The functional significance of mitochondrial respiratory chain supercomplexes

Abstract

The mitochondrial respiratory chain (MRC) is a key energy transducer in eukaryotic cells. Four respiratory chain complexes cooperate in the transfer of electrons derived from various metabolic pathways to molecular oxygen, thereby establishing an electrochemical gradient over the inner mitochondrial membrane that powers ATP synthesis. This electron transport relies on mobile electron carries that functionally connect the complexes. While the individual complexes can operate independently, they are in situ organized into large assemblies termed respiratory supercomplexes. Recent structural and functional studies have provided some answers to the question of whether the supercomplex organization confers an advantage for cellular energy conversion. However, the jury is still out, regarding the universality of these claims. In this review, we discuss the current knowledge on the functional significance of MRC supercomplexes, highlight experimental limitations, and suggest potential new strategies to overcome these obstacles.

Mitochondrial respiratory chain complexes can associate into supramolecular structures termed respiratory supercomplexes. This review discusses their structure, assembly and potential physiological functions.

The mitochondrial respiratory chain

As the most versatile form of chemical energy in cells, ATP is synthesized in large quantities by mitochondrial activity. Biochemical reactions, many occurring in the cytoplasm, convert carbohydrates and lipids to mitochondrial acetyl‐CoA. This central metabolite is processed in the mitochondrial matrix by the tricarboxylic acid cycle (TCA; Fig 1). Electrons originating from these TCA reactions are passed over to the typically four multiprotein complexes (CI‐CIV) of the electron transport chain. In a process termed oxidative phosphorylation (OXPHOS), the electrons are transported through the mitochondrial respiratory chain (MRC), where they finally reduce molecular oxygen to form water (Fig 1). The energy of this reaction is converted into an electrochemical gradient across the inner mitochondrial membrane (IMM). In a subsequent step, protons flow from the mitochondrial intermembrane space (IMS) back into the matrix, powering the ATP synthase (CV), which condenses ADP and phosphate to form ATP (Fig 1).

Figure 1. The tricarboxylic acid cycle (TCA) and mitochondrial respiratory chain.

Upper Panel: Acetyl‐CoA is a central metabolite produced via various metabolic reaction pathways. It can subsequently be utilized in the TCA, were among others several molecules of NADH and FADH₂ are generated. Lower Panel: Electrons from these metabolites can be subsequently utilized by the mitochondrial respiratory chain to convert energy in the form of ATP. Structures of the individual respiratory chain complexes and the electron carrier cytochrome c are shown, as well as respective redox reactions (PDB CI: 6ZKC; PDB CII: 1ZP0; PDB CIII: 6Q9E; PDB CIV and CytC: 5IY5; PDB CV: 5ARA). IMM, inner mitochondrial membrane; IMS, intermembrane space; OMM, outer mitochondrial membrane.

NADH feeds two electrons into the NADH dehydrogenase (CI), the first and largest of the MRC multiprotein complexes in mammals, consisting of 45 subunits and comprising a mass of approx. 1 MDa (Kampjut & Sazanov, 2020; Fig 1). Electrons from NADH are thereby transferred to oxidized quinone (Q), a reaction coupled to proton extrusion from the matrix into the IMS via CI's antiporter subunits (Jones et al, 2017; Fig 1). Q is an abundant mitochondrial lipid that can be reversibly reduced and oxidized, respectively, and freely diffuses in the IMM. Interestingly, some species, including the model organism Saccharomyces cerevisiae, do not contain a conventional CI, but rather possess non‐proton‐translocating peripheral membrane NADH dehydrogenases (Nde1, Nde2, and Ndi1; Joseph‐Horne et al, 2001). In addition, electrons extracted from succinate can reduce Q via the succinate dehydrogenase (CII; Fig 1). Whereas CI pumps protons, no H⁺ is translocated by CII. Of note, Q can also be reduced by several other membrane‐bound dehydrogenases (e.g., glycerol 3‐phosphate dehydrogenase). The reduced quinone (QH₂) subsequently transfers two electrons to cytochrome bc ₁ reductase (CIII), which exists as an obligate homodimer (CIII₂; Fig 1) in all species (Stephan & Ott, 2020). CIII couples electron transfer from QH₂ to cytochrome c (Cytc) with the shuttling of H⁺ into the IMS via a mechanism called Q‐cycle (Mitchell, 1975; Osyczka et al, 2005). In one such Q‐cycle, two QH₂ molecules are oxidized to Q, releasing 4 H⁺ into the IMS and reducing two Cytc molecules (Fig 1). Of note, another Q is reduced to QH₂ in this process, which can subsequently be used for another round of the Q‐cycle. The protein Cytc acts as a mobile electron carrier, with a central, covalently attached heme group that can reversibly bind one electron. Cytc transfers electrons from CIII to Cytc oxidase (CIV), where molecular oxygen is reduced to water (Fig 1). In this last step of the electron transport chain, another 4 H⁺ are pumped into the IMS by oxidizing 4 Cytc molecules (Wikstrom, 1977). Finally, the electrochemical gradient across the IMM formed by the transfer of protons from the mitochondrial matrix into the IMS is used by the ATP synthase (CV), which phosphorylates ADP with inorganic phosphate to ATP, whereby the flow of protons back into the matrix powers ATP synthesis (Mitchell, 1975) by a rotary mechanism of the ATP synthase (Fig 1).

