Cryo-EM structure and kinetics reveal electron transfer by 2D diffusion of cytochrome c in the yeast III-IV respiratory supercomplex

Significance

In the last steps of food oxidation in living organisms, electrons are transferred to oxygen through the membrane-bound respiratory chain. This electron transfer is mediated by mobile carriers, such as membrane-bound quinone and water-soluble cytochrome c. The latter transfers electrons from respiratory complex III to complex IV. In yeast, these complexes assemble into III₂IV_1/2 supercomplexes, but its role has remained enigmatic. This study establishes a functional role for this supramolecular assembly in the mitochondrial membrane. We used cryo-EM and kinetic studies to show that cytochrome c shuttles electrons by two-dimensional diffusion, sliding along the surface of III₂IV_1/2. The structural arrangement of III₂IV_1/2 supercomplexes suggests a mechanism to regulate cellular respiration.

Abstract

Energy conversion in aerobic organisms involves an electron current from low-potential donors, such as NADH and succinate, to dioxygen through the membrane-bound respiratory chain. Electron transfer is coupled to transmembrane proton transport, which maintains the electrochemical proton gradient used to produce ATP and drive other cellular processes. Electrons are transferred from respiratory complexes III to IV (CIII and CIV) by water-soluble cytochrome (cyt.) c. In Saccharomyces cerevisiae and some other organisms, these complexes assemble into larger CIII₂CIV_1/2 supercomplexes, the functional significance of which has remained enigmatic. In this work, we measured the kinetics of the S. cerevisiae supercomplex cyt. c-mediated QH₂:O₂ oxidoreductase activity under various conditions. The data indicate that the electronic link between CIII and CIV is confined to the surface of the supercomplex. Single-particle electron cryomicroscopy (cryo-EM) structures of the supercomplex with cyt. c show the positively charged cyt. c bound to either CIII or CIV or along a continuum of intermediate positions. Collectively, the structural and kinetic data indicate that cyt. c travels along a negatively charged patch on the supercomplex surface. Thus, rather than enhancing electron transfer rates by decreasing the distance that cyt. c must diffuse in three dimensions, formation of the CIII₂CIV_1/2 supercomplex facilitates electron transfer by two-dimensional (2D) diffusion of cyt. c. This mechanism enables the CIII₂CIV_1/2 supercomplex to increase QH₂:O₂ oxidoreductase activity and suggests a possible regulatory role for supercomplex formation in the respiratory chain.

Aerobic organisms obtain energy by linking oxidation of food to the synthesis of adenosine triphosphate (ATP). An intermediate step in this process is translocation of protons across a membrane by a series of integral membrane proteins collectively known as the respiratory chain. In eukaryotes, the respiratory chain is found in the inner mitochondrial membrane. Transmembrane proton translocation is driven by electron transfer from NADH and succinate to O₂, which renders the mitochondrial matrix more negative (n side) and the intermembrane space more positive (p side). The resulting electrochemical proton gradient is used for ATP production by ATP synthase and for transmembrane transport processes.

In Saccharomyces cerevisiae, NADH donates electrons to dehydrogenases Nde1, Nde2, and Ndi1 (1), which reduce membrane-bound quinone (Q) to quinol (QH₂). Succinate dehydrogenase also contributes to the pool of reduced QH₂ in the membrane. QH₂ diffuses in the hydrophobic core of the lipid bilayer to donate electrons to cytochrome (cyt.) c reductase (also known as cyt. bc₁ or complex III), which forms an obligate homodimer, CIII₂ (reviewed in refs. 2–5). Within CIII, electrons are transferred from QH₂ to cyt. c₁, which is a component of CIII, in a series of reactions known as the proton-motive Q-cycle (Fig. 1A) (3, 5). In brief, binding of QH₂ in a site near the p side, called Q_P (or Q_o), is followed by transfer of one of its electrons, first to an FeS center and then to cyt. c₁. This oxidation of QH₂ is linked to release of its two protons to the p side of the membrane. The second electron from the resulting semiquinone (SQ^•-) in the Q_P site is transferred via hemes b_L and b_H to bound Q in the Q_N (or Q_i) site, reducing it to SQ^•-. Repetition of this sequence of events leads to reduction of the SQ^•- in the Q_N site to QH₂, which abstracts two protons from the n side of the membrane and subsequently equilibrates with the reduced QH₂ pool.

