Skip to main content
PMC home page
eLife logoLink to eLife
. 2016 Nov 10;5:e21290. doi: 10.7554/eLife.21290

Functional asymmetry and electron flow in the bovine respirasome

Joana S Sousa 1, Deryck J Mills 1, Janet Vonck 1, Werner Kühlbrandt 1,*
Editor: Stephen C Harrison2
PMCID: PMC5117854  PMID: 27830641

Abstract

Respirasomes are macromolecular assemblies of the respiratory chain complexes I, III and IV in the inner mitochondrial membrane. We determined the structure of supercomplex I1III2IV1 from bovine heart mitochondria by cryo-EM at 9 Å resolution. Most protein-protein contacts between complex I, III and IV in the membrane are mediated by supernumerary subunits. Of the two Rieske iron-sulfur cluster domains in the complex III dimer, one is resolved, indicating that this domain is immobile and unable to transfer electrons. The central position of the active complex III monomer between complex I and IV in the respirasome is optimal for accepting reduced quinone from complex I over a short diffusion distance of 11 nm, and delivering reduced cytochrome c to complex IV. The functional asymmetry of complex III provides strong evidence for directed electron flow from complex I to complex IV through the active complex III monomer in the mammalian supercomplex.

DOI: http://dx.doi.org/10.7554/eLife.21290.001

Research Organism: Other

Introduction

Mitochondria are intricate membrane organelles found in virtually all eukaryotic cells, where they serve a number of essential physiological functions. Their central role is to provide energy to the cell in the form of ATP by oxidative phosphorylation. The mitochondrial respiratory chain consists of four complexes (I–IV), which transfer electrons from NADH and succinate to molecular oxygen. A part of the energy gained in electron transfer is used to pump protons across the inner mitochondrial membrane. The resulting proton gradient is utilized by the mitochondrial ATP synthase to generate ATP. The respiratory chain complexes reside mostly, if not entirely, in the mitochondrial cristae (Kühlbrandt, 2015), which are deep invaginations of the inner membrane into the matrix.

NADH:ubiquinone oxidoreductase, also known as complex I, is the largest assembly in the electron transfer chain. Mammalian complex I comprises 44 different subunits, including two copies of subunit SDAP, and therefore consists of a total of 45 subunits (Vinothkumar et al., 2014). The 14 core subunits are conserved from prokaryotes to mammals (Walker, 1992). The characteristic L-shape of complex I arises from the association of three different units. The dehydrogenase and hydrogenase-like units constitute the matrix arm and are responsible for the transfer of electrons from NADH to ubiquinone (Sazanov and Hinchliffe, 2006). As a third unit, the membrane-embedded transporter assembly pumps four protons from the matrix to the cristae lumen per catalytic cycle (Galkin and Terenetskaya, 1999; Leif et al., 1995). Succinate:ubiquinone oxidoreductase, or complex II, is the only complex in the electron transfer chain that does not translocate protons, but merely feeds electrons into the process. The cytochrome bc1 complex (complex III) is a symmetrical dimer, with cytochrome b, the Rieske iron-sulfur protein and cytochrome c1 as core subunits (Yang and Trumpower, 1986). The Rieske subunit extends across both monomers, stabilizing the dimer that is essential for function. In mammals, complex III contains a total of 11 subunits per monomer, of which eight are supernumerary (Schägger et al., 1986). Complex III transfers electrons from ubiquinone to cytochrome c, a small soluble electron carrier protein in the cristae lumen. Finally, cytochrome c oxidase, also known as complex IV, transfers electrons from cytochrome c and catalyzes the reduction of molecular oxygen to water. Mammalian complex IV has three core subunits (COX1, COX2 and COX3) and 14 subunits in total (Balsa et al., 2012; Kadenbach et al., 1983).

The structural organization of the complexes that carry out oxidative phosphorylation in the inner mitochondrial membrane has been subject to numerous investigations. For many years it was assumed that the respiratory chain complexes exist as separate units in the fluid lipid bilayer of the inner membrane and interact by random collision (Hackenbrock et al., 1986). The discovery of supercomplexes by blue-native polyacrylamide gel electrophoresis (BN-PAGE) after solubilization with mild detergents (Schägger and Pfeiffer, 2000) gave rise to the plasticity model, which postulates that respiratory complexes can exist both free in the membrane and as larger supramolecular entities (Acin-Perez et al., 2008; D'Aurelio et al., 2006). Since then, several stoichiometric supercomplexes have been identified, amongst which the respirasome (supercomplex I1III2IV1) is the most prominent. The respirasome contains all components required to transfer electrons from NADH to molecular oxygen (Schägger and Pfeiffer, 2000).

The possible functional and structural roles of supercomplexes have been hotly debated. At the functional level, advantages due to partitioning of the quinol pool and substrate channeling have been postulated and are supported by several independent studies (Bianchi et al., 2003; Lapuente-Brun et al., 2013). Other results are inconsistent with a partitioning of the quinol pool (Blaza et al., 2014). A reduction in the level of reactive oxygen species (ROS) that are generated as side products of electron transfer reactions in the respiratory chain has also been suggested as a possible role (Maranzana et al., 2013; Panov et al., 2006; Seelert et al., 2009). From the point of view of protein structure, supercomplexes have been proposed to confer stability to complex I or assist in its assembly (Marques et al., 2007; Schägger et al., 2004). In line with this, a model for the generation of supercomplexes was proposed, where the assembly of catalytic subunits of the complex I NADH:dehydrogenase module occurs at a late stage to activate the supercomplexes (Moreno-Lastres et al., 2012). However, recent complexome profiling studies failed to detect supercomplexes containing immature complex I, suggesting that the respirasome forms by association of fully assembled component complexes (Guerrero-Castillo et al., 2016).

Several mitochondrial disorders are associated with impaired respirasome formation. Genetic mutations that impair the assembly of complex III result in a loss of complex I and combined complex III/I defects (Acin-Perez et al., 2004; Bruno et al., 2003; Lamantea et al., 2002; Schägger et al., 2004). Complex IV deficiencies associated with a reduction of complex I levels in mouse and human cells have also been reported (D'Aurelio et al., 2006; Diaz et al., 2006; Vempati et al., 2009). The secondary loss of complex I upon impaired expression of complexes III and IV has been taken to mean that complex I stability depends on physical interaction with other complexes in the respiratory chain, since pharmacological inhibition was not sufficient to reduce complex I levels to the same extent (Acin-Perez et al., 2004; Diaz et al., 2006). Recent studies suggest however that low levels of cytochrome bc1 and cytochrome c oxidase (as well as cytochrome c) result in an accumulation of reduced quinone, which would trigger reverse electron transfer (RET) and generation of superoxide by complex I. This could result in oxidative damage and complex I degradation (Guaras et al., 2016). When ROS production by RET was inhibited, complex I levels were restored (Guaras et al., 2016).

Structures of the isolated respirasome at estimated resolutions of 20–30 Å have been obtained by negative-stain electron microscopy (EM) (Schäfer et al., 2007), single-particle cryo-EM (Althoff et al., 2011) and electron cryo-tomography (Dudkina et al., 2011). However, the map resolution was insufficient to fit the component complexes precisely, or to detect new functionally relevant features. Moreover, the recently published high-resolution cryo-EM structures of bovine complex I (Vinothkumar et al., 2014; Zhu et al., 2016) contribute to a comprehensive view of the respirasome. Here, we report a cryo-EM map of the supercomplex I1III2IV1 from bovine heart mitochondria at 9 Å resolution. We show specific protein-protein contacts between the three respiratory chain complexes within the respirasome. Our structure reveals a functionally asymmetric complex III, in which one monomer preferentially catalyzes the reduction of cytochrome c in a defined supramolecular organization of the electron transport chain.

Two cryo-EM studies of the respirasome from ovine and porcine heart mitochondria have been published since this manuscript was submitted (Gu et al., 2016; Letts et al., 2016b). Both are at significantly higher resolution than our structure, but neither of them observe the functional asymmetry of complex III, which we regard to be the most important functional insight from the respirasome structure. A detailed comparison has been added to the discussion.

Results

Isolation of mitochondrial supercomplexes solubilized with PCC-a-M

Since the discovery of mitochondrial supercomplexes (Schägger and Pfeiffer, 2000), digitonin has been the detergent of choice for their solubilization and characterization. Digitonin is very suitable for the isolation of supercomplexes I1III2IV1 and I1III2 (Figure 1A), and density gradient centrifugation has been the preferred method for their purification. However, to separate the two digitonin-solubilized supercomplexes on a preparative scale by this method proved to be challenging, due to their small difference in density.

Figure 1. Isolation and single-particle cryo-EM of the bovine supercomplex I1III2IV1.

(A) BN-PAGE of bovine heart mitochondria (BHM) solubilized with digitonin (lane 1) or PCC-a-M (lane 2); PCC-a-M-solubilized complex after exchange to A8-35 (lane 3); density gradient fraction of supercomplex I1III2IV1 in A8-35 (lane 4). (B) Electron micrograph of bovine respirasomes in vitrified buffer, recorded with a Falcon III direct detector in integrating mode on a FEI Tecnai Polara electron microscope operating at 300 kV. (C) 2D class averages produced by reference-free 2D classification of 156,536 particles in RELION 1.4.

DOI: http://dx.doi.org/10.7554/eLife.21290.002

Figure 1.

Figure 1—figure supplement 1. Assessment of sample quality.

Figure 1—figure supplement 1.

(A) In-gel activity assay for NADH:dehydrogenase in BN-PAGE gel. Dark bands indicate the presence of active complex I (B) 2D BN/BN-PAGE of mitochondria solubilized with PCC-a-M confirms presence of complexes I, III and IV in the top band of the first dimension BN-PAGE lane. (C) 3D classification and reconstruction of 4000 and 42,000 negatively stained supercomplex particles after initial solubilization with digitonin or PCC-a-M. The purified digitonin-solublized sample contains a mixture of supercomplexes I1III2IV1 and I1III2, while solubilization with PCC-a-M produces pure I1III2IV1.
DOI: http://dx.doi.org/10.7554/eLife.21290.003
Figure 1—figure supplement 2. 3D classification and refinement.

Figure 1—figure supplement 2.

3D classification with RELION 1.4 reveals a range of different populations, including free complex I (salmon), I1III2IV1 lacking subunits of complex I matrix arm (green and gold) and I1III2 (violet and gold). Maps with similar features are shown in the same colors.
DOI: http://dx.doi.org/10.7554/eLife.21290.004
Open in a new tab

PCC-a-M (trans-4-(trans-4’-propylcyclohexyl)cyclo-hexyl-α-D-maltoside) is a recently developed mild, non-ionic detergent (Hovers et al., 2011). In an initial screen, we identified a narrow concentration range in which PCC-a-M solubilizes mitochondria efficiently, while preserving the interactions between complexes I, III and IV (Figure 1A). Inspection by 2D BN/BN-PAGE and in-gel activity assays (Figure 1—figure supplement 1A,B) indicated that this procedure yielded a sample that was active and highly enriched in supercomplex I1III2IV1. The smaller supercomplex I1III2 was not detected on the gel.

