*5.3. ROS as Signals*

It is now established that ROS can act as second messengers [265–274] by modulating the expression of several genes involved in signal transduction. The effects related to ROS signals are different and even opposite, depending on the concentration. ROS may be oncogenic and promote proliferation, invasiveness, angiogenesis and metastasis [275,276], and accordingly ROS generation increases in various cancers. Nevertheless, ROS can also promote anti-tumorigenic signaling and trigger oxidative stress-induced tumor cell death.

Many of the ROS-mediated cellular signals paradoxically protect the cell against oxidative stress, but above a given threshold ROS become harmful and induce oxidative stress [268]. Not all ROS are equally suitable for signal transduction: the OH· radical is too aggressive to participate in catalyzed reactions, whereas O<sup>2</sup> <sup>−</sup> and H2O<sup>2</sup> are known as important signaling molecules. Moreover, other reactive species are involved in redox signaling, such as nitric oxide, hydrogen sulfide and oxidized lipids [274].

Because some ROS such as H2O<sup>2</sup> are oxidants, they modify the redox state of signaling proteins by oxidizing specific sulfhydryl groups [274,277,278] in enzyme-catalyzed reactions.

Several transcription factors contain redox-sensitive cysteine residues at their DNAbinding sites [279,280]. The ensuing formation of disulfide bonds acts as a redox switch to transduce the response of the protein to redox signaling [271,281].

Following the oxidation of signaling proteins by ROS, other post-translational changes (e.g., phosphorylation, acetylation, ubiquitination, and SUMOylation [270]) may occur in the same protein and in other proteins of the signaling cascade.

Nrf2 is an important example of mitochondrial ROS signaling which leads to nuclear gene expression changes [282]; in the presence of ROS, NRf2 is transferred from the cytoplasm to the nucleus, where it binds the DNA antioxidant response element of genes involved in the antioxidant response (e.g., heme oxygenase, NRF-1 and other inducers of mitochondrial biogenesis) [283].

Signaling proteins modified by ROS include phosphoprotein phosphatases (PTPs), Ras, large G-proteins, serine/threonine kinases of the MAPK families, transcription factors as AP-1, NFκB, p53 and others. A different effect is exerted on these proteins: PTPs are inhibited, while nuclear transcription factors are activated [284]. Different combinations of activation or deactivation of these proteins may drive the cell to either survival or death.

Mitochondrial ROS production is involved in preconditioning, by which low doses of a noxious stimulus promote a protective response against future injury [285–287]. Some protective pathways converge on the inhibition of mPTP opening during reperfusion [287].

Mitochondrial ROS are also important signals which rule the inflammation by activating the inflammasome [288,289]; treatment with antioxidants could scavenge excessive ROS and attenuate inflammatory responses by suppressing inflammasome activation [290]. ROS also regulate autophagic processes including mitophagy [291].

#### 5.3.1. Regulation of Mitochondrial ROS in Cell Signaling

Mitochondrial ROS levels depend on the dynamic equilibrium between their rates of production and removal.

Studies on ROS generation under different substrate, inhibitor conditions and different oxygen tensions in rat liver mitochondria lead to the conclusion that only CI may be a significant source of ROS at physiological O<sup>2</sup> concentration [292]. The factors directly associated with respiratory activity (e.g., the redox potential of the NAD+/NADH couple and the proton-motive force) regulate the reduction in oxygen [268]) but these factors are in turn dependent on other parameters (cf. also Section 5.2).

Moreover, mitochondria enhance ROS generation in response to external stimuli, such as TNFα [293], hypoxia ([294], cf. Section 5.3.2), serum deprivation [295] and oxidative stress itself (ROS-induced ROS release, Zorov et al. [296]). Several proteins, such as p53, p66Shc, the Bcl-2 family and Romo-1 [267] control mitochondrial ROS generation.

An interplay of mitochondria with non-mitochondrial ROS generating systems also occurs, as with the NADPH oxidases (NOX) family [297–299].

