Mitochondrial complex I ROS production and redox signaling in hypoxia

Abstract

Mitochondria are a main source of cellular energy. Oxidative phosphorylation (OXPHOS) is the major process of aerobic respiration. Enzyme complexes of the electron transport chain (ETC) pump protons to generate a protonmotive force (Δp) that drives OXPHOS. Complex I is an electron entry point into the ETC. Complex I oxidizes nicotinamide adenine dinucleotide (NADH) and transfers electrons to ubiquinone in a reaction coupled with proton pumping. Complex I also produces reactive oxygen species (ROS) under various conditions. The enzymatic activities of complex I can be regulated by metabolic conditions and serves as a regulatory node of the ETC. Complex I ROS plays diverse roles in cell metabolism ranging from physiologic to pathologic conditions. Progress in our understanding indicates that ROS release from complex I serves important signaling functions. Increasing evidence suggests that complex I ROS is important in signaling a mismatch in energy production and demand. In this article, we review the role of ROS from complex I in sensing acute hypoxia.

Graphical abstract

Oxygen sensing in acute hypoxia. The astrocytes in the central nervous system and retina, the glomus cells in the carotid and aortic bodies, and the pulmonary artery smooth muscle cells respond to acute hypoxia-induced alteration in hydrogen peroxide (H₂O₂) production from complex I. The subsequent increase in intracellular calcium ion concentration [Ca²⁺] elicits the release of neurotransmitters and other mediators that produce a physiologic response. Created with BioRender.com.

Highlights

• •Mitochondria complex I is structurally and functionally dynamic and regulates OXPHOS.
• •Complex I ROS is critical in redox and stress signaling.
• •Complex I ROS release is key in peripheral and central O₂ sensing.
• •Reversible modification of conserved cysteine residues in complex I mediates acute hypoxia-induced responses.

1. Introduction

Cellular energy generation is a complex multistep process involving the catabolism of metabolic fuels (e.g., carbohydrates, lipids, and proteins) to produce adenosine 5′ trisphosphate (ATP) [1]. ATP serves as the global energy currency of cells in addition to other functions including signaling. ATP is produced on demand by substrate-level phosphorylation and OXPHOS [2,3]. Mitochondrial complex I NADH:ubiquinone oxidoreductase or NADH dehydrogenase) is an important electron entry point for OXPHOS. Complex I is the first enzyme complex of the ETC. It oxidizes NADH generated by the tricarboxylic acid (TCA) cycle, mitochondrial fatty acid β-oxidation (FAO), and amino acid catabolism in the mitochondrial matrix [[4], [5], [6]]. Following oxidation of NADH, paired electron transfer reduces ubiquinone (UQ) to ubiquinol (UQH₂). UQH₂ shuttles electrons from complex I to reduce oxygen (O₂) to water (H₂O). Electron transfer through complex I is coupled with transmembrane proton (H⁺) pumping, which generates Δp necessary for ATP synthesis (Fig. 1). In addition to regenerating NAD⁺ following oxidation of NADH, complex I ROS. ROS is a generic term used to define a range of oxidant molecules with a broad spectrum of biological functions including signaling [7]. Hereinafter, ROS refers to superoxide anion (O₂^•‾), hydrogen peroxide (H₂O₂) or their derivatives and, when possible, we will refer to the specific type of ROS meditating the response. NAD⁺ and ROS are important complex I metabolites that play key roles in regulation of cellular metabolism. Complex I adopts structural and conformational changes to regulate ROS production, and modulate the activity of the ETC.

Fig. 1.

Mitochondrial electron transport chain. Electrons (e⁻) following oxidation of reducing equivalent (NADH) are transferred through series of iron-sulfur clusters (orange spots) in complex I to reduce ubiquinone (UQ) and pump protons (H⁺). Complex II oxidizes FADH₂ to reduce UQ. Electrons are then transferred by UQ₂ to complex III, then to cytochrome c and finally to oxygen (O₂) at complex IV. During this process protons pumped by complexes I, III and IV generative protonmotive force (Δp) utilized by complex V for the synthesis of ATP.

Complex I is important in maintaining metabolic homeostasis by sensing and initiating responses to mitochondrial and extramitochondrial stress, such as hypoxia [[8], [9], [10]]. Hypoxia signaling has been shown to involve mitochondrial ROS. Studies now highlight the importance of complex I ROS in acute hypoxia signaling. The role and contribution of complex I to hypoxia signaling is an emerging area of research with potential for therapeutics and management of many disease conditions.

1.1. Oxidative phosphorylation

Mitochondrial ATP production proceeds via coupling series of reduction-oxidation (redox) reactions in the ETC with the phosphorylation machinery through a process known as OXPHOS. Together, OXPHOS consists of the ETC and complex V (ATP synthase, F₁F₀-ATPase). The ETC consists of four enzyme complexes (complex I – IV) and mobile electron carriers (the hydrophobic UQ and the hydrophilic cytochrome c), which are located in the inner mitochondrial membrane (IMM) [[1], [3], [6]]. These enzyme complexes include complex I, complex II (succinate:ubiquinone oxidoreductase or succinate dehydrogenase), complex III (ubiquinol:cytochrome c oxidoreductase or cytochrome bc₁ complex), and complex IV (cytochrome c oxidase) [[1], [3], [6]]. These enzyme complexes catalyze redox reactions, and are involved in generation of the Δp that drives synthesis of ATP by complex V (Fig. 1) [[1], [6]]. This Δp is the driving force for several mitochondrial functions beyond ATP synthesis, such as ROS production, uptake and homeostasis of ions (including calcium Ca²⁺, sodium Na⁺, iron Fe²⁺, magnesium Mg²⁺, manganese Mn²⁺) [[11], [12], [13], [14]], import and export of proteins [[15], [16], [17]], biomolecule synthesis [18], and mitochondrial dynamics and modeling [19]. The Δp is critical for many facets of health by influencing cell homeostasis directly or indirectly through the regulating of various processes including but not limited to redox and pH microenvironments, cell proliferation, thermogenesis, immune and inflammatory responses, and apoptosis signaling [[20], [21], [22], [23]].

1.2. ETC redox reactions

Complex I oxidizes NADH to reduce UQ to UQH₂ in a proton coupled electron transfer reaction [[24], [25], [26], [27], [28]]. UQ can also be reduced by complex II upon oxidation of succinate, linking the ETC and the TCA cycle. UQ can be reduced by other non-canonical ETC oxidoreductases including electron transfer flavoprotein:ubiquinone oxidoreductase, glycerol 3-phosphate dehydrogenase, proline dehydrogenase, choline dehydrogenase, dihydroorotate dehydrogenase, sulfide:quinone oxidoreductase [[6], [29], [30], [31], [32]]. The redox status of the UQ pool, which is the ratio of UQH₂ to UQ serves as a sensor of ETC efficiency. Alterations in the UQ pool redox status is important in complex I ROS production since changes in the UQH₂/UQ ratio will affect the direction of electron flow and ROS production [30,33]. UQH₂ is oxidized by complex III which, facilitated by the UQ cycle transfers electrons singly to cytochrome c. Complex IV oxidizes cytochrome c, ultimately reducing cellular O₂, the terminal electron acceptor to H₂O. The difference in the redox potentials of NADH (Eo′=−340 mV) and O₂ (Eo′=+810 mV) energetically favors flux of electrons through the ETC from complex I down the electrochemical gradient to complex IV [29,34].

1.3. Generation of protonmotive force (Δp)

Electron flux through complex I, III and IV is coupled with H⁺ translocation from the matrix across the IMM to the intermembrane space (IMS) [1,3,23]. Complex I and III pump H⁺ at a stoichiometry of 4H⁺/2e⁻, while complex IV pumps at lower stoichiometry of 2H⁺/2e⁻ [1,3]. The Δp consists of an electrochemical potential (Δψ_m, of about −120 to −200 mV) and a pH gradient (ΔpH, of 0.3–0.8 pH units) under physiologic conditions [1,[35], [36], [37]]; [23,38]. The Δp acts as a transducer, serving to store energy that drives protons into the proton-conducting channel of F₀ domain of complex V. This produces a rotatory motion that is transmitted to the catalytic head of the F₁ domain through the central stalk of complex V. Rotation of the catalytic head of F₁ domain produces a conformation change that catalyze the phosphorylation of adenosine 5′ diphosphate (ADP) in the presence of orthophosphate to produce ATP [23,39,40]. OXPHOS generates about 90 % of ATP used in cellular metabolism, of which complex I contributes about 40 % of the total Δp [1,2,[1], [2], [6], [41], [42], [43]].