For further purification, the supercomplex was transferred from PCC-a-M to amphipol A8-35, in which density gradient centrifugation has been shown to work well (Althoff et al., 2011). Gradient centrifugation in amphipol A8-35 produced essentially pure supercomplex I1III2IV1 (Figure 1A). In particular, supercomplex I1III2 was not detectable. Image analysis of negatively stained samples confirmed that supercomplex I1III2IV1 purified from membranes solubilized with PCC-a-M was pure, whereas with digitonin, 42% of the particles had lost complex IV (Figure 1—figure supplement 1C).

Structure determination by single-particle cryo-EM

Cryo-EM grids of supercomplex I1III2IV1 in amphipol A8-35 were prepared immediately after purification and electron micrographs were recorded (Figure 1B). A total of 156,519 particles was picked manually and classified in RELION 1.4 (Figure 1C). Class averages with recognizable views of the supercomplex were selected, and the resulting 137,606 particles were refined, using an earlier low-resolution cryo-EM map (Althoff et al., 2011) low-pass filtered to 60 Å as initial reference. The resulting overall average map had a nominal resolution of 8.1 Å (Figure 1—figure supplement 2). The map definition varied considerably across the structure, suggesting that it might be an average of several different assemblies or conformations. Multiple rounds of 3D classification indeed revealed a high degree of compositional and conformational heterogeneity. The most homogeneous class of 17,094 particles (class 1; 11% of the total) yielded a structure of the respirasome at a final resolution of 9.1 Å. Notwithstanding its slightly lower nominal resolution as a consequence of the smaller number of particles, the features of this map are very much clearer than in the global average. A second structure of supercomplex I1III2IV1 with distinctly different features was resolved to 10.4 Å (class 2; 6% of the total). Lastly, a third class with 12,042 particles lacking complex IV (class 3; 8% of the total) was refined to 9.9 Å (Figure 1—figure supplement 2).

Defined protein-protein contacts

The three component complexes I1, III2 and IV1 of the respirasome are well-resolved in class 1 (Figure 2A) and class 2. At an estimated local resolution of 8.6 Å (Figure 2—figure supplement 1), the transmembrane region of class 1 indicates clear densities for most of the 132 membrane-spanning α–helices in the supercomplex (Figure 2B). Atomic models for bovine complex I (Vinothkumar et al., 2014), III (Iwata et al., 1998) and IV (Tsukihara et al., 1996) were docked into the 3D maps. The fit was excellent for all three complexes in all map regions. A good fit of all three complexes was also obtained for class 2, whereas class 3 lacked the density for complex IV and thus corresponded to supercomplex I1III2.

Figure 2. Protein-protein contacts in the respirasome.

Most contacts are mediated by supernumerary subunits of complexes I, III and IV. (A) Side views (upper row) and views from the matrix (bottom row) of class 1 cryo-EM map filtered to 8.6 Å with docked atomic models of complexes I (blue) (Vinothkumar et al., 2014), III (green) (Iwata et al., 1998) and IV (yellow) (Tsukihara et al., 1996). (B) Slices through the map at positions shown in (A) indicate contact points between the three complexes in the membrane. Core subunits are shown in grey and supernumerary subunits in color.

DOI: http://dx.doi.org/10.7554/eLife.21290.005

Figure 2.

Figure 2—figure supplement 1. Local map resolution.

Figure 2—figure supplement 1.

(A) Side (left) and matrix view (right) of class 1 map colored according to local resolution by RESMAP (Kucukelbir et al., 2014). The best-resolved map regions are at the interface between complexes I and III. (B) Soft mask (red) used to determine the resolution of the transmembrane domain in class 1 of supercomplex I1III2IV1. (C) Resolution estimated by gold-standard Fourier Shell Correlation (0.143 threshold). FSCs were calculated from two reconstructions obtained from half datasets of the entire map volume with a tight, soft global mask (blue) or for the masked membrane region shown in (B) (red).
DOI: http://dx.doi.org/10.7554/eLife.21290.006
Open in a new tab

The presence of supercomplex I1III2 and of free complex I (15% of the initial data set; Figure 1—figure supplement 2) in the cryo-EM sample was surprising, since biochemical analysis and negative-stain EM had both shown that the amphipol-solubilized complex was stable and free of these assemblies (Figure 1—figure supplement 1). Evidently, supercomplex I1III2IV1 partly dissociated on the cryo-EM grid prior to freezing, either as the result of an increase in ionic strength due to evaporation, or of surface forces at the air-water interface.

The class 1 map shows that the three component complexes are in close contact with each other at defined points primarily near the membrane surfaces (Figure 2B). With one exception, all inter-complex contacts involve supernumerary subunits. The most extensive interactions occur between complexes I and III. In the matrix, subunit B22 of complex I is in touch with subunit 1 of complex III at a minimum distance of 4 Å. In the membrane region, subunit B14.7 of complex I approaches subunit 8 of cytochrome bc1 near both membrane surfaces to within sidechain contact (Figure 3). In the centre of the supercomplex, subunit B14.7 establishes clear contacts with complex III within the hydrophobic interior of the respirasome.

Figure 3. Central position of complex I supernumerary subunit B14.7.

Figure 3.

Open in a new tab

(A) Horizontal slice through 8.6 Å map, with fitted models for complexes I (light blue), III (light green) and IV (yellow). Subunit B14.7 of complex I is dark blue and subunit 8 of complex III is dark green. (B) Detailed side view of interface shows close contacts between complex I B14.7 and complex III subunit 8 in the hydrophobic interior of the supercomplex. Distances are between α–carbons in polypeptides of adjacent complexes.

DOI: http://dx.doi.org/10.7554/eLife.21290.007

Contacts between complexes I and IV are mediated by subunit ND5 of complex I that approaches complex IV subunit COX7C to within an α–carbon distance of around 10 Å at the membrane surface on the matrix side. ND5 appears to be the only one of the 14 complex I core subunits that is involved directly in supercomplex formation. The interface between complexes III and IV is formed by subunits 1, 9 and 10 of the bc1 complex and COX7A1 of cytochrome c oxidase. Subunit 9 of complex III has previously been shown to be required for stable interaction with cytochrome c oxidase (Zara et al., 2007).

Conformational and compositional variability

A significant portion of the particles clustered in class 2, where the mutual arrangement of complexes III and I is different from that in class 1, whereas the interaction of complex I with complex IV is unchanged (Figure 4A). Classes 1 and 2 are related by a 25° rigid-body rotation of the whole complex III dimer relative to complex I, around an axis roughly perpendicular to the membrane plane. This rotation breaks the protein-protein contacts of complex III with complexes I and IV observed in class 1. The particles that cluster in class 2 therefore appear to represent a different form of the respirasome. There were no significant new interactions between complex III and complexes I and IV in this class, implying that it may be less stable. In class 2, the matrix arm of complex I has moved away from the center of the supercomplex by a rotation of 3–4° around an axis parallel to the membrane (Figure 4B). Structures of active and deactive states of complex I have been recently presented, which differ by a similar movement of the matrix arm (Zhu et al., 2016). Class 2 may thus represent a subpopulation of the supercomplex in which complex I is in the deactive state. Alternatively, class 2 may result from rearrangement or destabilization of the supercomplex during purification. The mutual arrangement of complexes I and III in class 3 is the same as in class 1, confirming that the interaction between these two proteins is sufficiently stable to fix them in this specific conformation (Figure 4A and B).

Figure 4. Conformational and compositional heterogeneity of the respirasome.

Figure 4.

Open in a new tab

(A) Top view of density maps of class 1 (dark cyan) and class 2 (pink) of the respirasome and supercomplex I1III2 (violet). Classes 1 and 2 differ by a 25° rigid-body rotation of complex III in the membrane plane relative to complex I. Complex IV occupies identical positions in both maps. The arrangement of complex I and III in class 3 is similar to that in class 1. (B) Side views of classes 2 and 3 overlayed with the class 1 map (dark cyan mesh) indicate flexibility of the complex I matrix arm. In class 2 the matrix arm is displaced by 3–4° away from the center of the map. (C) Consensus map of the respirasome at 8.1 Å with weak density at the distal part of the matrix arm. (D) Complex I matrix arm in the consensus map with fitted atomic model of complex I with the 51 kDa subunit in dark blue seen from the front (left) and side (right).

DOI: http://dx.doi.org/10.7554/eLife.21290.008

Several of the classes obtained (~30% of the initial dataset) have no density for the NADH:dehydrogenase module of complex I, or even lack the whole matrix arm (Figure 1—figure supplement 2). While these classes might reflect a partial loss of this module during purification or cryo-EM grid preparation, we cannot rule out the possibility that they represent biologically relevant assembly, disassembly or recycling intermediates of the respirasome. A recent study of respirasome biogenesis (Moreno-Lastres et al., 2012) has suggested that supercomplexes form by association of individual subunits of the respiratory complexes or their intermediate assemblies, rather than from the fully assembled respiratory chain complexes. The incorporation of the NADH:dehydrogenase module of complex I has been presented as the final step in the assembly process (Moreno-Lastres et al., 2012). The study reported a significant delay in the integration of the 51 kDa subunit in this module, which is one of the core subunits of the N-catalytic unit of complex I. These observations are in agreement with our 8.1 Å global average (Figure 4C,D), in which the density for the 51 kDa subunit is weak and could explain why the NADH:dehydrogenase module is absent in several of the 3D class averages (Figure 1—figure supplement 2). However, other studies suggest that supercomplexes form by association of fully assembled component complexes (Guerrero-Castillo et al., 2016), which would thus not account for the intermediate structures we observe.

The presence of supercomplex assemblies lacking N module subunits is equally consistent with different turnover rates of complex I subunits. It has been shown that several of the nuclear-encoded subunits in the mature enzyme can be exchanged against newly imported copies, in parallel to the de novo assembly of the complex (Dieteren et al., 2012; Lazarou et al., 2007). This holds true for most subunits of the N-catalytic unit (Dieteren et al., 2012; Lazarou et al., 2007), which are particularly susceptible to oxidative damage. A direct replacement of these subunits would provide an efficient means of repairing oxidative damage of complex I (Dieteren et al., 2012; Lazarou et al., 2007). These classes therefore might represent intermediates of the supercomplex, generated during its assembly or for complex I regeneration.

Functional asymmetry and electron flow in complex III

The most striking and unexpected feature of the respirasome map is that one of the two membrane-extrinsic iron-sulfur domains of the complex III dimer shows clear density (Figure 5A). The iron-sulfur domain is the active part of the Rieske protein subunit of the bc1 complex that carries electrons from the QP site via its iron-sulfur cluster to heme c1, where they are picked up by the small, soluble cytochrome c protein for transfer to complex IV (Kühlbrandt, 2015). Interestingly, the iron-sulfur domain that bridges the gap in the electron transfer path between the hemes of one complex III monomer belongs to the Rieske protein of the opposite monomer. In order to transfer electrons to c1, the iron-sulfur domain has to move on a hinge by 10–15 Å (Iwata et al., 1998). Such a movement would make the iron-sulfur domain invisible in a 9 Å map. The hinge movement of the iron-sulfur domain is essential for function since when it is blocked, no electrons are transferred to cytochrome c1 (Darrouzet et al., 2000). The fact that one of the two iron-sulfur domains in the supercomplex is resolved means that one of the two complex III monomers is inactive. The inactive monomer is the one in contact with the well-defined iron-sulfur domain. This monomer is exposed to the lipid bilayer on most of its periphery, except for the hydrophobic surface area that is involved in complex III dimer formation. The active monomer near the disordered iron-sulfur domain is surrounded by complex I, complex IV and the inactive complex III monomer (Figure 5A).