Connectivity of mitochondrial respiratory chain complexes

Despite detailed insights into the fundamental structural and catalytic properties of the individual MRC complexes, available since last century, their overall structural organization and physiological implications of that organization have been extensively revisited over the last 20 years and are still debated today (Milenkovic et al, 2017). Historically, two opposing models for how the complexes are functionally connected were proposed early on, namely, the “solid‐state” and the “liquid‐state” (also referred to as “fluid‐state”) models. The solid‐state model postulates that all components of the MRC (including Q and Cytc) form a single unit, which can catalyze the entire process of mitochondrial respiration. Thereby, Q and Cytc mediate electron transfer via enclosed pathways, never exchanging with the exterior of the unit (Ragan & Heron, 1978). On the contrary, the liquid‐state model proposes that individual complexes reside independently in the IMM. Likewise, Q and Cytc diffuse freely and randomly to transfer electrons between the individual complexes (Hochli & Hackenbrock, 1976; Hackenbrock et al, 1986). The liquid‐state model received extensive experimental support, mostly owing to the observation that individual MRC complexes could be purified in active forms and that the small mobile electron carriers are freely diffusing and in large stoichiometric excess over MRC complexes. However, the solid‐state model gained new experimental support. Through the use of mild detergents for mitochondrial membrane solubilization and the establishment of blue native polyacrylamide gel electrophoresis, it was demonstrated that MRC complexes form supramolecular assemblies, termed respiratory chain supercomplexes (SCs) (Cruciat et al, 2000; Schägger & Pfeiffer, 2000). SCs were shown to exist in varying stoichiometry in different species, ranging from bacteria to plants, fungi, and animals. Because they exist widely in multiple life forms, they must confer specific, selectable advantages to organisms. In this review, we provide a comprehensive overview of the various forms of mitochondrial SCs, their biogenesis, and their structure and discuss the current understanding of the functional roles of SCs.

Structures and assembly of different respiratory chain complexes

Overview of different supercomplex configurations

Only a small fraction of CI is present as a free complex in bovine heart mitochondria, and in some studies using cell cultures, the free form of CI is even undetectable (Lobo‐Jarne et al, 2018), whereas the rest is found in various SC assemblies with varying stoichiometry (Schägger & Pfeiffer, 2001) (Fig 2A). The most prevalent of these assemblies in human cells is the respirasome, consisting of CI + CIII₂ + CIV and a size of approx. 1.7 MDa, accounting for approximately half of CI in SCs (Schägger & Pfeiffer, 2001; Gu et al, 2016; Letts et al, 2016; Wu et al, 2016; Lobo‐Jarne & Ugalde, 2018; Fig 2A). Other variants of the respirasome, which have been proposed to contain further copies of CIV (CI + CIII₂ + CIV_2‐4), present only minor amounts of CI‐containing SCs. The second most prevalent assembly of CI is the CI + CIII₂ configuration (Schägger & Pfeiffer, 2001; Fig 2A). Of note, although the stoichiometry of these assemblies is the same between yeasts, plants, and mammals, the structures and frequencies differ (Davies et al, 2018). Evidence for the existence of even larger assemblies than the CI + CIII₂ + CIV₄ respirasome comes from cryo‐EM studies in HEK293T cells, reporting a symmetrical respiratory megacomplex with the stoichiometry CI₂ + CIII₂ + CIV₂ (Guo et al, 2017). Although the authors also speculate that two units of CII could potentially bind to this megacomplex, a corresponding unassigned density could not be observed in their data, and so far, no direct experimental evidence for the participation of CII in supercomplex assembly of metazoa or plants was reported (Caruana & Stroud, 2020). However, a recent study using the ciliate protist Tetrahymena thermophila reports a CI + CII + CIII₂ + CIV₂ SC, which has important implications in the curvature of the IMM in this organism (Mühleip et al, 2023). Of note, this arrangement shows significant differences compared to known mammalian SCs.

Figure 2. Respiratory chain supercomplex assemblies.

(A) Stoichiometry of several mitochondrial respiratory chain supercomplexes. (B) Main interaction surfaces of mammalian respirasomes (CI + CIII₂ + CIV, PDB: 5j4z). Close‐up of interactions between CI (green) and CIII (blue) subunits NDUFA11, UQCRB, UQCRQ, and UQCRH (left panel), as well as NDUFB4 and NDUFB9 with UQCRC1 (lower right panel) are shown, as well as the interaction surface between CIII and CIV (red), formed by the CI subunit ND5 and the CIV subunit COX7C (upper right panel). Visualization is adapted from (Letts et al, 2016). (C) Interaction surfaces of the Saccharomyces cerevisiae respiratory chain supercomplex CIII₂ + CIV (PDB: 6YMX). The inner membrane protein–lipid–protein interaction between Rip1 (CIII, blue, cardiolipin (CL) and Cox5 (CIV, red) are shown (left panel), as well as the matrix‐situated protein–protein interaction between Cor1 and Cox5a). Visualization was modified from (Berndtsson et al, 2020). (D) Structure of the mammalian respiratory chain supercomplex CIII₂ + CIV in the unlocked configuration (PDB: 7O3C, upper panel). The lower panel shows a transparent visualization of the same structure to display interactions between the CIII (blue) subunit Sub9 and the CIV (red) subunit SCAF1.

Supercomplexes are also formed between CIII and CIV without CI, with the stoichiometries CIII₂ + CIV₂ or CIII₂ + CIV_, respectively (Fig 2A). Whereas CIII forms an obligate homodimer (CIII₂), CIV can occur in its monomeric, as well as in its dimeric (CIV₂) form.

