*2.1. Organization of the Respiratory Chain: Historical Outline*

The advancement of science is long and tortuous: it often happens that novel discoveries, rather than providing further clarifying information on the existing basic knowledge, bring doubts and contradictory issues. This happened for the discovery of the respiratory chain supercomplexes (SCs) and for the elucidation of their function. Therefore, we perform a concise historical survey of the respiratory chain, since some early findings came out before the present knowledge, but their significance was not understood.

As we mentioned in the previous section, Hatefi et al. [13] accomplished the systematic resolution and reconstitution of four functional respiratory complexes from mitochondria, and proposed that the overall electron transfer from substrates to oxygen results from both intra-complex and inter-complex redox reactions: intra-complex electron transfer takes place in the "solid" state of redox components (e.g., flavins, FeS clusters, cytochromes) having fixed steric relation, whereas inter-complex electron transfer operates by rapid diffusion of the mobile components acting as co-substrates, i.e., CoQ and cyt. *c* [14]. In the following years, this proposal was confirmed by the kinetic analysis of Kröger and Klingenberg [25,26], leading Hackenbrock et al. [27] to postulate the *Random Collision Model of Electron Transfer*:

"Electron transport is a diffusion-coupled kinetic process; Electron transport is a multicollisional, obstructed, long-range diffusional process; The rates of diffusion of the redox components have a direct influence on the overall kinetic process of electron transport and can be rate limiting, as in diffusion control . . . It is concluded that mitochondrial electron transport is a diffusion-based random collision process, and that diffusion has an integral and controlling effect on electron transport."

The random collision model was accepted by most researchers in this field. Nevertheless, the accumulation of further experimental evidence obtained with newly developed

techniques has led to the proposal of a different model of supramolecular organization based upon specific interactions between individual respiratory complexes [28].

Indeed, a supramolecular assembly of respiratory components had been already present in the pioneering studies of Chance and Williams [29], who described the respiratory chain as an aggregate of coenzymes (flavins and cytochromes) in solid-state, embedded in a protein matrix.

Surprisingly, evidence for a non-random arrangement of respiratory complexes also derived from the same early investigations of Hatefi's group, reporting isolation of CI+CIII units [30], suggesting that such units preferentially assemble in the native membrane. These aggregates were considered physiological by their discoverers, indicating the existence of supramolecular units of electron transfer. It is therefore clear that SCs, as well as the entire NADH oxidase (known as the respirasome, see later) had been isolated and reconstituted. The publication of the fluid mosaic model of biomembranes [31] was applied also to mitochondrial membranes, thus favoring the random collision model of Hackenbrock. Thus, Hatefi's idea of specific associations among complexes was overlooked.

The idea of fixed associations, however, was never completely abandoned: in fact, some authors reported the possible existence of specific associations between respiratory complexes, either fixed [32] or dynamic [33].

At the turning of the century, Schägger and Pfeiffer [34] applied the technique of Blue-Native Gel Electrophoresis (BN-PAGE) to digitonin-solubilized yeast and mammalian mitochondria and found electrophoretic bands of high molecular weight indicating specific associations among respiratory complexes. According to their discoverers, these associations represent the state of the respiratory chain under physiological conditions. The same study also provided evidence for a dimeric ATP synthase complex.

The supramolecular arrangement of the respiratory complexes is well established [35,36], although the functional role of respiratory SCs is still controversial and their relationship with a random distribution of the individual complexes partially remains to be clarified [37–40].

#### *2.2. Distribution and Composition of Respiratory Supercomplexes*

All SCs exhibit highly ordered structures, thus they are not the result of artificial protein–protein interaction due to the extraction and solubilization procedures [41].

In most studies, SCs appear to contain the three "core" respiratory complexes, i.e., CI, CIII, and CIV, whereas the other respiratory enzymes may be randomly distributed in the lipid bilayer [12]. These "core" complexes have the common feature of carrying out proton translocation and of having subunits encoded by mitochondrial DNA.

Respiratory SCs have been described and characterized in the mitochondria of several mammalian tissues and in many other organisms, also including plants, fungi and bacteria [42].

The supramolecular structural organization of the "core" respiratory complexes CI, CIII, and CIV is conserved in all higher eukaryotes. The SC I1III2IV1–4 contains all three complexes required for the complete oxidation of NADH by molecular oxygen, and for this reason it was called "respirasome". SCs containing only two components are also normally found in large amounts, as the SC I1III<sup>2</sup> in which CIV is not present, as well as SCs III2IV. In bovine heart mitochondria, only a small aliquot of CI is present in free form in the presence of digitonin [43]; thus, it is believed that all CI is bound to CIII in the native membrane.

A higher level of complexity in the organization state of SCs may be due to the presence of additional protein components unrelated to electron transfer complexes (e.g., Shy1, SCAF1, Rcf2, HIGD2A) occasionally found by BN-PAGE [44].

Wang et al. [45] described a multifunctional mitochondrial fatty acid β-oxidation (FAO) complex that is physically associated with SCs. Analysis of genetic defects in both fatty acid oxidation and OXPHOS [46] revealed that some patients with primary FAO deficiencies exhibit secondary OXPHOS defects. These metabolic interrelations support the view that OXPHOS proteins and FAO are physically associated, and that these interactions are critical for both functions.

The other respiratory enzymes not comprising the "core" of the proton translocation machinery appear not to be associated in SCs [47]. In most studies of BN-PAGE, CII was not found associated with other complexes of the respiratory chain. In addition, mitochondrial glycerol phosphate dehydrogenase is absent in the respirasome.

Fang et al. [48] observed that mitochondrial function is hampered by dihydroorotate dehydrogenase deficiency; the enzyme was shown to physically interact with respiratory complexes CII and CIII by immunoprecipitation and BN/SDS/PAGE analysis.

Sulfide-quinone oxidoreductase and sulfite oxidase, involved in sulfide oxidation, which transfer electrons to the respiratory chain, were also described to be associated with SCs containing CIV [49].