In tissues and individual cells, O₂ gradients exist that affect oxidative metabolisms and health [[44], [45], [46]]. Changes in O₂ gradient affects ETC function [[47], [48], [49]]. However, OXPHOS can operate under limited O₂ as low as 2–10 % O₂ (physiologic hypoxia), but is inhibited at near anoxic level, ≤1 % O₂ (pathologic hypoxia) [43,47,50,51]. Low O₂ tension (<2 %) impedes ETC function and may limit OXPHOS [43,51]. Important here is the “apparent” k_M of complex III. While the k_M measured for complex IV is in the low micromolar to high nanomolar range, various cellular factors such as prevailing concentrations of nitric oxide can modulate this, such that the apparent k_M can be as high as 1–10 μM [52,53]. Dependency on O₂ for ATP synthesis necessitates a system of O₂ sensing to inform adoption of alternative mode of energy generation such as anaerobic respiration. Failure to quickly meet the energy demand may necessitate induction of physiologic changes such as cardiopulmonary responses including angiogenesis, conservation of energy through metabolic suppression, or behavioral response to avoid anoxic/hypoxic environments [[54], [55], [56], [57]]. Earlier studies have established that complex III is critical in chronic hypoxia signaling [54,58,59]. However, recent studies indicate that acute hypoxia sensing operates through a disparate mechanism that may involve signals upstream of complex III [[60], [61], [62], [63], [64], [65]]. This review highlights recent progress in unveiling the role of complex I in acute hypoxia signaling. Understanding the role of complex I in acute hypoxia signaling is important in unveiling the mechanisms of many pathologies and the discovery of targeted medical intervention for such conditions.

2. Complex I: functional complexity

2.1. Complex I structure

Complex I is the largest (∼980 kDa) of the ETC protein complexes having 45 protein subunits encoded by 44 genes, 37 by the nuclear deoxyribonucleic acid (DNA, nDNA) and 7 by the mitochondrial DNA (mtDNA) [27,28,66]. The 7 mtDNA-encoded subunits (ND1, ND2, ND3, ND4, ND5, ND6 and ND4L), and the 7 nDNA-encoded “core” subunits (NDUFV1, NDUFV2, NDUFS1, NDUFS2, NDUFS3, NDUFS7 and NDUFS8) form the conserved central subunits responsible for the core bioenergetic function of complex I. The remaining proteins (about 30 in mammalian mitochondria), referred to as accessory or supernumerary subunits, are required for assembly and structural stability [27,[66], [67], [68]]. Complex I is structurally asymmetric and has a modular L-shaped architecture consisting of a hydrophilic peripheral matrix arm and a hydrophobic membrane arm. The peripheral matrix arm comprises the N module and Q module, and consists of the 7 nDNA-encoded core subunits [27,67]. The N module contains the NADH docking cavity (site I_F), and contains a tightly but noncovalently bound flavin mononucleotide (FMN) cofactor as the electron acceptor. The N module is connected to the Q module, which contains the UQ-binding pocket (site I_Q) by a series of iron-sulfur (Fe–S) clusters (N3, N1b, N4, N5, N6a, N6b, N2) that is ∼100 Å long [27,67,69]. The peripheral arm interfaces at an angle of ∼100° with the membrane arm (P module) at the UQ-binding pocket [67]. The membrane arm contains 3 highly hydrophobic proton pumping subunits, ND2, ND4 and ND5; a fourth proton pump is suggested to exist but is yet to be characterized [70].

2.2. NADH:ubiquinone oxidoreduction

Complex I can adopt two structurally and catalytically different states, the active form (A-form), which can reversibly transition to the deactive, dormant form (D-form) [28,[71], [72], [73], [74]]. In the absence of NADH and UQ, complex I slowly but spontaneously transitions to the D-form, whereas with the addition of NADH it rapidly transitions to the A-form catalyzing redox reactions at high rates [28,72]. Oxidation of NADH regenerates NAD⁺, which rapidly dissociates from the docking site while FMNH₂ transfers two electrons to N3 of the Fe–S clusters [69]. The electrons are transferred through the Fe–S clusters of the Q module to reduce UQ via proton-coupled electron transfer reaction. Reduction of UQ occurs at the UQ-binding pocket (site I_Q) located at the interface of the peripheral and membrane arms (Fig. 2A). This pocket is ∼20 Å above the membrane interface and within 12 Å of N2. It is accessed by a long narrow channel (Q-channel) long enough to contain most of the approximately 50 Å-long isoprenoid tail of UQ-10 [[75], [76], [77], [78]]. The mechanism of UQ reduction at this site is still under investigation. However, reduction of UQ was recently proposed to occur through a “concerted two-electron two-proton transfer” reaction. Here, one electron comes from N2, and another electron together with a proton comes from the conserved Tyr87 of NDUFS2, while NDUFS2 His38 supply the second proton [77,79]. This overcomes the thermodynamically unfavorable energetics of one-electron sequential reduction of UQ to semiubiquinone radical (UQH^•−), and finally to UQH₂. A transient tyrosyl radical formed in the reaction is eventually reduced following arrival of an electron from the Fe–S cluster [79]. The flux of electrons upon oxidation of NADH₂ to UQH₂ describes the catalytic activity of complex I in the forward electron transfer (FET) mode. Operation of complex I in FET is important in maintaining a high NADH/NAD⁺ ratio and a reduce UQ pool as well as generating Δp (Fig. 2B). Thus, complex I is important in intracellular redox-sensing through NADH/NAD⁺ ratio.

Fig. 2.

Directional flow of electrons through complex I produce ROS. A Complex I consists of the peripheral and membrane arms. The peripheral arm contains the NADH oxidation site at docking to flavin mononucleotide (FMN, orange) from which, electrons are transferred through a series of iron-sulfur cluster subunits (N1a, N3, N1b, N4, N5, N6a, N6b and N2, yellow) to the ubiquinone (UQ, purple) at the UQ-binding site, neighboring the NDUFS2 subunit. B Electron flow (dashed red lines) from NADH to reduce UQ to UQH₂ through forward electron transfer (FET). Low protonmotive force (Δp) and low UQH₂/UQ during FET promotes the generation of relatively low level of ROS from FMN (site I_F). C Favored by high Δp and high UQH₂/UQ, electrons are transferred from reduced UQH₂ during reverse electron transfer (RET) to generate bursts of ROS from the UQ-binding site (site I_Q).

2.3. UQ binding and proton translocation

Conformational changes in complex I is affected by interaction with other metabolites. For example, the entry and docking of UQ in the binding pocket results in a local structural and conformational change in this region. Electron flux through the Fe–S cluster, and the subsequent UQ reduction chemistry, induces an electrostatic interaction that produces a global conformational change of the peripheral arm [24,70,[80], [81], [82]]. The induced conformational change in the peripheral arm, hydration and protonation of residues at the UQ-binding pocket is propagated to the proton pumps ∼200 Å away in the peripheral arm to drive proton translocation from the matrix across the IMM [24,70,[80], [81], [82], [83], [84]]. UQ reduction at this site can be specifically inhibited using a number of structurally different compounds including piericidin, rotenone, and capsaicin [85]. Inhibition of this site using rotenone stops proton translocation and dissipates Δp [83,84,86]. Additionally, inhibition of UQ reduction results in buildup of electrons, reduction of the Fe–S cluster and leakage of electrons to O₂ leading to the formation of ROS [29,87].