The respirasomes

A common feature of respirasomes is the curving of the membrane arm of CI around the CIII dimer, which can already be observed in the isolated CI + CIII₂ complex (Letts et al, 2019). The overall structure of these SCs was similar to the respirasome configuration, but showed at least six open states (inactive form) of CI and one closed state (active form) (Letts et al, 2019). In structural studies for both ovine and bovine respirasomes, two different classes had significant variations of the CIV position (Letts et al, 2016; Sousa et al, 2016). For the ovine respirasome structure, “loose” (active) and “tight” (inactive) conformations were shown, where CIV contacts both CI and CIII in the tight variant but only CI in the loose form (Letts et al, 2016). Class 1 of the bovine respirasome is similar to the tight form and represents the most prevalent population and stable form of the respirasome (Gu et al, 2016; Letts et al, 2016; Letts & Sazanov, 2017). Two main interaction sites are found between CI and CIII₂, namely, one in the membrane between NDUFA11 and the UQCRB, UQCRQ, and UQCRH subunits of CIII, and a second one in the matrix between NDUFB4, NDUFB9, and CIII subunit UQCRC1 (Fig 2B). Interestingly, the contacts formed between CI and CIII within the respirasome involve so‐called supernumerary subunits. Compared to the 14 conserved CI subunits of bacterial origin, these supernumerary subunits are assumed to be of eukaryotic origin (Gray, 2012). As no CI‐containing supercomplexes have been identified in bacteria so far (however, CIII₂ + CIV₂ supercomplexes were described, Kaila & Wikström, 2021), it is feasible that certain CI supernumerary subunits evolved to form respirasomes (Letts & Sazanov, 2017). On the contrary, the contacts between CI and CIV are formed by the evolutionarily conserved core subunit ND5 of CI and the CIV subunit COX7C (Gu et al, 2016; Letts et al, 2016; Fig 2B).

Different states of the respirasome can be observed when looking at the four major rotational axes of its structure, including the hinge between the matrix and membrane arms of CI, the pivot between CI and CIII₂, the rotation along the two‐fold‐symmetry axis of CIII₂, and the internal asymmetric rotation of CIV (Letts et al, 2016; Letts & Sazanov, 2017). Whereas the rotation between the matrix and membrane arms of CI may be associated with the regulation of enzymatic activity or may represent different stages in the catalytic cycle (Letts et al, 2016; Agip et al, 2018; Blaza et al, 2018; Kravchuk et al, 2022), physiological implications of other conformational states remain unknown. It might be that these different conformations present intermediate states of respirasome assembly or that some configurations (e.g., the loose form of the respirasome) are sample preparation artifacts (Letts et al, 2016).

The exact mechanism of respirasome formation, particularly involving SC assembly factors, is still a matter of debate. The CIV subunit COX7A has two tissue‐specific isoforms, namely, COX7A1 in the heart and skeletal muscles and COX7A2 in the liver. However, both isoforms can be found in heart‐derived mitochondria, whereas COX7A2 is found in liver mitochondria only. Further, a homolog of COX7A2, COX7A2L, or SCAF1 was described, initially thought to be essential for both respirasome and CIII₂ + CIV SC formation (Ikeda et al, 2013; Lapuente‐Brun et al, 2013). It was postulated that SCAF1 is required to form respirasomes in all tissues except in the heart and skeletal muscle (Cogliati et al, 2016). However, overwhelming evidence shows that SCAF1 is only required for CIII₂ + CIV SC formation but not for the respirasome (Davoudi et al, 2016; Jha et al, 2016; Pérez‐Pérez et al, 2016; Sun et al, 2016; Williams et al, 2016). Although SCAF1 is essential for the assembly of the CIII₂ + CIV SC (as discussed in more detail below), it does not contribute to the assembly of the respirasome, as the formation of the CIII₂ + CIV SC is not necessary for respirasome formation (Pérez‐Pérez et al, 2016). Likewise, SCAF1 was not detected in the structure of the respirasome. However, a recent analysis of human cells by complexome profiling has led to the identification of two distinct respirasome forms that present slightly different electrophoretic mobility and contain either SCAF1 or COX7A2 isoforms (Fernández‐Vizarra et al, 2022).

Supercomplex CI + CIII₂ and CIII₂  + CIV

Although the configuration of CI + CIII₂ is observed much less frequently in mammalian cells as compared to the respirasome (approx. 17% of total complex I in mitochondria compared to approx. 54% for the respirasome) (Schägger & Pfeiffer, 2001), this arrangement is widely observed across kingdoms (Davies et al, 2018).

Recently, high‐resolution structures of T. thermophila respiratory chain complexes have been solved, including CI + CIII₂ (Zhou et al, 2022). Interestingly, the overall arrangement is similar to that of mammalian CI + CIII₂ SCs as discussed above. However, significant differences in the interaction sites were observed, with more contacts in both the matrix and the IMM, as well as an IMS bridge present in T. thermophila SCs. Thereby, these interactions cause structural and functional symmetry breaking in CIII₂. The authors further suggested that no free CI is present in T. thermophila and that the CI + CIII₂ represents the minimal organization of CI (Zhou et al, 2022). Moreover, very recent cryo‐EM structures of Arabidopsis thaliana CI + CIII₂ showed a plant‐specific additional interaction site between the complexes compared to ovine CI + CIII₂ structures, potentially contributing to the unusual stability of the plant SC (Klusch et al, 2023). Similar plant‐specific interaction surface between CI and CIII₂ was also shown for Vigna radiata (Maldonado et al, 2023).