On the other hand, control of ROS signaling may be operated by their removal by a series of enzymatic systems [300], including various isoforms of superoxide dismutase, catalyzing the dismutation of superoxide, whereas catalase, glutathione peroxidase and peroxiredoxins remove H2O2. In addition, an integrated set of thiol systems in mitochondria prevents oxidative damage [271] and may transduce redox signals.

Peroxiredoxins (Prxs) are small dithiol proteins constituting a major family of peroxidases (six isoforms in mammals). Due to their high sensitivity to oxidation by H2O2, Prxs are sensors and transducers of H2O<sup>2</sup> signaling [301] by transferring their oxidation state to effector proteins. The mitochondrial matrix contains large amounts of Thioredoxin-2 [302,303] and Prx3 [304].

The appropriate mitochondrial redox environment is greatly ascribed to glutathione [305–307] which is abundant and versatile to counteract hydrogen peroxide and lipid hydroperoxides, being a cofactor of glutathione peroxidase or glutathione-*S*-transferase. The mitochondrial GSH:GSSG ratio generally exceeds 100 and is widely used as an indicator of the redox status [308].

Protein glutathionylation is a regulatory mechanism that occurs through thiol disulfide exchange by protein disulfide isomerase [309,310]; the primary sistem catalyzing de-glutathionylation of protein-mixed disulfides with glutathione is glutaredoxin [311]. Mitochondrial proteins are highly susceptible to reversible *S*-glutathionylation, a posttranslational modification which is favored by the unique physico-chemical properties of the mitochondrion [312], which contains a variety of *S*-glutathionylation targets playing important physiological roles: if these reactions are defective, pathological consequences are expected to occur.

#### 5.3.2. Hypoxia and ROS Production

Among the many signaling functions, ROS appear as hypoxic signaling molecules. It seems paradoxical that a decrease in the level of the required substrate, O2, would result in an increase in ROS production. However, there is large evidence of its occurring [313–315].

Indeed, ROS appear to activate the hypoxia-inducible factor HIF-1α. Rho0 cells, which lack mtDNA and thus electron transport, do not show the HIF-1 DNA-binding activity following hypoxia [316]. Therefore, mitochondrial respiration is required for propagation of the hypoxic signal. Exogenous or endogenous H2O<sup>2</sup> stabilize HIF-1α, thus inducing its activation during normoxia [316], suggesting that HIF stabilization and activation may be induced by ROS.

Many studies demonstrate activation of HIF-1α by ROS [317,318], that are supposed to act by inhibiting the prolyl-4-hydroxylase that addresses the factor to proteolytic digestion. Once stabilized, HIF-1α binds to hypoxia-responsive elements in the DNA and stimulates the expression of a large array of genes [319].

The oxidized cyt *c* level in the IMS seems to be essential to control mitochondriallyderived ROS [320,321]; accordingly, cyt. *c* in the IMS oxidizes superoxide produced by CIII to O2, thus preventing H2O<sup>2</sup> production [321,322].

Inhibition of cytochrome *c* oxidase by hypoxia [323] enhances ROS formation by the respiratory chain; most likely because of the increased membrane potential [324] or by inducing a more reduced state of cyt. *c*.

The ROS produced by mitochondrial CIII seem to be critical for hypoxia signaling [294,325–328]. Hernansanz-Augustin et al. [329], however, showed that CI is involved in the superoxide burst under acute hypoxia in endothelial cells.

In addition, it should be pointed out that the plasma membrane NADPH oxidase [313,330] also seems to be involved in the enhanced ROS production during hypoxia.

#### *5.4. ROS and Mitochondrial Quality Control*

A major device for physiological control of the cells is the removal of damaged molecules; in particular, oxidized proteins are ubiquitinated and directed to the proteasome, a member of the ATP-dependent AAA+ proteases, where they are completely digested [331,332].

Mitochondrial proteins may particularly suffer from oxidative damage since they are close to the main sites of their generation, so that they must be removed in order to maintain mitochondrial integrity [333,334]. Internal mitochondrial proteins are translocated to the outer mitochondrial membrane, where they are ubiquitinated and then presented to the proteasome for degradation.

