Structure and Mechanism of Respiratory III-IV Supercomplexes in Bioenergetic Membranes

Because subunit Cox5 is located at the interface of cyt. bc1 and CytcO in the supercomplex33−35,37 (Figure 2A), it is positioned at the diffusion path of cyt. c. Expression of the two interchangeable isoforms of Cox5, i.e., Cox5A and Cox5B, correlates with the expression of the two cyt. c isoforms, iso-1 and iso-2, respectively; Cox5A and iso-1 cyt. c are expressed under normoxia, while Cox5B and iso-2 cyt. c are expressed under hypoxia.174,273,274 This correlation may be coincidental, but we discuss briefly its possible consequences. The supercomplex structure was essentially the same with either Cox5A or Cox5B,33 and no effects were observed on the supercomplex activity. In addition, the maximum catalytic activity of CytcO and its affinity for both cyt. c isoforms and O2 were unaffected upon replacement of Cox5A by Cox5B.177 However, the supercomplex activity was measured at a cyt. c:supercomplex ratio of >103,34 where the electron-transfer rate saturates at a maximum value, kcat. It is possible that at the much smaller cyt. c:supercomplex ratio of ∼2–3, found in S. cerevisiae mitochondria in vivo (cf. ref (37)), an effect on the intercomplex electron transfer would be observed depending on cyt. c and Cox5 isoforms. In other words, it cannot be excluded that electron transfer between cyt. bc1 and CytcO within the supercomplex is regulated by altering the pairwise expression levels of Cox5 and cyt. c isoforms.

Cytochrome c has also been shown to bind to Rcf1.232,234,275,276 The original suggestion that Rcf1 could be found at the interface of complexes III and IV prompted us to suggest that formation of a putative Rcf1–cyt. c co-complex would play a similar role to that of cyt. cy, i.e., mediate electron transfer via a membrane-associated cyt. c.234 However, this particular consequence of the Rcf1–cyt. c interaction appears less likely in S. cerevisiae in view of the putative binding of Rcf1 to Cox3/Cox13 (see above), and the position of these subunits at the distal edge of the supercomplex, rather than between cyt. bc1 and CytcO (Figure 2A). On the other hand, assuming that Rcf1 would bind at the same position as Rcf2,33 cyt. c binding to an Rcf1–CytcO cocomplex would position the cyt. c near the cyt. c-binding cleft defined by CytcO subunits Cox12 and Cox2. Interaction of cyt. c with Rcf1 at this position would result in increasing the affinity for cyt. c to CytcO to allow electron transfer between complexes III and IV via two transiently bound cyt. cs, as discussed previously.37,264 Similarly, interaction of cyt. c with HIGD1 in mammalian mitochondria has also been observed and discussed.227,235,277

As outlined above, the Rcf proteins appear to support a range of functions in respiration, one of which involves binding of cyt. c. However, additional data is needed to fully understand the functional significance of the cyt. c–Rcf1 interactions at the molecular level.

When considering complexes III and IV, the answer to the question above is rather trivial in the case of the Gram-positive actinobacteria, which do not harbor any water-soluble cyt. c. We therefore focus the discussion on the mitochondrial III2IV1/2 supercomplexes. A discussion of a functional significance of these mitochondrial supercomplexes is complicated by the variability in their composition, the variable distribution of free complexes and supercomplexes in different organisms,54,278 and the differences in relative orientation of the respiratory complexes within the supercomplexes, i.e., the flexibility in the interaction surfaces of the supercomplex components among different species (Figure 7). Nevertheless, it is well established that supercomplexes do form in a wide range of organisms and are likely to have functional significance. As already indicated above, various physiological roles of supercomplexes have been discussed (e.g., refs (23,53−55,279,280)), and below we summarize some specific suggestions with a focus on cyt. bc1–CytcO supercomplexes.

The lack of well-defined structural changes of the respiratory enzymes upon association into supercomplexes, and the differences in the relative orientation of the components in different organisms (Figure 7) suggest that formation of supercomplexes does not result in changes in functionality of individual components. Changes in turnover activity of individual respiratory complexes upon forming supercomplexes have been reported, but they are typically too small to yield any functionally relevant changes in the overall electron flux through the respiratory chain (see refs (53,54)). Furthermore, as outlined above, the electron flux through the respiratory chain in vivo is typically lower than the kcat values of the components. Therefore, formation–dissociation of the mitochondrial supercomplexes is unlikely to comprise a universal mechanism to modulate function through changes of the activity of complexes III or IV themselves.