Fig. 1.

Reactions catalyzed by CIII₂ and CIV and possible models for electron transfer between them. (A) Q_N and Q_P indicate the two quinol/quinone-binding sites. The black and blue arrows indicate electron-transfer and proton-transfer reactions, respectively. In CIV, H⁺_p and H⁺_s indicate protons that are pumped or used as substrate for reduction of O₂ to H₂O, respectively. Only reactions in one-half of the CIII₂ dimer are indicated. These reactions occur independently in the two halves of the dimer. Transfer of electrons between CIII₂ and CIV via the 3D diffusion of cyt. c is depicted. (B) Electron transfer from CIII₂ to CIV via 2D diffusion of a cyt. c associated with the CIII₂CIV₂ supercomplex. (C) Electron transfer from CIII₂ to CIV via a bridge formed by more than one bound cyt. c.

Electrons from cyt. c₁ within CIII are transferred to the water-soluble mobile one-electron carrier cyt. c, which resides in the mitochondrial intermembrane space (Fig. 1A). Reduced cyt. c binds to the p side surface of cytochrome c oxidase (also known as cyt. aa₃, or complex IV), where its electron is transferred first to CIV’s dinuclear copper site, Cu_A, then to heme a, and then to the binuclear heme a₃ and Cu_B catalytic site. Transfer of a second electron from a second cyt. c to CIV leads to reduction of both heme a₃ and Cu_B, allowing O₂ to bind to the heme a₃ iron. This O₂ is reduced to water after transfer of four electrons from four cyt. c molecules to the catalytic site. Each electron transfer to the catalytic site is linked to uptake of one proton from the n side of the membrane and pumping of one proton across the membrane from the n side to the p side (Fig. 1A) (reviewed in refs. 6, 7).

Each respiratory complex can function independently of the others, and early studies supported a model in which the complexes diffuse independently in the mitochondrial inner membrane (8) (Fig. 1A). However, more recently, variable fractions of the respiratory complexes have been shown to form larger supercomplexes consisting of two or more components of the respiratory chain (9–13). These supercomplexes can be isolated with preserved enzymatic activity (14, 15), and the overall arrangement of complexes within supercomplexes has been determined in cells from mammals, yeast, and plants (16). In addition, respiratory supercomplexes with different compositions and stoichiometries of components have been isolated using mild detergents, and their high-resolution structures have been determined by single-particle electron cryomicroscopy (cryo-EM) (reviewed in ref. 17).

In S. cerevisiae, essentially all CIVs are part of supercomplexes (18) composed of CIII₂ flanked by either one copy or two copies of CIV (9, 10, 18–26). Recent studies showed only minor structural changes in the individual complexes on association (27–29), which suggests that CIII-CIV binding does not result in functional differences in the components, and that supercomplex formation alters only their proximity.

Electron transfer from CIII to CIV within the supercomplex requires cyt. c (28). The cyt. c docking sites in CIII and CIV are 60 to 70 Å apart, which is too far to allow direct electron transfer through a single stationary cyt. c on the supercomplex surface. Thus, electron transfer between the two complexes must occur via one of three scenarios (Fig. 1 A–C): 1) diffusion of cyt. c between CIII and CIV via the bulk solvent, referred to as three-dimensional (3D) diffusion (8); 2) lateral diffusion of cyt. c along the supercomplex surface (26, 30–33; also see 34), referred to as two-dimensional (2D) diffusion; or 3) two or more cyt. c molecules that bind simultaneously on the supercomplex surface to bridge CIII and CIV. Here we explored these possibilities to address the functional significance of supercomplex formation with combined structural and functional studies of the S. cerevisiae supercomplex with added cyt. c. The kinetic data exclude the possibility that electron transfer between CIII and CIV by cyt. c involves 3D diffusion of reduced cyt. c between its CIII and CIV binding sites. The structural data show an ensemble of states in which cyt. c is bound to CIII, to CIV, or at intermediate positions between the two on the supercomplex surface. Collectively, these results indicate that electron transfer between CIII and CIV occurs along the surface of the supercomplex, suggesting a mechanism to regulate the redox state of the cyt. c pool by altering the CIII:CIV ratio and through association/dissociation of supercomplexes on changing environmental conditions.