Seminal studies by Chance and his group reported that electron flux through complex I in isolated mitochondria can operate conditionally in reverse [[88], [89], [90]]. Here, electrons may backup from the UQ pool in the presence of highly reduced UQ and at the expense of Δp in the reverse electron transfer (RET) mode [72,74,[91], [92], [93]]. Thus, increase in cellular NADH/NAD⁺ ratio favors RET by complex I (Fig. 2C) [94]. Mitochondria maintain a pool of NAD that is distinct from the rest of the cell, since NAD cannot diffuse across the IMM thus, allowing the NADH/NAD⁺ ratio to regulate mitochondrial redox homeostasis and OXPHOS [95,96]. A high mitochondrial NADH/NAD⁺ ratio exerts a negative feedback regulation of glycolytic enzymes, such as glyceraldehyde 3-phosphate dehydrogenase and pyruvate dehydrogenase complex (PDC) [97]. The high mitochondrial NADH/NAD⁺ ratio also alters the α-ketoglutarate/citrate ratio and limits entry of acetyl-CoA into the TCA cycle. This results in diminution in the generation of NADH, and flavin adenine dinucleotide (FADH₂) from TCA cycle [96,97]. Modulation in the generation of reducing equivalents, entry of electrons following oxidation of NADH into the ETC, and maintenance of the NADH/NAD⁺ ratio makes complex I a “gatekeeper” of OXPHOS. Changes in mitochondrial NADH/NAD⁺ ratio has been reported during ischemia-reperfusion, dysfunctional cellular metabolism; and it is also a hallmark of diabetes, cancer, neurodegeneration, stroke, and heart failure [[95], [98], [99]]. Many of these pathologies display dysregulated complex I activity, and mitochondrial ROS homeostasis.

3. Complex I ROS

3.1. Production of ROS by complex I

ROS are produced by various enzymatic and non-enzymatic processes in the cell [100,101]. O₂ exists as a paramagnetic biradical with two electrons occupying separate π* orbitals with parallel spins. This feature imposes spin restriction, which makes single-electron reduction of O₂ thermodynamically favored compared with paired-electron reduction [29,102,103]. Premature single-electron reduction of O₂ leads to the production of O₂^•‾, which undergoes rapid dismutation spontaneously or catalyzed by superoxide dismutase (SOD) into H₂O₂ [[102], [103], [104], [105], [106]]. Single-electron reduction of H₂O₂ in the presence of transition metals, such as Fe²⁺ and Cu²⁺ yields hydroxyl radical (^•OH) via the Fenton reaction [104,[106], [107], [108], [109]]. H₂O₂ is not a radical, and crosses membranes through aquaporin-mediated diffusion compared with O₂^•‾ or ^•OH, which have limited membrane permeability due to a negative charge or reactivity, respectively [29,91,102,105,110]. The pKa of O₂^•‾ is ∼5, such that at physiologic pH around 1 % is present as perhydroxyl radical (HO₂^•) which can freely cross membranes, although the importance of this in the overall context of mitochondrial ROS signaling is unclear.

3.2. Sites and modes of ROS production

Among the different (≥41) ROS production sites within the cell, the mitochondrion is a major source of ROS, having up to 16 different sites of ROS production associated with substrate catabolism and OXPHOS [111,112]. Complex I has high ROS production capacity. The FMN-containing NADH docking pocket (site I_F) and site I_Q are the canonical sites of ROS production [29,91,104,113,114]. Complex I predominantly produce O₂^•‾, but also can directly produce H₂O₂ [115]. The ratio of O₂^•‾ to H₂O₂ that is produced is influenced by various factors including substrate type and concentration, NADH₂/NAD⁺ ratio, direction of electron flux, supercomplex formation, polarity of the IMM, redox state of the UQ pool, and O₂ concentration [115,116]. Low (≤3 μM) NADH results primarily in O₂^•‾ production, while high (≥50 μM) NADH levels results in ≥60 % of the ROS being produced as H₂O₂ [115]. Complex I can produce measurable rates of ROS in FET or RET modes. In FET, site I_F produces ROS due to leakage of electrons from the reduced flavin (FMNH or FMNH₂) to O₂. The rate of ROS production depends on high NADH/NAD ratio and the steady-state concentration of FMNH or FMNH₂ [29,113,117] (Fig. 2B). However, complex I ROS production capacity by RET is greater [29,91,104,113,114,118]. Conditions that favor ROS production by RET include high UQ₂/UQ ratio, and presence of Δp. During ischemia (hypoxia), the ETC is hyperpolarized, succinate accumulates due to high UQ₂/UQ ratio and large Δp (Fig. 2C). Upon reperfusion (reoxygenation) there is rapid oxidation of the accumulated succinate leading to rapid ROS burst from complex I [91,[119], [120], [121], [122], [123]]. In RET, both O₂^•‾ (∼65 %) and H₂O₂ (∼35 %) are produced [115]. Recent studies show that site I_Q is the main contributor to ROS production in RET mode (RET ROS) in different tissues including heart [91,119,121,122,124], brain [125]; [null]; [null], kidney [124], skeletal muscles [113,126], and liver [127]. However, site I_F has been reported to also contribute significantly to RET ROS [91,128]. Importantly, in mitochondria oxidizing complex I substrates (glutamate and pyruvate with malate) rotenone-inhibition of electron flux through site I_Q stimulates ROS production from site I_F, and other proximal sites in the TCA cycle [29,113].

ROS production by RET is modulated by the degree and duration of O₂ deprivation [[129], [130], [131], [132], [133]]. The Siesjö group first reported damaging effect of incomplete ischemia (with cerebral blood flow less than 10 %) compared to complete ischemia on rat brain, which they hypothesized to be due to the formation of ROS upon reperfusion [134,135]. Kang et al. [130] have shown that ischemia-reperfusion is more consequential than ischemia alone, and majority of the damage from ischemia-reperfusion injury occurs in the first few minutes of reperfusion. RET ROS production occurs during reoxygenation following ischemic/hypoxic exposure when Δψ_m is not dissipated, but not in anoxic exposure that dissipates the Δψ_m [131]. Additionally, RET ROS production is highly sensitive to O₂ concentration [91,136,137]. Hoffman and Brookes [136] reported an apparent Km of 0.9 μM O₂ for complex I RET ROS production compared with 2.0 μM O₂ for complex III outer ubiquinone-binding site (site III_Qo). This indicates that complex I RET ROS production is more sensitive to O₂ concentration compared with complex III, which also has a high native rate of ROS production capacity and is important in sustained hypoxia signaling.

Following exposure to hypoxia complex I spontaneously transitions to the D-form [73,137] (Fig. 3). Structural rearrangement of site I_Q subunits upon transition to the D-form disrupts electron flux between the N2 terminal and site I_Q similar to the inhibitory effect of rotenone (site I_Q inhibitor) on electron flux [82,83,138]. Transitioning to the D-form is associated with increased rate of NADH-dependent ROS production that is time- and pH-dependent [73,139]. In the D-form, alkaline pH (9.4) unlike physiological pH (7.4) caused significant increase in O₂^•‾ production in rat heart submitochondrial particles. Acute hypoxia increased O₂^•‾ production in the first 10 min of exposure in bovine aortic endothelial cells [140], and after 30 min in mouse hippocampal slices [73]. In the D-form, reoxygenation with addition of succinate showed slower rate of ROS production with an initial lag phase compared with the A-form in rat brain mitochondrial [137,141]. Thus, transitioning to the D-form may serve to protect against oxidative damage due to greater magnitude of ROS release (as bursts) following ischemia-reperfusion in the A-form [142,137]. Furthermore, studies have demonstrated a redox-dependent reversible dissociation of the tightly but noncovalently bound FMN from the N module of the peripheral arm of brain complex I [143,144,133]. This has been reported to be independent of O₂ concentration or ROS release and may serve to maintain the native protein conformation [144,133].

Fig. 3.

Active and deactive forms of complex I. Complex I reversibly transitions from active to deactive states during hypoxia. Hypoxic conditions and limited NADH induce complex I to transition from active to deactive forms, changing protein conformation and altering enzymatic function.