A similar problem is associated with identifying specific effects of supercomplex formation on the “stability” of the components, which has been suggested in the past, although mainly for complex I (reviewed in refs (53,54)). As pointed out by Milenkovic et al.,53 many of the studies addressing this issue are based on observation of correlations of effects on function, structure and morphology, and it is at present not possible to deduce any specific mechanistic effects at a molecular level.

Blaza et al.281 proposed that formation of supercomplexes is a consequence of the very high protein density of the inner mitochondrial membrane (∼2/3 protein); formation of supercomplexes would outcompete irreversible, unspecific aggregation of respiratory complexes with other membrane components.53,281 However, as also noted by these authors, in mammalian mitochondria only 15–30% of CytcO is part of supercomplexes.54 This equilibrium of free complexes and supercomplexes indicates that association of respiratory complexes to form supercomplexes is realized through relatively weak reversible interactions. Because a reversible equilibrium of supercomplexes and free complexes could not block irreversible formation of aggregates between respiratory complexes and other membrane proteins, we consider this role of supercomplexes to be less likely.

In S. cerevisiae, a larger fraction (∼90%) of the CytcO population is part of supercomplexes.72 An equilibrium constant between supercomplex-bound and free CytcO in the order of 10 suggests that also in S. cerevisiae, the III2IV1/2 supercomplexes are held together by weak interactions. This conclusion is further supported by the necessity to use weak detergents for isolation of supercomplexes (e.g., digitonin or glyco-diosgenin, GDN) and the observation that they dissociate into components upon addition of n-dodecyl-β-d-maltoside (DDM).37 Thus, also in S. cerevisiae the cyt. bc1–CytcO interactions are reversible and could not outcompete irreversible nonspecific aggregation with other membrane-bound proteins.

Another suggestion for the role of supercomplexes originates from an observation of the preference for respiratory complexes for specific membrane topology.282 Fedor and Hirst283 suggested that formation of supercomplexes would ensure an even distribution of the respiratory complexes in the membrane, a plausible proposal that could be tested experimentally in future studies.

Formation of supercomplexes has been suggested to decrease the amount of produced reactive oxygen species (ROS) (e.g., refs (73,284)). Here, we briefly discuss this proposed role in the framework of effects at a molecular level. This discussion requires a definition of the term ROS as it does not describe a single chemical entity, but rather a range of molecules or ions that are formed upon incomplete reduction of O2 (i.e., reduction by <4 electrons), including superoxide, peroxide, and hydroxyl radicals.285 The reactivity of these species differs and therefore the term ROS only depicts a generally reactive molecule or ion. Reduction of O2 by one electron at a time yields first the superoxide anion (O2•–), which is the precursor of other ROS.285,286 The main sites of initial O2•– formation in mitochondria are at complexes I and III.285,286

The amount formed O2•– at a specific redox site at a particular O2 concentration is determined by the relative rates of O2•– formation (“side reaction”) and the rate by which the electron is transferred from that site to the next acceptor in the electron-transfer chain (physiological reaction). When assuming that formation of supercomplexes would yield less ROS, the implicit assumption is that the electron-transfer rate away from the ROS-forming site would be slower for individually diffusing complexes than for supercomplexes.

Data from studies of model systems suggest that the amount of ROS at complex I decreases upon supercomplex formation.284 However, Fedor and Hirst283 recently showed that QH2 produced by complex I in supercomplexes is oxidized to Q more rapidly outside the supercomplex than by the acceptor within the supercomplex (complex III). In other words, electrons from complex I are removed more rapidly in the absence than in the presence of supercomplexes. As a consequence, formation of supercomplexes that involve complex I would not per se result in decreasing the fraction of reduced ROS-forming sites at complex I.

A postulate that formation of supercomplexes composed of cyt. bc1 and CytcO would yield less ROS implies that association of the components would result in a faster reoxidation of cyt. bc1 because ROS is mainly formed at cyt. bc1. Indeed, as discussed above, reduction–oxidation of cyt. c is the rate-limiting step of electron transfer from QH2 (complex III) to O2 (complex IV) in S. cerevisiae. Therefore, a decrease in this transfer rate upon dissociation of the III2IV1/2 supercomplexes would result in a larger fraction of reduced complex III, which could result in accumulation of electrons at the QP site where nonphysiological reduction of O2 to O2•– is most likely to take place.81 Hence, we consider it possible that O2•– production is indeed lowered upon formation of III2IV1/2 supercomplexes.