Results

Isolation of the Yeast CIII₂CIV_1/2 Supercomplex.

The S. cerevisiae CIII₂CIV_1/2 supercomplex was purified using a FLAG tag on the Cox6 subunit of CIV (35) (SI Appendix, Fig. S1 A and B). Absorption spectroscopy of the preparation (SI Appendix, Fig. S1C and D) gave difference spectra similar to previously reported results (28) and yielded a heme a:b:c ratio of ∼1.7:2:1, which is consistent with a mixture of CIII₂CIV₁ and CIII₂CIV₂ supercomplexes, as observed previously (27–29). Furthermore, the heme b:c ratio of ∼2:1 indicates that the only heme c present was from the bound cyt. c₁ heme of CIII and not from copurifying soluble cyt. c. Association of soluble cyt. c with CIII and CIV is known to depend on the ionic strength of the surrounding solution (36). Consequently, all subsequent experiments were done in buffer with 150 mM KCl at near-physiological ionic strength.

Activity of CIII and CIV.

To probe the catalytic activity of CIII₂ and CIV within the CIII₂CIV_1/2 supercomplex, the activity of each component was measured separately. The supercomplex was maintained in solution by the mild detergent glyco-diosgenin (GDN). The activity of CIII₂ was measured at 90 ± 20 e⁻/s by following the reduction of cyt. c by decylubiquinol (DQH₂) spectrophotometrically while CIV was inhibited by cyanide. The activity of CIV was measured at 450 ± 20 e⁻/s using a Clarke electrode to follow the reduction of oxygen by cyt. c, which was maintained in a reduced state by ascorbate. The activities of free CIII₂ and CIV were then measured after disruption of the supercomplex into CIII₂ and CIV components by incubation with the detergent n-dodecyl β-d-maltoside (DDM) (37) (SI Appendix, Fig. S2). These activities were 60 ± 15 e⁻/s and 370 ± 30 e⁻/s, respectively, showing that the exchange of detergent and dissociation of the supercomplexes had only minor effects on this maximal activity of the individual complexes when cyt. c is at saturating concentrations. The error estimated for all measurements is ±SD for three technical replicates from each of two independent preparations.

Supercomplex Activity.

The QH₂:O₂ oxidoreductase activity of the supercomplex was measured by monitoring the O₂-reduction rate on addition of DQH₂ in the presence of varying amounts of S. cerevisiae cyt. c (Fig. 2A and SI Appendix, Fig. S3). As observed previously, the supercomplex is not able to reduce O₂ with DQH₂ in the absence of exogenous cyt. c (28) (Fig. 2A). Thus, even when CIII₂ and CIV are held in close proximity within a supercomplex, electron transfer between the two components does not occur without cyt. c. At a cyt. c:supercomplex ratio of ∼1 (∼20 nM cyt. c), the turnover rate was 15 ± 1 e⁻/s (Fig. 2B).

Fig. 2.

Catalytic activity. (A) The CIII₂CIV_1/2 supercomplex (SC) QH₂:O₂ oxidoreductase activity at different concentrations of S. cerevisiae or equine cyt. c. (B) The supercomplex QH₂:O₂ oxidoreductase activity at 20 nM cyt. c with intact supercomplex (SC) and dissociated supercomplex after incubation in DDM (SC+DDM). The cyt. c oxidation-O₂ reduction activity was measured for CIV within a supercomplex (CIV in SC), for isolated CIV in GDN (from the CIV peak shown in SI Appendix, Fig. S1A), or after incubation in DDM (CIV in DDM). (C) The QH₂:O₂ oxidoreductase activity of the supercomplex at 50 µM cyt. c with intact supercomplex (SC) and after dissociation of the supercomplex in DDM (SC+DDM). The maximum activities of CIII and CIV are also shown. Error bars are ± SD for three technical replicates from each of two independent preparations (except the 200-nM point in A, which is an average of a total of five replicates). For some points the error bars are smaller than the symbols. The solid lines were generated using Michaelis–Menten equations with K_M values 6 nM and 1.7 µM (yeast cyt. c) or 0.2 µM and 15 µM (equine cyt. c).