#### **3. Supercomplexes May Provide a Kinetic Advantage to Electron Transfer**

The discovery of SCs led to the proposal that these aggregates form to improve substrate channeling or enhance catalysis in inter-complex electron transfer. Substrate channeling is the direct transfer of an intermediate between the active sites of two enzymes which catalyzes reactions which occur one after the other [50]; in the respiratory chain, this means that electrons are directly transferred between two consecutive enzymes by alternate reduction and re-oxidation of an intermediate which is not diffused in the medium. In such a case, inter-complex electron transfer cannot be distinguished from intra-complex electron transfer. Therefore, the mobile intermediates predicted to exhibit substrate-like behavior in the random collision model, i.e., CoQ and cyt. *c*, would be buried in the interface between two consecutive complexes within the SC.

In the CoQ region, the occurrence of electron transfer by channeling of CoQ between adjacent CI and CIII contrasts with the so-called CoQ "pool behavior", previously sustained by the kinetic analysis of Kröger and Klingenberg [25]. To better understand the following sections, we believe it necessary to briefly mention their kinetic analysis. These authors showed that steady-state respiration can be described as a simple two-enzyme system, where the first enzyme reduces ubiquinone (Vred) while the second one oxidizes ubiquinol (Vox), kinetically behaving as a homogeneous pool. Thus, the total CoQ molecules must randomly link any number of the dehydrogenase(s) with any number of the oxidase(s); the overall steady-state activity (Vobs) is related to Vred and Vox as shown by the so-called pool equation: (Equation (3))

$$\text{Vobs} = (\text{Vred} \times \text{Vox}) / (\text{Vred} + \text{Vox}) \tag{3}$$

It has been shown that cyt. *c* also exhibits "pool behavior " [51].

In the following sections, the relations existing between the two models and the ensuing controversial findings will be analyzed. The main challenge to study the physiological role of SCs is to selectively disrupt respiratory SCs in the cell [40].

#### *3.1. Structural Evidence Suggesting Channeling*

#### 3.1.1. Molecular Structure of Supercomplexes

If channeling occurs between CI and CIII by the common intermediate CoQ, the redox groups involved in CoQ reduction by CI and CoQH<sup>2</sup> re-oxidation in CIII must be in close contact in order to form a driving pathway containing CoQ itself; similar reasoning applies to cyt. *c* between CIII and CIV. An essential requirement for this condition is the availability of a high-resolution map of the molecular structure of the SCs.

In the initial studies, purified SCs were analyzed by negative-stain electron microscopy and single-particle cryogenic electron microscopy (cryo-EM) [52,53]. Pseudo-atomic models of the respirasome (SC I1III2IV1) were produced by fitting the known X-ray structures of the component complexes to the 3D maps of the respirasome. Evidence for channeling in the respirasome stems from the observed distinct arrangement of the three component complexes, in which the CoQ-binding sites in CI and in CIII face each other and are divided by a gap within the SC membrane core, which most likely contains lipids. Althoff et al. [52] proposed that CoQ may cover a trajectory through such a 13 nm-long gap.

Further high-resolution studies by cryo-EM of mammalian mitochondrial SCs [54–57] may be useful to predict the CoQ pathway between CI and CIII, as shown in Figure 2.

**Figure 2.** Bovine mitochondrial supercomplex (SC) I<sup>1</sup> III<sup>2</sup> IV<sup>1</sup> . Fitted model by single particle cryo-EM. Side view (on the left) and view from the IMS (on the right) showing two CL (cardiolipin) molecules (in space filling mode) in the cavity of each monomer of CIII formed by cytochromes *c*<sup>1</sup> and *b*. Image modified from Mileykovskaya and W. Dowhan [58]. The enzyme subunits are drawn as ribbon representations obtained from modified PDB ID code: 2YBB. Blue, CI; red, CIII dimer; green, CIV.

In theory, channeling within the SC may occur through either close docking of the active sites with real tunneling of substrate among proteins, or covering relatively long distances presumably by substrate-restricted diffusion (microdiffusion) within the space between the two active sites; both alternatives share a channeling which necessarily occurs between two fixed sites of the same SC. Microdiffusion is kinetically distinguished from bulk diffusion (pool behavior), where the interaction of the substrate molecules stochastically occurs by collisional encounters with a large number of possible sites in distinct respiratory complexes attained by random diffusion.

Information on channeling within the SC molecular structure must be compatible with the known CoQ reduction mechanisms by CI and its re-oxidation by CIII dimers. Ubiquinol oxidation occurs via a cyclic mechanism known as the Q-cycle [59,60]. On considering the functional asymmetry of the CIII dimer [61], it is suggestive to consider that in the supercomplexed CIII dimer only one monomer is active for NADH-dependent respiration and CoQ-channeling, since the CoQ reducing site I CI and the Qo site where ubiquinol is reoxidized are close to each other. On the contrary, the other monomer, which lacks such constrictions, might be easily available to interact with the CoQ pool at the (distal) Q<sup>i</sup> site.

The group of Kuhlbrandt [57] showed that the CoQ reduction site of CI in the SC is 11 nm far from the proximal CIII monomer, whereas the distal one is 18 nm away, thus suggesting that the latter is functionally inactive. The linear CI arrangement with the active CIII monomer and with CIV is strongly in favor of channeling for electron transfer between complexes in the respirasome by substrate microdiffusion.

On the other hand, the respirasome structure reported by Letts et al. [54] indicates that both CoQ interaction sites in CI and CIII are separated and easily accessible to the membrane, and are likely to provide no limit to free CoQ diffusion in the inner mitochondrial membrnae (IMM).