In the above discussion, we consider a fully functional respiratory chain. However, in the native membrane, new respiratory complexes are continuously produced, and at a given time there are also partly assembled respiratory complexes with incompletely connected electron-transfer chains. These partly assembled complexes could accumulate electrons at their redox sites, which upon interaction with O2 may form ROS. It is possible that association of these partly assembled complexes with other fully functional partner complexes to form supercomplexes287 would provide a route for dissipation of these reducing equivalents. In so doing, the probability for ROS formation from partly assembled respiratory complexes would be diminished.

As already discussed above, early hypotheses suggesting “substrate channeling”, i.e., direct transfer of confined Q/QH2 or cyt. c between respiratory complexes within a supercomplex, have been rejected.38,53,88,269,281,283 Yet, supercomplexes have been proposed to allow a “more efficient” transport of electrons and an increase in the “efficiency” of respiration allowing higher “yields” of energy conservation (see e.g., refs (25,36,56,76,282,288)). Therefore, a consideration of effects of supercomplex formation on “efficiency” and “yield”, terms frequently used in the discussions, requires a definition of these terms and a more detailed analysis.

The free energy available at each respiratory complex (energy input) is defined by the difference in standard redox potentials of the electron donor and acceptor, the concentration ratio of reduced and oxidized donor, as well as the concentration ratio of reduced and oxidized acceptor. The free energy conserved at each respiratory complex (energy output) is determined by the number of protons transferred across the membrane and the charge separation upon oxidation of the electron donor and reduction of the acceptor. The term efficiency typically depicts the ratio of free energy output and free energy input in a given system. An assumption that association of respiratory–chain complexes into supercomplexes results in an increased efficiency of respiration implies that the efficiency of at least one component would increase. However, as discussed above, changes in the charge-separation stoichiometry of individual complexes are unlikely to occur upon association into supercomplexes and therefore the overall efficiency of the system is not expected to change upon forming supercomplexes.

The terms “yield” and “efficiency” are in principle equivalent but are often used in different context. The former is often used to depict the amount of ATP formed for a given amount of oxidized substrate of the respiratory chain (cf., the so-called P/O ratio). This parameter is also determined by the efficiency of each component, including the ATP synthase and, hence, it is not expected to change upon association of respiratory complexes into supercomplexes.

It is relevant to note that the yield of ATP formation is also dependent on proton leaks across the membrane. Proton leaks often occur at protein–membrane interface surfaces, which become smaller upon association of respiratory complexes into supercomplexes. However, the protein–protein interaction surface upon formation of a supercomplex comprises only a very small fraction of the sum of all protein–membrane interaction surfaces of all membrane proteins of the inner mitochondrial membrane. Therefore, the effect of decreasing the protein–membrane interaction surface upon forming supercomplexes would most likely not result in increasing the yield of ATP production. That said, it is clear that an intricate web of regulatory pathways in mitochondria controls energy conservation in respiratory complexes and the overall P/O ratio, depending on environmental conditions.289 These regulatory pathways may also involve formation and dissociation of supercomplexes. However, changes in the energy-conversion efficiency or yield cannot simply be a direct consequence of changing the distance between respiratory complexes to form supercomplexes.

If “more efficient” incorrectly alludes to an increase in the electron-transfer rate between respiratory complexes, the suggestion that supercomplex formation would result in “more efficient” electron transfer is plausible, at least when considering association of complexes III and IV (see above).

