Progress in understanding the molecular oxygen paradox – function of mitochondrial reactive oxygen species in cell signaling

T-cell activation

T-cells form a critical arm of the adaptive immune response. Antigen stimulation induces rapid T-cell division and differentiation into effector cells. T-cell activation is also accompanied by a dramatic shift in metabolism. Quiescent T-cells adopt a catabolic state where carbon is metabolized through the Krebs cycle, forming electron carriers that are oxidized to make ATP for basic housekeeping functions (Sena et al., 2013). T-cell activation results in a shift towards anabolic metabolism, where available carbon in the Krebs cycle is utilized to provide the constituents required for rapid cell division (Sena et al., 2013). ATP demands are met by an increase in glucose uptake and glycolysis. T-cell activation relies on the cross-linking of T-cell receptors (TCR), which induces a phosphorylation cascade that triggers the expression of factors involved in T-cell activation (Macian, 2005). Intriguingly, TCR activation is also associated with an increase in cell ROS production (Devadas et al., 2002). Through a set of ingenious experiments, Sena and coworkers found that this source of ROS was mitochondria, in particular complex III (Sena et al., 2013). Using mitochondria-targeted redox-sensitive GFP (mito-roGFP), it was found that T-cell activation results in increased mitochondrial H₂O₂ formation (Sena et al., 2013). Supplementation of cells with mitochondria-targeted antioxidants curtailed T-cell activation, which was also associated with attenuation of interleukin-2 production (Sena et al., 2013). In addition, T-cell-specific reduction of complex III curtailed mitochondrial ROS production and the induction of the adaptive immune response (Sena et al., 2013). It is important to note, however, that this study does not directly identify which mitochondrial ROS, O₂˙⁻ or H₂O₂, is required for T-cell activation. However, H₂O₂ is the most likely candidate because it has been found to modulate a number of transcription factors and proteins involved in various cellular processes through redox modification of cysteine residues (otherwise known as redox switches). ROS from complex III was required for NFAT activation and interleukin-2 production, two factors that are vital for T-cell activation and proliferation (Sena et al., 2013). In another study it was found that O₂˙⁻ production by complex I also stimulates NF-κB and AP1 signaling pathways, which are required to induce the expression of interleukins (Kaminski et al., 2010). Another intriguing aspect associated with T-cell activation is the upregulation of uncoupling protein-2 (UCP2), which plays a putative role in regulating mitochondrial O₂˙⁻/H₂O₂ production through the induction of proton return to the matrix. Antigen stimulation of T-cells is accompanied by an upregulation in UCP2 mRNA and protein expression (Chaudhuri et al., 2016). Quantitative expression analyses have revealed that UCP2 is mainly expressed in organs, cells, and tissues involved in the immune response (Rupprecht et al., 2012). Although speculative, it is possible that an increase in UCP2 expression after T-cell stimulation may be required to desensitize H₂O₂ signals once it has stimulated cell proliferation. This would allow mitochondria to control how much O₂˙⁻/H₂O₂ is formed, ensuring that the activated T-cell can take advantage its signaling properties whilst avoiding oxidative distress.

Adipocyte differentiation

White adipocytes are integral endocrine cells and organismal energy reservoirs. Adipocyte differentiation from mesenchymal stem cells (MSC) hinges on the activation of the transcription factors, C/EBPα and PPARγ. Stimulation of adipogenesis requires the activation of mammalian target of rapamycin (mTOR) complex-1 (mTORC1), which is mediated through the insulin and insulin-like growth factor stimulation of Akt signaling (Ricoult and Manning, 2013). Following its activation, mTORC1 elicits downstream effects that promote adipocyte differentiation (e.g. induction of C/EBPα and PPARγ). It has been found that mitochondrial respiration and ROS production also increase during adipogenesis (Wilson-Fritch et al., 2003; Imhoff and Hansen, 2010). In one particular study, it was found that stimulation of mTORC1 results in an increase in mitochondrial metabolism and biogenesis, which was also associated by an increase in O₂˙⁻/H₂O₂ production (Tormos et al., 2011). Differentiation of MSC correlated strongly with an increase in H₂O₂ levels, which could be inhibited with mitochondria-targeted antioxidants (Tormos et al., 2011). Moreover, MSC cultures treated with mitochondria-targeted antioxidants also diminished the expression of several transcriptional regulators that facilitate adipocyte differentiation (Tormos et al., 2011). Finally, supplementing MSC cultures exposed to mitochondria-targeted antioxidants with galactose oxidase/galactose, which produces H₂O₂, facilitated adipocyte differentiation and lipid accumulation. It was also shown that the elevation of cellular O₂˙⁻/H₂O₂ was an early event during adipogenesis pointing to its potential function in serving as a co-signal for differentiation (Tormos et al., 2011). It should be emphasized at this point that cellular H₂O₂ is well documented to play a role in cell differentiation (Weinberg et al., 2015). By knocking down Rieske iron sulfur protein (RISP), which prevents the accumulation of semiquinone and O₂˙⁻ formation, the authors were able to show that the source of ROS was complex III of the electron transport chain (Tormos et al., 2011). Thus, it would appear that mitochondrial H₂O₂ plays a role in reinforcing signaling programs in differentiating adipocytes. Hydrogen peroxide has also been found to induce the phosphorylation and activation of Akt and the autophosphorylation of insulin receptor, both of which play a part in the induction of adipocyte differentiation (Papaconstantinou, 2009; Sadidi et al., 2009). Thus, it is possible that mitochondrial H₂O₂ release serves as a positive feedback signal that prolongs the induction of adipocyte differentiation cascades. Interestingly, white adipocytes also express UCP2, which may be required to desensitize the ROS signal emitted by mitochondria (Anedda et al., 2008). A fairly recent study found that UCP2 displays variable expression in pluripotent stem cells and its upregulation limits H₂O₂ production, inhibiting cellular differentiation (Zhang et al., 2011). Thus, it is possible that UCP2 is either utilized to curtail adipogenesis or is expressed after differentiation to quench the ROS signal.

