Mitochondrial Structure and Bioenergetics in Normal and Disease Conditions

The OXPHOS machinery is embedded in the IMM, together with a variety of other mitochondrial proteins. Indeed, the IMM is one of the most protein-rich lipid bilayers in biological systems, with a protein/lipid mass ratio of ~75:25 [415]. Many protein complexes which localize in this compartment are not distributed randomly, but tend to cluster in specific regions, according to their individual function. As discussed previously, the IMM can be divided into two subdomains: the inner boundary membrane (IBM), opposite to the OMM, and the cristae membrane (CM), the alias for invaginations of the membrane in the matrix, connected by the cristae junctions.

Results obtained with quantitative immunogold-EM on mammalian mitochondria and yeast cells demonstrated that the preferential location of OXPHOS complexes is the CM, but that both subcompartments are dynamic and the distribution of mitochondrial proteins can change according to the physiological state of the cell [416,417]. In particular, it has been observed that OXPHOS enzymes—or intermediates of them—might localize in different regions of the membrane in different stages of maturation. With the only exception of CII, the respiratory chain enzymes are composed of both nuclear- and mitochondrial-encoded subunits, and their biogenesis is the result of the coordination between two temporally and spatially separated protein synthesis machineries. Consequently, it was proposed that, while proteins synthesized in the matrix are translated and inserted directly in the cristae, subunits that must be imported from the cytosol are preferentially inserted in the IBM [417]. More recent analyses by super-resolution microscopy and quantitative cryo-immunogold-EM have helped to determine where specifically CIII, CIV and CV subunits are translated and inserted in the yeast inner membrane [418]. This study confirmed that, under steady-state conditions, the mature forms of CIII, CIV and CV localize mainly in the CM, while early stages of assembly are more enriched in the IBM. Indeed, mitoribosomes translating COB mRNA (the very first step of CIII assembly) and Cbp3 and Cbp6 (markers for the early assembly of the enzyme) were found more present in the IBM, while Cbp4 (a marker of the early-to-mid-assembly phase) was already less enriched in this subcompartment. Conversely, the integration of the Rieske protein, the last step of maturation, which takes place after CIII dimerization, happened predominantly in the CM.

Similarly, Stoldt and colleagues investigated the localization of markers of the early (Pnt1, Cox18, Coa1 and Cox20) and the late (Pet100) phases of CIV assembly, finding the earliest stages of maturation enriched in the IBM and the late phases in the CM. By contrast, the entire assembly pathway of CV seems to occur mainly in the CM [418].

The picture of the OXPHOS machinery as individual enzymes sitting in the IMM is a simplified vision of what happens in the mitochondria of living cells. A second level of complexity is added by the formation of supercomplexes (SC), stable structures composed of the association of the respiratory chain enzymes.

The development of blue native PAGE techniques [419] allowed the separation and detection of both the individual complexes and the supercomplexes, composed of different combinations of CI, CIII2 and CIV. However, many questions remained open in the field. Firstly, it was necessary to characterize the type and strength of these inter-complex interactions, in order to hypothesize a realistic model of IMM organization. Secondly, it was necessary to attempt to explain the biological and physiological function of these structures. This second aspect will be one of the central questions of this thesis.

The models to explain the organization of the respiratory chain enzymes have changed over time. The first proposal, known as the “fluid state” model, describes the mitochondrial complexes as individual and independent units that float in the inner membrane and collide randomly with each other, allowing the passage of electrons. According to this theory, cytochrome c would be diffusing in three dimensions, in order to transport electrons between complexes, and this transient and casual formation of each SC species by collision would allow the cell to adapt to different energy demands. This model was the most accepted one during the 1980s, following the presentation of the random collision model of Hackenbrock [420]. The fluid model is supported by the very high protein-to-lipid ratio in the membrane, which allows the frequent interaction of the enzymes, and by scanning calorimetry [421] and freeze-fracture electron microscopy [422] studies showing that the intramembrane particles are randomly distributed in the IMM. Moreover, independent studies showed that CoQ is a mobile carrier [423,424], confirming an essential part of this theory. On the other hand, however, both blue native PAGE and electron microscopy experiments in different systems, such as mammals, bacteria, yeasts and plants, suggest more stable interactions between these particles [425,426,427,428,429]. In the original experiments, it was necessary to solubilize the membranes with a strong detergent in order to isolate the single enzyme, while milder detergents such as digitonin used in blue native studies preserve the various species of SC intact.