The position of the sites of ROS production in complex I (site I_F at the tip of the N module and site I_Q close to the matrix side of the IMM) results in release of ROS into to the matrix. Studies are now elucidating the importance of complex I site-specific ROS emission in physiology and disease, such that we are beginning to understand what the physiologic consequence would be for ROS made at site I_F compared with I_Q. The impact of ROS from complex I also will be influenced by the duration of release. Key among the protein targets of complex I RET ROS are inner membrane proteins such as cytochrome bc_I complex, complex II flavoprotein, and the electron transferring flavoprotein-ubiquinone oxidoreductase. Others are matrix aconitase, and the 2-oxoacid dehydrogenase complexes including the ROS generating flavin sites of PDC and oxoglutarate dehydrogenase complex (OGDC) [29,[145], [146], [147], [148]]. PDC and OGDC play crucial role in mitochondrial redox sensing through a negative feedback response to complex I [145,[148], [149], [150], [151]]. PDC and OGDC are entry points for pyruvate and amino acids into the TCA cycle, respectively. Ischemia-reperfusion has been shown to inhibit the activity of PDC in canine [152], and rat brain mitochondria [153]. Similarly, the activity of OGDC was reported to be inhibited by ischemia-reperfusion in rat cardiac mitochondria [154]. Inhibition of the activity of these enzymes especially OGDC, will result in reduced supply of NADH [155]. It can thus, be suggested from the reports of Bleier et al. [145], and Tretter and Adam-Vizi [148] that complex I RET ROS-induced inhibition of aconitase, PDC, and OGDC may be protective, serving to limit the supply of reducing equivalents from the TCA cycle to the ETC. Limiting the supply of reducing equivalents to the ETC will in turn reduce ROS production.

Complex I is also a target of ROS-induced inhibition and oxidative thiol-modifications [130,142,156,157]. RET ROS causes oxidative thiol-modification of several of the cysteine residues of complex I resulting in diminution of the catalytic activity of the enzyme complex [158,130,156,159]. Glutathionylation of complex I subunits including NDUFS1 and NDUFV1 leads to inhibition of complex I activity and ROS production [160]. Moreover, the D-form is more susceptible to oxidative thiol-modification [142]. Exposure of the ND3 Cys39 is generally accepted as a marker of transition of complex I into the catalytically inactive D-form. However, a recent report [161] demonstrated that in the catalytically active A-form, about 65 % of Cys39 is exposed compared with about 93 % in the catalytic D-form. ND3 Cys39 is a critical residue, which when exposed in the D-form is readily modified by ROS resulting in inhibition of complex I activity [138,142,162].

Complex I may be distributed individually as single free (uncomplexed) units or may form higher order respiratory supercomplexes in association with complex III referred to as supercomplex, or with complex III and IV referred to as respirosomes [163,164]. Formation of higher order respiratory supercomplexes has been reported to improve the enzyme stability, modulate the catalytic function and reduce formation of ROS [[163], [164], [165]]. In addition, formation of higher order respiratory supercomplexes has been shown to reduce exposure of ND3 Cys39 [161]. Adoption of different conformations, dissociation of FMN and transitioning to the D-form may serve to regulate OXPHOS, maintain redox status, and protect against oxidative damage [null]; [null]; [28]. Thus, complex I shows a high degree of autoregulation thereby modulating the activities of the ETC and ROS production.

4. Complex I ROS signaling

Complex I ROS signaling involves the oxidation of key cysteine residues to trigger downstream signaling [166]. Cysteine is one of the least abundant amino acid residues, but is the most highly conserved and is often seen in functionally important sites of proteins [167,168]. Cysteine accounts for about 2 % of cellular amino acid content, and constitutes the highest concentration of thiols in mitochondria [34,169]. Bak et al. [170] reported ∼1500 reactive cysteine residues on ∼450 mitochondrial proteins; while Danielson et al. [171] reported 130 cysteine residues within murine complex I. Complex I cysteine residues play critical role in diverse functions such as protein assembly and structure, metal binding, catalytic activity and regulation, redox sensing, and signaling [34,91,111,171,172]. Among these are the redox-sensitive cysteines with thiol-containing side chains, which are critical in sensing changes in redox status [173,174].

The thiol groups in redox-sensitive cysteines are deprotonated, and present as highly reactive thiolate anions (RS⁻) at physiologic conditions compared to the protonated (RSH) and largely nonreactive cytoplasmic cysteines [173]. Nominally, the higher pH in mitochondrial matrix would drive a less protonated state, which may render intra-mitochondrial cysteines particularly susceptible to redox reactions [175,176]. Oxidative thiol modifications are generally reversible, making these thiol cysteines act as redox switches. Reversible oxidative modifications of cysteines include but not limited to S-nitrosylation, S-glutathionylation, and sulfenic acid formation [177,178]. The thiolate anions are readily oxidized by ROS to the sulfenic acid intermediate which modulates the effector protein structure and function [34,111,173,174]. However, inordinate and sustained ROS release can progressively oxidize the sulfenic acid intermediates to sulfinic and to the sulfonic acids which, irreversibly deactivates the protein resulting in loss of function [34,174,179]. It is important to note that redox signaling involves reversible oxidative modification of cysteine thiols by H₂O₂ to sulfenic acid, whereas O₂^•‾ causes formation of thiyl radical at a rate constant as high as 10³ M⁻¹ s⁻¹ without specificity. However, higher rate constants (>10⁹ M⁻¹ s⁻¹) for the dismutation of O₂^•‾ by SOD to H₂O₂ and O₂ compared with cysteine reactions limits the signaling function of O₂^•‾ [180]. Overall, ROS signaling depends on location, duration and quantity of production wherein transient and low-level ROS production may be beneficial or physiologic (oxidative eustress), whereas inordinate and sustained release may be detrimental or pathologic (oxidative distress) [[181], [182], [183]]. Localization of ROS release restricts the effect within cellular microdomains. Additionally, antioxidant systems scavenge ROS thereby restricting and fine-tuning the signaling function. Dysregulation of ROS release in favor of prooxidants may cause biomolecular damage associated with the ischemia-reoxygenation injury associated with heart attack and stroke [120,184], neurodegeneration, senescence [[185], [186], [187]], and dysfunction signaling.

4.1. Complex I ROS in adaptive O₂ signaling

Complex I ROS is important in adaptive signaling and have been shown to promote sensory behavior in response to oxidative stress by activating different transcription cascades [188,189]. Complex I RET ROS is critical in maintaining homeostasis, and response to different stressors. Complex I RET ROS initiates metabolic reprogramming by redirecting glycolytic intermediates to the pentose phosphate pathway for the maintenance of NADPH levels and elicits transcriptional changes in response to thermal stress in Drosophila melanogaster [190,191]. Activation of RET ROS signaling has also been demonstrated to diminish with age, and loss of this signaling is detrimental for survival [190,191]. Induction of transient neuronal complex I H₂O₂ release has been reported to enhanced stress resistance and fitness resulting ultimately in the extension of lifespan in Caenorhabditis elegans by activation of PMK-1/p38 MAP kinase and SKN-1/NRF-2 [192].

4.2. Complex I ROS in acute O₂ sensing

Organisms are exposed to local and global O₂ transients, and dependence on O₂ for OXPHOS necessitates mechanisms for detecting (within seconds to minutes) acute changes in O₂ gradients. The ability to sense and respond to changes in O₂ levels is described as O₂ sensing [193]. Acute changes in O₂ gradients elicits physiologic responses such as hypoxic ventilatory response and hypoxic pulmonary vasoconstriction in mammals (HPV) [61,65,[194], [195], [196], [197], [198], [199]], and avoidance locomotory response in C. elegans [57,200]. In mammals, acute O₂ sensing mechanisms is found in strategically located tissues including the O₂-sensitive neuron-like type I cells, also called glomus or chief cells of the carotid body (CB), and aortic body. These are the major systemic chemoreceptor organs that detect hypoxia and hypercapnia [63,64,[201], [202], [203]]. The adrenal medulla (AM) chromaffin cells and pulmonary artery smooth muscle cells (PASMCs) are components of the peripheral chemoreceptor organs forming the homeostatic O₂-sensing system [[61], [62], [63], [64]].