Mitochondrial dysfunction due to loss of cellular protein homeostasis (proteostasis) is a hallmark of aging and aging-related degeneration disorders [335], such as Alzheimer's disease and Parkinson's disease. Thus, mitochondrial proteins should be finely tuned. Although mitochondria form a proteasome-exclusive compartment [336], it is the cytosolic ubiquitin–proteasome system that plays a major role in the quality control of mitochondrial proteins.

The ATP-stimulated Lon protease in the mitochondrial matrix is devoted to the selective degradation of oxidized proteins, as for example the selective degradation of the Cox 4-1 subunit during hypoxia [337,338]. Defects in protein degradation have been involved in the age-related accumulation of oxidized proteins [339,340].

Also, entire cellular organelles undergo turnover and are finally directed to autophagy by digestion in the lysosome compartment [341–343].

Non-selective autophagy as well as selective mitophagy are triggered by ROS in response to several stressing signals [291,344,345]. Damaged mitochondria promote ROS generation, and excessive ROS can trigger mitophagy so as to remove impaired mitochondria and reduce ROS levels. Therefore, mitophagy helps to maintain cellular homeostasis under oxidative stress. Low ROS levels can trigger mitophagy in a mitochondrial fission-dependent way, if these levels are insufficient to trigger non-selective autophagy [346]. Therefore, a very specific and selective signaling cascade initiated by ROS has been suggested.

#### *5.5. Supercomplexes Protect Complex I from ROS Damage and Limit ROS Generation*

Our group first suggested that SC arrangement represents the missing link between oxidative stress and energy deficiency [96]. We speculated that oxidative stress dissociates the SCs, with loss of electron channeling and causing electron transfer to depend only upon the diffusion-coupled collisional encounters of the free ubiquinone molecules with the partner complexes.

#### 5.5.1. Supercomplex Association Protects from ROS Damage

As predicted by Lenaz and Genova [96], SC dissociation also induces disassembly of CI and CIII and consequent loss of their catalytic activity. Therefore, the alteration of electron transfer may stimulate ROS generation. Here we briefly summarize the experimental evidence supporting this hypothesis.

Analysis of the SCs in patients with an isolated deficiency of single complexes [347] and in cultured cell models harboring cytochrome *b* mutations [348–350] showed that SCs enhance the stability of CI. Genetic mutations causing loss of CIII prevent SC formation and determine a secondary loss of CI, and consequently primary CIII assembly deficiencies appear as joint CIII/I defects. On the other hand, however, the absence of CI does not affect CIII stability.

It was recently suggested that the absence of CIII blocks CI biogenesis by preventing the incorporation of the NADH module rather than decreasing its stability [351].

Most evidence, however, is in line with the idea that misassembly of CIII affects CI stability because of their physical interaction within the SC. Misassembled CIII would prevent SC formation, and a lack of SC would induce an enhanced ROS generation from CI (see Section 5.5.2), with consequent damage to CI itself [350,352] which is vulnerable to oxidative stress both directly and through the lipid peroxidation, particularly of CL [353]. According to Guaras et al. [120], saturation of CoQ oxidation capacity keeps a highly reduced state of CoQ itself and induces reverse electron transport from reduced CoQ to CI; the resulting local superoxide generation oxidizes specific CI proteins, triggering their degradation and the disintegration of the complex.

An ischemia/reperfusion injury enhances ROS generation in mitochondria which promotes the opening of the nonselective mPTP [354]. The mPTP opening further compromises cellular bioenergetics and increases mtROS, resulting in SC disintegration and ultimately leading to cell death.

A reconstitution study [71] showed that lipid peroxidation before formation of CI-CIII proteoliposomes abolishes the SC formation, as determined by flux control analysis. It is likely that the distortion of the lipid bilayer induced by peroxidation provokes the dissociation of the SC.