The SC structure, resolved at up to 3.8 Å in four distinct states, suggests that CoQ oxidation may be rate limiting because of unequal access of CoQ to the active sites of CIII<sup>2</sup> [62]. CI changes between "closed" and "open" conformations, while a key transmembrane helix

rotates in a striking way. Moreover, the CI state influences the conformational flexibility within CIII2, demonstrating crosstalk between the enzymes.

According to Letts et al. [62], under high turnover conditions, individual CIs may overcome the capacity for CoQ pool balance, so that the local CoQH<sup>2</sup> pool may attain higher concentration than in the bulk membrane. By ensuring close association of CI and CIII<sup>2</sup> in SCs, but still allowing free CoQ exchange with the bulk, the SCs would prevent any local accumulation of CoQH2, without limiting CI's activity by making it entirely dependent on the adjacent CIII<sup>2</sup> for the CoQH<sup>2</sup> re-oxidation.

The structural analysis of SCs allows analogous considerations for cyt. *c* between CIII and CIV.

The structural evidence of purified SCs by cryo-EM adds structural reasons for the looseness of CIII–CIV interactions in mammalian mitochondria. In ovine mitochondria, Letts et al. [54] identified two distinct arrangements of SC I1III2IV1: a major "tight" form and a minor "loose" form. In both structures, the density for CIV is weaker relative to that for CI and CIII, indicating the greater conformational flexibility of CIV. In the tight respirasome, CIV contacts both CI and CIII, whereas CIV in the loose respirasome has defined contacts only with CI. Tight and loose architectures may represent independent structures or may interconvert each other, according to distinct stages of assembly or disassembly. The presence of two respirasome forms with distinct CIV linkages was also detected by Sousa et al. [57] and ascribed to instability of the purified SC.

#### 3.1.2. Supercomplexes Are Dynamic Structures: The Plasticity Model

Together with kinetic evidence by flux control analysis in favor of CoQ channeling (See Section 3.2.2), we also postulated [63] that the SC I1III<sup>2</sup> may be in equilibrium with free CI and CIII. In fact, we found that NADH-cyt. *c* reductase activity in reconstituted proteoliposomes can occur either by channeling or by random diffusion, depending on the protein density in the lipid bilayer of the proteoliposomes.

Acín-Pérez et al. [64] in a study of purified SCs showed by BN-PAGE that other types of associations exist besides the respirasome, such as CI+CIII or CIII+CIV. They therefore suggested that a variety of associations between respiratory complexes is likely co-exists in vivo with free complexes and proposed an integrated model, the plasticity model, for the organization of the respiratory chain. According to their view, the previous contrasting models, solid vs. fluid, are only the two extremes of a dynamic range of molecular associations between respiratory complexes [64]. The plasticity model proposes that both CoQ and cyt. *c* are sequestered into distinct sub-domains and are both trapped into SCs and freely diffusible. This promiscuity can stem either from the dissociation of SCs or from the escape of CoQ or cyt. *c* from the supramolecular assemblies [36,65].

The plasticity model and the dynamics and stability of mitochondrial SCs [37] are still debated issues. The putative factors involved in association/dissociation of SCs, such as membrane potential and post-translational changes, are reviewed by Genova and Lenaz [42]. However, even if indirect evidence shoulders the SC dynamic assembly and disassembly, up to now no direct demonstration exists that SCs physiologically form and dissociate in vivo.

#### 3.1.3. The Role of Lipids: Cardiolipin in Supercomplexes

SCs have 2–5 nm gaps at the transmembrane interfaces of the individual complexes (Figure 2) that are presumably filled with lipids. Wu et al. [56], in their high-resolution structure of the purified SC I1III2IV<sup>1</sup> from porcine heart, localized many phospholipid molecules including cardiolipin (CL) involved in the protein–protein interactions [66]. Substantial progress was made by studying Barth syndrome, a genetic syndrome characterized by severe cardiopathy, in which cardiolipin remodeling is altered due to the mutation of the gene *Tafazzin*; in Barth syndrome patients SCs are unstable, leading to a functional impairment characterizing the disease [67]. In immortalized lymphoblasts from Barth's

syndrome patients the amount of SCs is decreased, as well as the amount of individual CI and CIV [68].

Direct involvement of CL in the formation of SC III–IV was demonstrated in genetically manipulated strains of *S. cerevisiae* in which the CL content can be regulated in vivo. Yeast mutants lacking CL have normal amounts of individual complexes, but lack the stable SC III2IV<sup>2</sup> as observed in the wild type parental strain [69]. Since *S. cerevisiae* lacks CI, these studies could not provide information on SCs containing CI.

In contrast with CL depletion that destabilizes SCs, the loss of phosphatidylethanolamine (PE) favors the formation of larger SCs between CIII and CIV in *S. cerevisiae* mitochondria. The reason why CL and PE, even if both associate in non-bilayers, act in an opposite way on SC stability may depend on the different charge, since, at physiological pHs CL is anionic and PE is zwitterionic. Using yeast mutants of PE and phosphatidylcholine (PC) biosynthesis, Baker et al. [70] showed a specific requirement for mitochondrial PE, but not PC, in CIII and CIV activities, but not for their formation. Neither PE nor PC were involved in respiratory SC formation, emphasizing the specific requirement of CL in SC assembly.

Our laboratory has stressed the importance of CL in stabilizing CI-containing SCs. Reconstitution of binary CI/CIII proteoliposomes from bovine heart mitochondria at high lipid to protein ratio (30:1 w:w) prevents formation of SC I1III<sup>2</sup> [71]. However, SC I1III<sup>2</sup> and related NADH-cyt. *c* reductase activity are maintained if these high-lipid proteoliposomes are enriched with 20% CL (w:w), resembling the percent content of CL in the mitochondrial membrane (M. Kopuz, Y. Birinci, S. Nesci, G. Lenaz and M.L. Genova, unpublished data). Likewise, the dilution of native protein-bound CL in the excess exogenous phospholipids prevents SC assembly, but the effect is reversed by increasing the CL content of the liposomes, therefore shifting again the equilibrium to CL binding to protein.