The CB and AM working in concert constitute the CB-AM axis, and function to maintain cardiorespiratory homeostasis [63]. The glomus and chromaffin cells are neural crest-derived electrically excitable cells, that can generate action potentials repetitively [63,64,204]. These cells express a variety of O₂-sensitive ion channels, particularly the potassium ion (K⁺) channels including the voltage-gated K⁺ channels (K_V: K_V 1.5, K_V 1.2, K_V 2.1 and K_V 3.1), maxi-K channels, and background K⁺ channels [63,64,[204], [205], [206], [207]]. These K⁺ channels are reversibly inhibited by acute hypoxia-induced H₂O₂ release resulting in membrane depolarization [63,64,204,[206], [207], [208]]. The hypoxia-induced membrane depolarization activates voltage-dependent Ca²⁺ channels, initiating influx of extracellular Ca²⁺ which, in turn triggers exocytotic release of neurotransmitters [63,64,202,204,208]. The release of neurotransmitters (including ATP and acetylcholine) by the glomus and chromaffin cells stimulates the central respiratory and autonomic centers through the afferent sensory fibers to initiate HPV [63,64,202,204,208,209]. This cascade of events is the basis of the “membrane model” of acute O₂ sensing by peripheral chemoreceptor organs. However, reports indicate that acute hypoxia signaling proximal to the acute O₂ sensitive-K⁺ channels is multimodal, involving processes other than those of the membrane model and have mitochondrial metabolism as a unifying factor [63,64]. Key among these processes is the alteration in mitochondrial metabolism resulting in accumulation of succinate, NADH and lactate, and increased ROS production [63,65,210].

These metabolites may act as upstream signals that modulate the activities of the O₂ sensitive-K⁺ channels. NADH but not NAD⁺, ADP, ATP, Na⁺, or Mg²⁺ have been reported to inhibit the activity of rabbit PASMCs large conductance Ca²⁺-activated K (K_C) channels [211]. The inhibition of PASMCs K_C by NADH was hypothesized to be due to change in the redox state of the channel rather than direct binding of NADH on the channel, suggesting that the K_C may be acting as a redox sensor [211]. In support, Archer et al. [212] showed evidence of a redox-based O₂ sensor in the PASMCs K⁺ channels that depends on mitochondrial ROS release. Use of rotenone or antimycin A (specific inhibitors of mitochondrial complex I and III, respectively) mimicked acute hypoxia by causing a change in the redox status which, in turn reversibly inhibited PASMCs K⁺ channels [212]. Reports show that in response to moderate hypoxia, lactate production is increased as a protective signal to modulate vasodilation [213]. Lactate signaling activates the glomus cell and potentiates hypoxia-induced activation [210]. Lactate activation of mouse and rat glomus cells in response to acute hypoxia was shown to be Ca²⁺ dependent and results in a dose-dependent reversible increase in matrix and IMS ROS release [210]. However, responsiveness to hypoxia requires complex I since ablation of Ndufs2, which encodes the core nuclear-encoded 49 kDa protein necessary for complex I assembly and activity, results in loss of responsiveness to acute hypoxia [27,[60], [61], [62],214,215]. These studies suggest that complex I ROS is key in O₂-sensing.

4.3. Complex I ROS signaling in peripheral acute O₂-sensing

Complex I activity is important in the O₂-sensing function of CB and AM. The CB and AM have unique metabolic and redox status that facilitates their O₂-sensing function. Studies have shown that CB unlike other neural cells have elevated succinate levels [62,216]. Mitochondria of the CB are atypical, expressing both high and low O₂ affinity cytochrome a₃, which are required for O₂-sensing [217,218]. The O₂-sensing of glomus cell is due in part to hypoxia-inducible factor subunit 2 alpha (Hif2α)-dependent expression of atypical nuclear-encoded mitochondrial subunits including NDUFA4l2, COX4i2, and COX8b, which are responsible for the accelerated oxidative metabolism and strict O₂-dependent complex IV activity [65,219]. Hypoxia induces a decrease in complex IV activity and causes a buildup of electrons along the ETC. The resultant high UQH₂/UQ ratio leads to RET ROS and altered complex I activity [63,65]. Thus, hypoxia produces a reversible increase in ROS in the cytosol and IMS, but a decrease in matrix ROS [60,[62], [63], [64], [65],214]. Similar to the reported decrease in matrix ROS, previous studies show that acute hypoxia produced a rapid and reversible reduction in the levels of matrix H₂O₂ in neonatal but not juvenile rat chromaffin cells [220] PASMCs [61,221]. However, using microfluorimetry with genetically-encoded probes, Jiménez-Gómez et al. [214] showed that acute hypoxia produced a reversible increase in IMS and matrix ROS levels in complex I-deficient glomus cells. These studies suggest possible tissue-specific mitochondrial microdomains and compartmentalization of ROS production in response to acute hypoxia.

ROS signaling will depend on the magnitude and duration of ROS release, and the redox poise of the environment [183,222,223]. However, the diffusion of H₂O₂ is limited by antioxidant systems including catalase, and the peroxiredoxin thioredoxin couple [222,[224], [225], [226]]. To overcome this limitation, H₂O₂ may use redox relays mediated by peroxiredoxins and or peroxidases to transmit signals across cellular microdomains or compartments. Here, peroxiredoxins are oxidized to sulfenates and disulfide forms, which in turn oxidize the redox targets [[227], [228], [229], [230], [231]]. Peroxiredoxin, especially mammalian mitochondrial peroxiredoxin 3 plays a key role in mitochondrial ROS redox relays. Peroxiredoxin 3 has been reported to react with about 90 % of mitochondrial H₂O₂ due to relative higher abundance and reactivity [232,233] hence, may be important in the redox relay of complex I H₂O₂ that is release into the matrix. Other peroxiredoxins are also important in functioning as a redox relay. The cytosolic peroxiredoxin 2 acts as a H₂O₂ receptor, which transmits the oxidative equivalents to the signal transducer and activator of transcription 3 (STAT3) [228]. The resultant disulfide-linked dimers and tetramers of STAT3 then regulates the transcription machinery in response to stress [231]. Peroxiredoxins are also important in the “floodgate model” of redox signaling. The peroxiredoxin are proposed in the floodgate model to be inactivated (oxidized to the sulfinic and sulfonic moieties) locally by H₂O₂. This leaves only a remote peroxiredoxin in the redox-active sulfenic form to transmit the redox signal to target proteins [234]. H₂O₂-induced signal transduction is therefore, predicated on the presence of robust antioxidant systems and redox-sensitive proteins particularly the thiol cysteine redox switches in the domain of release. These serve to maintain the localization, rapidity and reversibility of the H₂O₂-induced redox events. Thus, the proximity of mitochondrial ROS release to the redox switches is important in signaling.

Complex I has been demonstrated as the primary source of H₂O₂ signaling in response to acute hypoxia in the peripheral O₂-sensitive chemoreceptor organs. Studies have shown that rotenone treatment in the glomus cells [60,[62], [63], [64], [65]], PASMCs [61], and pulmonary artery myocytes [235,236] abolished the H₂O₂ signal and responsiveness to acute hypoxia. This strongly suggests that ROS release from complex I is important in acute hypoxia signaling. In confirmation of complex I as the source of the H₂O₂ signal, Ndufs2 has been shown to have critical regulatory role necessary for acute hypoxia response in peripheral O₂-sensitive chemoreceptor organs [61,62,64]. NDUFS2 is redox-sensitive being reduced during acute hypoxia. It is required for acute hypoxia-induced intracellular Ca²⁺ concentration ([Ca²⁺]_i) release, and induction of HPV [60,61,64]. Ablation, inactivation or deficiency of Ndufs2 results in loss of acute hypoxia response in peripheral O₂-sensitive chemoreceptor organs and absence of HPV [[60], [61], [62],64,214]. Reversible oxidation of NDUFS2 cysteine residues by complex I H₂O₂ was shown to be responsible for acute hypoxia response [61].

4.4. Putative astrocyte complex I ROS in central acute O₂-sensing

The brain is sensitive to hypoxia due to the high energy demand and minimal energy reserves [[237], [238], [239]]. Additionally, the central nervous system (CNS) is subject to varying levels of oxygenation due to variable degree of blood supply, which cannot be adequately monitored by the peripheral O₂-sensing chemoreceptor organs [239]. Furthermore, the brain shows selective neuronal vulnerability where specific neurons within different regions in the brain such as the cerebellar granule, hippocampus, and amygdala are highly sensitive to oxidative stress [240,241]. These regions coincidentally, have a relatively high astrocyte density [242]. Astrocytes exhibit heterogeneity within and between regions of the CNS indicating functional dynamism in response to the needs imposed on them [243,244]. Astrocytes are a specialized subset of glial cells, which are closely integrated into neural networks having remarkable adaptive plasticity enabling them to support neuronal viability, metabolism, and function in addition to regulating CNS homeostasis [244,245]. Astrocytes in the different brain regions show varying sensitivity to O₂ deprivation, and H₂O₂ exposure. Among different brain regions, hippocampal astrocytes show more sensitivity to O₂ deprivation compared with cortical or striatal astrocytes in rats [246,247]. Contrastingly, cortical astrocytes show more sensitivity to H₂O₂ compared with hippocampal or striatal astrocytes [246].