In the yeast S. cerevisiae that lacks a canonical CI, the ratio between the two SC species CIII₂ + CIV₂ or CIII₂ + CIV reflects the cellular energy demands under different growth conditions (Schägger & Pfeiffer, 2000). In particular, the abundance of SC CIII₂ + CIV₂ is enhanced in cells propagating in respiratory media and is mainly controlled by the amount of CIV available (Schägger & Pfeiffer, 2000). The first elucidated high‐resolution structure for mitochondrial CIII₂ + CIV came from Saccharomyces cerevisiae (Hartley et al, 2019; Rathore et al, 2019). The association between CIII₂ and CIV occurs via protein–protein interactions between the CIII subunit Cor1 and the CIV subunit Cox5, stabilizing the complex from the matrix side, and a protein–lipid–protein interaction between Cor1, cardiolipin, and the CIV subunit Rip1 (Fig 2C). Whereas alanine substitutions in Cor1 abolished the formation of SCs, the reduction of cardiolipin content via deletion of the cardiolipin synthase Crd1 had no direct consequence on SC formation (Pfeiffer et al, 2003), showing that the protein–protein interaction between Cor1 and Cox5 is necessary and sufficient for the formation of CIII₂ + CIV SCs in yeast (Berndtsson et al, 2020). However, the yeast and mammalian CIII₂ + CIV differ substantially in their structure. In yeast, CIV binds to CIII₂ at the same site as CI does in the respirasome, resulting in different binding modes between CIII and CIV in yeast and mammals (Rathore et al, 2019). CIII₂ + CIV of mammals was observed in two different conformations, a locked and an unlocked state (Vercellino & Sazanov, 2021). In the unlocked state, CIV is peripherally attached to CIII_2, and almost 180° rotated compared to the yeast form of this assembly. Whereas the unlocked conformation contained both fully mature CIII₂ and CIV, the locked form harbors assembly intermediates of CIII_2.

The assembly of mammalian CIII₂ + CIV has been shown to follow a specific sequence. The assembly factor SCAF1 binds to a premature form of CIII₂ and drives the docking of CIV to form an intermediate‐locked assembly. This is followed by the maturation of CIII₂, forming the locked conformation. Subsequently, CIV shifts to the unlocked state, in which the newly added subunits of the mature CIII₂ form contacts with CIV (Vercellino & Sazanov, 2021; Fig 2D).

The yeast proteins Rcf1 (and its mammalian homolog HIGD2A), as well as Rcf2 (yeast‐specific), were previously suggested to mediate assembly and stability of CIII₂ + CIV SCs (Chen et al, 2012; Strogolova et al, 2012; Vukotic et al, 2012; Zhou et al, 2018). However, Rcf1 was later shown to interact with CIV, thereby regulating the late stages of its assembly (Su et al, 2014; Garlich et al, 2017). In addition, a recent study showed that Rcf1 and Rcf2 regulate respiration of yeast cells, without significantly affecting the assembly of SCs (Dawitz et al, 2020). It was also shown that Rcf1 and Rcf2 are not stoichiometric subunits of CIV, and thus, they were also not present in the cryo‐EM structure of yeast SCs under physiological conditions (Rathore et al, 2019). However, Rcf2 could be detected in SC structures under hypoxia‐mimicking conditions (Hartley et al, 2020). This is in line with the observation that Rcf1 and Rcf2 regulate CIV activity, in particular under energy‐demanding conditions (Dawitz et al, 2020). These results suggest that the primary functions of Rcf1 and Rcf2 are related to the assembly of individual complexes and regulating their function rather than specifically mediating SC formation. In mammals, other relevant HIGD family proteins are HIGD1A and HIGD1C. HIGD1A impacts the assembly of CIII (Timón‐Gómez et al, 2020), and in addition, it binds to Cytc and CIV in the respirasome to enhance electron transfer (Hayashi et al, 2015; Guerra‐Castellano et al, 2018). In agreement, HIGD1A overexpression improves cellular respiration in models of mitochondrial disorders (Nagao et al, 2020). Differently, HIGD1C is almost exclusively expressed in glomus cells, where it also interacts with CIV to negatively regulate its activity and facilitate oxygen sensing (Timón‐Gómez et al, 2022).

Physiological functions of supercomplexes

The physiological functions of SCs remain controversial, despite more than two decades of research (Schägger & Pfeiffer, 2000). Several hypotheses regarding their physiological significance were made and will be discussed in this section.

Free diffusion versus substrate channeling

The random collision model suggests that the electron transport within mitochondria is a diffusion‐based process. The mechanistic role of Q to transport electrons from NADH or succinate oxidation to CIII was established in the 1960s, when it was demonstrated that a surprisingly high concentration of Q was found in the IMM, where it behaves as a pool, binding electrons (and protons) to the Q headgroup, and thereby helps to connect electron flow in the MRC (Fig 3A). Because Q is a substantially hydrophobic molecule, owing to the attachment of varying numbers of isoprenyl moieties (Stefely & Pagliarini, 2017), the coenzyme is solved in the lipidic core of the membrane, where it diffuses in the plane of the lipid bilayer. In contrast, Cytc is a protein of roughly 15 kDa molecular weight that carries a covalently attached heme moiety enabling the transport of one electron at a time. Cytc belongs to the class of peripheral membrane proteins. In the absence of transmembrane segments, membrane association stems from its attachment to the IMM by electrostatic interactions with negatively charged phospholipids, particularly cardiolipin, the signature phospholipid of mitochondria (Kinnunen et al, 1994). Moreover, biophysical experiments suggested that in addition to electrostatic interactions, also an alternative form of membrane interaction could exist. In this model, an acyl chain of cardiolipin might get inserted into the folded Cytc, giving rise to a tight membrane‐bound form (Rytomaa & Kinnunen, 1995). However, Cytc release from mitochondria during apoptosis is a highly conserved and important event, necessitating that the protein is detached from the IMM (Ott et al, 2007). This propensity of Cytc, namely to interact with the IMM via various mechanisms and the well‐documented fact that it can be detached from the membrane to act as a signaling molecule during cell death execution (Garrido et al, 2006), complicates the interpretation of how the protein diffuses inside mitochondria. However, the general consent is that Cytc is freely moving in the IMS (Hochman et al, 1982), either moving in two dimensions at the level of the IMM, in three dimensions in the IMS, or a mixture of both, whereby the actual ionic strength might dictate the degree of membrane association (Fig 3B).