Addition of reactive oxygen species (ROS) to mitochondria affects the respiratory activity due to cardiolipin peroxidation, since CL is required for the optimal activity of respiratory complexes [72,73]. We showed by flux control analysis (cf. Section 3.2.2) that the SC I1III<sup>2</sup> in proteoliposomes disappears if lipid vesicles are peroxidized before protein reconstitution [71]. Evidently, the distortion of the lipid bilayer induced by peroxidation and the alteration of the phospholipid annulus originally present in the purified SC I1III<sup>2</sup> determines its dissociation.

CL deficiency was reported to cause a defect in respiratory function and a decrease in ATP synthesis [74], an indirect demonstration of CL role in the structure of the SCs. Rieger et al. [75] investigated the effect of ALCAT1 overexpression. This enzyme catalyzes the incorporation of polyunsaturated fatty acids into CL. The resulting CL species are more susceptible to oxidative damage, leading to increased ROS production and mitochondrial dysfunction. Using galactose as sugar supply, cells must employ OXPHOS for ATP synthesis, shown as a significant increase in basal- and ADP-linked respiration; ALCAT1 overexpression prevents the galactose-induced increase in respiration.

Likewise, a genetic defect of CL synthase induces loss of SC assembly and of respiratory activity as well as other mitochondrial defects [76].

#### *3.2. Kinetic Evidence for Channeling in the Coenzyme Q Region*

#### 3.2.1. Rate Advantage Imposed by SCs

The overall rate in a diffusion-coupled pathway is always less than that of the ratelimiting step and only approaches the latter when it is widely different (i.e., much slower) compared to the rate of the other step(s) in the pathway [25]. On the contrary, the rate is expected to be equal to the rate of the limiting step if channeling of the intermediate substrate(s) occurs.

Ragan and Heron [77] investigated in detail the reconstitution of CI in the respiratory chain; they showed that purified CI and CIII, when mixed as concentrated solutions, reversibly associate in a 1:1 molar ratio to form an SC I+III endowed with NADH-cyt. *c* reductase activity. The study demonstrated a stoichiometric behavior for this activity, revealing the existence of an active SC formed by CI and CIII: electron transfer is fast within the SC and conversely very slow from the SC to free CIII via the CoQ pool.

On the other hand, prevailing CoQ-pool behavior could be induced and CI and CIII could be made to operate independently of each other if the SCs were functionally dissociated into "free" complexes, by raising the amount of phospholipid and ubiquinone in the concentrated mixture. Under this condition, electron transfer from CI to CIII is still ensured by CI+CIII<sup>2</sup> units that however dissociate and reform at rates exceeding the rates of electron transfer in the individual complexes.

According to Heron et al. [78], the mobility of CI and CIII in the membrane is lost and the complexes are frozen in their SC assembly when phospholipids are not in excess with respect to the content needed to form a lipid annulus around the protein. Such frozen state favors a stable orientation of the site of CoQ reduction by CI with respect to the site of oxidation by CIII. They also showed that protein-bound CoQ<sup>10</sup> leaks out of the SC I+III in the lipid bilayer when an extra phospholipid is present in the proteoliposomes; the consequent decrease in activity could be reversed by adding additional quinone. As a corollary, we may deduce that a large amount of ubiquinone found in the natural membrane is needed to maintain the CoQ<sup>10</sup> content in the SC unit when it is formed.

Our laboratory provided a direct evidence that the phospholipid amount affects the choice of channeling with respect to CoQ-pool behavior [79]. A proteoliposome system obtained by fusing a crude mitochondrial fraction enriched in CI and CIII with different amounts of phospholipids and CoQ<sup>10</sup> allowed us to discriminate between these two mechanisms. The experimental NADH-cyt. *c* reductase activity was compared with the theoretical values obtained from the pool equation of Kroger and Klingenberg [25], showing overlapping results in the range from 1:10 to 1:40 (*w*/*w*) protein to lipid ratios. However, pool behavior was not shown at low (1:1 *w*/*w*) lipid to protein ratio. Moreover, the observed rates of NADH-cyt. *c* reductase were higher than the theoretical values; this 1:1 ratio corresponds to the mean nearest distance between respiratory complexes in mitochondria.

Several studies point out that SC disorganization by several causes is accompanied by decreased electron transfer and energetic efficiency (cf. also Section 3.1.3).

Recently Garcia-Poyatos et al. [80] in Enriquez's group investigated the physiological role of SCs, generating two null allele zebrafish lines for supercomplex assembly factor 1 (SCAF1). The *scaf1*−*/*<sup>−</sup> fish showed altered OXPHOS activity due to the disrupted interaction of C III and CIV.

Solsona-Vilarrasa et al. [81] studied the effects of increased cholesterol levels, as occurring in vivo in alcoholic and non-alcoholic fatty liver, or by in vitro cholesterol enrichment of mouse liver mitochondria. Cholesterol feeding caused oxidative stress and mitochondrial GSH (mGSH) depletion, which lead to liver steatosis and damage. The overload of cholesterol in mitochondria disrupts mitochondrial functional performance and the organization of respiratory SC assembly, thus potentially contributing to oxidative stress and liver injury.

Tomkova et al. [82] showed that Tamoxifen-resistant cells show a significant decrease in mitochondrial respiration accompanied by a decrease in mitochondrial respiratory SC and significantly increased levels of mitochondrial superoxide (See Section 5).

Balsa et al. [83] showed that endoplasmic reticulum (ER) stress and glucose deprivation stimulate mitochondrial OXPHOS and formation of respiratory SCs by action of protein kinase R-like ER kinase (PERK). Accordingly, PERK genetic ablation or pharmacological inhibition abolishes nutrient and ER stress-mediated increase in SC levels and reduce OXPHOS-dependent ATP synthesis.

