs (51, 125, 130), although this remains controversial (36). Alternatively, it has been shown that allosteric activation of the ANT by AMP turns it into a H+ channel (28). Notably, AMP is elevated in I/R, and inhibitors of adenylate kinase (which catalyzes the reaction 2ADP ↔ ATP + AMP) are of therapeutic benefit in cardiac I/R (81, 168). In addition, AMP-dependent protein kinase is known to activate endothelial NOS (eNOS) (3, 32), thereby leading to generation of NO· and possibly additional effects on mitochondria, MAP kinases, and other signaling pathways. Although at first glance mitochondrial uncoupling might be seen as detrimental to cardiac function because it would inhibit ATP generation, there may be benefits, too. Uncoupling would decrease Δψ-dependent Ca2+ accumulation and would also decrease ROS generation. Indeed, several groups have demonstrated the efficacy of mitochondrial uncoupling (either chemical or by transfection of UCPs) as a protective strategy in I/R injury (12, 58, 117, 199). In summary, Fig. 7B outlines the roles of [Ca2+]m overload, ROS generation, and diminished ATP in the progression of cardiac mitochondrial dysfunction in I/R.
EXAMPLES OF PATHOLOGICAL [Ca2+]M OVERLOAD—NEURONAL EXCITOTOXICITY
Excitotoxicity is a phenomenon in which neuronal cells undergo necrosis or apoptosis in response to overexposure to excitatory amino acids such as glutamate (166). In particular, prolonged activation of the N-methyl-d-aspartate (NMDA) receptor leads to massive Ca2+ influx, resulting in [Ca2+]c overload and cell death. Presently there are two schools of thought regarding the mechanisms of excitotoxicity: the mitochondrial Ca2+ hypothesis and the nuclear poly-ADP-ribose polymerase (PARP) hypothesis, with both theories emphasizing a key role for ROS.
The mitochondrial Ca2+ hypothesis of excitotoxicity centers on [Ca2+]m overload as a central event in cell death (50, 135). In support of this, inhibition of [Ca2+]m accumulation with the protonophore FCCP prevents excitotoxic cell death (191). Moreover, during the progression to excitotoxic cell death, a late secondary increase in [Ca2+]c occurs, termed “delayed Ca2+ deregulation” (DCD) (134, 135). Although the mechanism of DCD is still under debate, it is agreed that the speed of DCD is strongly dependent on the magnitude of [Ca2+]m accumulation, suggesting that mitochondrial Ca2+ release causes DCD (134).
Parallel to the mitochondrial Ca2+ hypothesis, a role for PARP-1 in excitotoxicity has been established by showing that PARP-1 inhibitors or genetic knockouts prevent excitotoxicity (149, 194). The mechanism of PARP-1-mediated excitotoxicity has been proposed to involve NMDA-receptor-mediated Ca2+ influx, leading to activation of neuronal NOS (nNOS) and excessive production of NO· (43). Overload of [Ca2+]c also activates the protease calpain (33), which converts xanthine dehydrogenase (XDH) to xanthine oxidase (XO), leading to ROS generation. Interestingly, mitochondria also contain a protease that can convert XDH to XO, and this enzyme is released under conditions of mitochondrial stress (162). Subsequently, the generation of ONOO− (from NO· and O2−·) damages DNA and both directly and indirectly activates PARP-1 (149, 194, 207). PARP-1 then catalyzes the hydrolysis of NAD+ into nicotinamide plus PAR, causing depletion of NAD+ (2). This results in cellular energy failure and necrotic cell death. Intriguingly, it has been shown that PARP-1 activation, possibly via generation of non-protein-bound PAR polymers, leads to a loss of Δψm and the release of both cytochrome c and apoptosis-inducing factor (AIF) (220). In addition, NO· has been shown to elicit synaptic glutamate release (120) and may therefore contribute to further excitotoxicity in neighboring cells.
The interplay between PARP-1 and Ca2+ in excitotoxicity is shown in Fig. 8 and demonstrates the central role of mitochondria. The first major signal on NMDA receptor activation is an overload of [Ca2+]c, which leads to generation of ROS/RNS. However, the amount of ROS/RNS generated via cytosolic Ca2+-dependent pathways appears to be insufficient for cell death, because blockade of [Ca2+]m uptake averts excitotoxicity (191). Therefore, Ca2+ stimulation of mitochondrial ROS may also be important. In addition, ONOO− is a potent trigger of the PT pore and cytochrome c release in Ca2+ loaded mitochondria (21, 144) but cannot induce PT alone (i.e., without Ca2+) (21). Thus it appears that the concerted effects of RNS and Ca2+ on mitochondria are required for cell death to occur—another demonstration of the two-hit hypothesis (Fig. 1).
Further interaction between PARP-1 and ROS could occur at the level of NAD(P)+/NAD(P)H. Because GR requires NAD(P)H to reduce oxidized glutathione (GSSG) back to GSH, a depletion in NAD(P)+ would lead to a depletion in the reducing substrates NAD(P)H and a subsequent shift in GSH-to-GSSG ratio. PT pore opening is dependent on the redox status of both the NAD(P)H and GSH pools, being greater in the oxidized state (38). In addition, NAD(P)+ depletion may impact on mitochondrial ATP synthesis, and this has profound implications for a cell such as the neuron that relies heavily on mitochondrial ATP for the maintenance of PM ion gradients. Overall, it appears that [Ca2+]m overload and subsequent mitochondrial dysfunction are unifying factors in excitotoxicity, playing key roles in cell death. As discussed in the next section, mitochondria are emerging as the target of choice for therapeutic intervention in neurodegenerative disease.
