Cellular and mitochondrial mechanisms of atrial fibrillation

The kinetics of mitochondrial Ca²⁺ uptake are still a matter of debate. It has been suggested that mitochondria take up and release Ca²⁺ during each heart beat (Fig. 3a, Model I: Oscillator) or that they may act as integrators, taking up Ca²⁺ gradually in response to increased cytosolic Ca²⁺, resulting in an increase in steady state [Ca²⁺]_m (Fig. 3a, Model II: Integrator) [75]. Difficulties in measuring [Ca²⁺]_m with membrane-permeable fluorescent ester indicators likely contribute to the uncertainty as to which model is correct [75]. There is, however, evidence for both models in combination (Fig. 3b, c) [6, 69, 74]. It appears that [Ca²⁺]_m transients can occur with each cytosolic Ca²⁺ transient, and that diastolic [Ca²⁺]_m increases progressively at higher stimulation frequencies or cellular Ca²⁺ loading. Whereas the oscillator model seems to be relevant at lower stimulation frequencies, the integrator model is thought to dominate at higher frequencies, which is important for matching ATP production to demand. Moreover, peak Ca²⁺ uptake in the mitochondria appears to occur faster than cytosolic Ca²⁺ increase, supporting the concept of a Ca²⁺ microdomain between the SR and mitochondria that is necessary for efficient Ca²⁺ uptake [58, 74].

The major mitochondrial Ca²⁺-influx and -efflux pathways are summarised in Fig. 4. Mitochondrial Ca²⁺ microdomains necessary for Ca²⁺ influx are spatially distinct from Ca²⁺ efflux mechanisms, and this is likely to be important for efficient Ca²⁺ signalling and excitation-energetics coupling [62]. The mitochondrial matrix is separated from the cytosol by two membranes which need to be passed for exchange of metabolites, adenine nucleotides and cations, including regulators of mitochondrial activity such as Ca²⁺. A voltage-dependent anion channel (VDAC) allows Ca²⁺ to pass across the outer mitochondrial membrane, and thus plays a role in mediating mitochondrial Ca²⁺ uptake [104, 123]. The relative low permeability of the inner mitochondrial membrane (IMM) to ions is required to maintain the strongly negative membrane potential between the intermembrane space and the mitochondrial matrix (∆Ψ_m ≈ − 180 mV). Ions can only pass into the mitochondrial matrix via specialised channels and transporters. Several mechanisms have been described for mitochondrial Ca²⁺ uptake, all driven by the large ∆Ψ_m across the IMM. The best-characterised pathway is the mitochondrial Ca²⁺ uniporter (MCU) [61, 112]. The MCU is regulated by divalent cations, ruthenium compounds and adenine nucleotides. The Ca²⁺-binding protein MICU1 (mitochondrial Ca²⁺ uptake 1) is likely required for mitochondrial Ca²⁺ uptake and may act as an auxiliary regulatory protein, determining the [Ca²⁺]_m threshold for Ca²⁺ uptake and preventing mitochondrial Ca²⁺ overload. Low Ca²⁺ affinity (K_m ≈ 10–20 µM) is a critical property of the MCU, making mitochondrial proximity to microdomains of high [Ca²⁺]_i (e.g. RyR2) a requirement for effective mitochondrial Ca²⁺ uptake. The importance of this proximity has been demonstrated by Lu and colleagues in rabbit ventricular myocytes transfected with a genetically encoded mitochondrial Ca²⁺ sensor (Mitycam); it was found that mitochondrial Ca²⁺ uptake is greater and faster near SR Ca²⁺ release sites (Z-line of sarcomere), compared to the middle of the sarcomere (M-line), where SR Ca²⁺ release is limited and MCU density is lower [69]. The role of the mitochondrial MCU complex during basal metabolism versus stress conditions has been extensively discussed by others but the exact involvement of the MCU during pathophysiology remains to be elucidated [41, 47].

Two other potential avenues of Ca²⁺ entry into mitochondria are rapid-mode mitochondrial Ca²⁺ uptake (RaM) and mitochondrial RyRs (mRyR1), as reviewed by De Stefani et al. [113]. RaM was found in heart mitochondria after its initial discovery in liver [23]. However, RaM is thought to be negligible during physiological [Ca²⁺]_i changes in cardiac myocytes because of its slow recovery from inactivation. In contrast, mRyR1 allows Ca²⁺ sequestration at low cytosolic Ca²⁺ concentrations and therefore it could contribute to physiological beat-to-beat Ca²⁺ uptake in the heart [17]. Furthermore, there are data showing that mitochondrial RyR1 may also be important in slower uptake of inositol 1,4,5-triphosphate receptor (IP₃R)-mediated SR Ca²⁺ release [106, 107].