As the cyt. c:supercomplex ratio is increased gradually from 1 to 10 (20 to 200 nM cyt. c), the turnover rate increases only slightly, from ∼15 to ∼20 e⁻/s. However, as the cyt. c:supercomplex ratio is increased beyond 10, turnover increases more quickly, reaching a maximum value of ∼60 e⁻/s at a cyt. c:supercomplex ratio of 2,500 (50 µM cyt. c). This maximum rate is only slightly lower than the maximum turnover rate of CIII₂ alone (∼90 e⁻/s) but is much lower than the rate of CIV (∼450 e⁻/s), suggesting that supercomplex activity is limited by the turnover rate of CIII₂ (Fig. 2C). The rate decreases slightly at a cyt. c:supercomplex ratio of 5,000 (100 µM cyt. c), the highest ratio used (Fig. 2A). This decrease may be due to cyt. c binding a “nonproductive” site at CIV (38) that interferes with cyt. c diffusion between the productive binding sites at CIII and CIV.

The low and high cyt. c:supercomplex regimes for the cyt. c titration in Fig. 2A yield apparent K_M values of ≤6 nM and ∼1.7 µM, respectively. These values are distinct from the dissociation constant, K_d, for cyt. c binding, because in the QH₂:O₂ oxidoreductase experiment cyt. c mediates electron transfer between two functionally-independent enzymes. Steady-state turnover of isolated S. cerevisiae CIV with isoform-1 cyt. c yielded biphasic Eadie–Hofstee plots with K_M values of ∼100 nM and ∼30 µM at low and high cyt. c:supercomplex ratios, respectively (39, 40). These K_M values are much larger than the apparent K_M values from Fig. 2A, indicating that the QH₂:O₂ oxidoreductase activity of the supercomplex is not rate-limited by 3D diffusion of cyt. c to CIV.

Measurement of QH₂:O₂ oxidoreductase activity with nonphysiological equine cyt. c yielded apparent K_M values of ∼0.2 µM and ∼15 µM (Fig. 2A), which is consistent with a weaker affinity of the mammalian cyt. c for the S. cerevisiae supercomplex. The larger of these apparent K_M values is similar to the ∼14 µM K_M value obtained previously for equine cyt. c binding to S. cerevisiae CIV alone (40). Furthermore, with an equine cyt. c:supercomplex ratio of 1, the QH₂:O₂ turnover rate for the intact supercomplex was only ∼3 e⁻/s (Fig. 2A), similar to that of the disrupted supercomplex in DDM using S. cerevisiae cyt. c (Fig. 2B). These data suggest that with equine cyt. c, which binds the supercomplex weakly, activity is rate-limited by the binding of cyt. c to CIV via 3D diffusion. The maximum supercomplex activity at an equine cyt. c:supercomplex ratio of 2,500 was ∼30 e⁻/s (Fig. 2A), consistent with an earlier study (28).