Astrocytes have a metabolic signature distinct from neurons. Astrocytes have a functional ETC, but depend more on glycolysis compared with neurons [[248], [249], [250], [251], [252]]. Furthermore, astrocytes unlike neurons have the capacity for FAO, and oxidize fatty acids as an alternative fuel to augment energy supply during metabolic stress [[253], [254], [255]]. FAO has been shown to increased ROS release in mouse astrocytes [255]. Astrocytes reportedly have an abundance of complex I in the free form that are not assembled as supercomplex in contrast to neurons where complex I are assembled into supercomplex [251]. The free complex I contributes to the higher ROS production capacity of astrocytes compared with neurons [251]. Moreover, neurons have lower antioxidant levels and are more sensitive to oxidative stress than astrocytes. Additionally, neurons that produce more mitochondria-derived ROS relative to other neurons are more vulnerable to oxidative stress [[256], [257], [258]]. The distribution and function of astrocytes within specific chemoreceptor sites in the brain have been suggested to mediate the response to ROS release and redox stress [[259], [260], [261], [262]]. Astrocytic mitochondrial ROS has been reported to modulate neuronal metabolism and function, and also alter organismal behavior [248,263,264].

To maintain neuronal homeostasis, astrocytes function as key chemosensors for CO₂ and pH, which are linked to Ca²⁺ signaling [[265], [266], [267]]. Astrocytes are positioned closer to and are part of the blood-brain barrier, and thus are subject to greater fluctuations in pO₂ in circulation compared to neurons, which are buffered from this by glial cells [268,269]. Studies are now highlighting the role of astrocytes as central O₂-sensing chemoreceptors in addition to other functions. O₂-sensing has been demonstrated in vivo and in cultured cortical, hippocampal, midbrain and brainstem astrocytes [270]. In their study, Angelova et al. [270] showed that hypoxia resulted in elevated [Ca²⁺]_i and increased exocytosis of ATP-containing vesicles. Their study showed that astrocytic O₂-sensing is dependent on mitochondrial ROS and Δψ_m. Using MitoSOX, a mitochondria-targeted fluorescent probe, they Angelova et al. [270] showed an immediate increase in mitochondrial ROS in astrocytes in response to hypoxia. The observed increase in ROS was reduced by pretreatment with 0.5 μM carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP, a mitochondrial uncoupler) or 100 nM MitoQ (a mitochondrial-targeted antioxidant). Treatment with 0.5 μM FCCP or 100 nM MitoQ in vivo and in cultured astrocytes abolished the hypoxia-induced increase in [Ca²⁺]_i. Simultaneous measurement of Δψ_m and [Ca²⁺]_i in cultured rat astrocytes using Rh123 and fura-2, respectively, showed hypoxia induces a decrease in Δψ_m that precedes increase in [Ca²⁺]_i [270]. Supporting the central O₂-sensing role of astrocytes [271], showed that nucleus of the solitary tract in neonatal rat brain responded to acute hypoxia by altering Ca²⁺ activities. Astrocytes of the ventral surface of the medulla oblongata have also been shown to act as central O₂-sensing chemoreceptors. In these astrocytes, acute hypoxia inhibits the O₂-induced suppression of sensor cation channel transient receptor potential (TRP) ankyrin 1 (TRPA1) activity. Complementing the report of Angelova et al. [270], rat cortical astrocytes in vivo have been shown to respond to decreases in brain perfusion with increased frequency and duration of [Ca²⁺]_i signals, which are required for compensatory sympathetic and cardiovascular response to cerebral ischemia [272]. Acute hypoxia-induced mitochondrial ROS release was demonstrated to contribute to activation of plasma membrane-inserted TRPA1 resulting in Ca²⁺ influx, which leads to ATP release from these astrocytes [239,273]. It is therefore, possible that complex I may contribute to mitochondrial ROS release necessary for increased [Ca²⁺]_i, subsequent activation of inositol triphosphate (IP₃) receptors, and ultimately to vesicular ATP release in astrocytes [270,274,275].

Angelova et al. [270] demonstrated that astroglia have a lower pO₂ threshold for activating [Ca²⁺]_i signaling of about 17 mmHg compared with about 37 mmHg for glomus cells. This suggests that astrocytes may detect O₂ levels below the threshold for glomus cells. Supporting the findings of Angelova et al. [270], Barioni et al. [276] showed that intermediolateral nucleus astrocytes contribute to spinal O₂ sensing in rats. Their study showed that acute hypoxia activated [Ca²⁺]_i signal through ROS-dependent TRP channels in addition to other mechanisms. The spinal O₂ sensing capacity was reported to be similar to that of the carotid bodies [276].

The central respiratory O₂ sensing function of astrocytes has been suggested to be in part due to the mitochondrial complex I-derived ROS [251]. Specifically [251], showed that higher rate of H₂O₂ release by astrocytes compared with neurons is not due to contributions from other major nonmitochondrial ROS sources such as xanthine oxidase, NADPH oxidases, and nitric oxide synthase. Interestingly, astrocytes that utilize glycolysis for energy metabolism showed similar activities for different ETC enzymes including complex I compared with neurons. However, the higher rate of ROS release in glycolytic astrocytes may be partly due to the lower abundance of NDUFS1 subunit in astrocytes [251]. NDUFS1 expression was demonstrated to be necessary for assembly of supercomplexes, higher electron transfer efficiency, and lower rates of ROS release. Corroborating this [277], showed reduced Ndufs1 expression in myocardium of human heart failure patients and in mice with myocardial infarction. However, overexpression of Ndufs1 prevented the hypoxia-induced high ROS rates in mice [277]. Furthermore, a biallelic mutation in nuclear Ndufs1 in a human patient has been reported to cause reduced complex I N-module stability resulting in impaired supercomplex formation, reduced electron transfer rate between N4 and N5 subunits of the Fe–S cluster, upregulation of glycolysis, and increased ROS production [278]. Altogether, it appears that the lower abundance of NDUFS1 may be a default redox poise of astrocyte complex I, which predisposes it to higher rates of ROS release. However, whether astrocyte complex I ROS release is the proximal signal in central O₂-sensing remains to be conclusively demonstrated.

Similar to the brain, the retina has high metabolism with low functional O₂ reserve and is considered one of the most O₂-sensitive tissues [279,280]. Retinal mitochondrial-dense photoreceptor cells canonically depend on oxidation of glucose via glycolysis for energy metabolism and the products, lactate and aspartate are then oxidized by Müller glia cells [281]. In addition, the free fatty acid receptor 1 senses availability of fatty acids and activates FAO energy metabolism [282]. Interestingly the glycolytic retinal photoreceptor cells are vulnerable to O₂ deprivation, and ischemia-reoxygenation damage that is typical of complex I RET ROS [280,283,284]. Recent studies demonstrate that mitochondrial ROS following ischemia-reperfusion is responsible for the death of retinal ganglion cells, which can be inhibited by metformin or mito-TEMPO [284]. Response to acute hypoxia in the retina involves promoting blood flow by mechanisms yet to elucidated [280]. In the optic nerve, astrocytes have also been proposed to function as hypoxia sensors in the unmyelinated nerves [285].