Steroidogenesis in the adrenal cortex and circadian rhythms

Peroxiredoxins are a family of H₂O₂ clearing enzymes that are well conserved from archaea to humans (Perkins et al., 2014). There are six known mammalian isozymes (PRX1-6), with PRX3 and PRX5 found in the matrix of mitochondria (Zhu et al., 2012). Aside from H₂O₂ degradation, PRX proteins have also been found to mediate cell signals. This can be achieved either by (1) serving as a redox relay where the disulfide formed on PRX is then transferred to a transcription factor (Sobotta et al., 2015) or (2) via the floodgate model where overoxidation of PRX prolongs its deactivation allowing for an accumulation of H₂O₂ for signaling (Jarvis et al., 2012). Although the former is intriguing, for our purposes here we will focus on mechanism number two since it involves mitochondria. The normal catalytic cycle of PRX involves the H₂O₂-mediated oxidation of a catalytic thiol, which forms a sulfenic acid (SOH) moiety that is quickly resolved by a second cysteine residue. This results in the formation of a disulfide bridge that is reduced in an NADPH- and TRX-dependent fashion. If H₂O₂ levels are high enough, SOH can be oxidized to a sulfinic acid (SO₂H) (Rhee and Kil, 2017). Although SO₂H groups can be reduced back to an active cysteine thiol by sulfiredoxin (SRX), the reaction is slow. Thus, the overoxidation of PRX can prolong its deactivation allowing H₂O₂ levels to accumulate eliciting a cell signaling response (Rhee and Kil, 2017). Once sulfiredoxin successfully reduces SO₂H and reactivates PRX, cellular H₂O₂ can be cleared desensitizing the signal.

The reaction mechanism described above mostly applies to the cytosolic PRX isoforms, PRX1 and PRX2 (Rhee and Kil, 2017). However, recent work has found that this concept also applies to mitochondrial PRX isoform, PRX3. Kil and colleagues found that the over-oxidation of PRX3 is instrumental for the mitochondrial H₂O₂ mediated regulation of corticosterone production in the adrenal cortex (Kil et al., 2012). The production of glucocorticoids is activated by the adrenocorticotropic hormone (ACTH) response – ACTH secreted from the pituitary gland activates steroidogenesis in the adrenal cortex (Kil et al., 2012). Induction of steroidogenesis requires the production of cAMP, which stimulates the synthesis of corticosterone from cholesterol by activating cytochrome P450 (CYP) enzymes (Kil et al., 2012). Intriguingly, one of these CYP enzymes, called CYP11B1, is found in mitochondria where it catalyzes the last step of corticosterone synthesis. CYP enzymes are notorious sources of ROS. It was also found that oxidative stress can hinder steroidogenesis in the adrenal cortex by decreasing the availability of steroidogenic acute regulatory protein (StAR), which plays a vital role in the final step of corticosterone biosynthesis (Diemer et al., 2003). High ROS also activates p38 mitogen activated protein kinase, which inhibits StAR expression (Abidi et al., 2008). Bringing this full circle, Kil and coworkers found that, during steroidogenesis, the H₂O₂ formed by CYP11B1 in the matrix of mitochondria overoxidizes PRX3 (Kil et al., 2012). This allows H₂O₂ accumulation and diffusion into the cytosol where it activates p38 inhibiting StAR and corticosterone production (Kil et al., 2012). Therefore, H₂O₂ formed by CYP11B1 during steroidogenesis serves as a feedback inhibitor for corticosterone production. One potential issue associated with these findings is that PRX3, unlike PRX1, is highly resistant to overoxidation (Cox et al., 2009). A viable way to explain how PRX3 is deactivated is that the H₂O₂ levels become quite high during corticosterone production, allowing for PRX3 hyperoxidation (to our knowledge the absolute concentration of H₂O₂ in the adrenal cortex during steroidogenesis has not been quantified). This could indeed be the case since in the same study the authors found that ~20% of the PRX3 pool was in the sulfinic acid form whereas it was not detected in other tissues (Kil et al., 2012).