We consider electron transfer between complexes III and IV via cyt. c diffusion and discuss two scenarios: (i) freely diffusing complexes III and IV where after reduction at cyt. bc1, cyt. c equilibrates with the cyt. c pool in the intermembrane space and electrons are transferred to CytcO from this cyt. c pool (Figure 10A); (ii) electron transfer from cyt. bc1 to CytcO by 2D diffusion along the surface of a CIII2CIV1/2 supercomplex (Figure 10B). According to scenario (i), the redox state of the cyt. c pool in the intermembrane space is determined by the relative rates of cyt. c reduction at cyt. bc1 and oxidation at CytcO. According to scenario (ii), the redox state of the cyt. c pool is determined by the equilibrium constant of cyt. c bound to the supercomplex surface and free cyt. c in the bulk solution, i.e., the probability that a surface-associated cyt. c in the reduced state is replaced by a bulk oxidized cyt. c. In addition, cyt. c from the cyt. c pool may transiently interact and exchange electrons with any of the complexes or the bound cyt. c during the 2D electron transfer. Nevertheless, the reduction level of the cyt. c pool is expected to depend on the fractions cyt. bc1 and CytcO that are part of a supercomplex because the nature of the electronic link changes upon supercomplex formation/dissociation. As proposed by Moe et al.,37 the scenario suggests yet another possible functional role of supercomplex formation, i.e., to alter the reduced:oxidized ratio of cyt. c. Because cyt. c is involved in an intricate web of cellular interactions,290,291 there may be a link between assembly of cyt. bc1 and CytcO into supercomplexes, changes in environmental conditions, and cellular redox-signaling pathways.

Yet another possibility is that formation of supramolecular assemblies is not directly linked to functional properties of the respiratory chain. Cytochrome c is a positively charged dipolar molecule, which resides in an environment containing negatively charged proteins.247 Association of cyt. c with the supercomplex surface by electrostatic interactions may be necessary to outcompete nonspecific reversible binding to other negatively charged proteins and membrane surfaces in the intermembrane space. Formation of supercomplexes that allow electron transfer by 2D diffusion along the supercomplex surface could thus be a consequence of the electrostatic binding of cyt. c to cyt. bc1 and CytcO.

The discussion above leaves us with a question: why do mitochondria use a soluble, diffusible cyt. c rather than a membrane-anchored counterpart? In this context, it is interesting to recapitulate that R. prowazekii, the closest known microbe relative of mitochondria,257,259 harbors only a membrane-anchored cyt. cy homologue.257,259 We speculate that if the role of cyt. c is only to shuttle electrons between cyt. bc1 and CytcO, then at a minimal cyt. c concentration, the highest possible electron-transfer rate is maintained by a membrane-anchored cyt. c. However, evolution has given also other, regulatory functions to cyt. c, such as, e.g., being a messenger in apoptosis,203,291 which is linked to the redox properties of this electron carrier and may require a water-soluble, diffusible variant. A “best of both worlds” scenario, e.g., in S. cerevisiae, would therefore be to keep the same electron-transfer mechanism as that in R. prowazekii by association of cyt. c with a cyt. bc1–CytcO supercomplex surface, but to use a water-soluble cyt. c that can also sustain other mitochondrial functions.

Respiratory supercomplexes are found in a wide range of organisms. Structures of the bacterial and mitochondrial III2IV1/2 supercomplexes show a great variability in their overall composition and relative orientations of the components, which suggests that the only common structural characteristics of the supramolecular assemblies is proximity of the components. Cryo-EM structures of the III2IV1/2 supercomplexes show that the components are connected via a small number of protein–protein interactions as well as interfacial cardiolipin, and the structures of cyt. bc1 and CytcO remain essentially unaltered upon association. Collectively, the data suggest that the functional role of the supramolecular assemblies is to minimize the distance between the components. We suggest that this organization supports a mechanism that allows electron transfer by 2D diffusion of cyt. c across the merged negatively charged surface of the supercomplex.37 The consequence of electron transfer by 2D diffusion upon forming a supercomplex is a change in the fraction of reduced/oxidized cyt. c in the intermembrane space, which may be sensed by multiple regulatory pathways of the cell. Alternatively, the 2D diffusion mechanism may be a consequence of tight binding of cyt. c to cyt. bc1 and CytcO in order to outcompete nonspecific interactions between cyt. c and negatively charged proteins and membrane surfaces in the intermembrane space. In actinobacteria, electron transfer from complex III to complex IV is conducted via the diheme cyt. cc domain of subunit QcrC. In these supercomplexes, there is an additional effect from the intricate intertwining and shared structural domains, which suggests that the supercomplex functions as a single unit.

This unit also comprises novel key structural features such as an FeS domain that is locked at a fixed position in complex III and a complex III “lid” that shapes a novel proton pathway orifice in complex IV. Future studies will hopefully reveal the functional significance of these novel structural features and offer further general insights into the functional significance of respiratory supercomplexes at a molecular level.