Figure 5. Substrate and electron flow in the respirasome.

Figure 5.

Open in a new tab

(A) Respirasome map with fitted atomic models (blue, complex I; green, complex III; yellow, complex IV) seen from the crista lumen. Only one of the two Rieske iron-sulfur domains of complex III is resolved (red), indicating that the complex III monomer associated with it is inactive. The active complex III monomer is light green, the inactive monomer is dark green. The red circles indicate the position of the ubiquinol binding sites of complex III in the intermembrane side of the membrane. The dashed black circle indicates the position of the quinol binding site of complex I on the matrix side. (B) Oblique view from the lumenal side of the surface-rendered respirasome map, with routes of electron and substrate transfer. Red and blue straight arrows indicate substrate and electron transfer, respectively. Respiratory chain reactions are indicated by curved arrows. Electrons from NADH pass through complex I and reduce quinone to quinol in the quinol binding site (dashed black circle). Reduced quinol diffuses in the membrane, binding preferentially to the QP site of the central, active monomer of complex III. Quinol is oxidized to quinone, transferring one electron to heme bL for quinone reduction at the QN site, and another electron to the iron-sulfur cluster of the Rieske protein in the b position. The iron-sulfur domain moves from the b to the c1 position and reduces heme c1. Heme c1 reduces cytochrome c, which diffuses to complex IV where it donates electrons for reduction of O2 to H2O.

DOI: http://dx.doi.org/10.7554/eLife.21290.009

It is generally assumed that both monomers of the complex III dimer participate equally in mitochondrial electron transport. However, for the bacterial cytochrome bc1 complex, which is a simpler version of mitochondrial complex III and likewise dimeric, it has been shown that there is no difference in the electron transport activity of the bc1 complex if one of its monomers is permanently inactivated by mutagenesis (Castellani et al., 2010). Therefore, in bacteria only one half of the bc1 dimer is required for electron transfer, and this most likely holds true for the mitochondrial complex as well. In the bacterial complex, the choice of the active monomer is thought to be random (Covian and Trumpower, 2008), whereas in the respirasome, the active monomer is clearly the one in the centre of the supercomplex. At present we do not know what constrains the movement of one of the two iron sulfur domains in the supercomplex. Either electrostatic or steric effects in the asymmetrical environment of the respirasome may be responsible. In an unconstrained complex III dimer, both iron-sulfur domains should be mobile, and hence invisible in the map.

Discussion

Electron and substrate transfer

Under normal turnover conditions, the two monomers of complex III are not active simultaneously (Castellani et al., 2010). A model for electron transfer within complex III has been proposed, in which the activation of either monomer is random and both monomers switch between active and inactive states (Covian and Trumpower, 2008). This model implies that the two branches for electron transfer in complex III are equivalent and, in particular, both iron-sulfur domains will move between their b and c1 positions (Iwata et al., 1998) for reduction of heme c1 and, subsequently, cytochrome c.

In the various crystal structures of complex III, the iron-sulfur domains occupy different positions, either as a result of crystal contacts or, more significantly, in response to the occupancy of the QP site (Berry et al., 2013). In single particle cryo-EM conformational changes are not subject to such contacts. Therefore, in the absence of inhibitors, and in light of the known mechanism of electron transfer, a cryo-EM map of complex III should reveal two equally unresolved iron-sulfur domains, as a result of averaging particles in different states. In our map, only one iron-sulfur domain is unresolved, and hence active. The observed asymmetry is thus incompatible with an alternating activation of the two complex III monomers and implies that each monomer has a different function. Our data suggest that in the respirasome, the central monomer of complex III preferentially catalyzes quinol oxidation. As electrons can equilibrate rapidly between the two bL hemes (Castellani et al., 2010), quinone reduction may occur in either monomer, although it is possible that the distal monomer has no function, except as a scaffold.

In the respirasome, the quinol oxidation site of the active complex III monomer faces the site on complex I where the reduced quinol is released. The shortest distance between the two sites is about 11 nm (Figure 5A). The quinol oxidation site of the inactive monomer is on the far side of the dimer, at a distance of at least 18 nm from the complex I site, assuming that the quinol cannot pass through the complex III dimer. A quinol diffusing through the lipid bilayer from complex I to complex III would thus encounter the active monomer first. The linear arrangement of complex I, the active complex III monomer and complex IV in the respirasome seems to be particularly favourable for efficient substrate and electron transfer. Significantly, only one of the complex III monomers is well-placed to accept reduced quinol from complex I, which may explain why only this monomer is active. Substrate channeling within the respirasome has been extensively debated. Results inconsistent with the existence of distinct quinone pools and substrate channeling have been reported (Blaza et al., 2014), but substrate channeling between complexes I and III is supported by most studies of metabolic flux control (Bianchi et al., 2003; Lapuente-Brun et al., 2013) and in recent reviews discussing the results of Blaza et al (Enríquez, 2016; Lenaz et al., 2016). The asymmetry observed for complex III in our supercomplex structure is in agreement with these results and provides further support for electron transfer by substrate channeling in the respirasome.

While it is clear that all functional complex III in mitochondria is dimeric, it is not known at this stage what proportion of the complex III dimers is incorporated into respirasomes. The functional asymmetry of complex III and its iron-sulfur domains in the supercomplex provides strong evidence for a spatially organized flow of electrons and substrates in the respiratory chain from complex I to complex IV (Figure 5B). Turning off one of two electron transfer paths through complex III that are, in principle, equally likely might also offer a kinetic advantage in the electron transfer to complex IV via cytochrome c. The cytochrome c binding site on the active bc1 monomer is about 10 Å closer to complex IV than that of the inactive monomer. Complex IV needs four electrons, and hence four cytochromes c, to produce one molecule of water. The proximity of the binding sites, which would minimize the time for each of the four cytochrome c transfer steps, may be an important advantage.

Our observation that in the respirasome only one of two possible electron transfer branches is active is reminiscent of the two near-symmetrical electron transfer chains in photosystem II, one of which has been turned off in the course of evolution to avoid wasteful electron transfer to two identical quinol electron acceptors (Nelson and Yocum, 2006). In some ways, oxygen reduction in complex IV is the reverse of the water oxidation reaction in photosystem II, and it might be equally important in both cases to control the flow of four electrons into, or away from, the reaction site by turning off one of two possible transfer paths.

Supercomplex assembly, complex I stability and human health

It is widely known that dysfunctional electron transfer complexes of the mitochondrial respiratory chain cause severe and, at present, incurable genetic disorders. Many of these conditions can be traced to point mutations in subunits of mitochondrial complex I (Rodenburg, 2016) or III (Meunier et al., 2013). Many mutations result in reduced levels or activities of either complex I or III in mitochondrial membranes. However, point mutations (Bruno et al., 2003) or truncation (Lamantea et al., 2002) of the mitochondrially encoded cytochrome b subunit of complex III that cause progressive exercise intolerance and lactic acidosis in human patients were found to not only reduce the levels of complex III but also that of complex I. Biochemical studies of similar mutants established that the depletion of complex III levels in the membrane results in a secondary loss of complex I (Acin-Perez et al., 2004; Schägger et al., 2004). More recently, stable expression of complex I in the absence of complex III has been demonstrated through direct or indirect inhibition of ROS production by RET (Guaras et al., 2016). Inhibition of cytochrome bc1 activity however was insufficient to induce complete degradation of complex I (Acin-Perez et al., 2004; Guaras et al., 2016), suggesting that physical interaction between the complexes has an at least partially stabilizing effect. These findings point to a central role of the respirasome in complex I stability.

The nuclear-encoded, supernumerary complex I subunit B14.7 (Figure 3) evidently has a major part in the formation and stability of the respirasome. A point mutation in this subunit is associated with complex I deficiency in patients with fatal infantile lactic acidemia (Berger et al., 2008), and a disruption of the gene of the homologous protein in the fungus Neurospora crassa results in incomplete assembly of complex I (Nehls et al., 1992). At present it is not known whether these mutations affect the assembly of the respirasome, but on the basis of our structure it seems more than likely.

Mammalian respirasome structures

While this manuscript was under review, two other cryo-EM structures of mammalian respirasomes were published (Gu et al., 2016; Letts et al., 2016b). These studies used mitochondria from porcine or ovine, rather than bovine, heart. The ovine complex I has been described as being particularly stable and suitable for structural studies (Letts et al., 2016a) and the same most likely holds true for supercomplexes. At resolutions of 5–6 Å, both structures corroborate most of the protein-protein contacts we observe in the bovine complex. Several additional subunits were found to participate in the stabilization of the supercomplex, which were not clearly visible at our map resolution. A detailed comparison is possible only for the ovine complex, since neither the map nor the coordinates of the porcine supercomplex have been released.

Of the two new studies, only one (Letts et al., 2016b) found a second significant respirasome conformation at an overall resolution of 6.7 Å. The two states are referred to as the tight and loose conformation. The tight conformation of the ovine complex is essentially identical to class 1 of the bovine respirasome, but the arrangement of component complexes in the loose state differs considerably from our class 2. In the bovine respirasome, complex III rotates by 25° relative to complex I, while the position of complex IV remains unchanged. In contrast, in the loose ovine respirasome, complex IV swings away from complex III. Complex III rotates in the same direction as in our class 2, but to a lesser extent. In both the bovine and the ovine respirasome, interactions between complex III and IV are disrupted in these minor classes. In the case of the ovine complex, the proportion of particles in the loose state was found to increase with time (Letts et al., 2016b). This suggests that the loose state of the ovine respirasome and our class 2 of the bovine complex result from incipient denaturation during and after purification. The differences observed between these two classes most likely reflect different degrees of respirasome instability, which may be species-specific, or due to differences in supercomplex purification.

Surprisingly, the two iron-sulfur domains in the complex III dimer are both resolved in the porcine and ovine respirasome, whereas one of them is disordered in our structure of the bovine complex. Therefore neither of the two other structures shows a mobile and therefore catalytically active Rieske protein, as we observe in the bovine respirasome. Letts et al propose that due to the close proximity between complex IV and the Rieske protein of the outer monomer of complex III, the iron-sulfur domain of the outer complex III monomer cannot undergo the conformational changes required for electron transfer. This is consistent with our findings, although neither structure shows any direct protein-protein contacts that would inhibit domain movements. Letts et al conclude that quinone oxidation is likely to take place in the central complex III monomer. This agrees with our model, which is however based on firm experimental evidence of an unresolved, and hence mobile, iron-sulfur domain. Letts et al also propose that quinone reduction in complex III should be catalyzed by the distal monomer; however the respirasome structures provide no evidence for this. Further studies should clarify the exact transfer path of electrons and quinol substrates through the mammalian respirasome.

Conclusions

The 9 Å structure of the mammalian mitochondrial supercomplex shows a well-defined arrangement of the respiratory chain complexes I, III and IV from the inner mitochondrial membrane. The ordered structure of one of the two iron-sulfur domains in the complex III dimer indicates that, contrary to expectation, one of the two electron transfer pathways in complex III is always active, whereas the other monomer is inactive. The active complex III monomer is in a pivotal position between the quinol binding site of complex I and the cytochrome c binding site of complex IV, and hence for directed electron flow through the respirasome. Far from being a random collection of close-packed respiratory chain complexes, the mitochondrial respirasome has not only a clear role for the long-term stability of its component complexes, in particular complex I, but also in optimizing electron flow in cellular respiration.