The second and opposite model proposed was the “solid state” model, which supported the vision that the different activities and redox centers were contained in an undissociated protein matrix. These structures were also thought to contain Q and cyt c [430,431]. This view, however, was challenged by the observation that the isolated complexes were functional and that they could diffuse within the lipid bilayer, as well as activity rate values being compatible with a mechanism of random collision [432]. The “solid state” model remained the most accepted view until the development of the native electrophoretic techniques described above, which showed the co-existence of supercomplexes of different sizes and the individual respiratory chain complexes. From a functional point of view, the proximity of the enzymes and the molecular carriers involved would increase the interactions and prevent the intermediates from escaping or being sequestered by other enzymes for use in secondary metabolic pathways [433]. Indeed, in the model in which the different supramolecular species coexist and trap the mobile electron carriers, the association of CI, III and IV and, on the other side, CIII and IV would define two different functional CoQ populations: CoQ dedicated to transferring electrons originating from NADH (CoQNADH), which is trapped in SC containing CI, and free CoQ in the inner mitochondrial membrane for use by CII and other enzymes that use FAD (CoQFAD) [434]. However, this system is not compatible with the kinetics of the CoQ pool, and recent additional studies indicated that cytochrome c diffuses freely in the IMS in yeast [433] and that only one CoQ and cytochrome c pools are present, accessible to all the enzymes [435].

An additional interesting analysis has been proposed after inserting AOX (an alternative oxidase) in vitro, a cyanide-insensitive quinol oxidase originally from Trypanosoma brucei, in bovine heart mitochondria [436]. AOX receives electrons from ubiquinone, bypassing CIII and CIV. In the case of two separate pools of ubiquinone, the “respirasome pool” should continue to provide a substantial flux of electrons through the supercomplex. Instead, a competition between respirasomes and AOX was observed in increasing the AOX supplement, confirming the theory that ubiquinone is not channeled but can move freely in the IMM.

In conclusion, putting together the information available today, the most realistic proposal is a middle-ground and more dynamic model in which various species of supercomplexes and free respiratory complexes coexist. For now, the model that explains this phenomenon is the so-called “dynamic aggregate” or “plasticity” model, which proposes a dynamic interchange of the complexes in their “free” and associated states into SC in response to varying energetic demands. A premise for the model is that the complexes are pre-assembled individually before associating with each other [437,438].

Structures of the I1III2IV and I1III2 SC from mammals and plants have been resolved by single-particle electron cryo-EM [439,440,441,442] and by electron cryo-tomography (cryo-ET) [443,444]. The yeast Saccharomyces cerevisiae lacks CI, and complexes III and IV form two SC species III2IV1 and III2IV2, the structures of which have been recently solved [445,446].

In mammals, the main supercomplex species are III2IV1, I1III2 and the respirasomes I1III2IV1−n. The existence of a “Megacomplex” with a I2III2IV2 stoichiometry, which might also bind CII, was reported in cultured human cells [441]. Additionally, CIV and CV can form homodimers and CIV is involved in the formation of other high-molecular weight structures not yet identified, but visible in blue native PAGE experiments [395,447,448]. “Megacomplexes” containing also CII have been proposed as well, but their existence is debated.

Since Saccharomyces cerevisiae lacks CI [449], yeast-based studies of the respirasomes have been conducted in Yarrowia lipolytica [444]. In this model, CI and CIII interactions are comparable to what has been observed in mammals, while CIV was found in various positions around CIII2, but not in the position most commonly observed in the porcine heart respirasomes, contacting the ND5 module of CI (Figure 14).

In plants, on the other hand, CIV was never observed in the respirasomes and the only apparent supercomplex was I1III2. Interestingly, supercomplexes are not universal and simpler organisms such as Escherichia coli, which express CI, CII and CV homologues, but not CIII, do not present such structures [450]. In these bacteria, the different complexes have been shown to not co-localize together and to not sub-localize in specific domains over the membrane surface [451].