The lower apparent yeast cyt. c K_M values for the supercomplex (Fig. 2A) compared with CIV alone suggest that cyt. c binds more strongly to the supercomplex than to the independent CIV and CIII₂. This difference in K_M is consistent with the supercomplex presenting a large binding surface for cyt. c between CIII₂ and CIV. To test this hypothesis, we measured the QH₂:O₂ oxidoreductase activity of a supercomplex preparation after addition of DDM to disrupt the supercomplex into free CIII₂ and CIV (37). With approximately one cyt. c per supercomplex (20 nM of each) the QH₂:O₂ oxidoreductase rate decreased from ∼15 e⁻/s to ∼5 e⁻/s on disruption of the supercomplex (Figs. 2 and 3), showing that an increase in the average distance between CIII₂ and CIV results in a decrease in the turnover rate. Indeed, the supercomplex’s QH₂:O₂ oxidoreductase activity of ∼15 e⁻/s is higher than the cyt. c:O₂ oxidoreductase of ∼3 e⁻/s for CIV alone at the same cyt. c:CIV ratio. In measurements of the supercomplex activity, cyt. c is only partially reduced, while in measurements of isolated CIV activity, it is fully reduced. Consequently, the 3 e⁻/s activity for CIV is an upper limit, and the difference between supercomplex activity and CIV activity is even more pronounced than the experiment indicates. The activity of CIV was the same whether obtained from CIV in GDN (from the CIV peak shown in SI Appendix, Fig. S1A) or from CIV in DDM (Fig. 2B), which shows that DDM does not interfere with cyt. c binding. The QH₂:O₂ oxidoreductase activity of dissociated supercomplex (∼5 e⁻/s, SC + DDM) was slightly larger than the cyt. c:O₂ oxidoreductase activity of CIV (∼3 e⁻/s), presumably due to incomplete dissociation of supercomplexes on addition of DDM, which is shown in SI Appendix, Fig. S2 as a small shoulder at the position of the supercomplex. The fraction of undissociated supercomplexes was estimated to be 10 to 15% from the integrated intensity of the shoulder relative to the peaks attributed to CIII₂ and CIV. The cyt. c:O₂ oxidoreductase activity of CIV within an intact supercomplex (∼7 e⁻/s) was higher than that of isolated CIV (∼3 e⁻/s) at a cyt. c:CIV ratio of 1, which is consistent with the hypothesis that a large cyt. c-binding surface in the supercomplex facilitates cyt. c binding.

Fig. 3.

The QH₂:O₂ oxidoreductase activity of the supercomplex in GDN and DDM. The oxidoreductase rate was determined from the slope of the graph. Measurements were done with two different aliquots from the same supercomplex preparation in 0.01% GDN.

These data indicate that cyt. c diffusion to CIV limits the QH₂:O₂ oxidoreductase activity for freely diffusing CIII₂, CIV, and cyt. c at a cyt. c:supercomplex ratio of 1. In contrast, with intact supercomplex, accumulation of cyt. c on the surface of the supercomplex facilitates electron transfer between CIII and CIV. With a cyt. c:supercomplex ratio of 1 to 10 (20 to 200 nM cyt. c), this mechanism allows for a supercomplex activity of 15 to 20 e⁻/s (Fig. 2A), which is higher than that of CIV alone (Fig. 2B), because at 20 nM cyt. c, the CIV turnover activity is rate-limited by 3D diffusion of cyt. c.

Structure Determination by Cryo-EM.

To investigate the association of cyt. c with the surface of the supercomplex structurally, samples of the two were mixed at a ∼1:12 ratio (11 µM supercomplex with 130 µM cyt. c), and the resulting preparation was frozen on EM grids for analysis by cryo-EM (SI Appendix, Table S1 and Fig. S4). Cryo-EM resulted in 3D maps of supercomplexes consisting of both CIII₂CIV₁ and CIII₂CIV₂ (∼2:1 CIII₂CIV₁:CIII₂CIV₂). As reported previously (27–29), in these structures the CIII₂ dimer in the supercomplex is flanked by one or two CIV monomers (SI Appendix, Fig. S4B). Similar to the earlier studies, no asymmetry between the two halves of the CIII₂CIV₂ structure could be detected, and thus the CIII₂CIV₂ and CIII₂CIV₁ datasets were combined for focused refinement of a CIII₂CIV₁ map to 3.7-Å resolution (Fig. 4 A, i and SI Appendix, Fig. S4 C–E). An atomic model of CIII₂CIV₁ from previous studies (27–29) fit well in the map (Fig. 4 A, ii and SI Appendix, Fig. S5A). Local refinement of the CIII₂ region led to small improvements in resolution for CIII₂, while local refinement of the CIV region led to a large improvement in resolution for CIV from ∼5 Å to 4 Å (SI Appendix, Figs. S6 and S5 B and C). No density was present for Rcf2, which was found in a recent CIII₂CIV_1/2 supercomplex structure (29).

Fig. 4.