These studies suggest that electron transfer between CI and CIII in NAD-linked respiration can take place either by CoQ channeling within the SC I1III<sup>2</sup> or by a less efficient random diffusion behavior, depending on the state of the membrane lipids: nevertheless, it appears that channeling is preferred under physiological conditions.

#### 3.2.2. Metabolic Flux Control Analysis

Our laboratory first exploited flux control analysis [84] to establish the functional association between individual respiratory complexes in beef heart mitochondria [63]. Using specific inhibitors to establish the extent of metabolic control exerted by each individual complex over the entire NADH-dependent respiration, we demonstrated that CI and CIII are both rate-limiting, since they exhibit flux control coefficients (FCC) close to 1, thus behaving as a single enzymatic unit (e.g., SC I1III2); we therefore concluded that electron transfer through CoQ is accomplished by channeling between CI and CIII [63]. Using the same method for succinate oxidation we found that CII, but not CIII, is rate-limiting, confirming the notion that CII does not form SCs and that the oxidation of succinate follows pool behavior.

A decade later, Blaza et al. [85], on the basis of flux control analysis, criticized the evidence for channeling in the respiratory chain: accordingly, rotenone competes with CoQ and, consequently the CI inhibition extent is affected by the additional quinone analog used as substrate in the CI assay, but absent in the NADH aerobic oxidation assay. Indeed, using rotenone and CoQ<sup>1</sup> as substrate they find an FCC for CI exceeding 1 that is certainly artefactual. However, in our study we used decyl-ubiquinone (DB) rather than CoQ1. DB has much lower affinity than CoQ<sup>1</sup> for CI [86], thus exerting lower competition with rotenone and lower influence on the inhibition efficiency measured by us [63]. Moreover, the FCC values found by Blaza et al. [85] using alternative inhibitors of CI (e.g., piericidin and diphenyleneiodonium, not competitive with CoQ) are unrealistically low, indicating the presence of a rate-limiting step downhill. Most likely, the limiting amount of cyt. *c* in their sample would shift the main control of the chain to the cyt. *c* region. Significantly, in Blaza's study, addition of exogenous cyt. *c* to the mitochondrial samples enhanced the FCC from 0.19 to 0.67.

Overall, we must emphasize that our major clue [36] to demonstrate association of CI to CIII was not so much the high FCC of CI but the high FCC of CIII in NADH oxidation, which is incompatible with a CoQ-pool model where CI and CIII independently float in the membrane. Flux control of CIII was not performed in the study by Blaza et al. [85]. Such high FCC value of CIII over NADH-dependent respiration was measured by inhibitor titration with mucidin, and is clearly not artefactual because the corresponding FCC of CIII measured by using the same inhibitor in succinate oxidation is low (as expected from the rate-limiting role of CII and the lack of SC II+III). It is evident that the FCC of CIII is close to 1, indicating a rate-limiting step in CIII, only if the preceding COQreductase is assembled with CIII in an SC, which is the case only for CI, but not for CII. In addition, the high FCC that CIII shows over NADH-cyt. *c* reductase activity in reconstituted proteoliposomes is strongly lowered when the SC I1III<sup>2</sup> is dissociated [71] by reconstitution in excess phospholipids or in peroxidized phospholipids.

The few other studies addressed to SCs using metabolic control analysis have confirmed the possibility that the respiratory chain is arranged as functionally relevant supramolecular structures [87–89].

#### 3.2.3. Coenzyme Q Compartmentalization

The possibility that the CoQ pool may not be homogeneous has been advanced long time ago. Gutman and Silman [90] observed partial additivity and competition between succinate oxidation and NADH oxidation. On the basis of their observations (and those of others) [91], they proposed a compartmentalization of the CoQ pool, in which two sub-domains partially interact through a spill-over diffusion-mediated mechanism.

Gutman [92] also investigated NADH and succinate oxidation in submitochondrial particles (SMP), as well as the rates of energy-dependent reverse electron transfer from succinate to NAD<sup>+</sup> and of forward electron transfer from NADH to fumarate, concluding that:

"The electron flux from succinate dehydrogenase to oxygen (forward electron transfer towards Complex III) or to NADH dehydrogenase (reverse electron transfer) employs the same carrier and is controlled by the same reaction" whereas "the electron transfer from NADH to oxygen does not share the same pathway through which electrons flow in the NADH-fumarate reductase".

In other words, CI and CII (reverse electron transfer) are linked by a different pathway with respect to CI and CIII (forward electron transfer), even if CoQ is a common intermediate for both pathways. We can clearly interpret this non-homogeneity of the CoQ pool with respect to succinate and NADH oxidation as deriving from sequestration of CoQ in the SC I+III in contrast with the free CoQ molecules that connect CII and CIII.

In past years, several other reports have suggested that the CoQ pool is not homogeneous and questioned its universal validity [93]. Similar observations were made by other authors [94] who observed different functional CoQ pools in mitochondria.

Lapuente-Brun et al. [95] demonstrated that the physical association of CI and CIII determines a preferential pathway for electrons mediated by a dedicated subset of CoQ molecules. This compartmentalization would prevent significant cross talk and mixing up between NADH oxidation (CI-dependent) and succinate oxidation (dependent on CII) or other flavoenzyme-dependent oxidations. Moreover, the CIII units within the SC I+III are exclusively dedicated to oxidation of NADH whereas free CIII is dedicated to oxidation of other substrates using the free CoQ pool (e.g., fatty acids through ETF, succinate, glycerol-3-phosphate and choline).

On the contrary, Blaza et al. [85] showed that the steady-state rates of aerobic NADH and succinate oxidation were only partly additive in bovine heart submitochondrial particles SMP; moreover, cytochromes *b*H, *b*L, *c* and *c*<sup>1</sup> in the same cyanide-inhibited particles were reduced to a similar extent by either NADH or succinate or a mixture of the two substrates. Blaza et al. [85] interpreted the results as a demonstration that a single homogeneous pool of CoQ molecules exists that receives electrons indifferently from CI and CII. Note that these data were obtained on the whole pathway of electrons from ubiquinol to oxygen and therefore also comprise the steps through cyt. *c* and CIV. It is therefore not possible to discriminate whether the pool behavior is due to the homogeneity of the CoQ-pool or of the cyt. *c* pool.