In the same study published by Kil and colleagues, it was also found that PRX3 oxidation displays a circadian pattern (Kil et al., 2012). In the adrenal cortex, the sulfinic acid form of PRX3 was at its lowest at 12 h; this coincided with the mitochondrial uptake of SRX and a decrease in phosphorylated p38 (Rhee and Kil, 2016). After 24 h, SRX levels decreased, which was matched by an increase in PRX3 hyperoxidation and p38 activation. Mitochondrial import of SRX relies on HSP90 and its subsequent removal is facilitated by Lon protease (Rhee and Kil, 2016). Comparable results were obtained with brown adipose and heart tissue, suggesting that various tissue types utilize PRX3 oxidation and H₂O₂ signaling as a common mechanism for the circadian rhythm-mediated regulation of cell physiology.

Glucose stimulated insulin release

Insulin is likely the most discussed hormone in the scientific and public sphere since lack of production is associated with the development of two of the most common metabolic disorders: type 1 and type 2 diabetes. This is simply because it serves as the linchpin for the regulation of overall body metabolism after a meal. It has been known for some time that insulin release from pancreatic β-cells is stimulated by a rise in blood glucose levels. Briefly, glucose is imported into the cytoplasm of β-cells after a meal and is metabolized through glycolysis and oxidative phosphorylation forming ATP, inducing insulin release (Jensen et al., 2008). Other secretagogues, like fatty acids and amino acids, have also been found to potentiate insulin release, mostly during the second phase insulin secretion (Andrikopoulos, 2016). Insulin secretion is also sensitive towards redox signals, in particular mitochondrial H₂O₂ (Leloup et al., 2009). Indeed, exposure of β-cells to high glucose (5–16.7 mm) induces a robust increase in cellular ROS and insulin release, a response that can be blunted by exogenous antioxidants (Leloup et al., 2009). Glucose-stimulated insulin release could also be simulated using respiratory chain inhibitors that increase cellular O₂˙⁻/H₂O₂ levels (Leloup et al., 2009). In addition, it was found that the O₂˙⁻ generator, paraquat, and exogenous H₂O₂ can also promote insulin release (Mailloux et al., 2012ab). The finding that H₂O₂ can serve as an insulin secretagogue is supported by the presence of UCP2 in pancreatic β-cells, which has been shown in several studies to attenuate glucose-stimulated insulin release and can account for up to ~70% of cellular respiration (Affourtit et al., 2011; Robson-Doucette et al., 2011). UCP2-mediated regulation of glucose-stimulated insulin release stems from its putative function in proton leaks. In this scenario, the UCP2-induced return of protons to the matrix diminishes the protonmotive force, lowering O₂˙⁻/H₂O₂ release by the respiratory chain. It was found later that during the initial uptake and metabolism glucose, UCP2 is maintained in an inactive state inducing a spike in H₂O₂ levels in cultured insulinoma cells and pancreatic islets, promoting insulin release (Mailloux et al., 2012ab). The increase in cellular H₂O₂ also feeds back on UCP2 activating proton leaks, desensitizing the insulin secretion signal (Affourtit et al., 2011). Inducible leaks through UCP2 and control over ROS production was later found to be associated with changes in the S-glutathionylation of the protein (discussed in detail below in the section ‘Regulation of mitochondrial hydrogen peroxide production’) (Mailloux et al., 2012ab). Indeed, the S-glutathionylation agents, diamide and BioGEE, were able to stimulate insulin secretion via the modification of UCP2 (Mailloux et al., 2012ab). Recent work has also shown that ROS and UCP2 also play a role in controlling glucagon release from pancreatic α-cells, indicating that the pancreas utilizes a common regulatory mechanism to control the secretion of two very important metabolic hormones into the blood stream (Allister et al., 2013). Although it is unknown if S-glutathionylation events are required to control insulin release in vivo, results collected so far indicate redox signals are required to modulate glucose-stimulated insulin release and the release of glucagon from α-cells during starvation.