Complex I has been shown to be important in astrocyte function. Mutation in different subunits of complex I leads to various neurodegenerative conditions. Mutations in various complex I subunits including NDUFS1 [251], NDUFS2 [[286], [287], [288]], NDUFS4 [22,289,290], ND1 [291], ND3 [292,293] results in defective complex I assembly, altered mitochondrial activity and dysregulation of ROS production. Further, loss of Ndufs4 results in significant diminution of astrocyte complex I activity [294]. Similarly, exposure to pharmacologic inhibitors of complex I diphenyleneiodonium or rotenone inhibited hypoxia sensing and HPV in rat pulmonary myocytes implicating RET ROS in O₂-sensing [235,236]. Furthermore, using suppressors of site I_Q electron leak (S1QEL), which do not cause ROS production or impair mitochondrial function, Brand et al. [295] showed that site I_Q is a major source of RET ROS in astrocytes exposed to ambient and low O₂ tension conditions. In support of the role of complex I ROS in signaling acute changes in O₂ tension, Read et al. [296] demonstrated that S1QEL inhibited the rapid increase in ROS production and the subsequent elevation of [Ca²⁺]_i following normoxic exposure in human ductus arteriosus smooth muscle cells (DASMC). The DASMCs contain O₂ sensors that respond to elevation of pO₂ leading to constriction of the ductus arteriosus following the first breath, and inhibition of ROS production from complex I with S1QEL prevented closure of rabbit ductus arteriosus in vivo [296]. They [296] demonstrated that inhibition of complex I ROS release specifically, and not complex III ROS is responsible for O₂ signaling. While astrocytes function to protect neurons and retinal photoreceptor cells from ROS, excessive ROS release due to mutations in complex I subunits may overwhelm the Ca²⁺ or ROS-buffering capacity of glial mitochondria, thereby activate these astrocytes. The activated astrocytes then release chemokines and inordinate levels of ROS ultimately leading to neuronal and retinal degeneration [297,298]. Thus, elucidating the involvement of complex I in astrocyte O₂-sensing will be valuable in understanding the pathophysiology of oxidative neuronal and retinal degeneration.

5. NDUFS2 residues mediate acute hypoxia behavioral response

Altogether, mitochondrial complex I activity is key in peripheral and central respiratory O₂ sensing tissues. Critical for the assembly and function of complex I is the ubiquitously expressed NDUFS2, which is located at the interface of the membrane and matrix arm and constitutes part of the ubiquinone binding pocket together with ND1 and NDUFS7 [62,75,78,215,287,299]. Loss of Ndufs2 has been shown to induce the degradation of complex I and disappearance of other subunits including NDUFS1, NDUFV2, NDUFS4 and NDUFB8 [62,299,300]. Demonstrating the importance of NDUFS2 in mitochondrial complex I physiology, Cabello-Rivera et al. [301] reported that loss of NDUFS2 resulted in perinatal death, defective proliferation of neural progenitor cells, and dysfunctional differentiation of neurons and oligodendrocytes in NDUFS2-knockout mice. In support, the disruption of NDUFS2 was also reported to induce a metabolic shift that downregulated complex I driven respiration and promoted glycolysis in neuronal mitochondria [302]. Disruption of NDUFS2 in human embryonic kidney cell line 293 (HEK293) resulted in reduction in growth, complex I-supported respiration, glycolytic capacity, cell membrane integrity but increased complex II-supported respiration and ROS production [300]. Disruption of Ndufs2 resulted in altered ROS homeostasis and loss of acute hypoxia behavioral response [[60], [61], [62]] whereas inactivation of Ndufs4 resulted only in about 50 % decrease in complex I activity without loss of hypoxia response [60,62]. This suggests that ROS from complex I may be acting locally, producing redox events within complex I putatively on NDUFS2 to elicit acute hypoxia responses.

NDUFS2 contains a number of cysteine residues that undergo oxidative modification resulting in altered complex I activity. The location and interaction of NDUFS2 cysteine residues with neighboring redox active amino acids influences the extent to which NDUFS2 can modulate redox events. The privileged location of NDUFS2 brings it in contact with other subunits and exposes it directly to ROS from complex I especially from site I_Q. NDUFS2 contacts and interacts with a number of other core and accessory subunits including but not limited to ND1, ND3, NDUFS7, NDUFS3, NDUFS6, NDUFA5 and NDUFA10-13 [60,303] (Fig. 2A). These interactions create a local electrostatic environment produced by the charge differentials of different residues in these subunits. This electrostatic environment permits fine-tuning of redox active residues, which in turn induces modification of other neighboring redox active residues [166,304,305]. Also, charged modifications of residues produce conformational transitions of the individual subunits that is propagated to other subunit leading to change in catalytic activity of the complex I [70,77,306]. For example, the loop between the first and second β-strands in NDUFS2 (β1- β2 loop) which contains highly conserved histidine residues critical for UQ binding and catalysis shows great conformational dynamism [5,307]. Changes in the conformation of this loop may induce alteration in complex I activity [308] that may lead to altered ROS homeostasis.

NDUFS2 cysteine residues including Cys146, Cys326, and Cys347 are reported to be susceptible to direct oxidative modification resulting in conformational change of the subunit and altered complex I activity [158,130]). The Cys146 residue is buried inside NDUFS2 in contrast to Cys326 and Cys347 residues, which are surface-exposed and located at the same α-helix of NDUFS2 [130]. The surface exposed Cys326 and Cys347 undergo oxidative modification following ischemia-reperfusion in rats, which may indirectly lead to oxidation of Cys158 and Cys188 at the N2 cluster of NDUFS7 [130]. Redox proteomics assay for S-glutathionylation of mouse gastrocnemius muscles after fatigue show that NDUFS2 Cys347 was among the cysteines of complex I subunits that was oxidatively modified [309]. Similarly, using quantitative redox proteomics analysis with GELSILOX of hypoxic mitochondrial protein of COX10 knockout mouse skin fibroblast cells, Guarás et al. [33] show that the Cys347 detected is mainly in the reduced form and undergoes oxidation following reoxygenation (Fig. 4A).

Fig. 4.

An evolutionarily conserved redox-sensing cysteine in NDUFS2. A NDUFS2 interacts with the final Fe–S cluster N2, critically stabilizing its assembly and stability to support electron flow through Complex I. The Cys347 (Cys366 in C. elegans) residue is susceptible to redox modification and important in oxygen-sensing. The position of Cys347 residue is highly conserved across NDUFS2 structure. B Cys347 and surrounding residues are highly conserved in listed species (indicated by color, paired with A).

Recent studies using both genetic and pharmacologic interventions demonstrate the involvement of NDUFS2 subunit in complex I ROS-mediated behavioral response to acute hypoxia in C. elegans [57]. Acute hypoxia in C. elegans produces avoidance behavior characterized by rapid and transient increase in locomotory speed and turning angle [310]. In C. elegans mutants, GAS-1 and ISP-1 were identified by Zhao et al. [200] as mitochondrial components required for acute hypoxia response. Mutation in gas-1, the C. elegans ortholog of Ndufs2, increased ROS levels and entirely abolished the locomotory response to acute hypoxia. To confirm ROS is mediating the behavioral response, sod-2 (encoding primarily mitochondrial manganese-dependent SOD) mutant C. elegans showed partially inhibited locomotory response to acute hypoxia. However, mutation of all SOD genes produced a complete inhibition of acute hypoxia response [200]. C. elegans has five sod genes: sod-1, sod-2 and sod-4, which are the primary cytoplasmic, mitochondrial and extracellular sod genes, respectively; while sod-3 and sod-5 are inducible mitochondrial and cytoplasmic genes, respectively. Loss of any individual SOD may be compensated by presence or upregulation other SOD genes [311,312]. Therefore, the partially inhibited locomotory response to acute hypoxia with sod-2 mutants reported by Zhao et al. [200] may be due compensation by other SOD genes.

Comparable to the loss of acute hypoxia response with sod-2 mutants reported by Zhao et al. [200], Onukwufor et al. [57] also reported that loss of sod-2 and sod-2/-3 abolished activation of acute hypoxia locomotory response. The two mitochondrial isoforms of SOD, sod-2 and sod-3 have been reported to associate with supercomplexes in C. elegans and in mouse mitochondria [311]. Loss of sod-2 resulted in damage to proteins specific to the ETC, dissociation of supercomplexes, decrease in NDUFS2, and reduction in complex I function [311]. Similarly, previous reports show ROS-induced reduction of complex I and mitochondrial aconitase activity in sod-2 mutant mouse heart and brain mitochondria [313,314]. This is supported by reports that loss of SOD results in absence of the acute hypoxia behavioral response since SOD dismutates the precursor O₂^•‾ to H₂O₂ and O₂. This supports the finding by Onukwufor et al. [57] that the acute hypoxia avoidance behavior in C. elegans is mediated by specifically by H₂O₂. These reports show a conserved protective location of SOD at sites of high ROS emission that serves to maintain a redox environment conducive for ROS signaling. Therefore, O₂^•‾ may be primarily a precursor for H₂O₂ and not necessarily involved directly in signaling. Together these studies strongly suggest that H₂O₂ produced locally at complex I may be the proximal signal in acute hypoxia sensing.