Mitochondrial Ca²⁺ extrusion pathways comprise both Na⁺-dependent and Na⁺-independent mechanisms. However, in cardiac mitochondria, the mitochondrial Na⁺–Ca²⁺ exchanger (mNCX or NCLX [Na⁺/Ca²⁺/Li⁺ exchanger]) is suggested to be the predominant Ca²⁺ extrusion mechanism [24, 100]. Moreover, lethality observed in NCLX-KO mice was attributed to mitochondrial Ca²⁺ overload, which consequently results in cardiac remodelling and dysfunction [71]. It is widely accepted that NCLX is electrogenic and exchanges 3 Na⁺ ions per extruded Ca²⁺ ion, favouring Ca²⁺ extrusion due to the highly negative ∆Ψ_m [90]. To maintain intra-mitochondrial Na⁺ concentration, Na⁺ is removed from the mitochondrial matrix by a Na⁺–H⁺ exchanger (mNHE) located in the IMM. Of note, NCLX develops half-maximal activity in the physiological range of cytosolic Na⁺, making NCLX sensitive to physiological Na⁺ fluctuations and the cytosol-matrix Na⁺ gradient, as previously reviewed [14, 16, 90, 122].

It has been shown in non-ischaemic ventricular myocytes that almost all ATP is derived from mitochondrial oxidative phosphorylation, with the remainder from glycolysis and GTP formation in the Krebs cycle. During oxidative phosphorylation, four multi-protein complexes (Complexes I–IV) which make up the aforementioned ETC, located in the IMM, catalyse electron transport from NADH and FADH₂ to oxygen. The energy released by electron transport allows Complexes I, III and IV to pump protons from the mitochondrial matrix into the intermembrane space; the resulting proton gradient across the IMM contributes to the electrochemical gradient (∆Ψ_m), which fuels ATP synthesis by F₁/F_o ATP-synthase (also known as Complex V) (Fig. 4).

It is thought that there is a complete turnover of the myocardial ATP pool in less than a minute under normal conditions and, depending on heart rate, this may be as quick as every 10 s [111]. Therefore, there is need for tight regulation of mitochondrial ATP production to meet varying metabolic demands resulting from beat-to-beat changes in cardiac workload. Two major regulators of oxidative phosphorylation have been proposed: (1) the products of ATP hydrolysis itself (ADP and P_i) and (2) Ca²⁺. The classical hypothesis was based on measurements of oxygen consumption in isolated mitochondria [64]. ADP and P_i directly stimulate the F₁/F_o-ATP synthase, thus acting as a “pull” on oxidative phosphorylation and causing NADH oxidation. This hypothesis, however, has been challenged by the finding that large changes in cardiac work and ATP consumption in vivo are not associated with measurable changes in ADP/P_i [9]. An increased workload requires increased time-averaged [Ca²⁺]_i. This also acts to increase [Ca²⁺]_m. [Ca²⁺]_m stimulates substrate flow through the Krebs cycle by directly activating key enzymes, thereby increasing production of NADH and FADH₂, fuelling oxidative phosphorylation and increasing ATP production [18, 58]. In addition, mitochondrial Ca²⁺ is thought to directly stimulate the F₁/F_o ATP-synthase and possibly Complex III of the ETC, thus increasing the rate of oxidative phosphorylation [89, 114]. IP₃R, activated by endothelin signalling, may also play an important role in the bioenergetics of cardiac myocytes by inducing mRyR1-mediated Ca²⁺ entry and consequently stimulating mitochondrial ATP production. This is thought to be due to close interaction between the SR and mitochondria, facilitated by mitofusin 2 (Mfn2) [106, 107].