To overcome this uncertainty, we carried out experiments on succinate and NADH oxidation by exogenous cyt. *c* in KCN-inhibited mitochondria, thus shortening the electron route for both oxidation reactions by only including CoQ and CIII as redox partners [36]. Under such conditions, NADH and succinate oxidation by cyt. *c* was completely or almost completely additive and close to the theoretical sum of the two individual reactions, thus suggesting that two separate CoQ-compartments may exist.

Notably, we also obtained similar results (Tioli G, Falasca AI, Lenaz G and Genova ML, data communicated at EBEC2016-Session P4) using proteoliposomes enriched in CI, CII, and CIII. In these proteoliposomes, the additivity of NADH and succinate oxidation decreased at increasing CIII inhibition by mucidin, thus substantiating the hypothesis that the CoQ molecules in the SC I+III are in a dissociation equilibrium with the molecules in the pool [96], cf. next section.

Fedor and Hirst [97] in another study incorporated an alternative quinol oxidase (AOX) into mammalian heart mitochondrial membranes to have another competing pathway for ubiquinol oxidation. Since AOX strongly increased the rate of aerobic NADH oxidation, they concluded that the quinol generated in SCs by CI is more rapidly reoxidized in the CoQ pool by AOX than by CIII inside the SC. Based on these results, Hirst [98] concludes that quinone and quinol diffuse freely in and out of SC: no substrate channeling occurs, since it is not required to support respiration.

In our opinion, this criticism is inconsistent for the same reasons discussed above for concomitant NADH and succinate oxidation in presence of a CIII inhibitor, since Fedor and Hirst used KCN to block electron transfer through CIV: under this condition electrons from ubiquinol are forced to follow the alternative pathway by dissociation into the CoQ pool (cf. Section 3.3.1). It is intuitive that channeling is largely predominant during high turnover of the respiratory chain, whereas at low turnover, CoQ diffusion in the pool would preferentially take place.

Recently, our analysis was strongly supported by Szibor et al. [99] who expressed AOX in mouse mitochondria and observed that electron input by succinate (CII), but not by NADH (CI), almost completely reduces the Q pool (>90%) irrespective of the respiratory state. AOX enhances the forward electron flux from CII and decreases reverse electron transport. Interestingly, AOX does not act on CI substrates, except in the presence of a respiratory inhibitor (as proven by Fedor and Hirst [97]).

Recently, Enriquez's group [100] also presented results confirming that the SC of CI and CIII allows the partial segregation the CoQ pool and allows substrate channeling. Their results with different cellular models suggest that when CI is not superassembled, as in CIII-KO cells, CoQ occurs in a unique pool, whereas CI superassembly triggers the formation of two partially different CoQ pools.

#### *3.3. If Electron Transfer Occurs via Channeling, What Is the Role of Coenzyme Q Pool?*

Since a mobile pool of CoQ in the mtIM coexists with CoQ sequestered in SCs, we may ask whether this pool is a mere reservoir of an excess of CoQ molecules without a specific function or whether the CoQ pool is in any way required for physiological electron transfer and/or for additional functions.

#### 3.3.1. Dissociation Equilibrium of Bound Coenzyme Q

Compelling evidence that electron transfer from CI to CIII occurs physiologically by channeling/restricted diffusion stems from the observation that CI is almost totally associated in an SC with CIII. However, we reasoned that an excess of entrapped CoQ in the pool is also required for channeling to occur [96]. In fact, the CoQ molecules bound in the SC, that ensure electron transfer directly from CI to CIII are in dissociation equilibrium with the CoQ pool. Therefore, the extent of CoQ binding to the SC depends on the equilibrium with free CoQ in the pool and on the size of the pool itself. The existence of this equilibrium is widely confirmed by a series of studies, such as the previously reported findings by Heron et al. [77] on CI to CIII association, by the saturation kinetics for CoQ exhibited by the integrated activity of CI and CIII (NADH-cyt. *c* oxidoreductase) [101] and by the decrease in respiration in mitochondria fused with phospholipids causing subsequent dilution of the CoQ pool [102].

In conclusion, besides acting as a diffusible substrate of several flavin reductases not involved in SC organization (cf. next section), free CoQ molecules are a reservoir for binding to the SC. Studies on respiration under pathological conditions [103,104] showed that SC assembly is required for proper respiration, even if activity of the individual complexes is normal. On the other hand, reconstitution in vitro as described in the previous Section 3.2, suggests that respiration can occur in both modes (channeling and diffusion), if CoQ is available. In a proteoliposome system where CI was reconstituted together with an alternative oxidase (AOX) and CoQ10, and in absence of SC, Jones et al. [105] found high rates of NADH oxidation through AOX, demonstrating that CI can deliver electrons through the CoQ pool.

The dynamic character of CoQ bound within the SC, which is in dissociation equilibrium with the free pool of CoQ in the membrane, is the major point favoring controversy concerning channeling. We propose that the CoQ molecules trapped in a lipid microdomain within the SC and are channeled from CI to CIII during electron transfer at steady state, without significant exchange with the free pool; however, CoQ dissociation from the SC to the pool becomes significant when electron transfer in the respiratory chain is slow or blocked by an inhibitor. In this respect, the plasticity model would be a functional rather than structural feature of the respiratory chain, a possibility taken into account by Enriquez himself [37]. According to this idea, the SCs are stable structures that do not readily dissociate under physiological conditions, while CoQ is said to behave in a highly dynamic fashion: CoQ channeling and diffusion shall occur in proportions depending on the turnover rates of the respiratory chain.