Materials and methods

Isolation and purification of supercomplexes from bovine heart mitochondria

Bovine heart mitochondria were prepared by differential centrifugation as described (Krause et al., 2005). Mitochondria were solubilized with 1% (w/v) digitonin (Calbiochem) at a detergent-to-protein ratio of 28:1 (Schäfer et al., 2006) or with 0.11% (w/v) PCC-a-M (Glycon) at a detergent-to-protein ratio of 1:1 by incubation for 1 hr at 4°C. Detergent exchange to amphipol A8-35 and protein purification were performed as described before (Althoff et al., 2011).

BN-PAGE and in-gel activity assays

Proteins were separated on 3–12% polyacrylamide linear gradient gels by BN-PAGE and 2D BN/BN-PAGE (Wittig et al., 2006). Functionally active supercomplexes were detected by in-gel activity assays for NADH:dehydrogenase and cytochrome c oxidase (Grad and Lemire, 2006; Kuonen et al., 1986). Briefly, for complex I, gels were incubated in buffer containing 100 mM Tris, pH 7.4, 0.5 mM NBT (p-nitrotetrazolium blue) and 100 μM β–NADH for 1 hr. For complex IV, gels were incubated for 10–12 hr in 50 mM sodium phosphate buffer at pH 7.2, 0.5 mg/ml DAB (3,3’diaminobenzidine tetrahydrochloride), 1 mg/ml equine heart cytochrome c and 20 U/ml bovine liver catalase.

Data collection

3 μl of a respirasome sample were applied to freshly glow discharged Quantifoil R2/2 holey carbon grids (Quantifoil Micro Tools, Germany) that had been pretreated in chloroform for 1–2 hr. The grids were blotted for 10 s at 90% humidity and 10°C in an FEI Vitrobot plunge freezer. Immediately before freezing, 0.6 µl of 1 M KCl were added to avoid protein aggregation. Cryo-EM images were collected on a FEI Tecnai Polara operating at 300 kV aligned as previously described (Mills et al., 2013). The microscope was equipped with a Falcon III direct electron detector. Images were recorded manually at a nominal magnification of 59,000x yielding a pixel size at the specimen of 1.77 Å. The camera system recorded 32 frames/s. Videos were collected for 1.5 s with a total of 46 frames with a calibrated dose of about 1.5 e-/Å2 per frame, at defocus values between −1.3 and −4.3 μm.

Image processing

A set of 3592 micrographs was collected. Whole-image drift correction of each movie was performed using the algorithm developed by Li (Li et al., 2013). Particles were picked manually using EMAN Boxer (Ludtke et al., 1999), and the micrograph-based CTF was determined using CTFFIND3 (Mindell and Grigorieff, 2003) in the RELION 1.4 workflow (Scheres, 2012). The initial dataset contained 156,519 particle images (288 pixels x 288 pixels). Particles were subjected to two-dimensional reference-free classification in RELION 1.4 (Scheres, 2012). Visual selection of particle classes with interpretable features resulted in a dataset of 137,606 particle images for the first 3D consensus refinement. The earlier low-resolution respirasome map (Althoff et al., 2011) was low-pass filtered to 60 Å and used as an initial model for the 3D refinement in RELION 1.4. Individual frames were B-factor weighted and movements of individual particles were reversed by movie frame correction in RELION 1.4 (Scheres, 2014). The resulting dataset of polished particles was used for 3D classification and the best 3D classes were selected for further processing. UCSF Chimera (Pettersen et al., 2004) was used for visualization of cryo-EM maps and docking of atomic models. Figures were drawn with UCSF Chimera.

Data deposition

The cryo-EM maps were deposited in the Electron Microscopy Data Bank with accession code EMD-4107, EMD-4108 and EM-4109 for the cryo-EM maps of class 1, 2 and 3, respectively and the structure coordinates of the complexes fitted to class 1 were deposited in the Protein Data Bank with accession number 5LUF.

Acknowledgements

We thank Matteo Allegretti for help with cryo-EM data acquisition and image processing, and Natalie Bärland for help with particle picking. This work was funded by the Max Planck Society and the Cluster of Excellence Frankfurt ‘Macromolecular Complexes’ (DFG Project EXC 115).

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Funding Information

This paper was supported by the following grants:

  • Max-Planck-Gesellschaft to Werner Kühlbrandt.

  • Cluster of Excellence Frankfurt DFG Project EXC 115 to Werner Kühlbrandt.

Additional information

Competing interests

WK: Reviewing editor, eLife.

The other authors declare that no competing interests exist.

Author contributions

JSS, Wrote the article, Acquisition of data, Analysis and interpretation of data.

DJM, Supported the cryo-EM work, Drafting or revising the article.

JV, Conception and design, Analysis and interpretation of data, Drafting or revising the article.

WK, Wrote the article, Conception and design, Analysis and interpretation of data.

Additional files

Major datasets

The following datasets were generated:

Sousa JS,Mills DJ,Vonck J,Kuehlbrandt W,2016,Cryo-EM map of bovine respirasome,http://www.ebi.ac.uk/pdbe/entry/emdb/EMD-4107,Publicly available at the EBI Protein Data Bank (accession no: EMD-4107)

Sousa JS,Mills DJ,Vonck J,Kuehlbrandt W,2016,Cryo-EM of bovine respirasome,http://www.ebi.ac.uk/pdbe/entry/emdb/EMD-4108,Publicly available at th EBI Protein Data Bank (accession no: EMD-4108)

Sousa JS,Mills DJ,Vonck J,Kuehlbrandt W,2016,Cryo-EM of bovine respirasome,http://www.ebi.ac.uk/pdbe/entry/emdb/EMD-4109,Publicly available at the EBI Protein Data Bank (accession no: EMD-4109)

Sousa JS,Mills DJ,Vonck J,Kuehlbrandt W,2016,cryo-EM of bovine respirasome,http://www.rcsb.org/pdb/explore.do?structureId=5LUF,Publicly available at the RCSB Protein Data Bank (accession no: 5LUF)

The following previously published datasets were used:

Vinothkumar KR,Zhu J,Hirst J,2014,Electron cryo-microscopy of bovine Complex I,http://www.rcsb.org/pdb/explore/explore.do?structureId=4UQ8,Publicly available at the EBI Protein Data Bank (accession no: 4UQ8)

Iwata S,Lee JW,Okada K,Lee JK,Iwata M,Rasmussen B,Link TA,Ramaswamy S,Jap BK,1999,CYTOCHROME BC1 COMPLEX FROM BOVINE,http://www.rcsb.org/pdb/explore/explore.do?structureId=1BGY,Publicly available at the RCSB Protein Data Bank (accession no: 1BGY)

Tsukihara T,Aoyama H,Yamashita E,Tomizaki T,Yamaguchi H,Shinzawa-Itoh K,Nakashima R,Yaono R,Yoshikawa S,1996,STRUCTURE OF BOVINE HEART CYTOCHROME C OXIDASE AT THE FULLY OXIDIZED STATE,http://www.rcsb.org/pdb/explore/explore.do?structureId=1OCC,Publicly available at the RCSB Protein Data Bank (accession no: 1OCC)