Mitochondrial ROS also affect CI and CIV activity through CL peroxidation in beef heart SMP [353,355]. CL liposomes exogenously added to mitochondria from aged rats almost completely restored the activity of these enzyme complexes to the values of young control animals [356,357]. Other phospholipid classes, as well as peroxidized CL, could not mimic the CL effect.

#### 5.5.2. Supercomplex Association Limits ROS Generation

Indirect considerations are compatible with the possibility that loss of SC organization may enhance ROS generation by the respiratory chain [12,96,358]. There are two possible reasons to support this hypothesis: a tight assembly of the respiratory complexes within SCs may screen auto-oxidizable groups hindering their reaction with oxygen (cf. [55]), and/or enhanced electron flow from NADH in the chain because of channeling would keep the prosthetic groups in a more oxidized form, thus preventing their interaction with oxygen [250]. On the other hand, on succinate oxidation, the reverse electron flow through CI keeps its centers more reduced favoring production of superoxide.

The molecular structure of the individual complexes and of the SCs does not provide structural evidence of such screening, since the prosthetic groups of CI in the matrix arm do not seem to be in close contact with CIII [359,360]. However, the SC I1III2IV<sup>1</sup> from bovine heart has a different CI conformation with respect to the free complex, that appears to be bent in such way as to be closer to CIII [359]. This observation agrees with the notion that CI may undergo important conformational changes [361].

A different kind of indirect evidence on the protection exerted by SC organization against ROS generation comes from the observation that high mitochondrial membrane potential supports ROS generation, while uncoupling decreases ROS production [243,249]. Although these effects may have different explanations, they are compatible with the suggestion [87] that high membrane potential induces SC dissociation.

A direct demonstration that loss of SC organization enhances ROS production by CI was obtained in our laboratory [362] using a model system of reconstituted CI/CIII proteoliposomes at high lipid-to-protein ratio (30:1), where formation of the SC I1III<sup>2</sup> is prevented. In this system, the generation of superoxide was much higher than in a similar system reconstituted at a 1:1 ratio, which is rich in SCs. In addition, we also dissociated the reconstituted proteoliposomes by mild detergent treatment using dodecyl maltoside and observed that the dissociation of CI was accompanied by a three/four-fold increase in ROS generation. An increase in ROS also occurs in mitochondrial membranes after detergent treatment.

A significant finding supporting our in vitro conclusions comes from a study of mitochondrial CI supramolecular structure in neurons and astrocytes. Lopez-Fabuel et al. [363] observed that CI is largely assembled into SCs in neuronal mitochondria, whereas astrocytes have higher content of free CI. The presence of free CI in astrocytes correlates with the severalfold higher ROS production by astrocytes compared with neurons. Thus, regulation of ROS generation by CI assembly into SCs may contribute to the bioenergetic differences between neurons and astrocytes. Nevertheless, ROS levels were found to be controlled by an efficient antioxidant system in astrocytes, to regulate redox signaling [364]. By exerting this control in ROS levels, metabolic functions are finely tuned in both neural cells.

A study of CI activity in the basidiomycete *Ustilago maydis* [365] suggested that the contacts between CI, CIII<sup>2</sup> and CIV in the respirasome increase the catalytic efficiency of CI and regulate its activity to prevent ROS production.

The results of our proteoliposome study [362] are supported by many observations in cellular and animal models showing that SC dissociation is concomitant with enhanced ROS production.

Diaz et al. [352] showed enhanced ROS generation in mouse lung fibroblasts lacking the Rieske FeS protein of CIII and hence devoid of the SCs containing CI.

Mouse fibroblasts expressing the activated form of the *k*-ras oncogene and having low levels of high molecular weight SCs also produce higher ROS in comparison with wild type fibroblasts [366]. Hyperglycemia prevents SC formation, while increasing ROS levels in liver mitochondria of streptozotocin-diabetic rats. On the other hand, physical exercise was found to induce the chronic assembly of CIs into SCs in skeletal muscle [367], thus protecting mitochondria against oxidative damage.