#### 3.3.2. Electron Transfer between Free Complexes

Several flavoenzymes reduce CoQ as outlined in Section 2. Among these, the most widely investigated is CII. All evidence converges in the statement that electron transfer from CII to CIII takes place only through the CoQ pool. Accordingly, succinate oxidation kinetically follows pool behavior after extraction and reconstitution [101] and in intact mitochondria [106] in accordance with the notion that CII does not participate in SC formation (see previous sections). For the same reason, also energy-dependent reverse electron transfer from succinate to NAD<sup>+</sup> , taking place through sequential activity of CII and CI connected by CoQ, must take place by collisional interactions in the CoQ pool. The hyperbolic relation experimentally found by Gutman [92] between the rate of reverse electron transfer and succinate oxidase is in complete accordance with the pool equation.

Other CoQ reductases such as glycerol-3-phosphate dehydrogenase, ETF dehydrogenase, dihydroorotate dehydrogenase, choline dehydrogenase, sulfide dehydrogenase, that are present in minor amounts, are probably inserted in the respiratory chain by interaction through the CoQ pool [12], but kinetic evidence is poor.

Some knowledge exists for mitochondrial glycerol phosphate dehydrogenase (mt-GPDH). A study [107] demonstrated that in brown adipose tissue (BAT) mitochondria the inhibition curve of glycerol phosphate-cyt. *c* reductase is sigmoidal in the presence of myxothiazol or antimycin A. Such type of inhibition suggests the occurrence of a homogeneous CoQ pool between mtGPDH and CIII [26]. CIII reduction by mtGPDH in human neutrophil mitochondria was also shown to occur in absence of SC organization and of NAD-linked respiration [103], providing further evidence that mtGPDH functions through the CoQ pool. Preliminary studies by BN-PAGE (M.L. Genova, M. and H. Rauchova, unpublished) showed that mtGPDH is not apparently linked to any of the respiratory complexes. Mráˇcek et al. [108] demonstrated that mtGPDH associates into homo-oligomers but also in high molecular weight SCs of more than 1000 kDa whose composition is unknown, but not associated with CI, III, or CIV as shown by BN-PAGE analysis. No association was found between CI or other OXPHOS complexes and the electron transferring flavoprotein (ETF) that participates in fatty acid oxidation.

Indirect evidence, however, suggests that fatty acyl CoA oxidation proceeds through a separate CoQ pool with respect to NADH oxidation (cf. Section 3.5).

#### *3.4. Electron Transfer through Cytochrome c: Is There Evidence for Channeling?*

BN-PAGE clearly shows that a fraction of CIV, as well as CIII, participates in SC assembly (cf. Section 2). However, metabolic flux control analysis [63] showed that CIII has a high FCC, as well as CI, in NAD-linked respiration, whereas CIV has a low FCC; this means that CI and CIII behave as a functional enzyme unit, whereas CIV seems to be functionally independent.

These results cannot be easily explained [109]. It is true that most of CIV appears to be free in the BN-gels (cf. Section 2), and one might consider that the cytochrome *c* oxidase activity in the respirasome is masked by the large excess of active enzyme randomly distributed in the membrane. On the other hand, this would mean that channeling is not a major feature of electron transfer in the cyt. *c* region. The latter means that the free molecules of CIV are involved in electron transfer from NADH and implies that the molecules of CIV assembled in the respirasomes are not involved in the channeling of cyt. *c* [38,42], but do behave similarly to the free CIV units in electron transfer by random diffusion.

The conclusion that the absence of channeling in the cyt. *c* region of mammalian mitochondria is a physiological feature, is supported by the observation that the respirasome of potato tuber mitochondria is completely functional in cyt. *c* channeling, as detected by the same flux control analysis in mammalian mitochondria [71]. This difference may be ascribed to a tighter binding of cyt. *c* in the potato respirasome, as shown by BN-PAGE.

The structural evidence of purified SCs by cryo-EM adds structural reasons on the looseness of CIII–CIV interactions in mammalian mitochondria. In ovine mitochondria, Letts et al. [54] identified two distinct arrangements of SC I1III2IV1: a major "tight" form

and a minor "loose" form (resolved at the resolution of 5.8 Å and 6.7 Å, respectively). In both respirasome structures, CIV has a weaker density than the other complexes in the respirasome, indicating the greater conformational flexibility of CIV. In the tight respirasome, CIV contacts both CI and CIII, whereas CIV in the loose respirasome only interacts with CI. Tight and loose architectures may represent independent structural entities or may interconvert, indicating different stages of assembly or disassembly. The presence of two forms of the respirasome having distinct CIV linkages was also detected by Sousa et al. [57] and ascribed to instability of the purified SC.

In partial contrast with the above considerations, Lapuente-Brun et al. [95] demonstrated that at least part of CIV forms a functional SC with channeling of cyt. *c*, but the SC formation depends on the availability of the SC assembly factor SCAF1. They showed that when CIV participates in SCs due to the presence of SCAF1, a significant proportion of CIV activity is not utilized in cell respiratory activity. They also identified three CIV populations, one of which is dedicated exclusively to receive electrons from NADH oxidation (forming SC I+III+IV), presumably by cyt. *c*. channeling. The second population receives electrons from FAD-dependent enzymes (forming SC III+IV and presumably operating by cyt. *c* channeling), whereas the third major one is in free form and receives electrons from both NADH and FADH<sup>2</sup> oxidization, presumably using the free cyt. *c* pool. On the contrary, if CIV is maintained permanently detached from SCs by elimination of SCAF1, all CIV is in free form and electron transfer takes place via diffusion of a single pool of cyt. *c*.