Treatment with known complex I redox cyclers such as paraquat, and juglone generates ROS increased release. Treatment with 200 mM paraquat, 1 mM juglone, and 10 μM rotenone for 10, 2 and 60 min, respectively in the bacterial food abolished the locomotory response to acute hypoxia [200]. Interestingly, Onukwufor et al. [57] reported that treatment with lower doses of paraquat (1 mM) or rotenone (1 μM) 24 h prior to assay resulted in increased C. elegans locomotion. The observed differences in locomotory behavior may be as a result of dose- and time-dependent effect of the pharmacological agents. In C. elegans, the increase in ROS emission following mutation of isp-1, the ortholog of complex III Rieske Fe–S protein, was abolished by rotenone [200]. This recapitulates the hypothesis that complex I ROS may be the proximal signal mediating acute hypoxia response or that electrons from complex I are required for complex III ROS since rotenone, which is a specific inhibitor of site I_Q abolishes the locomotory behavioral response to acute hypoxia.

UQ and inhibitors of site I_Q including rotenone, piericidin A1, aureothin and pyridaben are believed to bind to His59 and Tyr108, key residues of NDUFS2 that form part of the amphipathic redox active site of Q-channel [75,78]. Using simulations of mouse complex I, rotenone binding has been shown to reorder the NDUFS2 β1-β2 loop of the Q-channel resulting in structural changes similar to that observed during transition from the open to closed states of complex I [78]. While the use of pharmacological inhibitors highlights the importance of NDUFS2; use of these toxins may produce off-target effects and does not provide spatial or temporal control. The use of optogenetics confers better spatiotemporal control in the investigation of complex I ROS signaling compared with the use of genetics or pharmacological inhibitors. With SuperNova, an optogenetic ROS-generating monomeric variant of KillerRed fused to NUO-1 (NADH:ubiquinone oxidoreductase-1; mammalian ortholog, NDUFV1, encoding complex I site I_F) Onukwufor et al. [57] showed that ROS emanating from complex I mediated a rapid reversible acute hypoxia-induced behavioral response in C. elegans. They [57] show that O₂^•‾ specifically, from the matrix arm of complex I is converted to H₂O₂ to signal change in the membrane arm to ultimately evoke the organismal behavioral response. Additionally, with SuperNova tagged to complex II or complex III, the magnitude of locomotory response to acute hypoxia was not recapitulated compared with SuperNova tagged to complex I [57]. This indicated the specificity and compartmentalization of local complex I ROS in signaling.

The acute hypoxia avoidance behavior was shown by Onukwufor et al. [57] to be due to reversible oxidative modification of a single cysteine, Cys366 (Cys347 in mammals) residue in complex I (Fig. 4b). They [57] reported that mimicking redox modification of a key cysteine residue (Cys366) of NDUFS2 to serine (Cys366Ser, non-oxidizable) or aspartate (Cys366Asp, oxidized) altered organismal behavior in response to ROS. The ROS generated following photoactivation of SuperNova produced avoidance behavior in C. elegans similar to that induced by paraquat [57]. This shows that reversible oxidative modification of a single NDUFS2 cysteine residue modulates complex I activity and mediates rapid organismal behavioral response to acute hypoxia.

The identification of NDUFS2 Cys366 in mediating acute hypoxia behavioral response opens opportunity for interventions in pathophysiologies involving complex I ROS metabolism. The conserved nature of NDUFS2 Cys366 (Cys347 in mouse, rat, bovine and human) (Fig. 4B), the importance of this residue in complex I stability and function, and proximity of Cys366 to the site of ROS production in complex I indicate a sentinel-like function as an “early warning sign” for hypoxia signaling. However, how the reversible modification of this single cysteine residue transmits signal out of the mitochondrial to the whole organism remains an important question for further investigation.

6. Concluding remarks

Recent advances into understanding O₂ sensing has made tremendous progress, particularly highlighting the role of mitochondrial complex I during acute hypoxia. This has been made possible by understanding the complex functional morphology, redox biology, and ROS production mechanisms in mitochondrial complex I. Here, we highlighted modalities of ROS production from complex I and the involvement in hypoxia signaling. Importantly, we present recent progress in the understanding of complex I H₂O₂ hypoxic signaling in peripheral and central O₂-sensing chemoreceptor tissues. Use of novel techniques such as live in vivo optogenetics to target ROS production and measurement to select sites with spatiotemporal precision. Understanding how complex I H₂O₂ signaling mediates behavioral response to acute hypoxia is of importance in future studies and in the development of new therapeutic interventions in conditions associated with hypoxia, ischemia-reoxygenation injuries in stroke, and other pathologies associated with altered ROS homeostasis including neurodegeneration and aging.

Author contributions

C.N.O. wrote the manuscript. C.N.O. and S.A.K. made figures. S.A.K., and A.P.W. edited the manuscript and figures. All authors read and agreed to the published version of the manuscript.

Funding

A.P.W. is supported by grants from National Institutes of Health (R01 NS092558, R01 NS115906).

Declaration of competing interest

The authors declare that they have no conflicts of interest that could influence the contents of this article.

Abbreviations

(OXPHOS)

Oxidative phosphorylation

(ETC)

electron transport chain

(Δp)

protonmotive force

(NADH)

reduced nicotinamide adenine dinucleotide

(ROS)

reactive oxygen species

(ATP)

adenosine 5′ trisphosphate

(NADH)

reduced nicotinamide adenine dinucleotide

(TCA) cycle

tricarboxylic acid

(FAO)

fatty acid β-oxidation

(UQ)

ubiquinone

(UQH₂)

ubiquinol

(O₂)

oxygen

(H₂O)

water

(H⁺)

proton

(NAD⁺)

oxidized nicotinamide adenine dinucleotide

(O₂^•‾)

superoxide anion

(H₂O₂)

hydrogen peroxide

(redox)

reduction-oxidation

(Ca²⁺)

calcium

(Na⁺)

sodium

(Fe²⁺)

iron

(Mg²⁺)

magnesium

(Mn²⁺)

manganese

(IMM)

inner mitochondrial membrane

(IMS)

intermembrane space

(Δψ_m)

electrochemical potential

(ΔpH)

pH gradient

(ADP)

adenosine 5′ diphosphate

(kDa)

kilodalton

(DNA, nDNA)

nuclear deoxyribonucleic acid

(mtDNA)

mitochondrial DNA

(FMN)

flavin mononucleotide

(Fe–S)

iron-sulfur

(A-form)

active form

(D-form)

deactive, dormant form

(UQH^•−)

semiubiquinone radical

(FET)

forward electron transfer

(RET)

reverse electron transfer

(PDC)

pyruvate dehydrogenase complex

(acetyl-CoA)

acetyl-coenzyme A

(FADH₂)

reduced flavin adenine dinucleotide

(SOD)

superoxide dismutase

(^•OH)

hydroxyl radical

(HO₂^•)

perhydroxyl radical

(RET ROS)

ROS production in RET

(OGDC)

oxoglutarate dehydrogenase complex

(RS⁻)

reactive thiolate anions

(RSH)

protonated thiol

(HPV)

hypoxic pulmonary vasoconstriction

(CB)

carotid body

(AM)

adrenal medulla

(PASMCs)

pulmonary artery smooth muscle cells

(K⁺)

potassium ion

(Hif2α)

hypoxia-inducible factor subunit 2 alpha

(STAT3)

signal transducer and activator of transcription 3

([Ca²⁺]_i)

intracellular Ca²⁺ concentration

(CNS)

central nervous system

(FCCP)

carbonyl cyanide p-trifluoromethoxyphenylhydrazone

(MitoQ)

mitochondria-targeted antioxidant

(TRPA1)

transient receptor potential (TRP) ankyrin 1

(IP₃)

inositol triphosphate

(mito-TEMPO)

mitochondria-targeted SOD mimetic

(S1QEL)

suppressors of site I_Q electron leak

(DASMC)

ductus arteriosus smooth muscle cells

(HEK293)

human embryonic kidney cell line 293

Data availability

No data was used for the research described in the article.