References

  1. Acín-Pérez R, Bayona-Bafaluy MP, Fernández-Silva P, Moreno-Loshuertos R, Pérez-Martos A, Bruno C, Moraes CT, Enríquez JA, Bayona-Bafaluy María Pilar, Enríquez JA. Respiratory complex III is required to maintain complex I in mammalian mitochondria. Molecular Cell. 2004;13:805–815. doi: 10.1016/S1097-2765(04)00124-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Acín-Pérez R, Fernández-Silva P, Peleato ML, Pérez-Martos A, Enriquez JA. Respiratory active mitochondrial supercomplexes. Molecular Cell. 2008;32:529–539. doi: 10.1016/j.molcel.2008.10.021. [DOI] [PubMed] [Google Scholar]
  3. Althoff T, Mills DJ, Popot JL, Kühlbrandt W. Arrangement of electron transport chain components in bovine mitochondrial supercomplex I1III2IV1. The EMBO Journal. 2011;30:4652–4664. doi: 10.1038/emboj.2011.324. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Balsa E, Marco R, Perales-Clemente E, Szklarczyk R, Calvo E, Landázuri MO, Enríquez JA. NDUFA4 is a subunit of complex IV of the mammalian electron transport chain. Cell Metabolism. 2012;16:378–386. doi: 10.1016/j.cmet.2012.07.015. [DOI] [PubMed] [Google Scholar]
  5. Berger I, Hershkovitz E, Shaag A, Edvardson S, Saada A, Elpeleg O. Mitochondrial complex I deficiency caused by a deleterious NDUFA11 mutation. Annals of Neurology. 2008;63:405–408. doi: 10.1002/ana.21332. [DOI] [PubMed] [Google Scholar]
  6. Berry EA, De Bari H, Huang L-S. Unanswered questions about the structure of cytochrome bc1 complexes. Biochimica Et Biophysica Acta (BBA) - Bioenergetics. 2013;1827:1258–1277. doi: 10.1016/j.bbabio.2013.04.006. [DOI] [PubMed] [Google Scholar]
  7. Bianchi C, Fato R, Genova ML, Parenti Castelli G, Lenaz G, Castelli GP. Structural and functional organization of Complex I in the mitochondrial respiratory chain. BioFactors. 2003;18:3–9. doi: 10.1002/biof.5520180202. [DOI] [PubMed] [Google Scholar]
  8. Blaza JN, Serreli R, Jones AJY, Mohammed K, Hirst J. Kinetic evidence against partitioning of the ubiquinone pool and the catalytic relevance of respiratory-chain supercomplexes. PNAS. 2014;111:15735–15740. doi: 10.1073/pnas.1413855111. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Bruno C, Santorelli FM, Assereto S, Tonoli E, Tessa A, Traverso M, Scapolan S, Bado M, Tedeschi S, Minetti C. Progressive exercise intolerance associated with a new muscle-restricted nonsense mutation (G142X) in the mitochondrial cytochrome b gene. Muscle & Nerve. 2003;28:508–511. doi: 10.1002/mus.10429. [DOI] [PubMed] [Google Scholar]
  10. Castellani M, Covian R, Kleinschroth T, Anderka O, Ludwig B, Trumpower BL. Direct demonstration of half-of-the-sites reactivity in the dimeric cytochrome bc1 complex: enzyme with one inactive monomer is fully active but unable to activate the second ubiquinol oxidation site in response to ligand binding at the ubiquinone reduction site. The Journal of Biological Chemistry. 2010;285:502–510. doi: 10.1074/jbc.M109.072959. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Covian R, Trumpower BL. Regulatory interactions in the dimeric cytochrome bc1 complex: The advantages of being a twin. Biochimica Et Biophysica Acta (BBA) - Bioenergetics. 2008;1777:1079–1091. doi: 10.1016/j.bbabio.2008.04.022. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. D'Aurelio M, Gajewski CD, Lenaz G, Manfredi G. Respiratory chain supercomplexes set the threshold for respiration defects in human mtDNA mutant cybrids. Human Molecular Genetics. 2006;15:2157–2169. doi: 10.1093/hmg/ddl141. [DOI] [PubMed] [Google Scholar]
  13. Darrouzet E, Valkova-Valchanova M, Moser CC, Dutton PL, Daldal F. Uncovering the [2Fe2S] domain movement in cytochrome bc1 and its implications for energy conversion. PNAS. 2000;97:4567–4572. doi: 10.1073/pnas.97.9.4567. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Diaz F, Fukui H, Garcia S, Moraes CT. Cytochrome c oxidase is required for the assembly/stability of respiratory complex I in mouse fibroblasts. Molecular and Cellular Biology. 2006;26:4872–4881. doi: 10.1128/MCB.01767-05. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Dieteren CE, Koopman WJ, Swarts HG, Peters JG, Maczuga P, van Gemst JJ, Masereeuw R, Smeitink JA, Nijtmans LG, Willems PH. Subunit-specific incorporation efficiency and kinetics in mitochondrial complex I homeostasis. Journal of Biological Chemistry. 2012;287:41851–41860. doi: 10.1074/jbc.M112.391151. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Dudkina NV, Kudryashev M, Stahlberg H, Boekema EJ. Interaction of complexes I, III, and IV within the bovine respirasome by single particle cryoelectron tomography. PNAS. 2011;108:15196–15200. doi: 10.1073/pnas.1107819108. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Enríquez JA. Supramolecular Organization of Respiratory Complexes. Annual Review of Physiology. 2016;78:533–561. doi: 10.1146/annurev-physiol-021115-105031. [DOI] [PubMed] [Google Scholar]
  18. Galkin ON, Terenetskaya IP. ‘Vitamin D’ biodosimeter: basic characteristics and potential applications. Journal of Photochemistry and Photobiology B: Biology. 1999;53:12–19. doi: 10.1016/S1011-1344(99)00115-3. [DOI] [PubMed] [Google Scholar]
  19. Grad LI, Lemire BD. Riboflavin enhances the assembly of mitochondrial cytochrome c oxidase in C. elegans NADH-ubiquinone oxidoreductase mutants. Biochimica Et Biophysica Acta (BBA) - Bioenergetics. 2006;1757:115–122. doi: 10.1016/j.bbabio.2005.11.009. [DOI] [PubMed] [Google Scholar]
  20. Gu J, Wu M, Guo R, Yan K, Lei J, Gao N, Yang M. The architecture of the mammalian respirasome. Nature. 2016;537:639–643. doi: 10.1038/nature19359. [DOI] [PubMed] [Google Scholar]
  21. Guarás A, Perales-Clemente E, Calvo E, Acín-Pérez R, Loureiro-Lopez M, Pujol C, Martínez-Carrascoso I, Nuñez E, García-Marqués F, Rodríguez-Hernández MA, Cortés A, Diaz F, Pérez-Martos A, Moraes CT, Fernández-Silva P, Trifunovic A, Navas P, Vazquez J, Enríquez JA. The CoQH2/CoQ ratio serves as a sensor of respiratory chain efficiency. Cell Reports. 2016;15:197–209. doi: 10.1016/j.celrep.2016.03.009. [DOI] [PubMed] [Google Scholar]
  22. Guerrero-Castillo S, Baertling F, Kownatzki, Wessels, Arnold, Brandt U, Nijtmans L. The assembly pathway of mitochondrial Respiratory chain complex I. Cell Metabolism. 2016: S1550-4131(16)30480-6. doi: 10.1016/j.cmet.2016.09.002. [DOI] [PubMed] [Google Scholar]
  23. Hackenbrock CR, Chazotte B, Gupte SS. The random collision model and a critical assessment of diffusion and collision in mitochondrial electron transport. Journal of Bioenergetics and Biomembranes. 1986;18:331–368. doi: 10.1007/BF00743010. [DOI] [PubMed] [Google Scholar]
  24. Hovers J, Potschies M, Polidori A, Pucci B, Raynal S, Bonneté F, Serrano-Vega MJ, Tate CG, Picot D, Pierre Y, Popot JL, Nehmé R, Bidet M, Mus-Veteau I, Busskamp H, Jung KH, Marx A, Timmins PA, Welte W. A class of mild surfactants that keep integral membrane proteins water-soluble for functional studies and crystallization. Molecular Membrane Biology. 2011;28:171–181. doi: 10.3109/09687688.2011.552440. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Iwata S, Lee JW, Okada K, Lee JK, Iwata M, Rasmussen B, Link TA, Ramaswamy S, Jap BK. Complete structure of the 11-subunit bovine mitochondrial cytochrome bc1 complex. Science. 1998;281:64–71. doi: 10.1126/science.281.5373.64. [DOI] [PubMed] [Google Scholar]
  26. Kadenbach B, Jarausch J, Hartmann R, Merle P. Separation of mammalian cytochrome c oxidase into 13 polypeptides by a sodium dodecyl sulfate-gel electrophoretic procedure. Analytical Biochemistry. 1983;129:517–521. doi: 10.1016/0003-2697(83)90586-9. [DOI] [PubMed] [Google Scholar]
  27. Krause F, Reifschneider NH, Goto S, Dencher NA. Active oligomeric ATP synthases in mammalian mitochondria. Biochemical and Biophysical Research Communications. 2005;329:583–590. doi: 10.1016/j.bbrc.2005.02.010. [DOI] [PubMed] [Google Scholar]
  28. Kucukelbir A, Sigworth FJ, Tagare HD. Quantifying the local resolution of cryo-EM density maps. Nature Methods. 2014;11:63–65. doi: 10.1038/nmeth.2727. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Kuonen DR, Roberts PJ, Cottingham IR. Purification and analysis of mitochondrial membrane proteins on nondenaturing gradient polyacrylamide gels. Analytical Biochemistry. 1986;153:221–226. doi: 10.1016/0003-2697(86)90084-9. [DOI] [PubMed] [Google Scholar]
  30. Kühlbrandt W. Structure and function of mitochondrial membrane protein complexes. BMC Biology. 2015;13:89. doi: 10.1186/s12915-015-0201-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Lamantea E, Carrara F, Mariotti C, Morandi L, Tiranti V, Zeviani M. A novel nonsense mutation (Q352X) in the mitochondrial cytochrome b gene associated with a combined deficiency of complexes I and III. Neuromuscular Disorders. 2002;12:49–52. doi: 10.1016/S0960-8966(01)00244-9. [DOI] [PubMed] [Google Scholar]
  32. Lapuente-Brun E, Moreno-Loshuertos R, Acín-Pérez R, Latorre-Pellicer A, Colás C, Balsa E, Perales-Clemente E, Quirós PM, Calvo E, Rodríguez-Hernández MA, Navas P, Cruz R, Carracedo Á, López-Otín C, Pérez-Martos A, Fernández-Silva P, Fernández-Vizarra E, Enríquez JA. Supercomplex assembly determines electron flux in the mitochondrial electron transport chain. Science. 2013;340:1567–1570. doi: 10.1126/science.1230381. [DOI] [PubMed] [Google Scholar]
  33. Lazarou M, McKenzie M, Ohtake A, Thorburn DR, Ryan MT. Analysis of the assembly profiles for mitochondrial- and nuclear-DNA-encoded subunits into complex I. Molecular and Cellular Biology. 2007;27:4228–4237. doi: 10.1128/MCB.00074-07. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Leif H, Sled VD, Ohnishi T, Weiss H, Friedrich T. Isolation and characterization of the proton-translocating NADH: ubiquinone oxidoreductase from Escherichia coli. European Journal of Biochemistry. 1995;230:538–548. doi: 10.1111/j.1432-1033.1995.tb20594.x. [DOI] [PubMed] [Google Scholar]
  35. Lenaz G, Tioli G, Falasca AI, Genova ML. Complex I function in mitochondrial supercomplexes. Biochimica Et Biophysica Acta (BBA) - Bioenergetics. 2016;1857:991–1000. doi: 10.1016/j.bbabio.2016.01.013. [DOI] [PubMed] [Google Scholar]
  36. Letts JA, Degliesposti G, Fiedorczuk K, Skehel M, Sazanov LA. Purification of ovine respiratory Complex I results in a highly active and stable preparation. Journal of Biological Chemistry. 2016a: jbc.M116.735142. doi: 10.1074/jbc.M116.735142. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Letts JA, Fiedorczuk K, Sazanov LA. The architecture of respiratory supercomplexes. Nature. 2016b;537:644–648. doi: 10.1038/nature19774. [DOI] [PubMed] [Google Scholar]
  38. Li X, Mooney P, Zheng S, Booth CR, Braunfeld MB, Gubbens S, Agard DA, Cheng Y. Electron counting and beam-induced motion correction enable near-atomic-resolution single-particle cryo-EM. Nature Methods. 2013;10:584–590. doi: 10.1038/nmeth.2472. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Ludtke SJ, Baldwin PR, Chiu W. EMAN: semiautomated software for high-resolution single-particle reconstructions. Journal of Structural Biology. 1999;128:82–97. doi: 10.1006/jsbi.1999.4174. [DOI] [PubMed] [Google Scholar]
  40. Maranzana E, Barbero G, Falasca AI, Lenaz G, Genova ML. Mitochondrial respiratory supercomplex association limits production of reactive oxygen species from complex I. Antioxidants & Redox Signaling. 2013;19:1469–1480. doi: 10.1089/ars.2012.4845. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Marques I, Dencher NA, Videira A, Krause F. Supramolecular organization of the respiratory chain in Neurospora crassa mitochondria. Eukaryotic Cell. 2007;6:2391–2405. doi: 10.1128/EC.00149-07. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Meunier B, Fisher N, Ransac S, Mazat J-P, Brasseur G. Respiratory complex III dysfunction in humans and the use of yeast as a model organism to study mitochondrial myopathy and associated diseases. Biochimica Et Biophysica Acta (BBA) - Bioenergetics. 2013;1827:1346–1361. doi: 10.1016/j.bbabio.2012.11.015. [DOI] [PubMed] [Google Scholar]
  43. Mills DJ, Vitt S, Strauss M, Shima S, Vonck J. De novo modeling of the F(420)-reducing [NiFe]-hydrogenase from a methanogenic archaeon by cryo-electron microscopy. eLife. 2013;2:e00218. doi: 10.7554/eLife.00218. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Mindell JA, Grigorieff N. Accurate determination of local defocus and specimen tilt in electron microscopy. Journal of Structural Biology. 2003;142:334–347. doi: 10.1016/S1047-8477(03)00069-8. [DOI] [PubMed] [Google Scholar]
  45. Moreno-Lastres D, Fontanesi F, García-Consuegra I, Martín MA, Arenas J, Barrientos A, Ugalde C. Mitochondrial complex I plays an essential role in human respirasome assembly. Cell Metabolism. 2012;15:324–335. doi: 10.1016/j.cmet.2012.01.015. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Nehls U, Friedrich T, Schmiede A, Ohnishi T, Weiss H. Characterization of assembly intermediates of NADH:ubiquinone oxidoreductase (complex I) accumulated in Neurospora mitochondria by gene disruption. Journal of Molecular Biology. 1992;227:1032–1042. doi: 10.1016/0022-2836(92)90519-P. [DOI] [PubMed] [Google Scholar]
  47. Nelson N, Yocum CF. Structure and function of photosystems I and II. Annual Review of Plant Biology. 2006;57:521–565. doi: 10.1146/annurev.arplant.57.032905.105350. [DOI] [PubMed] [Google Scholar]
  48. Panov A, Dikalov S, Shalbuyeva N, Hemendinger R, Greenamyre JT, Rosenfeld J. Species- and tissue-specific relationships between mitochondrial permeability transition and generation of ROS in brain and liver mitochondria of rats and mice. AJP: Cell Physiology. 2006;292:C708–C718. doi: 10.1152/ajpcell.00202.2006. [DOI] [PubMed] [Google Scholar]
  49. Pettersen EF, Goddard TD, Huang CC, Couch GS, Greenblatt DM, Meng EC, Ferrin TE. UCSF Chimera--a visualization system for exploratory research and analysis. Journal of Computational Chemistry. 2004;25:1605–1612. doi: 10.1002/jcc.20084. [DOI] [PubMed] [Google Scholar]
  50. Rodenburg RJ. Mitochondrial complex I-linked disease. Biochimica Et Biophysica Acta (BBA) - Bioenergetics. 2016;1857:938–945. doi: 10.1016/j.bbabio.2016.02.012. [DOI] [PubMed] [Google Scholar]
  51. Sazanov LA, Hinchliffe P. Structure of the hydrophilic domain of respiratory complex I from Thermus thermophilus. Science. 2006;311:1430–1436. doi: 10.1126/science.1123809. [DOI] [PubMed] [Google Scholar]
  52. Scheres SH. Beam-induced motion correction for sub-megadalton cryo-EM particles. eLife. 2014;3:e03665. doi: 10.7554/eLife.03665. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Scheres SH. RELION: implementation of a Bayesian approach to cryo-EM structure determination. Journal of Structural Biology. 2012;180:519–530. doi: 10.1016/j.jsb.2012.09.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Schäfer E, Dencher NA, Vonck J, Parcej DN. Three-dimensional structure of the respiratory chain supercomplex I1III2IV1 from bovine heart mitochondria. Biochemistry. 2007;46:12579–12585. doi: 10.1021/bi700983h. [DOI] [PubMed] [Google Scholar]
  55. Schäfer E, Seelert H, Reifschneider NH, Krause F, Dencher NA, Vonck J. Architecture of active mammalian respiratory chain supercomplexes. Journal of Biological Chemistry. 2006;281:15370–15375. doi: 10.1074/jbc.M513525200. [DOI] [PubMed] [Google Scholar]
  56. Schägger H, de Coo R, Bauer MF, Hofmann S, Godinot C, Brandt U. Significance of respirasomes for the assembly/stability of human respiratory chain complex I. Journal of Biological Chemistry. 2004;279:36349–36353. doi: 10.1074/jbc.M404033200. [DOI] [PubMed] [Google Scholar]
  57. Schägger H, Link TA, Engel WD, von Jagow G. Isolation of the eleven protein subunits of the bc1 complex from beef heart. Methods in Enzymology. 1986;126:224–237. doi: 10.1016/s0076-6879(86)26024-3. [DOI] [PubMed] [Google Scholar]
  58. Schägger H, Pfeiffer K. Supercomplexes in the respiratory chains of yeast and mammalian mitochondria. The EMBO Journal. 2000;19:1777–1783. doi: 10.1093/emboj/19.8.1777. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Seelert H, Dani DN, Dante S, Hauß T, Krause F, Schäfer E, Frenzel M, Poetsch A, Rexroth S, Schwaßmann HJ, Suhai T, Vonck J, Dencher NA. From protons to OXPHOS supercomplexes and Alzheimer's disease: Structure–dynamics–function relationships of energy-transducing membranes. Biochimica Et Biophysica Acta (BBA) - Bioenergetics. 2009;1787:657–671. doi: 10.1016/j.bbabio.2009.02.028. [DOI] [PubMed] [Google Scholar]
  60. Tsukihara T, Aoyama H, Yamashita E, Tomizaki T, Yamaguchi H, Shinzawa-Itoh K, Nakashima R, Yaono R, Yoshikawa S. The whole structure of the 13-subunit oxidized cytochrome c oxidase at 2.8 A. Science. 1996;272:1136–1144. doi: 10.1126/science.272.5265.1136. [DOI] [PubMed] [Google Scholar]
  61. Vempati UD, Han X, Moraes CT. Lack of cytochrome c in mouse fibroblasts disrupts assembly/stability of respiratory complexes I and IV. Journal of Biological Chemistry. 2009;284:4383–4391. doi: 10.1074/jbc.M805972200. [DOI] [PMC free article] [PubMed] [Google Scholar]
  62. Vinothkumar KR, Zhu J, Hirst J. Architecture of mammalian respiratory complex I. Nature. 2014;515:80–84. doi: 10.1038/nature13686. [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Walker JE. The NADH:ubiquinone oxidoreductase (complex I) of respiratory chains. Quarterly Reviews of Biophysics. 1992;25:253–324. doi: 10.1017/S003358350000425X. [DOI] [PubMed] [Google Scholar]
  64. Wittig I, Braun HP, Schägger H. Blue native PAGE. Nature Protocols. 2006;1:418–428. doi: 10.1038/nprot.2006.62. [DOI] [PubMed] [Google Scholar]
  65. Yang XH, Trumpower BL. Purification of a three-subunit ubiquinol-cytochrome c oxidoreductase complex from Paracoccus denitrificans. The Journal of Biological Chemistry. 1986;261:12282–12289. [PubMed] [Google Scholar]
  66. Zara V, Conte L, Trumpower BL. Identification and characterization of cytochrome bc(1) subcomplexes in mitochondria from yeast with single and double deletions of genes encoding cytochrome bc(1) subunits. FEBS Journal. 2007;274:4526–4539. doi: 10.1111/j.1742-4658.2007.05982.x. [DOI] [PubMed] [Google Scholar]
  67. Zhu J, Vinothkumar KR, Hirst J. Structure of mammalian respiratory complex I. Nature. 2016;536:354–358. doi: 10.1038/nature19095. [DOI] [PMC free article] [PubMed] [Google Scholar]
eLife. 2016 Nov 10;5:e21290. doi: 10.7554/eLife.21290.024