Proton leaks

Mitochondrial O₂˙⁻/H₂O₂ levels can be regulated via degradation pathways or control over production (Mailloux et al., 2016a,b). O₂˙⁻ and H₂O₂ are degraded by a series of redundant antioxidant systems, which were briefly discussed in the introduction and reviewed extensively elsewhere (Murphy, 2012). Control over its production also represents an important means of modulating how much O₂˙⁻/H₂O₂ is available in a cell. Mechanisms for regulation include (1) proton leaks, (2) supercomplex assemblies, and (3) redox signals (e.g. protein S-glutathionylation) (Figure 2). Of these mechanisms, the best characterized is proton leaks. Proton leaks are defined as the return of protons to the mitochondrial matrix, by-passing ATP synthase (Jastroch et al., 2010). The function of proton leaks in regulating mitochondrial ROS production has been extensively reviewed (Harper et al., 2008; Divakaruni and Brand, 2011; Mailloux et al., 2011). Thus, we will only discuss this concept in brief whilst trying to highlight some of the debate surrounding how leaks can regulate ROS production. The main proteins involved in leaks are adenine nucleotide transporter (ANT) and uncoupling proteins (UCP) 1-5 (Divakaruni and Brand, 2011; Ramsden et al., 2012). A non-Ohmic relationship exists between the strength of the proton gradient and mitochondrial ROS production, where a small increase in protonmotive force can lead to the augmentation of mitochondrial O₂˙⁻/H₂O₂ production (Korshunov et al., 1997). Enhancing proton return to the matrix can have the opposite effect, diminishing ROS production by mitochondria. This has been shown with protonophores such as FCCP, which uncouple respiration diminishing ROS production. By decreasing protonic back pressure, successful electron transfer through the respiratory chain to O₂ at complex IV increases, diminishing the amount of electrons available to form O₂˙⁻/H₂O₂ (Divakaruni and Brand, 2011). However, the use of proton leaks to govern how much ROS is actually formed by mitochondria has been a focal point for controversy for more than a decade. For example, it has been found that activation of leaks through UCP1 in brown fat mitochondria, either by oleate or cold acclimation, increases mitochondrial ROS production (Mailloux et al., 2012ab; Shabalina et al., 2014). It is also not entirely clear as to whether or not UCP2 (more ubiquitously expressed) and UCP3 (found exclusively in skeletal muscle) utilize leaks to prevent oxidative damage and control ROS production. For instance, it has been found that both putative uncouplers can eliminate lipid hydroperoxides (Lombardi et al., 2010). Some studies have found that UCP2 and UCP3 translocate calcium ions into the matrix of mitochondria whereas others have shown that UCP3 transports fatty acids (Schrauwen et al., 2003; Trenker et al., 2007). Moreover, recent work has suggested that mitochondrial O₂˙⁻/H₂O₂ production is independent of the strength of the protonmotive force as inhibition of the electron transport chain and decreasing the rate of respiration do not necessarily correlate with how much ROS is formed (Brand, 2016). In addition, increasing mitochondrial respiration rate and proton return can also augment ROS production (Mailloux et al., 2012ab). It is important to emphasize that numerous studies have shown that the uncoupling proteins, in particular UCP2 and UCP3, play vital roles in preventing oxidative stress and damage and fulfill some important physiological functions (e.g. optimization of fuel metabolism in muscle or satiety signaling and insulin release). However, whether these roles are related to the induction of proton leaks and control over mitochondrial O₂˙⁻/H₂O₂ release is still a matter for debate.

Figure 2:

Regulation of mitochondrial H₂O₂ release.

H₂O₂ formed during nutrient metabolism can be increased or decreased by proton leaks (i), supercomplex assemblies (ii), or redox signals (iii). (i) Induction of proton return by adenosine nucleotide translocator (ANT) or uncoupling proteins (UCP) lowers the strength of the protonmotive force across the mitochondrial inner membrane. This diminishes protonic back pressure on the electron transport chain, increasing electron flow through the chain and the successful reduction of O₂ to H₂O at complex IV, which decreases the number of electrons available for ROS production. (ii) Fully assembled respirasomes containing complex I, III, and IV with ubiquinone and cytochrome c diminish ROS production by increasing the efficiency of electron flow through the chain. Disassembly results in decreased efficiency in electron movement, over-reduction of electron donating sites, and increased ROS formation. (iii) The degree of protein S-glutathionylation in mitochondria is dictated by the concentration of H₂O₂ which alters the availability of GSH and GSSG. A decrease in the GSH/GSSG redox pair increases S-glutathionylation of ROS forming sites in mitochondria diminishing H₂O₂ formation.

Supercomplexes

Other criteria for controlling mitochondrial O₂˙⁻/H₂O₂ production include (1) substrate supply, (2) the type of substrate being oxidized, (3) redox state of electron donating sites, (4) concentration of electron donating sites, and (5) access to oxygen. In this regard, the assembly of supercomplexes, also called respirasomes, represents one potential mechanism that may modulate O₂˙⁻/H₂O₂ production by controlling electron movement between different sites of ROS production (Enriquez, 2016). The respiratory supercomplex theory was originally developed to explain the co-migration of multi-respiratory complex units composed of complexes I, III, and IV during blue native polyacrylamide gel electrophoresis (Schagger and Pfeiffer, 2000). It was estimated that most of the complex I pool is involved in forming adducts with complex III, which form a functional “respirasome” with complex IV (Schagger and Pfeiffer, 2000). In 2008, Acin-Perez and colleagues found that blue native bands corresponding to these so-called respirasomes were enzymatically active and able to transfer electrons from NADH at complex I to complex IV resulting in a measurable rate of O₂ consumption (Acin-Perez et al., 2008). Supercomplex assemblies have been found to vary widely in composition, leading to the development of the plasticity model for respirasome formation, which affects the overall respiratory efficiency of mitochondrial preparations (Moreno-Loshuertos and Enriquez, 2016). Recent work has also demonstrated that supercomplex assembly status also affects mitochondrial ROS production. For instance, O₂˙⁻/H₂O₂ production by complex I is higher when dissociated from complex III (Maranzana et al., 2013). Although little information exists on how supercomplex assemblies physically influence ROS production, it is possible that (1) dissociation of complex I from III makes FMN more accessible for electron transfer to O₂ forming O₂˙⁻/H₂O₂ (Maranzana et al., 2013) and (2) FMN in complex I is more amenable to over reduction due to diminished electron transfer to complex III increasing electron availability for the monovalent reduction of O₂. Thus, although speculative, it is likely that increased O₂˙⁻/H₂O₂ production by destabilized respirasomes is associated with increased electron availability and potentially increased access of O₂ to electron donating sites.