Other studies in yeast mutants lacking the SC assembly factor Rcf1 and thus devoid of SCs CIII-CIV [117,368,369], showed enhanced ROS generation. Since the yeast *S. cerevisiae* lacks CI, in this case we may consider the origin of the extra ROS being presumably CIII. Thus, a major function of SC assembly appears to be this limitation of ROS production [370]: in the above case the decreased ROS generation by the Rcf1-stabilized SC III2-IV<sup>2</sup> may be due to the more efficient electron transfer between CIII and CIV by cyt. *c*.

Several compounds including phospholipids, proteins, and certain chemicals are known to promote or stabilize mitochondrial SCs directly or indirectly [371], thus hindering ROS generation. Overexpression of the CL conversion enzyme ALCAT1 reduced SC formation and promoted ROS production, while preventing upregulation of coupled respiration [75]. These data suggest that the amount of ALCAT1 is critical for coupling mitochondrial respiration and metabolic plasticity.

The CL defect in Barth syndrome, a cardio-skeletal myopathy with neutropenia characterized by respiratory chain dysfunction, results in destabilization of the SCs, with higher levels of superoxide production in lymphoblasts from patients, compared to control cells [67,68]. Analogous results were obtained un studies on CL-lacking yeast mutants [372,373]. Moreover, Chen et al. [368] observed that yeast mutants, which cannot synthesize CL, exhibit increased protein carbonylation, an indicator of ROS.

Stein et al. [374] described a highly conserved 56-amino-acid microprotein named mitoregulin (Mtln). Mtln localizes to the mtIM where it binds CL and influences protein complex assembly. In cultured cells, Mtln overexpression increases mitochondrial SCs and mitochondrial activity while decreasing mitochondrial ROS.

The subunit composition of CI is crucial for SC formation and control of ROS production. Hou et al. [375] demonstrated that CI subunit AB1 (NDUFAB1), also known as mitochondrial acyl carrier protein, coordinates the assembly of respiratory CI, CII, and CIII, and SCs and is a crucial regulator of mitochondrial energy and ROS metabolism.

Jang and Javadov [376] observed that the pharmacological inhibition of CI and CII stimulated disruption of the respirasome accompanied by reduced ATP formation and increased ROS production. Overall, these studies provide biochemical evidence that the CI activity, and the NDUFA11 subunit are important for assembly and stability of the respirasome. The SDHC subunit of CII is not involved in the respirasome however the complex may play a regulatory role in respirasome formation.

Ramirez-Camacho et al. [377] found that mitochondria from reperfused hearts treated with *N*-acetyl-cysteine reduced oxidative stress and maintained SCs assemblies containing CI, CIII, CIV and the adapter protein SCAF1. The associations of the mitochondrial respiratory chain components into SCs could have pathophysiological relevance in metabolic diseases, as supramolecular arrangements, by sustaining a high electron transport rate, might prevent ROS generation [378].

#### **6. Physiological and Pathological Implications**

As we have pointed out in the Introduction, mitochondria are deeply involved in pathological processes. Mitochondrial dysfunction is often present in pathological states: at the end of this review, we intend to demonstrate that mitochondrial dysfunction is causative in the pathogenesis of most systemic diseases.

What are the signs showing mitochondrial dysfunction? There are many of them that are more or less specific. Among these we have changes of electron transfer, such as uncoupling, i.e., the occurrence of electron transfer not accompanied by ATP synthesis, changes of mitochondrial shape and number, alterations in quality control, changes in supramolecular structure (SCs) accompanied by decreased respiration and enhanced ROS generation, and finally opening of the mPTP and cell death.

We first briefly examine OXPHOS uncoupling. As for ROS generation cf. Section 5, for quality control see Section 5.4, and for the nature and role of the mPTP cf. Section 4.4. A possible role of SC disassembly as the initial pathogenetic event in many complex diseases will be described in Section 6.3.