Figure 3. Physiological functions of respiratory chain supercomplexes.

(A) Comparison between the free diffusion and substrate channeling model for coenzyme Q (Q) within respiratory chain supercomplexes. Arrows indicate the free exchange with the Q pool (free diffusion model, left panel) or the direct Q transfer in the form of substrate channeling (right panel). (B) Comparison of various postulated models for cytochrome c shuttling between complex CIII and CIV in respiratory chain supercomplex configuration. Arrows indicate the modality of cytochrome c transfer/exchange with the free pool, similar as described for the Q pool above. (C) Potential physiological functions of different respiratory chain supercomplex assemblies. Several of these assigned functions are controversially discussed. Arrows indicate a respective increase or decrease of the indicated physiological functions and pathophysiological effects.

The discovery of SCs, however, led to the hypothesis that respirasomes might act as solid‐state units, with each respirasome carrying its own Q and Cytc molecules and catalyzing the transfer of electrons from NADH to O₂ (Fig 3A and B). Flux control analyses take advantage of inhibitors to block a certain step in a reaction pathway. If the effect on the isolated step is equal to the effect on the overall pathway (and hence, the flux control coefficient is 1), this step has complete control over the pathway. Such an analysis has been performed for the NADH oxidation pathway with the inhibitors rotenone and mucidin, and it was shown that CI and CIII behave as a single functioning unit (Blanchi et al, 2004). In support of this model, Q and Cytc were found in respirasomes and shown to be able to transfer electrons from NADH to O₂ (Acín‐Pérez et al, 2008). It was therefore proposed that neither Q nor Cytc form homogeneous pools, but instead, each respirasome contains a dedicated small pool of Q and Cytc, which are not exchanged and allow a localized electron transfer via substrate channeling (Lapuente‐Brun et al, 2013; Fig 3A and B). Substrate channeling in complexes of enzymes that act sequentially in a pathway is a process where a specific substrate is transferred from one enzymatic activity to the next without allowing free diffusion of the substrate into the bulk solution. For example, the large E1‐E2‐E3 enzymes like pyruvate dehydrogenase or alpha‐ketoglutarate dehydrogenase employ substrate channeling by transferring a covalently bound substrate between the complexes' three enzymatic entities using a highly flexible domain.

Although the substrate channeling hypothesis represents an attractive model for the physiological functions of SCs, experimental evidence against this hypothesis emerged. First, repetitions of flux control experiments showed that the outcome of these experiments strongly depended on the respective inhibitor used (Blaza et al, 2014). As certain inhibitors even resulted in a biological meaningless flux control coefficient of > 1, this approach is not a reliable method to analyze physiological functions of SCs (Blaza et al, 2014; Milenkovic et al, 2017). The authors also measured substrate oxidation using NADH, succinate, or both in submitochondrial particles containing supercomplexes, prepared with and without the addition of Cytc to alter the rate‐limiting step for substrate oxidation. The same system with the same set of supercomplexes showed both additive and non‐additive kinetics (Blaza et al, 2014). These results reject the hypothesis of a Q pool partitioning in SCs. Similarly, spectroscopic measurements in living cells also demonstrated that Cytc does not face any barriers to free diffusion (Trouillard et al, 2011). In an additional study, an alternative QH₂ oxidase was inserted into mammalian heart mitochondrial membranes, which then presents a competing pathway for QH₂ oxidation (Fedor & Hirst, 2018). In the case of substrate channeling, the flux through the competing pathway should be negligible. However, the rate of cyanide‐insensitive NADH:O₂ oxidoreduction via the alternative QH₂ oxidase was significantly higher, revealing that CIII/CIV catalysis is rate‐limiting for both, NADH and succinate oxidation, and that Q is not channeled in SCs. In addition, structures of the mammalian respirasomes (Gu et al, 2016; Letts et al, 2016) and CIII₂ + CIV SCs (Vercellino & Sazanov, 2021), yeast CIII₂ + CIV_1‐2 SCs (Hartley et al, 2019; Rathore et al, 2019; Berndtsson et al, 2020), or plant CIII₂ + CIV SCs (Maldonado et al, 2021) do not contain barriers restricting diffusion of Q or Cytc away from the complexes, features that would be necessary for substrate channeling. Direct tunneling of electrons between CIII₂ + CIV in the MRC can also be excluded because of a too‐large distance (> 6 nm) between the respective Cytc binding sites (Rathore et al, 2019; Vercellino & Sazanov, 2021), allowing electron transfer only by a mobile pool of Cytc. However, bacterial SCs do contain covalently bound Cytc, clearly allowing substrate channeling to transfer electrons from CIII to CIV (Daldal et al, 2003; Gong et al, 2018; Kao et al, 2022).

Efficiency of electron transfer and competitive fitness

While substrate channeling in a strict definition can be excluded as a possibility to explain the occurrence of SC in mitochondria, it is still feasible that a bioenergetic advantage of SC formation can derive from an increased efficiency of electron transfer between the individual complexes (Fig 3C). One approach to test this hypothesis would be to delete factors essential for the formation of SCs and then determine electron transfer efficiencies. This has been performed for SCAF1, which, as stated above, does not affect the formation of respirasomes, but the assembly of CIII₂ + CIV SCs. The ablation of SCAF1 did not affect bioenergetics, but rather provoked changes in nutrient sensing (Pérez‐Pérez et al, 2016; Lobo‐Jarne & Ugalde, 2018). In contrast, bioenergetic differences were observed in a zebrafish model, where absence of SCAF1 caused abnormal fat deposition and reduction in fertility (García‐Poyatos et al, 2020). As further discussed below, controversial results were reported in mouse models, showing either measurable or no effects on energy metabolism upon deletion of SCAF1 (Lapuente‐Brun et al, 2013; Mourier et al, 2014; Shiba et al, 2017).