Controlling mitochondrial ROS production by redox signals

Protein cysteine thiols can be subjected to a range of different redox modifications that depend on the concentration of H₂O₂, glutathione (GSH) and glutathione disulfide (GSSG), H₂S, availability of nitric oxide (NO) and its adducts (like S-nitroso-glutathione or S-nitroso-coenzyme A), proximity of neighboring thiols allowing formation of intra- or inter-molecular disulfide bonds, or amides which can react with sulfenic acid to form sulfonamide. Most work in terms of understanding the impact of redox reactions on protein function has focused on the direct oxidation of cysteine thiols by H₂O₂ or NO. The direct oxidation of protein cysteine thiols by H₂O₂ is referred to as sulfenylation where a protein cysteine thiol (SH) is oxidized to a sulfenic acid (SOH). Notably, this reaction depends on the ionization of the sulfur group on cysteine forming a thiolate anion, a strong nucleophile that can attack H₂O₂. It has been postulated that H₂O₂ can negatively feedback on sites of production or on uncoupling proteins to control mitochondrial ROS release through the formation of SOH groups (Chouchani et al., 2016). However, very few proteins in mitochondria have been found to undergo sulfenylation. In addition, it was recently suggested that sulfenylation does not fit the criteria for regulation of proteins by covalent modification (Mailloux et al., 2016a,b). Indeed, to serve as an efficient regulatory system, the covalent modification needs to be specific, reversible, rapid, and fulfill some physiological role (Shelton et al., 2005). Hydrogen peroxide reacts slowly with cysteine residues and, to date, no known enzymes have been found to mediate these reactions nor do proteins display “sulfenylation motifs”. Moreover, SOH groups are highly unstable, reacting with various electrophiles forming different adducts (Nagy et al., 2007). A recent study showed that proteins are more likely to undergo S-glutathionylation and if a protein is sulfenylated it is rapidly resolved by neighboring amides to form sulfonamides (Forman et al., 2017). Sulfenylation can be reversed by the S-glutathionylation of SOH groups (Applegate et al., 2008). Sulfenic acid can also be oxidized by H₂O₂ forming sulfinic acid (SO₂H), which can be reversed by sulfiredoxin (Rhee and Kil, 2016). However, these are protective mechanisms rather than regulatory modifications because both prevent the irreversible oxidation of thiols to sulfonic acid (SO₃H). Collectively, it is unlikely that H₂O₂ directly regulates its own production in mitochondria as it does not fit the criteria of a signaling molecule. It is worthwhile noting that some proteins in the cell do exhibit sensitivity towards sulfenylation such as PRX proteins, which plays a role in either redox relay or floodgate signaling (discussed in detail in the section ‘Steroidogenesis in the adrenal cortex and circadian rhythms’). Modification of cysteine thiols by nitric oxide (S-nitrosylation) also suffers from the same issues, except a number of mitochondrial proteins have been found to undergo modification by NO. In addition, NO is well documented to serve as an inhibitor for complex IV of the respiratory chain. Thus, NO can still exert regulatory effects on the respiratory chain even though it still remains uncertain as to whether or not S-nitrosylation is a regulatory mechanism or a nonselective protein modification associated with cellular stress.

An emerging candidate for the post-translational control of proteins by redox signaling in the mitochondria is protein S-glutathionylation. This involves the conjugation and removal of a glutathione moiety from a protein. The first characteristic that sets S-glutathionylation apart from the other redox modifications is that it can proceed either spontaneously or enzymatically. The former is usually involved in protecting proteins during oxidative distress whereas the latter regulates protein function in response to normal physiological stimuli that alter local redox environments and buffering networks. The ratio of GSH/GSSG is usually kept quite high in the mitochondria, around 50–100, but this can be lowered in response to an increased H₂O₂ concentration (either higher than normal H₂O₂ formation by the respiratory chain or during the mitochondria-mediated quenching of cellular ROS). This increase in GSSG can induce the S-glutathionylation of proteins. Non-enzymatic S-glutathionylation has been documented to only occur when GSH/GSSG approaches 1 since some proteins have been found to have a K_OX of ~1 (Gallogly and Mieyal, 2007). This occurs under oxidative distress when H₂O₂ is high and glutathione is actively being used for its degradation. Therefore, during oxidative distress proteins are S-glutathionylated to protect from irreversible damage (Dalle-Donne et al., 2009). Importantly, this modification is reversed by glutaredoxins (Grx; discussed in more detail below) which remove GSH and reactivate the protein. However, some proteins have a K_OX with GSSG of ~50, meaning that under physiological conditions certain proteins can be S-glutathionylated spontaneously (Gallogly and Mieyal, 2007). In fact, some cellular proteins display basal S-glutathionylation under physiological conditions. For example, complex II (succinate dehydrogenase; SDH) has been found to be basally S-glutathionylated in heart tissue, which maintains its activity and diminishes the production of O₂˙⁻/H₂O₂ (Chen et al., 2007). 2-oxoglutarate dehydrogenase (OGDH) has also been found to undergo S-glutathionylation when GSH is abundant, a reaction thought to occur following the formation of a glutathionyl radical and its reaction with a protein cysteine thiol (Mailloux et al., 2016a,b). It has been suggested that, due to its highly folded nature, the mitochondrial inner membrane can generate highly variable redox micro-environments where the levels of GSH and GSSG can differ substantially (Drose et al., 2014). It is entirely plausible that this can occur when mitochondria transition between different states of respiration. For example, mitochondria display periodic changes in H₂O₂ and GSH levels when transitioning between state 3 and state 4 respiration (Cortassa et al., 2014). As mitochondria transition towards state 4 respiration, GSSG levels rise due to increased H₂O₂ production whereas induction of state 3 respiration has the opposite effect (Cortassa et al., 2014). This could have a strong influence on whether or not a mitochondrial protein can under spontaneous protein S-glutathionylation. Overall, nonspontaneous S-glutathionylation reactions are required to protect mitochondrial proteins, such as OGDH or complex I, from irreversible deactivation during oxidative distress. However, it may also be possible to spontaneously modify proteins with GSH if (1) the K_OX is high enough and (2) the micro-environment GSH/GSSG gradient favors a spontaneous reaction.