Interestingly, the percentage of free complexes vs. SC varies largely between different species. In bovine heart mitochondria, electron cryo-tomography indicated that only 56% of CI was associated with other complexes [444]. However, BN-PAGE-based quantifications of the same cell type estimated that only 14–16% of free CI was present [426]. On the other hand, blue native-PAGE experiments using human mitochondria show a minimal part of CI (~10% at the most, in the presence of digitonin) migrating as an individual complex [453]. In any case, there seems to be more of “free” CI in mitochondria from different tissues in mammalian organisms such as sheep, pigs or mice than in humans [437,454]. These differences might be due to various reasons, including artefacts introduced during the sample preparation or actual differences in the regulation of energy demands between different organisms.

In addition, in the same organism, not all the complexes engage in the formation of SC in the same proportion. Indeed, in humans, it has been estimated that while 85%–100% of CI is inserted in a supercomplex structure, only 55%–65% of CIII and 15%–25% of CIV are found in SC [426,453,455,456].

Structural analyses have also been important to define which subunits are involved in the complex–complex interaction. The association between CI and CIV seems to involve subunit COX7A of CIV, and ND5 and NDUFA9 of CI (Letts et al., 2016). COX7A, together with COX8B, could be involved also in the interaction with CIII, through binding of UQCRC1, UQCRQ, UQCR10 or UQCR11 (Letts et al., 2016) [440,444]. Concerning the connection between CI and CIII, subunit NDUFA11 of CI seems to face and contact the transmembrane region of CIII, close to UQCRB and UQCRQ [440] (Letts et al., 2016). Other CI subunits that might be involved in the interaction are NDUFAB1 and NDUFB9, contacting UQCRC1 of CIII.

The biological function of SC is still highly debated and various hypotheses have been proposed. The first theory proposes that their role is substrate channeling, increasing OXPHOS efficiency due to the physical proximity of all the machinery components [457]. The second possible function of SC might be to minimize ROS production. ROS are generated by the reduction of oxygen by electrons leaked from the ETC and are physiologically used as a signaling molecule [458]. However, if there is an imbalance between the excessive formation of ROS and limited antioxidant defenses, ROS can become deleterious and damage mtDNA, lipids and proteins. In this case, the formation of supra-structures might decrease electron or proton leakages during respiration. This theory is supported by two studies. In the first one, ROS production by CI has been shown to increase in two experimental conditions that inhibit the formation of CI and CIII interactions: treating bovine heart mitochondria or liposome-reconstituted supercomplex I-III with dodecyl maltoside, and reconstructing CI and CIII at a high phospholipid/protein ratio [459]. In the second study, the difference in SC formation in astrocytes and neurons was analyzed, showing how astrocytes, which present a higher percentage of free CI, are characterized by poorer mitochondrial respiration but higher ROS production [460].

The last hypothesis, which will be the basis of the first project presented in this thesis, is that the formation of SC is necessary for the assembly and/or stability of the single enzymes. This theory is supported by the observation that defects in one enzyme can lead to multi-complex deficiencies. Indeed, mutations in MTCYB, fundamental for CIII assembly, also induce CI deficiencies [461,462,463]. Similarly, patients presenting with mutations in CIII assembly factors, such as BCS1L, can display defects in both CI and CIV [464], while the suppression of CIV in mouse fibroblasts affects CI assembly or stability [465]. An explanation for this phenomenon was that CI was destabilized after its complete assembly by an active oxidative stress-triggered degradation [466,467,468]. However, a multistep model for SC assembly that sees CI intermediates binding CIII and CIV subunits before the completion of the mature enzyme was proposed [469]. In the same study, UQCRFS1 was shown to insert preferentially within the III2 + IV supercomplex, instead of dimeric CIII2.

Finally, an additional role of SC is reserved to the ATP synthase. CV does not seem to interact with the rest of the respiratory chain, but it is able to form dimers and more complex supra-structures such as tetramers and hexamers [470]. These supercomplexes localize specifically at the bottom of the cristae, where they appear to enforce a strong local curvature on the inner membrane, necessary for the formation of the invagination [471].

Being that the assembly of the monomeric forms of the OXPHOS enzymes requires the participation of a number of assembly factors, it has been proposed that additional proteins should be involved in the formation of supercomplexes.