Moreover, in many systems, including mammalian mitochondria, the respiratory chain complexes assemble in a variety of supercomplexes with varying degrees of stoichiometries, thereby complicating analyses of the bioenergetic efficiency of the different assemblies. Here, yeast offers a unique advantage by lacking CI and having CIII and CIV organized primarily in common supercomplexes. We recently used information from the cryo‐EM structure of S. cerevisiae CIII₂ + CIV_1‐2 (Rathore et al, 2019) to selectively disrupt SCs without affecting individual complexes. We genetically engineered mutants of Cor1, expressed under their endogenous promoter at physiological levels (Berndtsson et al, 2020). These point mutations caused the disruption of CIII₂ + CIV_1‐2 SCs as the only SC species present in S. cerevisiae, without affecting individual complexes. This selective disruption impaired electron transport by impacting the diffusion efficiency of Cytc between the individual complexes (Berndtsson et al, 2020). Increased concentrations of Cytc could correct for this defect. Interestingly, this decreased efficiency of electron transfer had no severe consequences on cellular physiology, including unaltered oxidative stress levels, cellular growth, and viability in isolated cultures. However, the disruption of SCs caused massively reduced competitive cellular fitness, which could be restored by the overexpression of Cytc. These results thus demonstrate that the formation of SCs in living aerobic cells increases the efficiency of cellular energy conversion, which promotes an evolutionary advantage that can be selected for (Berndtsson et al, 2020) (Fig 3C).

Results on increased electron transfer efficiency established by this work align well with recent mathematical modeling of this process (Stuchebrukhov et al, 2020). Interestingly, further work performed with yeast mitochondria in combination with cryo‐EM suggests that this kinetic advantage might be conferred by 2D diffusion of Cytc between CIII₂ and CIV in SCs rather than 3D diffusion between separated, individual complexes (Moe et al, 2021) (Fig 3B). In line with a previous proposal (Pérez‐Mejías et al, 2019), it was suggested that CIII₂ + CIV SC formation provides a negatively charged surface patch on top of this SC, on which the positively charged Cytc can “roll”. These structural properties can also be found in the mammalian version of this complex (Vercellino & Sazanov, 2021). Hence, this mechanisms might present an alternative to substrate channeling, for which diffusion of Cytc within the SC is favored by electrostatic interactions, but still allows a free exchange with the universal soluble Cytc pool (Moe et al, 2021; Vercellino & Sazanov, 2021, 2022; Fig 3B). However, the significance of an electrostatic interaction for 2D diffusion restricted to SCs depends on the ionic strength of the local environment and the lipid composition of the IMM. Currently, these aspects have not been firmly determined, but are essential to establish the relevance of such a 2D diffusion. Moreover, whether physiological or environmental factors could modulate the ratio of Cytc 2D versus 3D diffusion has not been investigated. In addition, limiting the 3D diffusion of Cytc impairs mitochondrial function and cellular viability (Toth et al, 2020). Of note, these results were obtained by reducing the mobility of the whole Cytc pool and not only a potential subpopulation interacting with SCs. In sum, current studies demonstrate that the formation of CIII₂ + CIV SCs increases the efficiency of electron transport via Cytc. However, the exact modalities and biophysical properties establishing this advantage remain to be investigated.

Mitochondrial respiratory complex stability and assembly

The structural interdependence between MRC complexes was first observed in cells from patients with mitochondrial diseases, where mutations affecting CIII or CIV frequently lead to a combined CI deficiency (Lamantea et al, 2002; Fernandez‐Vizarra et al, 2007), suggesting that SCs may play roles in CI stability. The observation that, in a CIII mutant, CI levels could be restored by treatment with antioxidants, while SC‐associated CI is not affected by high ROS levels (Diaz et al, 2012), suggests that only free CI is highly sensitive to ROS‐induced damage and degradation.

Another hypothesis on the physiological significance of SCs suggested that these large assemblies are scaffolds for the late stages of CI biogenesis involving the incorporation of the NADH module to the CI peripheral arm (Moreno‐Lastres et al, 2012). However, this hypothesis has been challenged by kinetic studies of CI assembly, where a small amount of fully assembled CI could be detected before the formation of SCs (Guerrero‐Castillo et al, 2017). More recently, it was shown that the lack of CIII prevents the incorporation of CI NADH module from completing the functional enzyme (Protasoni et al, 2020). Moreover, the association of the late assembly CI precursor with CIII and/or CIV was detected in mutant cell lines and animal models (Alston et al, 2018; Adjobo‐Hermans et al, 2020; Fang et al, 2021). Additionally, data supporting the interaction of CIII and CIV subunits and assembly modules directly in the SCs have been reported (Lazarou et al, 2009; Lobo‐Jarne et al, 2020; Timón‐Gómez et al, 2020). Taken together, increasing evidence supports the existence of distinct but interconnected assembly pathways for free complexes and SCs, a model referred to as the cooperative assembly model (Fernández‐Vizarra & Ugalde, 2022).