Decision letter

Editor: Stephen C Harrison1

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "Functional asymmetry and electron flow in the bovine respirasome" for consideration by eLife. Your article has been favorably evaluated by John Kuriyan as the Senior Editor and three reviewers, one of whom, Stephen C. Harrison (Reviewer #1), is a member of our Board of Reviewing Editors.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary and specific request:

This manuscript reports the structure of the bovine respirasome at 9 Å resolution. Its principal conclusion concerns a functional asymmetry: one of the iron-sulfur clusters is docked in such a way that electrons cannot be transferred, because the distance to the partner is too great. The authors conclude that one of the two is inactive. The structure is well determined at the stated resolution, but two structures at much higher resolution appeared after this manuscript was submitted. The reviewers request a revised manuscript that places the current conclusions in the context of those reports. In particular, the authors should address the likelihood of a different interpretation – some sort of alternating access model in which only one subunit can access the transfer site at a time, while the other docks in the observed, autoinhibited position. That is, the statement that the resolved Rieske center is immobile and therefore not functional is one possibility, but just because it has a fixed orientation in this structure of the inactive respirasome does not mean it cannot move in the active respirasome.

The following detailed comments may help with the revision.

Reviewer #2:

It is strange to show an FSC with phases randomized beyond 20 Å. Use of a phase randomized FSC to correct for effects from masking of maps is excellent, but the corrected FSC should be shown, not the phase randomized FSC.

Reviewer #3:

Introduction:

First paragraph: The entrance of electrons in the mitochondrial electron transfer chain is branched, therefore CI is the largest complex, but cannot be called the first, because other enzymes that deliver electrons to CoQ are equally first in their branch. Please correct.

First paragraph: It should be made clear that mammalian CI comprises 44 different subunits, one being duplicated. Quoting Carrol et al. 2002 could be misleading, since in this reference it was considered that CI comprises 45 different subunits. Latter it was demonstrated and generally accepted that NDUFA4 is not a CI subunit. It is advisable to clarify and to use the correct references.

First paragraph: The number of mammalian CIV subunits has been proposed to be 14.

Third paragraph: The issue of CoQ partitioning is still hotly debated. The authors correctly mention that Blaza et al. 2014 disagree with this idea but forget to mention that those authors' experimental setup was rigorously contested in two recent reports (Enriquez 2016. Annu. Rev. Physiol. 78, 533-561 & Lenaz et al., 2016. Biochim Biophys Acta 1857, 991-1000). This is relevant because the very interesting proposal by Sousa et al. for the mechanism of electron transfer within the respirasome agrees with the interpretation of Enriquez and Lenaz rather than that of Blaza.

Third paragraph: The best experimental support for the role of supercomplexes in minimizing ROS production was probably in the paper by Maranzana (Maranzana et al. 2013. Antioxid Redox Signal 19, 1469-1480), which is not quoted.

Third paragraph: The role of supercomplexes in stabilizing CI is noted in the Introduction, but in a way that may not reflect the current view of this question. The historical origin of this proposal is as follows: in 2004 two groups independently proposed that CI stability was compromised in the absence of CIII (Acin-Perez et al., 2004 & Schägger et al. 2004). Latter, the group of Carlos Moraes demonstrated that ablation of CIV or cyt c, also compromised the stability of CI (Diaz et al. 2006- Mol Cell Biol 26, 4872-4881 & Vempati, et al. 2009. J. Biol. Chem. 284, 4383-4391). More recently it was demonstrated that CI can be assembled functionally in the absence of CIII, CIV or cyt c by recovering the re-oxidation of CoQ (Guarás, 2016. Cell Reports 15, 197-209). This series of reports led to the conclusion that assembly of the individual respiratory complexes can proceed independently and requires neither the presence of other complexes nor interaction with them. In line with this conclusion, kinetic analysis showed that the formation of complexes happened before the formation of supercomplexes (Acín-Peréz et al. 2008. Mol. Cell 32, 529-539). This result does not preclude, however, that incomplete CI or CIII could interact with other complexes and form supercomplexes, as reported by several groups (Marques, I., 2007. Eukaryotic Cell 6, 2391-2405). In 2012 it was proposed that the assembly of the NADH-DH module of CI require that formation of a non-functional respirasome between partially assembled CI with CIII and CIV (Moreno-Lastres et al. 2012. Cell Metab.15, 324-335). This proposal contradicted that of (Acín-Peréz et al. 2008. Mol. Cell 32, 529-539). A very recent reevaluation concluded that CI, CIII and CIV are fully assembled before they associate into supercomplexes, refuting the proposal by Moreno-Lastres (Guerrero-Castillo S, Baertling F, Kownatzki D., Wessels H.J., Arnold S., Brandt U. and Nijtmans L. 2016. Cell Metab, in press).

Reviewer #3 gave this extended explanation because the authors depended heavily on Moreno-Lastres proposal in interpreting their findings; it may be better to be cautious and explore other potential interpretations.