#### *6.1. Uncoupling: When Proton Movement and ATP Synthesis Are Disjointed*

The term uncoupling in bioenergetics indicates that the H<sup>+</sup> pumping activity of respiratory complexes which forms the ∆*p* is not exploited to synthesize ATP from ADP and Pi by F1FO-ATPase. Uncoupling is often associated with another term, mitochondrial dysfunction, which suggests that mitochondria do not work properly and are unable to fully perform their bioenergetic task. Uncoupling can have variable extent. We can distinguish between mitochondrial uncoupling, due to the downhill H<sup>+</sup> flux through pathways outside the F1FO-ATPase, and intrinsic F-ATPase uncoupling, which means that the enzyme complex itself cannot match ATP synthesis to H<sup>+</sup> channeling, due to molecular defects. While the mitochondrial uncoupling has a recognized physiological role, even if many mechanisms remain to be clarified [257], as far as we are aware the F1FO-ATPase uncoupling only represents a basic biochemical symptom of pathologies [379].

#### 6.1.1. Mitochondrial Uncoupling Due to Dissipative Pathways

The mitochondrial uncoupling can be defined as the detachment of H<sup>+</sup> movement across the mtIM from the F1FO-ATPase activity [134]. Accordingly, the mitochondrial membrane potential generation is not used to build ATP. Even if it seems surprising, normally OXPHOS is not completely coupled, and the coupling extent depends on multiple factors. The mitochondrial uncoupling can be detected directly as a decrease in ∆*p* or indirectly as a decrease in the phosphorylation efficiency, namely in the ADP/O ratio and or in the respiratory control ratio (RCR), which corresponds to the ratios of State 3 (ADP-stimulated) and State 4 (basal) respiratory activities. State 4 respiration, after subtraction of the non-mitochondrial oxygen consumption detected upon respiratory chain inhibition, mirrors the oxygen consumption of the so-called proton leak. The latter consists in the futile cycle of H<sup>+</sup> that from the IMS flow downhill in the mitochondrial matrix through pathways independent of the F1FO-ATPase [380]. The proton leak extent can be modified by protein complexes, exogenous compounds or permeability changes due to

changes in the mtIM composition. Even if, at first, the mitochondrial uncoupling was only associated with mitochondrial dysfunction, at present, it emerges as key ruler of biological processes, as suggested by the occurrence of endogenous natural uncouplers, even if they can also have pathological meaning [257]. Accordingly, dissipative pathways normally occur in variable amounts in different cell types. In mammals, the most known physiological uncoupling occurs in the brown adipose tissue, where an integral membranebound protein (UCP-1) channels H<sup>+</sup> and dissipates the electrochemical gradient resulting in thermogenesis. Other uncoupling proteins (UCP-2 and UCP-3) rule insulin secretion in pancreatic β cells and fatty acid metabolism in muscle, brown adipose tissue and heart, respectively [381]. Additionally, the four-membered family of adenine nucleotide translocators (ANTs), which catalyzes the ATP/ADP exchange across the mitochondrial membrane, possesses uncoupling properties [382].

Lastly, mitochondrial uncoupling can be also produced by exogenous chemicals, protonophores, namely lipophilic weak acids which can cross the mtIM, or non-protonophores, able to activate latent proton leaks.

A severe mitochondrial uncoupling leads to ATP depletion and eventually cell death, while mild mitochondrial uncoupling can have positive effects. Mitochondrial uncoupling can play opposite roles: it can promote cell death but also helps to protect cells against cell death, according to the cell type, the uncoupler and the uncoupling extent [257]. Uncoupling should prevent an excessive ∆Ψ rise which would block electron transfer along the respiratory chain. The decrease in ∆Ψ and consequently in ∆*p* also decreases ROS generation and vice versa, an increased ROS generation is known to decrease proton leak. On the other hand, the ATP synthase inhibition leads to electron accumulation in the respiratory complexes and ROS overproduction, which leads to oxidative stress and mitochondrial dysfunction. Endogenous mitochondrial uncoupling could prevent excessive ROS production. Accordingly, mitochondrial uncouplers have been tried to treat diseases featured by oxidative stress such as diabetes, obesity, cardiovascular diseases, neurodegenerative and aging-related diseases [134].