Mitochondrial respiratory chain organization and metabolic adaptation

The plasticity model of mammalian MRC organization proposes that individual complexes assemble into SCs in a dynamic manner to adapt to changing cellular metabolic needs (Acín‐Pérez et al, 2008). However, despite the variety of MRC complexes distribution in free form and SCs observed in different cell types and tissues (Lapuente‐Brun et al, 2013), the data reflect steady‐state levels and the dynamic nature of the MRC organization has not been experimentally demonstrated. Alternatively, adaptation to the cellular energetic requirements could be provided by the existence of interconnected assembly pathways, where incorporation of late assembly catalytic modules into individual complex or SC pre‐assembled intermediates could modulate MRC organization (Fernández‐Vizarra & Ugalde, 2022). Although the mechanisms behind these models are yet to be fully disclosed, both scenarios suggest the involvement of SC assembly/disassembly factors acting as sensors or assembly checkpoints. As mentioned above, the only dedicated SC assembly factor identified to date is SCAF1 (Cogliati et al, 2016). There is a general agreement that in all cell lines and model organisms, the absence of or mutations in SCAF1 lead to the lack of SC CIII₂ + CIV formation, while allowing the assembly of respirasomes (Cogliati et al, 2016; Pérez‐Pérez et al, 2016; Lobo‐Jarne & Ugalde, 2018). It is also established that the stability of respirasomes and, particularly, larger SC structures is compromised to different extents depending on the cell type and tissue (Cogliati et al, 2016; Pérez‐Pérez et al, 2016; Williams et al, 2016; Lobo‐Jarne & Ugalde, 2018). On the contrary, the question of whether the changes in SC organization promoted by the absence of SCAF1 ultimately lead to a metabolic adaptation remains to be fully answered. Some studies concluded that SCAF1‐dependent MRC remodeling is not essential, neither to maintain mitochondrial bioenergetics nor to cope with acute cellular stresses (hypoxia or oxidative stress) in HEK293T cells (Pérez‐Pérez et al, 2016; Lobo‐Jarne & Ugalde, 2018). Other studies showed that in U2OS cells, ER stress and glucose deprivation stimulate mitochondrial bioenergetics and SC formation through protein kinase R‐like ER kinase (PERK) by enhancing SCAF1 levels through the PERK‐eIF2α‐ATF4 axis (Balsa et al, 2019). Likewise, increased SC levels were associated with mitochondrial efficiency and growth in severely hypoxic pancreatic cancer (Hollinshead et al, 2020). An inconsistent picture was also obtained from the study of animal models. These studies have shown that SCAF1 promotes SC formation as well as metabolic efficiency in zebrafish (García‐Poyatos et al, 2020) and in some mouse models (Lapuente‐Brun et al, 2013; Shiba et al, 2017) but not others (Mourier et al, 2014). One of these studies showed that SCAF1 KO mice exhibit lower blood glucose levels after insulin or pyruvate injection, which suggested that the protein could be involved in the regulation of glucose homeostasis (Shiba et al, 2017). However, whether the physiological manifestations correlated with or were the cause of altered SC formation was not fully defined. Work in zebrafish models showed that SCAF1‐deficient animals were smaller in size and had abnormal fat deposition, phenotypes that were rescued by doubling the food supply (García‐Poyatos et al, 2020). This phenotype rescue correlated with improved bioenergetics; however, possible SC‐independent roles of SCAF1 in fat metabolism in fish were not fully discarded.

Two more recent studies have added support to the concept of MRC organization‐dependent metabolic remodeling. One study showed that two separate MRC organizations co‐exist in human cells and post‐mitotic tissues, which the authors called C‐MRC and S‐MRC. These are defined by the preferential expression of the CIV COX7A isoforms, COX7A2 and SCAF1 (Fernández‐Vizarra et al, 2022), respectively. The SCAF1‐dependent S‐MRC organization is characterized by the presence of the SCAF1‐containing respirasome, which accounts for approximately 50% of total CIII and CIV levels. At the same time, the remaining CIII and CIV are equally distributed between the CIII₂ + CIV SC and free complexes. On the contrary, the COX7A2‐dependent C‐MRC organization displays a relatively low amount of the COX7A2‐containing respirasome, no CIII₂ + CIV SC, and abundant free CIII (~ 60% of total CIII) and CIV (~ 80% of total CIV). The exclusive presence of one configuration or the other in knock‐out cells of the corresponding isoform led to some changes in mitochondrial bioenergetics. However, no differences in respiratory parameters were observed between SCAF1 or COX7A2 knock‐out and wild‐type cells, where the two MRC organizations co‐exist. The increment in respiratory rate supported by the addition of succinate (a CII substrate) to permeabilized cells already respiring CI substrates was shown to be lower in cells expressing the S‐MRC compared to C‐MRC, suggesting that the electrons from CI saturate the downstream MRC segment (Q to CIV) at a greater extent in the S configuration. Preferential utilization of the NADH versus FADH₂ route parallels with the fraction of total CIII and CIV distributed in respirasomes. This suggests the existence of a kinetic advantage for electron transfer within SCs possibly linked to the proximity of electron carrier binding sites. However, NADH‐ and FADH₂‐linked respiration are not fully additive in either configuration, arguing against Q pool partitioning. Moreover, the lack of COX7A2 causes a decrease in total CIV amount and enzymatic activity by 50%, which, although it does not affect endogenous and coupled respiration, reduces spare respiratory capacity. This relevant confounding variable can contribute to the observed decrease in CII minus CI‐ and ATP‐driven respiration in S‐MRC cells. Lastly, the authors showed that stimulation of pyruvate dehydrogenase (PDH) activity to promote a switch toward oxidative metabolism in human fibroblasts reduces SCAF1 steady‐state levels and favors the C‐MRC configuration. This observation establishes a direct link between the MRC organization and cellular metabolism, controlled by PDH activity. Intriguingly, while the study proposes that a metabolic shift toward oxidative metabolism destabilizes SCAF1, data from human and mouse tissues indicate that SCAF1 expression is induced by PGC1α and enhances respiratory capacity (Benegiamo et al, 2022).