Results and Discussion:

Subsection “Isolation of mitochondrial supercomplexes solubilized with PCC-a-M”, second paragraph: The authors conclude that PCC-a-M does not allow detection of SC: I1III2 and infer that this SC is not observed because it is preserved as I1III2IV1. The authors should be aware that different detergents generate different micelles containing the SCs and modify their electrophoretic motility. In this sense Figure 1—figure supplement 1 does not help very much, since CIV is not detected in the first dimension in some of the samples (Figure 1A and Figure 1—figure supplement 1A) while it is detected in others (Figure 1—figure supplement 1B), and in the second dimension almost all the CIV seemed to be present as free complex. In addition, the authors find I1III2 by cryo-electron microscopy analysis in sufficient quantity to propose its structure (subsection “Defined protein-protein contacts”, first paragraph). They justify that observation by assuming that I1III2 is an artefact generated in vitro by cleavage of I1III2IV2 (in the second paragraph of the aforementioned subsection). This could be the explanation, but it is also possible that they co-purified both, but their detergent did not resolve them in a gel. In fact, they do not report the observation of free CIV that has to be present if the cleavage happened after purification. Finally, if the authors are interested in identifying properly the different components of the bands obtained with the new detergent they should use immunodetection approaches rather than Coomassie staining. There are good antibodies for that purpose.

Figure 3A shows three proteins: CI-B14.7, CIII-Subunit 8, and a third in yellow (CIV) that remains unidentified. Please identify it. Does it participate in the interaction between complexes I and IV or III and IV?

Subsection “Defined protein-protein contacts”, first paragraph. The minimum number of expected TMHs for the respirasome I1III1IV1 is 131 but the authors report only 114. Which ones are they unable to detect and why?

The authors report: "Several of the classes obtained (≈ 30% of the initial dataset) have no density for the NADH:dehydrogenase module of CI or even lack the whole matrix arm". Contrary to the argument given for the unexpected observation of I1III2 without IV, they do not consider these new classes as an artefact of the manipulation. Rather they propose that represent intermediates of supercomplex assembly, potentially in agreement with those postulated by Moreno-Lastres et al. (subsection “Conformational and compositional variability”, last paragraph). The model of for the assembly of CI and SCs proposed by Moreno-Lastres has been questioned, however (see above). Moreover, two different groups proposed that CI damage occurred under natural conditions at the NADH:DH module, and that this can be replaced without the elimination of the rest of the complex (Lazarou, et al. 2007. Mol Cell Biol 27, 4228-4237. Dieteren, et al. 2012 J. Biol. Chem 287, 41851-41860). If so, the observed structures could also be consistent with the prediction of Lazarou and Dieteren and might have come from degradation rather than as assembly intermediates. The data in this manuscript cannot determine whether those structures are natural or artefactual and whether they might represent intermediates of synthesis or degradation. The reviewer's suggestion is to present either all the potential alternative explanations or to simply remove the discussion of the nature of this structures. In particular, the suggestion that complex III acts as an assembly factor for CI (subsection “Supercomplex assembly, complex I stability and human health”, first paragraph) is not justified by the results presented nor by most of the recent literature.

The authors state that their proposed model of functional asymmetry and electron flow in CIII: "does not imply a partitioning of the quinol pool or substrate channeling". However, they also state: "A quinol diffusing through the lipid bilayer from complex I to complex III would thus encounter the active monomer first. The linear arrangement of complex I, the active complex III monomer and complex IV in the respirasome seems to be particularly favourable for efficient substrate and electron transfer." This is confusing, because the statements appear to be contradictory. On the one hand, the authors propose that the respirasome has a particular defined structure that favors oxidization of quinol not only the by CIII incorporated into the respirasome but also by a defined monomer. In the other, they state that this does not imply a kind of channeling. The reviewer believes that their proposed model entails preferential reduction of the ubiquinol within the respirasome rather than a more stochastic picture.

Subsection “Supercomplex assembly, complex I stability and human health”, second paragraph: The statement that deficiency in CIV does not have strong adverse effect on CI stability is incorrect (Diaz et al. 2006- Mol Cell Biol 26, 4872-4881& Vempati, et al. 2009. J. Biol. Chem. 284, 4383-4391).

Comments from reviewer #3 regarding comparison with the other published structures:

One of the two published structures found also two alternative conformations for the respirasome – one similar to the authors' Class 1 and the other more similar to Class 2. Two features characterize the differences between the two conformations: the twist in the position of CIII and the absence of contacts between CIII and CIV. In the Class 2 structure presented here, there is a similar twist in the relative position of CIII, but the authors do not state whether if the interaction between CIII and CIV is also absent. A comment on the similarities and differences between the two classes of structures of the respirasome reported here and published by Sazanov group would be of interest.

The Yang and Sazanov structures agree in the protein-protein contacts. The contacts proposed here are substantially similar but there are discrepancies. In particular, according to the structure proposed here, the interface between CIII and CIV is formed by CIII subunits 1 (UQCR1) and 9 (UQCR10) with COX7A1 of CIV, while Yang and Sazanov both say that CIII UQCR1 and UQCR 11 contact CIV subunit COX7A1. Is there an explanation for the discrepancy?

The Sazanov structure would permit an asymmetric flux of electrons within CIII in the respirasome. The present manuscript also presents an interesting model. It would be of great interest to discuss both, their similarities and their differences.

eLife. 2016 Nov 10;5:e21290. doi: 10.7554/eLife.21290.025

Author response

[…] Summary and specific request:

This manuscript reports the structure of the bovine respirasome at 9 Å resolution. Its principal conclusion concerns a functional asymmetry: one of the iron-sulfur clusters is docked in such a way that electrons cannot be transferred, because the distance to the partner is too great. The authors conclude that one of the two is inactive. The structure is well determined at the stated resolution, but two structures at much higher resolution appeared after this manuscript was submitted. The reviewers request a revised manuscript that places the current conclusions in the context of those reports. In particular, the authors should address the likelihood of a different interpretation – some sort of alternating access model in which only one subunit can access the transfer site at a time, while the other docks in the observed, autoinhibited position. That is, the statement that the resolved Rieske center is immobile and therefore not functional is one possibility, but just because it has a fixed orientation in this structure of the inactive respirasome does not mean it cannot move in the active respirasome.

We have added a comparison with the ovine and porcine structures at the end of the Discussion. The main map classes are very similar and the assignment of interacting subunits is essentially the same in all three structures. There are differences between the minor map classes of the ovine and our bovine supercomplexes, which may be either species-specific or purification dependent. It is likely that these minor particle classes are due to destabilization of the respirasomes.

The possibility of an alternating activation of complex III monomers is now discussed. A detailed explanation of why such a model disagrees with our structure is provided. The recent structures at higher resolution have both iron-sulfur domains resolved and do not provide further insights into the electron and substrate flow in the respirasome. The model proposed by Letts et al., (2016) is discussed in the context or our map.

The following detailed comments may help with the revision.

Reviewer #2:

It is strange to show an FSC with phases randomized beyond 20 Å. Use of a phase randomized FSC to correct for effects from masking of maps is excellent, but the corrected FSC should be shown, not the phase randomized FSC.

The FSC with randomized phases was removed from the graph. Only the corrected FSC is shown now.

Reviewer #3:

Introduction:

First paragraph: The entrance of electrons in the mitochondrial electron transfer chain is branched, therefore CI is the largest complex, but cannot be called the first, because other enzymes that deliver electrons to CoQ are equally first in their branch. Please correct.

The statement has been removed.

First paragraph: It should be made clear that mammalian CI comprises 44 different subunits, one being duplicated. Quoting Carrol et al. 2002 could be misleading, since in this reference it was considered that CI comprises 45 different subunits. Latter it was demonstrated and generally accepted that NDUFA4 is not a CI subunit. It is advisable to clarify and to use the correct references.

The number of CI subunits is now clearly stated and we cite Vinothkumar et al. 2014, where the SDAP subunit was shown to be duplicated:

“Mammalian complex I comprises 44 different subunits, including two copies of subunit SDAP, and therefore consists of a total of 45 subunits (Vinothkumar et al., 2014).”

First paragraph: The number of mammalian CIV subunits has been proposed to be 14.

The number of CIV subunits has been corrected and Balsa et al. 2012 is cited for identification of the 14th subunit (NDUFA4):

“Mammalian complex IV has three core subunits (COX1, COX2 and COX3) and 14 subunits in total (Kadenbach et al., 1983; Balsa et al., 2012).”

Third paragraph: The issue of CoQ partitioning is still hotly debated. The authors correctly mention that Blaza et al. 2014 disagree with this idea but forget to mention that those authors' experimental setup was rigorously contested in two recent reports (Enriquez 2016. Annu. Rev. Physiol. 78, 533-561 & Lenaz et al., 2016. Biochim Biophys Acta 1857, 991-1000). This is relevant because the very interesting proposal by Sousa et al. for the mechanism of electron transfer within the respirasome agrees with the interpretation of Enriquez and Lenaz rather than that of Blaza.

We now provide a more comprehensive description of the current knowledge on substrate channeling by referring to the views of Enríquez (2016) and Lenaz et al. (2016) in the revised Discussion:

“Substrate channeling within the respirasome has been extensively debated. Results inconsistent with the existence of distinct quinone pools and substrate channeling have been reported (Blaza et al., 2014), but substrate channeling between complexes I and III is supported by most studies of metabolic flux control (Bianchi et al., 2003; Lapuente-Brun et al., 2013) and in recent reviews discussing the results of Blaza et al. (Enríquez, 2016; Lenaz et al., 2016).”

Third paragraph: The best experimental support for the role of supercomplexes in minimizing ROS production was probably in the paper by Maranzana (Maranzana et al. 2013. Antioxid Redox Signal 19, 1469-1480), which is not quoted.

We thank the reviewer for this reference, which is now cited in the fourth paragraph of the Introduction.

Third paragraph: The role of supercomplexes in stabilizing CI is noted in the Introduction, but in a way that may not reflect the current view of this question. The historical origin of this proposal is as follows: in 2004 two groups independently proposed that CI stability was compromised in the absence of CIII (Acin-Perez et al., 2004 & Schägger et al. 2004). Latter, the group of Carlos Moraes demonstrated that ablation of CIV or cyt c, also compromised the stability of CI (Diaz et al. 2006- Mol Cell Biol 26, 4872-4881 & Vempati, et al. 2009. J. Biol. Chem. 284, 4383-4391). More recently it was demonstrated that CI can be assembled functionally in the absence of CIII, CIV or cyt c by recovering the re-oxidation of CoQ (Guarás, 2016. Cell Reports 15, 197-209). This series of reports led to the conclusion that assembly of the individual respiratory complexes can proceed independently and requires neither the presence of other complexes nor interaction with them. In line with this conclusion, kinetic analysis showed that the formation of complexes happened before the formation of supercomplexes (Acín-Peréz et al. 2008. Mol. Cell 32, 529-539). This result does not preclude, however, that incomplete CI or CIII could interact with other complexes and form supercomplexes, as reported by several groups (Marques, I., 2007. Eukaryotic Cell 6, 2391-2405). In 2012 it was proposed that the assembly of the NADH-DH module of CI require that formation of a non-functional respirasome between partially assembled CI with CIII and CIV (Moreno-Lastres et al. 2012. Cell Metab.15, 324-335). This proposal contradicted that of (Acín-Peréz et al. 2008. Mol. Cell 32, 529-539). A very recent reevaluation concluded that CI, CIII and CIV are fully assembled before they associate into supercomplexes, refuting the proposal by Moreno-Lastres (Guerrero-Castillo S, Baertling F, Kownatzki D., Wessels H.J., Arnold S., Brandt U. and Nijtmans L. 2016. Cell Metab, in press).