Reactive oxygen species (ROS) in the mitochondria result from the single-electron reduction of molecular oxygen (O₂) during oxidative phosphorylation, among other less significant pathways [91]. Since O₂ is preferentially partitioned in biological membranes where it can interact with electron carriers, such as in the ETC, mitochondria are considered to be a major source of ROS in the cell [10, 91]. Experimentally, 1–2% of oxygen consumed ends up in ROS production at the ETC, primarily at Complexes I and III, although this is estimated to be considerably less in vivo [91]. The first product of the monovalent reduction of O₂ is superoxide anion (^·O₂⁻). Complex I produces ^·O₂⁻ by two mechanisms: (1) a high NADH/NAD⁺ ratio results in fully reduced flavin mononucleotide (FMN), thus backing up the electrons, increasing the time for interaction with O₂; (2) reverse electron transport (RET) due to a reduced ubiquinone (CoQ) pool, where the rate of ^·O₂⁻ production is considered to be highest in the mitochondria [91]. Complex III releases ^·O₂⁻ at least partially into the intermembrane space, depending on local ubisemiquinone (Q^·−) concentrations in the bc₁ complex, whereas ROS production and release by Complex I is limited to the matrix [19]. ROS is generated at Complex III via the Q-cycle when, in a reduced state, a halt in the electron flow enables more time for O₂ to interact with the reduced electron carrier, Q^·− [116]. ROS production at Complex III, however, is regarded as inconsequential compared to rate of production at Complex I, unless pharmacologically induced by Antimycin A. Dihidrolipoamide dehydrogenase, a component of the metabolic enzymes α-ketoglutarate dehydrogenase (αKGDH) and pyruvate dehydrogenase (PDH), is also capable of ROS production in a NADH/NAD⁺-dependent manner [3, 91, 103]. However, under physiological conditions in working cardiac myocytes, α-KGDH is not a relevant source of ROS, but rather contributes to regeneration of reduced nicotinamide adenine dinucleotide phosphate (NADPH) through tight functional coupling to the nicotinamide nucleotide transhydrogenase [119]. Further potential sites for ROS production in the mitochondria have been suggested, as reviewed by Murphy [91]. ROS production increases when there is low ATP demand (state 4 respiration), causing a build-up of reduced NADH (electron donors) or when there is damage to the ETC [10]. Furthermore, ROS production in cardiac myocytes rises during increased workload, for example at higher stimulation frequencies [51, 56].

Net mitochondrial ROS emission from mitochondria is determined not only by the ROS formation rate, but also by ROS elimination. ^∙O₂⁻ is assumed to be present in low picomolar range due to its immediate dismutation by MnSOD [60, 91]. H₂O₂, the main ROS signal, is eliminated by peroxiredoxin and glutathione peroxidase, which require NADPH for regeneration, and also by mitochondrial catalase [16]. NADH generated by the Krebs cycle, and in particular, by α-KGDH, is converted to NADPH by nicotinamide nucleotide transhydrogenase (NNT), a process coupled to the proton gradient across the IMM [97, 119]. Furthermore, NADPH is regenerated by isocitrate dehydrogenase and malic enzyme, which also both derive substrates from the Krebs cycle (isocitrate and malate, respectively). Thus mitochondrial anti-oxidative capacity, ROS elimination and net mitochondrial ROS emission are largely dependent on the Krebs cycle turnover rate [16, 59]. Increased mitochondrial Ca²⁺ uptake and consecutively, enhanced Krebs cycle turnover rate maintain sufficient anti-oxidative capacity of the mitochondrial matrix during increased workload. Accordingly, myocytes from a heart failure model, in which [Ca²⁺]_m elevation in response to increased workload is limited, showed abnormal increases in mitochondrial ROS emission at higher stimulation frequencies [56].

The group of O’Rourke introduced the idea of redox-optimised ROS balance (R-ORB) [8, 30]. R-ORB conceptualises that the redox environment (RE; calculated from the oxidised and reduced states of mitochondrial redox couples) determines mitochondrial ROS levels, and that an intermediate redox state (maximal energy output, state 3 respiration) is accompanied by minimal ROS emission. At either extreme of the RE, ROS levels increase, albeit through different mechanisms; at extremely reduced RE, ROS formation excels due to increased electron slippage from the ETC and thus ROS production exceeds ROS scavenging. Conversely, in the case of an oxidative shift in the RE, for example during pathological increase in workload during heart failure, ROS emission will increase due to reduced ROS-scavenging capacity of the mitochondria, secondary to NNT reversal, as discussed later [97].

Mitochondrial Ca²⁺ overload leads to pathological opening of the mitochondrial permeability transition pore (mPTP), causing a profound decrease in mitochondrial membrane potential and ATP levels which leads to cell death [31, 44]. Interestingly, small, transient and low conductance openings of the PTP (tPTP) have also been identified; however, such openings are believed to be rare and are thought to occur under physiological conditions as a way to release excess mitochondrial Ca²⁺ [70]. The major role of the mPTP in cell death has been reviewed comprehensively elsewhere and will therefore not be the focus of this review [31, 44].