Protein S-glutathionylation is also enzymatically mediated. These reactions are catalyzed by glutaredoxin (Grx), a small heat stable thiol oxidoreductase with a thioredoxin fold that mediates the conjugation and removal of glutathione from a target protein. GRX1 was the first thiol oxidoreductase found to catalyze the deglutathionylation of protein targets (Mannervik and Axelsson, 1980). GRX1 deglutathionylates protein cysteine thiols through a simple nucleophilic displacement mechanism involving thiol disulfide exchange between the active site cysteine in GRX1 and the target protein (Gallogly et al., 2009). It was later found to also catalyze the forward reaction, conjugating glutathione to a protein cysteine thiol. Unlike the deglutathionylation of target proteins, less is known about the mechanism for protein S-glutathionylation by GRX1. Although still speculative, it is possible that GRX1 mediates protein S-glutathionylation by first being S-glutathionylated itself allowing for transfer of glutathione to a protein target. Recent work has shown that GRX1 can be S-glutathionylated, indicating that this may account for how GRX1 can conjugate glutathione to cysteine (Dong et al., 2016). This can also be demonstrated with the cellular glutathione probe, GRX1-roGFP2, which changes fluorescence following its GSSG-mediated S-glutathionylation. GRX1 is also found in the intermembrane space of mitochondria where it has been shown to play a part in mitochondrial protein import (Pai et al., 2007). By contrast, glutaredoxin-2 (GRX2) is found in the matrix of mitochondria. Although GRX1 and GRX2 show very low homology with one another, both enzymes utilize a common mechanism for the reversible S-glutathionylation of proteins (Gallogly et al., 2009). It has been found that GRX2-mediated S-glutathionylation or deglutathionylation reactions are highly sensitive to changes in the surrounding redox environment. A high GSH/GSSG ratio activates the deglutathionylase activity of GRX2 whereas high H₂O₂ and a low GSH/GSSG redox pair have the opposite effect (Mailloux et al., 2014). GRX2 also forms an inactive dimer by coordinating an iron sulfur center with two GSH that can be disassembled by O₂˙⁻ yielding two active GRX2 monomers and two GSH (which are required to form the 2Fe-2S cluster) (Lillig et al., 2005). The importance of GRX2 in mediating the reversible S-glutathionylation of proteins is underscored by the consequences of its deletion (reviewed in detail in Mailloux, 2016). Briefly, loss of GRX2 can compromise mitochondrial respiration and ATP production and alter O₂˙⁻/H₂O₂ production, which is associated with the development of heart disease, hypertension, cataracts, and perturbations in embryonic development. In addition to the glutaredoxin enzymes, it has also been found that glutathione S-transferase can catalyze the reversible S-glutathionylation of cellular proteins (Klaus et al., 2013). Pi and Mu-members of the GST family have been found to catalyze the S-glutathionylation of AMP kinase, the master energy gauge for the cell (Klaus et al., 2013). It remains to be determined if GST isoforms are also required to catalyze these reactions in mitochondria. The deglutathionylation of target proteins has also been found to be quite rapid ranging from ~10² to 10⁵m⁻¹ s⁻¹, which depends on the pKa of the target protein cysteine thiol and target proteins also have S-glutathionylation motifs, making this modification highly specific (Jensen et al., 2014). Thus, in contrast to sulfenylation or S-nitrosylation, protein S-glutathionylation reactions are rapid, reversible, highly specific, and sensitive to changes in H₂O₂ and the overall redox environment making this redox signal a strong candidate for regulating mitochondrial ROS production.