Cryo-EM analysis. (A) Cryo-EM map (i) and atomic model (ii) of CIII₂CIV₁ supercomplex. (B) Surface representation of CIII₁CIV₁ supercomplex with density for cyt. c in yellow. (C) Particle images could be separated into populations that show cyt. c bound to CIII₂ (i), in between CIII₂ and CIV (ii), and bound to CIV (iii). (D) An atomic model of yeast cyt. c (Protein Data Bank [PDB] ID code 1YCC) (72) (yellow) was fit into the structure from local refinement of the CIII₂-cyt. c region of the map (i) and compared to the position of cyt. c from a previous structure of yeast CIII₂-cyt. c (PDB ID code 1KYO) (42) (blue) (ii). (E) The atomic model of yeast cyt. c (yellow) was fit into the structure from local refinement of the CIV-cyt. c region of the map (i) and compared to a crystal structure of bovine CIV-cyt. c (PDB ID code 5IY5) (43) (blue) (ii). (F) Coulombic surface potential for the supercomplex. (Scale bars: 50 Å.)

In the CIII₂CIV₁ map, an elongated density corresponding to cyt. c is found along the supercomplex surface, bridging the cyt. c-binding sites in CIII and CIV (Fig. 4B). This density could have arisen due to multiple cyt. c molecules attached simultaneously to each supercomplex or to incoherent averaging of a single cyt. c molecule with variable positions along the supercomplex surface. To distinguish between these possibilities, the extended density was analyzed by principal component analysis, which is designed to detect continuous variability in protein structures (41) (SI Appendix, Fig. S7). The analysis separated particle images into three clusters that were used to generate and refine three separate 3D maps. The three maps show the supercomplex with cyt. c bound at different positions (Fig. 4C). These structures indicate that the elongated density in the consensus structure in Fig. 4B is due to a single cyt. c with a position that varies continuously between CIII and CIV. In the maps, cyt. c is seen bound at CIII (Fig. 4 C, i), positioned between CIII and CIV (Fig. 4 C, ii), or bound at CIV (Fig. 4 C, iii). The map with cyt. c bound to CIII was calculated from 26% of the particle images, and focused refinement of the CIII₂-cyt. c region shows clear density for a single well-defined cyt. c (Fig. 4 C, i and SI Appendix, Fig. S8). This map resembles a crystal structure of yeast CIII with cyt. c bound (42); however, a rigid body fit of cyt. c from the crystal structure into the current map indicates a slight shift in its position (Fig. 4D).

The map showing cyt. c bound at CIV was obtained from 33% of the particle images (Fig. 4 C, iii). Similar to a previous cryo-EM study (29), focused refinement resulted in large improvements in resolution for the CIV region of the map, which allowed docking of a single cyt. c (SI Appendix, Fig. S9). A recent crystal structure of mammalian CIV bound to cyt. c (43) could be fit into the map and closely matched the current density (Fig. 4E), indicating that the distance between cyt. c iron and Cu_A from CIV is similar in yeast and mammals.

The map with cyt. c approximately equidistant between CIII and CIV (Fig. 4 C, ii) was derived from 41% of the particle images. In this intermediate position, the density for cyt. c remains blurred, with diffuse density extending all the way between the cyt. c-binding sites at CIII and CIV. Comparison of the integrated density for cyt. c in this map to the two other maps, each with a well-defined single cyt. c bound at CIII or CIV, indicates a cyt. c occupancy of 1.1:1:1, respectively. Therefore, despite its low resolution, the cyt. c density in the intermediate position is not consistent with multiple cyt. c molecules bound to the supercomplex at the same time. Instead, it indicates that in the intermediate position between CIII and CIV, cyt. c retains positional variability that cannot be resolved by the principal component analysis. This observation suggests that there is an approximately equal probability of cyt. c binding CIII, binding CIV, or occupying the stretch of supercomplex surface between these two sites.

Cyt. c has a net positive charge and a dipole moment (44), with its heme c group positioned with its edge near the positively charged protein surface. The atomic models of cyt. c bound to CIII and CIV (42, 43) show that it receives and donates its electron with its positively charged surface facing its binding sites. The cyt. c positions detected by cryo-EM lie along a negatively charged path on the supercomplex surface (36, 45) (Fig. 4F). Together, these data indicate that supercomplex formation allows association of a single cyt. c with this negatively charged path. Cyt. c can then travel along this path while maintaining its positively-charged surface oriented toward the supercomplex surface, thereby allowing cyt. c to diffuse along the surface between its binding sites on CIII and CIV.