The various interactions between intracellular Ca²⁺ release sites and mitochondria, to what extent they are altered in AF and the effects of such changes on mitochondrial Ca²⁺ uptake remain unclear and therefore represent major targets for future investigation. Previous studies on mitochondrial activity in ventricular cells/tissue will undoubtedly serve as an invaluable basis for the design of atrial studies [56, 69]. Measurement of atrial mitochondrial (and cytosolic) Ca²⁺ at high pacing frequencies, as well as MCU function, is likely to be a crucial step in ascertaining the roles of Ca²⁺ handling dynamics between cytosolic and mitochondrial compartments and whether any alterations occur during—and contribute to—AF pathophysiology.

It will be important in future investigations to identify the weakest link(s), i.e. the “limiting factor”, in matching ATP supply to demand in long-term AF. Energetic substrate- and oxygen supply, metabolism-related enzyme activity and ETC activity all represent potential culprits, with aspects of the latter two being partially under Ca²⁺-controlled regulation, as discussed earlier. We suggest that the “limiting factor” may depend on the extent of electrophysiological and structural remodelling and, therefore, the stage of AF. In response to pathological elevations of cardiac workload in the mouse, the mitochondrial transhydrogenase (NNT) is able to support ATP production by reversing its direction, i.e. converting NADPH to NADH, the conformational aspects of which have recently been defined [53, 97]. This “compensatory” mechanism, however, is at the expense of mitochondrial ROS scavenging and thus net mitochondrial ROS emission increases. Whether a similar situation occurs in chronic AF, due to increased and unmet energy requirements, is currently unknown, but this represents a potentially important avenue of research for the future (Fig. 6). Measurements of mitochondrial membrane potential should be included in future investigations, as it has been suggested that initial Ca²⁺ overload conditions alter the potential, thereby leading to impaired ATP production [67].

The exact mechanism by which pro-arrhythmic cytosolic Ca²⁺ dynamics would impair mitochondrial function has yet to be fully resolved. However, we hypothesise that large alterations of cytosolic Ca²⁺, caused by increased Ca²⁺ leak from the SR, or upon defibrillation or during reperfusion after an ischaemic period, may result in a “push” condition at the ETC, as opposed to the aforementioned “pull”. This could increase the production of ^∙O₂⁻, thereby changing the redox environment to a state where rate of mitochondrial ROS production exceeds the rate of ROS scavenging, thus resulting in net mitochondrial ROS emission (Fig. 6). A recent study demonstrated that optimising mitochondrial activity pharmacologically with SS31 (a stabiliser of cardiolipin and, therefore, ETC activity) and by pharmacologically preventing increased Ca²⁺ influx can protect against the effects of tachypacing [120]. Initial findings in human also suggested CaMKII oxidation (and thus activation) by ROS as an important link between oxidative stress and AF [102]. As CaMKII can become constitutively active through this mechanism, this may represent another way in which oxidative stress could underlie chronic atrial pathology [36].

A further perpetuating cycle, involving mitochondria and suggested to play a role in myocardial pathologies, is the so-called ROS-induced ROS release (RIRR), as reviewed by Zorov et al. [131]; ROS activate the mPTP, inner membrane anion channel and ATP-sensitive K⁺ channel, thereby dissipating ∆Ψ_m [7, 126, 127, 130]. The consequence of this is that NADH and NADPH (via reverse NNT) must be used to restore ∆Ψ_m at the expense of the mitochondrial ROS-scavenging system, thus increasing mitochondrial ROS emission. Furthermore, when open, the mPTP and inner membrane anion channel are permeable to ROS, which can then propagate in the cytosol, potentiating ROS formation and release from neighbouring mitochondria [16, 127]. One would imagine the effect of diffusing ROS is greater in areas of high mitochondrial density and indeed electron microscopic analysis of human atria revealed mitochondrial aggregation in AF [108]. At times of high oxidative stress, for example in heart failure, cross-talk between different ROS sources likely amplifies the ROS signal [73]. Permanent collapse of ∆Ψ_m induces cell death; however, the metabolic oscillations, caused by increased ROS and RIRR, induce arrhythmia in ventricle [2]. Whether RIRR also underlies pro-arrhythmic activity in the atria during AF is still unresolved; however, this remains an interesting line of investigation.