Protein S-glutathionylation regulates mitochondrial O₂˙⁻/H₂O₂ production

2-Oxoglutarate dehydrogenase and pyruvate dehydrogenase

Complex I and III are historically viewed as the only sources of mitochondrial ROS. However, evidence collected over the past decade has shown that several other enzymes in mitochondria are also high capacity sites for O₂˙⁻/H₂O₂ formation. Indeed, it has been documented that skeletal muscle mitochondria can contain up to 12 ROS emitting sites. Of these sites it has been shown that 2-oxoglutarate dehydrogenase (OGDH) and pyruvate dehydrogenase (PDH) produce ~8× and ~4× more O₂˙⁻/H₂O₂ than complex I under experimental conditions where carbon is being oxidized by Krebs cycle enzymes (Quinlan et al., 2014). Similar observations have been made in liver and cardiac mitochondria where (1) OGDH and PDH are higher capacity sites for ROS formation than complex I and (2) OGDH is a more effective O₂˙⁻/H₂O₂ generator than PDH (Slade et al., 2017). It has also been found that purified OGDH and PDH both produce O₂˙⁻/H₂O₂ during reverse electron transfer (RET) from NADH (Ambrus et al., 2015; Mailloux et al., 2016a,b). Moreover, O₂˙⁻/H₂O₂ production during RET can occur at physiological NADH concentrations [0.1–1 μm, NADH/NAD⁺ ratio is 1/8 and concentration of NAD(H) is ~800 μm indicating physiological levels of NADH can stimulate ROS production by OGDH and PDH during reverse electron flow] and the rate of generation increases rapidly at higher NADH levels (Tretter and Adam-Vizi, 2004; Mailloux et al., 2016a,b). Overall, this would indicate that OGDH and PDH can form O₂˙⁻/H₂O₂ following reverse electron transfer from NADH, even during normal mitochondrial function. This would also indicate that both enzymes might be important sites of production under pathological conditions when NADH levels are increased due to a deficiency in complex I activity (Tretter and Adam-Vizi, 2005).

It has been known for some years that OGDH serves as a putative redox sensor (Gibson et al., 2005; McLain et al., 2011). Indeed, several studies over a decade ago had found that H₂O₂ can deactivate OGDH through oxidation of its lipoic residues (Tretter and Adam-Vizi, 1999). Afterwards it was uncovered that OGDH is also a source of O₂˙⁻/H₂O₂ leading to the postulate that H₂O₂ can feedback and limit its own production through the deactivation of the enzyme complex (McLain et al., 2011). Two studies showed that the reducing agent DTT or GRX1 could reactivate OGDH (Nulton-Persson et al., 2003; Applegate et al., 2008). Further inquiries into the modification of OGDH during oxidative stress showed that the lipoic acid residues on the E2 subunit were reversibly S-glutathionylated (Applegate et al., 2008; McLain et al., 2013). Considering that the vicinal lipoic acid thiols in OGDH are amenable to irreversible oxidation by H₂O₂, it was postulated that S-glutathionylation and its subsequent reversal was required to protect the enzyme complex from over-oxidation and reactivate it after the oxidative distress had subsided (McLain et al., 2011). It was also put forth that the S-glutathionylation of OGDH may also limit mitochondrial ROS production by diminishing NADH production and its subsequent oxidation by complex I (McLain et al., 2011). However, OGDH is also a major O₂˙⁻/H₂O₂ generator and thus its S-glutathionylation could also serve as a means of regulating H₂O₂ production. Our group recently showed that purified OGDH from porcine heart and OGDH in liver mitochondria can undergo S-glutathionylation, which is required to control O₂˙⁻/H₂O₂ release from the enzyme complex (Mailloux et al., 2016a,b; O’Brien et al., 2017). In two separate studies, it was found that the S-glutathionylation of the E2 subunit by either GSSG or diamide induces a significant decrease in O₂˙⁻/H₂O₂ release during 2-oxoglutarate oxidation (Mailloux et al., 2016a,b; O’Brien et al., 2017). Using purified OGDH, it was found that this decrease in O₂˙⁻/H₂O₂ production correlated strongly with a decrease in NADH production (Mailloux et al., 2016a,b). This would also diminish O₂˙⁻/H₂O₂ formation by the electron transport chain given that NADH production would be limited. Research into whether or not a high GSH/GSSG ratio or increased availability of GSH alone can affect ROS production led to the unique observation that reduced glutathione can actually augment O₂˙⁻/H₂O₂ release from OGDH through its S-glutathonylation. It was also found that GSH, at a higher concentration (>1 mm), could S-glutathionylate OGDH (Mailloux et al., 2016a,b). However, this event actually occurs on the E1 subunit, which augmented O₂˙⁻/H₂O₂ production by OGDH (Mailloux et al., 2016a,b). It should be noted here that OGDH produces O₂˙⁻/H₂O₂ from the E1 and E3 subunits; the former via thiamine radical formation and latter through its FAD prosthetic group (Nemeria et al., 2014). Thus, the GSH-mediated S-glutathionylation of the E1 subunit likely results in thiamine radical accumulation augmenting ROS production (Figure 3). The calculated EC₅₀ for the GSH-mediated increase in ROS production by OGDH was found to be ~2.125 mm for ROS genesis and the IC₅₀ for NADH production was ~2.819 mm (Mailloux et al., 2016a,b). In addition, purified GRX2 reversed these effects for both the purified form of OGDH and OGDH in permeabilized liver mitochondria. Therefore, when GSSG is high, it S-glutathionylates OGDH to limit O₂˙⁻/H₂O₂ production, which would be required to diminish production when ROS levels are high (Figure 3). Conversely, high GSH has the opposite effect, augmenting O₂˙⁻/H₂O₂ production when mitochondrial redox buffering capacity is reduced enough to control higher than normal ROS formation (Figure 4). Collectively, OGDH can serve as an overall redox buffering sensor, increasing and decreasing its ROS production rate in response to fluctuations in GSH and GSSG availability, which could influence the capacity of mitochondria to use H₂O₂ as a secondary messenger.