Direct targeting of drugs to mitochondria has long been a dream of pharmacologists, but only recently has it been realized that multiple pharmacological approaches that are not necessarily directed at the mitochondrion appear to converge on [Ca2+]m accumulation as an end point for cardio- and neuroprotection (129). An example is the NHE-1 inhibitor cariporide, which is entering stage 3 clinical trials and appears to work by preventing Na+ overload and subsequent [Ca2+]m overload (42, 48). Another example is ischemic preconditioning (IPC), in which a series of small ischemic events can protect against injury from a large ischemic event (215, 223). The sulfonylurea drugs (e.g., diazoxide) are proposed to work by mimicking the effects of IPC, opening a mitochondrial ATP-sensitive potassium () channel, inhibiting [Ca2+]m uptake and subsequent PT pore opening (195). Interestingly, ROS (possibly mitochondrial) are implicated in IPC signaling (223). In a similar manner, uncoupling mitochondria by chemical uncouplers (58) or by transfection UCPs (12, 117, 199) is both neuro- and cardioprotective, and these therapies are proposed to work by preventing [Ca2+]m accumulation, ROS generation, and PT pore opening. Downstream of [Ca2+]m overload, several strategies are aimed at preventing mitochondrial PT pore opening and cytochrome c release, including the PT pore inhibitor cyclosporin A, which is FDA indicated for ischemic heart disease.
An additional benefit of therapeutically targeting mitochondria comes from the increased cell functionality that results. A cardiomyocyte that is alive but not contracting is of very little use to the parent organism, and in this regard targeting mitochondria to maintain their integrity not only prevents cell death but also facilitates ATP synthesis, which is especially important in cells that are highly ATP dependent (e.g., myocytes and neurons). Clearly, the development of mitochondrially targeted drugs is an important research direction, and a range of strategies have been exploited to target therapeutic agents to mitochondria (182, 196, 209, 210).
One group of mitochondrially targeted molecules are prodrugs, which are activated by mitochondrial enzymes. The mitochondrial GSH pool plays an important role in cytoprotection (122), but to specifically investigate this compounds that selectively decrease mitochondrial GSH are required. In this regard, we synthesized 3-hydroxy-4-pentenoic acid (Fig. 9, 1), which is biotransformed solely by mitochondrial 3-hydroxybutyrate dehydrogenase to 3-oxo-4-pentenoic acid (Fig. 9, 2). 3-Oxo-4-pentenoic acid then reacts with GSH to give adduct 3 (Fig. 9) and thereby depletes mitochondrial GSH (175). 3-Hydroxy-4-pentenoic acid markedly potentiates t-BuOOH-induced cell death (175).
Another property employed for mitochondrial drug delivery is Δψm (150–220 mV, positive outside). Some positively charged “mitochondriotropic” compounds can cross the mitochondrial membranes and accumulate in the matrix (209). Examples are the triphenylphosphonium (TPP)-based compounds. 4-Iodobutyltriphenylphosphonium (Fig. 9, 4) is a reagent for selective labeling of thiols in mitochondria (35) and has been used to detect redox-dependent modification of mitochondrial proteins (Fig. 9, 5; Ref. 109). Cationic lipophilic phosphonium salts derived from phosphonium chloride have also been prepared and tested (209).
To protect mitochondria from the pathological effects of ROS, several mitochondrially targeted antioxidants have been developed, including a TPP+-linked vitamin E (Fig. 9, 6) that protects mitochondria from iron/ascorbate and t-BuOOH-induced oxidative damage (181). A TPP+-linked analog of ubiquinone (Fig. 9, 7) has also been prepared (97). Interestingly, this compound protected cells from H2O2-induced apoptosis but not from staurosporine- or TNF-α-induced apoptosis.
Other studies have questioned the utility of using TPP+ as a mitochondrial targeting strategy. Two TPP-based free radical scavengers, the TPP conjugate of Salen-Mn(III) complex of o-vanillin (EUK-134) (Fig. 9, 8) and the TPP conjugate of Tempol (Fig. 9, 9), delayed sodium selenite-induced apoptosis in HeLa cells but were no more effective than their nontargeted analogs (47). Clearly, more studies with a range of positively charged compounds are needed to determine the effectiveness of mitochondriotropics in combating mitochondrial dysfunction. In this regard, the compound 4-[2-(1H-indol-3-yl)-vinyl]-1-methylpyridinium (F16; Fig. 9, 10) accumulates in mitochondria and selectively inhibits proliferation of a range of cancer cell lines (52).
In addition to the mitochondriotropics, a range of high-molecular-weight mitochondrially targeted compounds have been studied. These include peptide nucleic acid (PNA) oligomers, vesicular carriers, DQAsomes, cationic liposomes, and mitochondrial leader sequence peptide-oligonucleotide derivatives (209). The goal of targeting experimental tools and therapeutic agents to mitochondria will continue to be an active area of investigation, hopefully yielding therapies for diseases involving mitochondrial dysfunction.
Overall there appears to be a multifactorial cross-talk among Ca2+, ATP, and ROS, centered on the mitochondrion (Fig. 6). Several of the key factors that mediate this cross-talk are also shown at the center of the triangle and include NO·, Na+, MAP kinases, and functions of the ER and SR. Elucidating the molecular switches that control this complex triangle is already proving, and will continue to be, an exciting challenge.
This work is supported by National Institutes of Health Grants HL-71158-01 (to P. S. Brookes), DK-061991 (to Y. Yoon), and NS-37710 and HL-33333 (to S.-S. Sheu) and American Heart Association 0050839T and New York State Department of Health Spinal Cord Injury Research Fund C017688 to S.-S. Sheu.
We thank Drs. Dieter Brdiczka and Thomas Gunter for comments on this manuscript. We also thank members of the University of Rochester Medical Center's Mitochondrial Research Interest Group for stimulating discussions.
- Copyright © 2004 the American Physiological Society