A study in humans has shown significant genetic variation in SCAF1 in several tissues, most prominently in skeletal muscle (Benegiamo et al, 2022). The most significant variant was an insertion in the 3′ untranslated region of the gene that creates a short‐repeated sequence and enhances SCAF1 expression. Human myotubes carrying this genetic variant have enhanced SC levels and respiratory capacity. One would expect that the S‐MRC would accumulate in these myotubes, but this was not determined. Nevertheless, an important finding was that this variant is associated with lower body fat and improved cardiorespiratory fitness in humans (Benegiamo et al, 2022). This observation was reproduced in a mouse model, where it was also shown that SCAF1 and SCAF1‐containing SCs are induced upon exercise, specifically in the skeletal muscle, in a manner independent of PGC1α (Benegiamo et al, 2022). Taken together, these studies support the conclusion that SCAF1 and SC levels correlate with metabolic fitness (Fig 3C), although the cellular context may modulate the nuances of this correlation.

Other proposed functions of respiratory chain supercomplexes

Most studies on SCs mention that the limitation of oxidative stress could be a function of these assemblies. However, only very limited evidence for such a role was provided. CI and CIII are the main production sites of reactive oxygen species in mitochondria (Murphy, 2009), and hence, it was postulated that the formation of the respirasome somehow alleviates oxidative stress from these sites. Indeed, an in vitro study suggests that the dissociation of CI + CIII₂ SCs increases oxidative stress (Maranzana et al, 2013). In line with these results, it was observed that the higher abundance of free CI in astrocytes compared to neurons correlates with higher oxidative stress found in astrocytes (Lopez‐Fabuel et al, 2016). Unfortunately, a molecular mechanism for such a potential reduction of reactive oxygen species by SC formation has not been demonstrated, and hence, this hypothesis is only based on correlations. A causative role of SC formation to limit oxidative stress remains to be determined.

Finally, it was also proposed that SCs prevent the aggregation of proteins in the IMM (Blaza et al, 2014). Although this is feasible, as the IMM presents one of the most protein‐dense regions of a cell, direct experimental evidence for this hypothesis is missing.

Summary

Since the discovery of SCs more than 20 years ago (Schägger & Pfeiffer, 2000), significant advances in our understanding of these gigantic assemblies have been made. In particular, the latest work on the structural characterization of SCs presented a giant leap in the field (Gu et al, 2016; Letts et al, 2016; Hartley et al, 2019; Rathore et al, 2019; Vercellino & Sazanov, 2021; Zhou et al, 2022; Klusch et al, 2023; Maldonado et al, 2023). Still, the physiological functions of SCs remain a matter of debate (Milenkovic et al, 2017). Recent reports have indicated that SC formation primarily enhances the efficiency of electron transport. The emerging picture favors a model in which SC formation, as such, does not modify the maximal activity of individual complexes (Mourier et al, 2014; Berndtsson et al, 2020), but optimizes their cooperation in the context of the overall pathway.

Supercomplex formation constitutes another example for a growing theme of gathering functionally related activities in common complexes or into proximal localization. This strategy, which is widely employed in life, is used extensively in mitochondria (Linden et al, 2020; Schulte et al, 2023) as exemplified by a supercomplex collecting enzymes mediating Q biosynthesis (Marbois et al, 2009), the establishment of membrane subdomains where early steps of MRC biogenesis are organized (Singh et al, 2020) and mitochondrial RNA granules (Antonicka et al, 2013; Jourdain et al, 2013), executing early steps of mtRNA maturation. However, while all these organizations might not be essential for these pathways to occur, they likely improve the overall biochemical efficiency, possibly only with a small, but nevertheless selectable, margin. While these small beneficial margins may be neglectable or even sometimes undetectable in experimental setups and under laboratory conditions, they have impacted selection during evolution and seem to affect performance in mammalian tissues. Particularly in the context of evolutionary pressure via competition in natural environments and in stress scenarios (e.g., high temperature, hypoxic conditions, limited nutrients), these subtle bioenergetic advantages of SC formation can be the basis for selection. Employing the recent structural insights into respiratory SCs should allow engineering of further sets of genetic mutants to fully unravel the intricacies and physiological impact of SC formation.

In need of answers.

• To fully unravel the functional interplay of respiratory chain complexes in SCs and its mechanistic benefit, it might be necessary to reconstitute purified SCs into proteoliposomes to study their performance in a minimal system, free from confounding secondary variables.

• The functional connectivity of the respiratory chain by mobile electron carriers is likely a key aspect of SC/respirasome formation. It would be important to determine the absolute and relative concentrations of Cytc or Q, and RCs in order to test how they might determine the dependency on SC/respirasome formation to achieve optimal electron transport rates.

• While this manuscript was under editorial proof, an article was published where a more than 50% reduction in SC and respirasome formation was achieved in a mouse model, without overly perturbing respiration or physiology (Milenkovic et al, 2023). Hence, it would be desirable for future studies to generate mutants in higher organisms, where the formation of SCs and respirasomes is completely abolished.

• Once these mutants are available, it would be very interesting to score how they impact stress tolerance and competitive fitness.

• The interdependency of assembly of the different respiratory chain complexes has suggested that mature complexes might facilitate assembly of the partner complex through direct contacts. However, assembly also occurs in the absence of the partner complex. To unravel the contribution for the co‐assembly pathways for overall assembly, efficiency will be an important aspect of future research.

Author contributions

Andreas Kohler: Conceptualization; writing – original draft; writing – review and editing. Antoni Barrientos: Conceptualization; writing – original draft; writing – review and editing. Flavia Fontanesi: Conceptualization; writing – original draft; writing – review and editing. Martin Ott: Conceptualization; writing – original draft; writing – review and editing.

Disclosure and competing interests statement

The authors declare that they have no conflict of interest.

EMBO reports (2023) 24: e57092