Figure 3:

GSH and GSSG have different effects on ROS release from 2-oxoglutarate dehydrogenase.

(A) At low ROS level, 2-oxoglutarate dehydrogenase (OGDH) electrons yielded from the combustion of 2-oxoglutarate are used to drive NADH and ROS formation. When H₂O₂ levels are higher than normal or a mitochondrion is experiencing mild oxidative distress, GSSG levels increase, resulting in the S-glutathionylation of the vicinal thiols on the E2 subunit of OGDH blocking electron transfer to the E3 subunit. This decreases ROS production by OGDH whilst simultaneously protecting the enzyme from irreversible oxidation. S-glutathionylation also decreases NADH production, which would lower ROS formation by the respiratory complexes. (B) At high enough levels, GSH can S-glutathionylate the E1 subunit of OGDH amplifying ROS production. The higher GSH levels can allow for easy buffering of any excess ROS formed during substrate oxidation. In addition, high GSH can maintain GRX2 in its inactive dimeric form. Activation of GRX2 by high ROS results in the deglutathionylation of OGDH and the restoration of NADH production.

Figure 4:

S-glutathionylation is required to control ROS production by pyruvate dehydrogenase.

(A) During forward electron flow from pyruvate to NAD⁺, pyruvate dehydrogenase (PDH) forms high amounts of ROS, which can be inhibited by the S-glutathionylation of the vicinal thiols on the lipoic acid in the E2 subunit. ROS emitted from PDH may be required for cell signaling and thus S-glutathionylation is required to desensitize the H₂O₂ signal. (B) PDH can form ROS following reverse electron transfer from NADH. During oxidative distress GSSG accumulates resulting in the spontaneous S-glutathionylation of PDH, which augments ROS production during reverse electron transfer. Oxidative distress is also associated with the deactivation of complex I, resulting in NADH accumulation which is then oxidized by PDH. High ROS production due to reverse electron flow when NADH is high and PDH is S-glutathionylated, results in GRX2 activation and the deglutathionylation of PDH. This restores PDH activity and lowers ROS production by reverse electron flow from NADH.

In regard to redox sensing less is known about PDH. However, recent work by our group and others found that it serves as a GSH and GSSG sensor much like OGDH (O’Brien et al., 2017). Using purified PDH and permeabilized muscle fibers, Fisher-Wellman and coworkers found that depletion of glutathione pools or inhibition of NADPH production augmented pyruvate-driven O₂˙⁻/H₂O₂ production (Fisher-Wellman et al., 2013, 2015). It was then postulated by the same group that PDH was able to sense changes in mitochondrial redox buffering capacity, potentially through S-glutathionylation, which alters ROS production (Fisher-Wellman et al., 2013). However, based on the experimental design it was unclear whether or not PDH was actually sensing any changes in mitochondrial redox buffering capacity or if the increase was simply due to the CDNB-mediated depletion of glutathione pools. Our group recently found that PDH is S-glutathionylated on its E2 subunit and to a lesser extent its E1 and E3 subunits (O’Brien et al., 2017). Gergondey and colleagues also recently showed that H₂O₂ and diamide can drive the S-glutathionylation of PDH (Gergondey et al., 2017). During pyruvate oxidation, it was found that the diamide or disulfiram mediated S-glutathionylation of purified PDH or PDH in liver mitochondria decreased O₂˙⁻/H₂O₂ production (O’Brien et al., 2017). This was also associated with a decrease in NADH formation which correlated with the S-glutathionylation of the E2 subunit of PDH. It was also found that the GSSG-mediated S-glutathionylation of purified PDH augmented ROS production during RET which could be reversed by GRX2 (O’Brien et al., 2017). Therefore, the S-glutathionylation of PDH serves as a mechanism for sensing changes in mitochondrial redox buffering capacity, which can sharply alter how much O₂˙⁻/H₂O₂ is formed by the enzyme complex. During forward electron flow from pyruvate to NAD⁺, S-glutathionylation of the E2 subunit is required to lower ROS production (Figure 4). Moreover, S-glutathionylation of PDH almost completely abolished NADH production, which would limit O₂˙⁻/H₂O₂ production by the electron transport chain. Overall, this could play an integral role in influencing mitochondrial H₂O₂ release and signaling to the rest of the cell given that PDH is a vital source of O₂˙⁻/H₂O₂ and an important entry point into the Krebs cycle for carbon yielded from glucose metabolism (Figure 4). In regard to production during RET, we postulate this may reflect pathological conditions as S-glutathionylation of PDH augmented NADH driven ROS production by eight times (O’Brien et al., 2017). This could contribute to oxidative distress when GSSG levels are high enough and NADH turnover by complex I is limiting, for example in dysfunctional mitochondria (Figure 4). Since GRX2 can reverse the GSSG-mediated S-glutathionylation of PDH and the sharp increase in O₂˙⁻/H₂O₂ production, it is likely that it is required to protect mitochondria from any potential oxidative distress associated with the over production of ROS (Figure 4).