Functional evidence for cyt. *c* channeling was also found in *S. cerevisiae* [110] mitochondria which are characterized by having all CIV bound to CIII in a supercomplex [111], thus preventing electron transfer through free CIV units. The structural evidence by single particle cryo-EM sustains channeling, since in the III2IV<sup>2</sup> SC (there is no CI in this yeast species) the distance between the binding sites of cyt. *c*, i.e., cytochrome *c*<sup>1</sup> of CIII and the CuA-subunit II of CIV, is considerably shorter than that in bovine mitochondria [112]. At odds with this structural evidence, however, Trouillard et al. [113] showed that the time-resolved oxidation of cyt. *c* by CIV in yeast mitochondria is a random-collision process.

Rydström Lundin et al. [114] studied the kinetics of aerobic quinol oxidation in *S. cerevisiae* mitochondria as a function of the respiratory SC factors Rcf1 and Rcf2 that mediate supramolecular interactions between CIII and CIV forming SCs III2IV1-2. They demonstrated that Rcf1 promotes formation of a direct electron-transfer pathway from CIII to CIV via a tightly bound cyt. *c*. Accordingly, in these mitochondria under steadystate conditions in the presence of added homologous cyt. *c*, the direct electron transfer through the bound cyt. *c* (i.e., cyt. *c* channeling), is faster than the equilibration of electrons with the cyt. *c* pool. Interestingly, when using heterologous cyt. *c* (from horse heart), electron transfer between CIII and CIV occurs only via the cyt. *c* pool. Realistic theoretical assumptions indicate that electron transfer between CIII and CIV can become rate limiting. Hence, there is a kinetic advantage of bringing CIII and CIV together in the membrane to form SCs [115]. In a yeast mutant defective in SC formation, but containing fully functional individual complexes, the diffusion of cyt. *c* between the separated complexes is delayed, thus reducing electron transfer efficiency [116].

Moreover, other results indicate that in the absence of Rcf1, the interactions required for stability of the SC III+IV are disrupted [117]. Consistently, Rydström Lundin et al. [114] demonstrated that direct electron transfer does not take place in the absence of Rcf1 and electrons are only transferred via the cyt. *c* pool in the Rcf1∆ strain.

In the next Section 3.5, we will discuss the possible physiological reasons why there is no major channeling of cyt. *c* in mammalian mitochondria, contrary to yeast or plant mitochondria.

#### *3.5. Supercomplexes and Regulation of Metabolic Fluxes*

Does CoQ channeling occur under physiological conditions? As discussed in the previous Section 3.3, the presence of a bottleneck downhill CoQ might induce interaction of the NADH pathway (via channeled CoQ in SC I1III2IVn) with the succinate pathway (CoQ pool behavior) by shifting the CoQ dissociation equilibrium. However, if a rate-limiting step is situated upstream, i.e., in or before the dehydrogenases, the reducing pressure of CoQ on its partner oxidases (particularly in the case of CIII assembled in the SC I1III2IVn) may not be present and the two routes would take place independently. On the other hand, in mammalian mitochondria, it appears that the two fluxes mix up in the cyt. *c* region where most evidence excludes cyt. *c* channeling (Section 3.4).

The existence of two different functional CoQ compartments, one for NADH oxidation an another one for oxidation of succinate and other FAD-linked substrates, has deep implications for the metabolic adaptation to the feeding state. The flux of electrons in the respiratory chain is conditioned by the orientation of metabolism to use different fuels. The use of different fuel molecules generates different proportions of NADH and FADH2, that require an optimal equilibrium between the corresponding routes of electron transfer in the respiratory chain. In a well-fed individual, the oxidative metabolism largely follows the glycolytic pathway and the Krebs cycle, so that the oxidation of NAD-linked substrates exceeds succinate oxidation (ca. 5:1) and the electron flux proceeds mainly through SC I1III2. On the other hand, metabolic adaptation of liver mitochondria to fasting forces fat mobilization and FAO [118]. In fact, an increase in the FADH2-dependent respiration, as in FAO, induces saturation of the CoQ pool reoxidation capacity and promotes reverse electron transport from ubiquinol to CI [119]. The resulting local generation of superoxide triggers protein degradation of CI subunits by oxidative damage, and consequent disintegration of the complex, as well as lipid peroxidation, disassembly of SCs and further instability of their enzyme components. Thus, we can infer that CoQ redox status acts as a metabolic sensor that fine-tunes the configuration and supramolecular organization of the respiratory chain in order to match the prevailing substrate profile [120]. In order to avoid such saturation of the CoQ pool oxidation capacity, fasting conditions require proper disassembly of SCs to achieve adjustment of respiratory SC proportions favoring the FAD-linked route. Similar adaptations may occur when the increase in FAO results in response to high-fat diet. Impairment of this adaptation may be relevant to pathological processes associated to obesity.

As postulated by Lapuente-Brun et al. [95], at the metabolic level, the fast kinetic adjustments of the mitochondrial respiration are followed by gene expression changes in the level of individual respiratory complexes.

Why appreciable channeling may not exist at the level of cyt. *c* in mammalian mitochondria? (cf. Section 3.4). Mammalian mitochondria have several dehydrogenases directing electrons to CoQ, however they usually have only one oxidase (CIV) receiving electrons from CIII via cyt. *c*. For this reason, it may be worth speculating that, whilst it may be useful to separate the major NADH-dependent flux from CI from those departing from succinate, fatty acid oxidation and other metabolic pathways by separating the CoQ compartments, there is no such need for cyt. *c* that is by and large receiving electrons univocally from CIII. This assumption is reinforced by the fact that, contrary to what is found in mammalian mitochondria, CIII and CIV of plant mitochondria are functionally operating as an SC and, at the same time, cyt. *c* is tightly bound in the SC, as revealed by flux control analysis [63] and BN-PAGE [71]. Not surprisingly, plant mitochondria are characterized by a high branching of the electron transfer pathways feeding electrons directly to cyt. *c* [121], which requires adjustment of the different routes as a response to physiological needs, as it happens in mammalian mitochondria at the CoQ level. In plant mitochondria, segmentation might be achieved by regulating different compartments of free and bound cyt. *c*.