Discussion

Together, our kinetic and structural data indicate that electron transfer between CIII and CIV in the supercomplex is mediated by 2D diffusion of a surface-associated cyt. c. This association is facilitated by a negatively charged path that increases the affinity of the positively charged cyt. c for the supercomplex (33, 46). While two of the three cyt. c positions observed by cryo-EM are consistent with earlier X-ray crystal structures (42, 43), the third position, intermediate between CIII and CIV, is unique to the CIII₂CIV_1/2 supercomplex. The nearly equal occupancy of the three positions presumably reflects the dynamics of the weakly interacting cyt. c that allow it to slide along the surface between the electron donor and acceptor sites. Thus, the association of CIII₂ and CIV to form a supercomplex does not enhance electron transfer by decreasing the distance that cyt. c diffuses in 3D (47), but rather enables a different mechanism for electron transfer with 2D diffusion of cyt. c.

Assuming that the 2D diffusion of cyt. c between its binding sites at CIII and CIV, which are ∼10 nm apart, occurs with a time constant of ∼50 ms (∼20 s⁻¹; Fig. 2A) at cyt. c:supercomplex ratios of 1 to ∼10, yields a 2D diffusion coefficient of 5 × 10⁻¹² cm²/s (L² = 4 D t). This value is ∼100 times smaller than the 2D diffusion coefficient obtained for cyt. c on a mitochondrial membrane surface (32, 48). The difference in coefficients suggests that the electrostatic interactions between cyt. c and the negatively charged supercomplex surface seen by cryo-EM are more specific than those between cyt. c and the average mitochondrial membrane.

At cyt. c:supercomplex ratios of 1 to ∼10, electron transfer between CIII and CIV involves a single associated cyt. c, as is evident from the minimal increase in rate as the concentration of cyt. c is increased in this range (Fig. 2A). In contrast, the increase in supercomplex activity with cyt. c:supercomplex ratios >∼10 indicates involvement of additional cyt. c molecules at higher ratios. The additional cyt. c could either bind transiently at the supercomplex surface to establish a connection via two or more bound cyt. c molecules or mediate electron transfer via 3D diffusion. Unfortunately, cryo-EM of the supercomplex with high concentrations of cyt. c is precluded by background in images due to the excess cyt. c. However, the decrease in supercomplex activity from ∼60 e⁻/s to ∼40 e⁻/s on dissociation of the supercomplex in DDM (Fig. 2C) supports the former explanation. Electron transfer from CIII to CIV via two cyt. c molecules resembles the stable supercomplex from Mycobacterium smegmatis, where a di-heme cyt. cc component of the supercomplex electronically connects CIII and CIV (49, 50). The concentration of cyt. c in the yeast intermembrane space was found to be 100 to 500 µM (8), with a cyt. c:CIV ratio of ∼5 (31, 51). Therefore, in vivo, electron transfer from CIII to CIV would be expected to involve at least one cyt. c associated with the supercomplex surface.

The binding of cyt. c to the supercomplex is in accordance with kinetic studies of S. cerevisiae mitochondria, which show a tightly bound cyt. c for each CIV in the mitochondrial inner membrane (51), as well as steady-state kinetic studies of electron transfer between CIII and CIV (52). This observation is also consistent with the finding that cyt. c copurifies with mouse supercomplexes (14). Furthermore, multiple interactions between cyt. c and CIII detected by NMR and isothermal titration calorimetry have been suggested to facilitate electron transfer to CIV via “sliding” of cyt. c within a supercomplex (46).

The functional role of supercomplexes has been debated extensively in the literature (12, 17, 34, 53–55). As is evident from the present study, changes in turnover activity of individual respiratory complexes on the formation of supercomplexes are too small to result in functionally relevant changes in the electron flux through the respiratory chain (34, 53). Furthermore, the interaction surfaces of the supercomplex components in different organisms are highly variable, suggesting that association of the components does not result in specific modulation of function. Thus, the only common feature of supercomplexes is the decreased distance between their components.

The 2D diffusion of cyt. c resembles the “substrate channeling” model of respiratory chains, in which cyt. c diffusion is restricted to the space surrounding the supercomplex. This model has been criticized based on the finding that cyt. c diffusion in S. cerevisiae is unrestricted (51); however, the scenario suggested by our data is that weak electrostatic interactions between the positively charged cyt. c and the negatively charged supercomplex surface lead to the surface association of cyt. c, which remains in equilibrium with the cyt. c pool (8, 51).