Reviewer #3 gave this extended explanation because the authors depended heavily on Moreno-Lastres proposal in interpreting their findings; it may be better to be cautious and explore other potential interpretations.

We thank the reviewer for bringing this point to our attention. Our revised Introduction now provides a more detailed and balanced description of the stability of complex I and how it depends on other respiratory complexes:

“Several mitochondrial disorders are associated with impaired respirasome formation. […] When ROS production by RET was inhibited, complex I levels were restored (Guaras et al., 2016).”

We also mention the recent results on the assembly pathway of supercomplexes:

“However, recent complexome profiling studies failed to detect supercomplexes containing immature complex I, suggesting that respirasomes form by association of fully assembled component complexes (Guerrero-Castillo, 2016).”

Results and Discussion:

Subsection “Isolation of mitochondrial supercomplexes solubilized with PCC-a-M”, second paragraph: The authors conclude that PCC-a-M does not allow detection of SC: I1III2 and infer that this SC is not observed because it is preserved as I1III2IV1. The authors should be aware that different detergents generate different micelles containing the SCs and modify their electrophoretic motility. In this sense Figure 1—figure supplement 1 does not help very much, since CIV is not detected in the first dimension in some of the samples (Figure 1A and Figure 1—figure supplement 1A) while it is detected in others (Figure 1—figure supplement 1B), and in the second dimension almost all the CIV seemed to be present as free complex. In addition, the authors find I1III2 by cryo-electron microscopy analysis in sufficient quantity to propose its structure (subsection “Defined protein-protein contacts”, first paragraph). They justify that observation by assuming that I1III2 is an artefact generated in vitro by cleavage of I1III2IV2 (in the second paragraph of the aforementioned subsection). This could be the explanation, but it is also possible that they co-purified both, but their detergent did not resolve them in a gel. In fact, they do not report the observation of free CIV that has to be present if the cleavage happened after purification. Finally, if the authors are interested in identifying properly the different components of the bands obtained with the new detergent they should use immunodetection approaches rather than Coomassie staining. There are good antibodies for that purpose.

The reviewer’s point regarding the possible effects of detergents on resolving membrane protein complexes on a gel is well taken. However, our negative stain EM analysis clearly shows that virtually all particles of the sample purified in PCC-a-M contain complex IV, whereas a substantial portion lacks this complex in the digitonin-solubilized sample (Figure 1—figure supplement 1B). We are therefore confident that our conclusions regarding the presence or absence of complex IV in the gels are correct.

CIV is quite small (~200 kDa) compared to the respirasome (1.7 MDa) and therefore hard to see on cryo-EM grids, especially at a defocus optimized for larger particles, which would make small particles even more difficult to detect. We did not pick particles of this small size for our analysis, which explains the absence of class averages of free CIV.

Figure 3A shows three proteins: CI-B14.7, CIII-Subunit 8, and a third in yellow (CIV) that remains unidentified. Please identify it. Does it participate in the interaction between complexes I and IV or III and IV?

The point of Figure 3 is to demonstrate the interaction between CI and CIII. It shows a segment of the respirasome map with complexes colored in blue (CI), green (CIII) and yellow (CIV), as in Figure 2 (and as indicated in the figure legend). Darker shades of blue and green are used to indicate the subunits involved in CI/CIII interaction in the membrane and on the membrane surface.

Other points of interaction between the complexes are mentioned explicitly in the Results under “Defined protein-protein contacts”. The map section in Figure 3 does show one of these contacts (between CI-ND5 and CIV-COX7C) but we prefer to keep the focus of this figure on the interactions between complex I and III as the most extensive contact region in the respirasome.

Subsection “Defined protein-protein contacts”, first paragraph. The minimum number of expected TMHs for the respirasome I1III1IV1 is 131 but the authors report only 114. Which ones are they unable to detect and why?

Thank you for pointing this out. We simply forgot to count the TMHs of the supernumerary CI subunits. In the revised manuscript this error has been corrected. The total number of TMHs is actually 132 (78 in CI, 26 in the CIII dimer and 28 in CIV). Our map resolves most TMHs in the respirasome, but not all, due to variations in local resolution (see Figure 2—figure supplement 1). In particular, all TMHS from CI are resolved, whereas some in complex III and most in complex IV are not. This concurs with our observation that complex IV is less firmly held in the respirasome than complex III.

The authors report: "Several of the classes obtained (≈ 30% of the initial dataset) have no density for the NADH:dehydrogenase module of CI or even lack the whole matrix arm". Contrary to the argument given for the unexpected observation of I1III2 without IV, they do not consider these new classes as an artefact of the manipulation. Rather they propose that represent intermediates of supercomplex assembly, potentially in agreement with those postulated by Moreno-Lastres et al. (subsection “Conformational and compositional variability”, last paragraph). The model of for the assembly of CI and SCs proposed by Moreno-Lastres has been questioned, however (see above). Moreover, two different groups proposed that CI damage occurred under natural conditions at the NADH:DH module, and that this can be replaced without the elimination of the rest of the complex (Lazarou, et al. 2007. Mol Cell Biol 27, 4228-4237. Dieteren, et al. 2012 J. Biol. Chem 287, 41851-41860). If so, the observed structures could also be consistent with the predicition of Lazarou and Dieteren and might have come from degradation rather than as assembly intermediates. The data in this manuscript cannot determine whether those structures are natural or artefactual and whether they might represent intermediates of synthesis or degradation. The reviewer's suggestion is to present either all the potential alternative explanations or to simply remove the discussion of the nature of this structures. In particular, the suggestion that complex III acts as an assembly factor for CI (subsection “Supercomplex assembly, complex I stability and human health”, first paragraph) is not justified by the results presented nor by most of the recent literature.

The reviewer is correct in that our respirasome maps cannot differentiate between complex I assembly and disassembly intermediates. Strong evidence for two distinct assembly models of the respirasome has been reported and we refer to these models in the revised Introduction. The revised manuscript still discusses our observations in the context of the model proposed by Moreno-Lastres 2012, but we now point out that the alternative model of Guerrero-Castillo (2016) does not support assembly intermediates as those described (subsection “Conformational and compositional variability”, second paragraph). In addition, we now discuss the model for direct exchange of CI subunits that suffer most from oxidative damage, which would obviate the need for de novo synthesis of the whole assembly (in the last paragraph of the aforementioned subsection), as proposed by Lazarou et al. (2007) and Dieteren et al. (2012). We discuss in particular how this model might relate to the 3D class averages we observe.

We explicitly do not discount the possibility that the incomplete classes result from partial disassembly of the respirasome during protein purification or cryo-EM grid preparation. In fact, we already pointed this out in the original manuscript. In the revised manuscript we emphasize this before mentioning any possible assembly, disassembly or recycling intermediates:

“Several of the classes obtained (~30% of the initial dataset) have no density for the NADH:dehydrogenase module of complex I, or even lack the whole matrix arm (Figure 1—figure supplement 2). While these classes might reflect a partial loss of this module during purification or cryo-EM grid preparation, we cannot rule out the possibility that they represent biologically relevant assembly, disassembly or recycling intermediates of the respirasome.”

The authors state that their proposed model of functional asymmetry and electron flow in CIII: "does not imply a partitioning of the quinol pool or substrate channeling". However, they also state: "A quinol diffusing through the lipid bilayer from complex I to complex III would thus encounter the active monomer first. The linear arrangement of complex I, the active complex III monomer and complex IV in the respirasome seems to be particularly favourable for efficient substrate and electron transfer." This is confusing, because the statements appear to be contradictory. On the one hand, the authors propose that the respirasome has a particular defined structure that favors oxidization of quinol not only the by CIII incorporated into the respirasome but also by a defined monomer. In the other, they state that this does not imply a kind of channeling. The reviewer believes that their proposed model entails preferential reduction of the ubiquinol within the respirasome rather than a more stochastic picture.

The statement has been removed and the fact that our structure agrees with quinone channeling is now stated clearly in the revised manuscript:

“The linear arrangement of complex I, the active complex III monomer and complex IV in the respirasome seems to be particularly favourable for efficient substrate and electron transfer. […] The asymmetry observed for complex III in our supercomplex structure is in agreement with these results and provides further support for electron transfer by substrate channeling in the respirasome.”

Subsection “Supercomplex assembly, complex I stability and human health”, second paragraph: The statement that deficiency in CIV does not have strong adverse effect on CI stability is incorrect (Diaz et al. 2006- Mol Cell Biol 26, 4872-4881& Vempati, et al. 2009. J. Biol. Chem. 284, 4383-4391).

The reviewer is correct; CIV deficiencies can induce a secondary loss of CI. The paper cited by us (D’Aurelio 2006) mentions, however, that high mutation levels of CIV are required to induce a secondary loss of CI in human cells. Such high mutation levels are not of practical relevance in our case and we have therefore removed this statement from the revised manuscript.

Comments from reviewer #3 regarding comparison with the other published structures:

One of the two published structures found also two alternative conformations for the respirasome – one similar to the authors' Class 1 and the other more similar to Class 2. Two features characterize the differences between the two conformations: the twist in the position of CIII and the absence of contacts between CIII and CIV. In the Class 2 structure presented here, there is a similar twist in the relative position of CIII, but the authors do not state whether if the interaction between CIII and CIV is also absent. A comment on the similarities and differences between the two classes of structures of the respirasome reported here and published by Sazanov group would be of interest.

The Yang and Sazanov structures agree in the protein-protein contacts. The contacts proposed here are substantially similar but there are discrepancies. In particular, according to the structure proposed here, the interface between CIII and CIV is formed by CIII subunits 1 (UQCR1) and 9 (UQCR10) with COX7A1 of CIV, while Yang and Sazanov both say that CIII UQCR1 and UQCR 11 contact CIV subunit COX7A1. Is there an explanation for the discrepancy?

The Sazanov structure would permit an asymmetric flux of electrons within CIII in the respirasome. The present manuscript also presents an interesting model. It would be of great interest to discuss both, their similarities and their differences.

In the revised manuscript, we have added a comparison of our maps and the recently published structures of porcine and ovine respirasome to the end of the Discussion (subsection “Supercomplex assembly, complex I stability and human health”, third paragraph). A detailed comparison with the porcine map was unfortunately not possible since neither the map nor the model coordinates are available at this point. We would like to call your attention to the following points:

Inspection of the released coordinates for the tight conformation of the ovine respirasome reveal distances between subunit 9 and COX7A1 of 16 Å, which is longer than in our bovine complex and would be incompatible with the interaction that we propose. This could be due to differences of the species, preparation or resolution.

Re-examination of our map in the light of the other published structures revealed that subunit 10 of CIII comes to within 5 Å of the N terminus of COX7A1, and thus also contributes to CIII/CIV interactions. We added this to the list of subunits participating in protein-protein contacts in the respirasome (subsection “Defined protein-protein contacts”, last paragraph).

A comparative analysis of our model with the one from Letts et al. (2016) is provided in the revised Discussion (last paragraph). The models agree on the participation of the central monomer from complex III in quinone oxidation, in our case based on the asymmetry of the iron-sulfur domains, in their case based on the proximity of the outer iron-sulfur domain to complex IV that might restrict its movement. Letts et al. further propose that quinone reduction should take place in the outer monomer. We do not find evidence in any of the structures that can support or contradict this hypothesis.

Articles from eLife are provided here courtesy of eLife Sciences Publications, Ltd

RESOURCES