The first proposed assembly factor was COX7A2L or SCAF1 (supercomplex assembly factor 1), an isoform of the CIV subunit COX7A, described originally to have a role in the inclusion of CIV in III2 + IV and I + III2 + IV1–4 [434]. However, following studies showed that the lack of COX7A2L affects the formation of the SC III2 + IV, but not of respirasomes [453,472,473,474]. Pérez-Pérez data also showed how COX7A2L binds to CIII2 early during its assembly, indicating a preferential interaction with this complex [472]. Thus, it might possess a role in the establishment of a checkpoint for the regulation of CIII2 levels and its incorporation into SC [453].

Another study showed how the incorporation of different COX7A isoforms might determine which CIV-containing species (monomer, dimer or respirasome) is formed [13]. Thus, the CIV subunit COX7A2 is replaced by COX7A2L in supercomplexes with both CIV and CIII, and by COX7A1 in CIV dimers. Moreover, other CIV tissue-specific subunit isoforms might play a role in this, as well as dimers, which appeared to be stabilized by the COX6A2 isoform rather than COX6A1. COX6A subunits are localized at the interface between the two monomers, a position that would favor the regulation of the dimerization. COX7A2 and COX6A1, instead, would favor the free CIV form.

Analyses conducted in Saccharomyces cerevisiae suggested the participation of Rcf1 and Rcf2 in the formation of respirasomes [448]. In yeast, the loss of Rcf1 affects CIV function and the correct insertion of Cox13 (equivalent to human COX6A). These data showed also deficiency in SC formation, but recent reports point to this being an indirect effect due to CIV defects [475,476,477]. While Rcf2 is specific to yeast, two orthologs of Rcf1 have been identified in humans: RCF1a/HIGD1A and RCF1b/HIGD2A. HIGD1A has been confirmed to bind CIV and it has been suggested to affect the interaction between CIV and cytochrome c, or to act in a still unclear way on the heme centers [306,307]. Another study found HIGD1A directly involved in CIII biogenesis, promoting the final incorporation of UQCRFS1, and suggested its participation in the formation of CIII-containing SC [478]. Concerning HIGD2A, instead, two independent groups demonstrated that the knockdown of this protein leads to impaired SC formation by the release of CIV from respirasomes [479,480]. However, HIGD2A was identified as a MTCO3 module assembly factor and it is possible that its impact on SC is a pleotropic effect due to impaired CIV biogenesis [478,481].

Concerning oligomerization of CV, the role of post-translational modifications in Atp20 has been investigated [482]. Specifically, the molecular mechanism seems to revolve around the phosphorylation of serine 62, which inhibits the dimerization of the ATP synthase.

Finally, also the lipid composition of the IMM and, in particular, the level of cardiolipin are important for the stabilization of SC. In yeast lacking Taz1, the acyltransferases involved in the remodeling of cardiolipin and CIII and CIV associations are destabilized [483]. In Barth syndrome patients, harboring mutations in Tafazzin (the human ortholog of Taz1), the assembly and stability of CIV and its SC forms are affected, with a secondary effect also on CI + III associations [484].

Mitochondria fulfil a variety of functions in the cell, including the well-known energy production via oxidative phosphorylation. For this reason, the failure of mitochondria to operate correctly is associated with a wide spectrum of genetic disorders. In this review, we have analyzed, in detail, the structure, the assembly pathway and the organization in supercomplexes of each component of the OXPHOS machinery and how defects in these enzymes are linked to diseases in humans.

Our review aimed to show the heterogeneity and the complexity behind the still growing group of pathologies normally identified as “mitochondrial diseases”, the pathophysiology of which is still often poorly understood. During the past decade, big advancements in genetic screening and in the study of mitochondrial biology and physiology have been made. However, the diagnosis of mitochondrial disorders is limited to about half of the suspected cases, the genotype–phenotype correlation is not always comprehended and effective therapies are still missing.

Future research will need to continue investigating the biology of this organelle, in order to better understand the molecular and cellular basis of mitochondrial diseases, and identify possible targets for new treatments.

This work was funded by the UK Medical Research Council core grant (MC_UU_00015/5) and the European Research Council (erc.europa.eu) Advanced Grant FP7-322424, NRJ-Institut de France, Fondazione Renato Comini ONLUS, and Telethon Italy grant GGP19007 to M.Z.

UK Medical Research Council core grant (MC_UU_00015/5) and European Research Council (erc.europa.eu) Advanced Grant FP7-322424, NRJ-Institut de France, Fondazione Renato Comini ONLUS, and Fondazione Telethon Italy grant GGP19007.