In addition to differences in cyt. c diffusion rates, 2D and 3D diffusion of cyt. c between CIII and CIV would lead to differences in the redox state of the cyt. c pool. With 2D diffusion, the redox state of the bulk cyt. c pool is determined by the equilibrium constant for exchange between the surface-associated cyt. c and the bulk cyt. c pool. For 3D diffusion, the redox state of the bulk cyt. c depends only on the relative turnover rates of CIII and CIV. Consequently, formation of supercomplexes should perturb the reduced:oxidized cyt. c ratio of the pool. Because cyt. c is involved in an intricate web of redox interactions (56, 57), regulated formation or dissociation of supercomplexes on changing environmental conditions may result in triggering of cellular signaling pathways.

Materials and Methods

Strain and Cell Growth.

The S. cerevisiae strain BY 4741 with a FLAG tag on CIV subunit Cox6 was used (35). Yeast cultures were grown in YPG (2% peptone, 1% yeast extract, and 2% glycerol) at 30 °C while shaking at 180 rpm.

Preparation of Mitochondrial Membranes.

Cells were harvested by centrifugation at 6,500 × g (5 min, 4 °C), washed with 50 mM potassium phosphate (KPi) buffer at pH 7, and then resuspended in 650 mM mannitol, 50 mM KPi buffer, and 5 mM EDTA, pH 7.4. Cells were lysed by mechanical disruption (Constant Systems cell disrupter) at a pressure of 35 kpsi, and cell debris was removed by centrifugation at 5,500 × g (20 min, 4 °C). The supernatant was then centrifuged at 120,000 × g for 1 h at 4 °C. The pellet was resuspended and homogenized in 100 mM KCl, 50 mM KP_i buffer, and 5 mM EDTA, pH 7.4; centrifuged at 120,000 × g for 30 min at 4 °C; and washed twice with 50 mM KPi buffer and 5 mM EDTA, pH 7.4 and centrifuged as in the previous step. Finally, the pellet was homogenized in 50 mM KPi buffer, pH 7.4; frozen in liquid nitrogen; and stored at −80 °C.

Isolation of Supercomplexes.

Membrane fragments were diluted to a total protein concentration of 2 mg/mL in 100 mM KCl, 50 mM KPi buffer, pH 7.4, and 0.5% (wt/vol) GDN (GDN101; Anatrace) and solubilized overnight at 4 °C. The membrane lysate was centrifuged at 140,000 × g for 90 min at 4 °C. The cleared lysate was diluted to a final GDN concentration of 0.2% and concentrated by centrifugation to roughly 70 mL with a 100-kDa molecular weight cutoff concentrator (Merck Millipore), followed by incubation with anti-FLAG M2 resin (Sigma-Aldrich) for 2 h at 4 °C. The anti-FLAG M2 resin with bound protein was washed with 20 mL of wash buffer (150 mM KCl, 20 mM KPi buffer pH 7.4, and 0.01% GDN) and eluted with 0.1 mg/mL FLAG peptide (Sigma-Aldrich), 150 mM KCl, 20 mM KPi buffer pH 7.4, and 0.01% GDN. Eluted supercomplex was concentrated with a 100-kDa molecular weight cutoff concentrator (Merck Millipore) and further purified by size exclusion chromatography with an Äkta Pure M25 (GE Healthcare) operated at 4 °C with UV detection at 280 nm and 415 nm. The sample was loaded on a Superose 6 Increase 10/300 GL column (GE Healthcare) equilibrated with 150 mM KCl, 20 mM KPi buffer pH 7.4, and 0.01% GDN. Fractions of 250 μL were collected, and fractions containing both CIII and CIV were pooled and concentrated as described above.

UV-Visible Difference Spectroscopy.

Optical absorption spectra were recorded with a Cary 100 UV-VIS spectrophotometer (Agilent Technologies). A small aliquot of sodium dithionite was used as a reducing agent to obtain a reduced minus oxidized difference spectrum. Peaks were fitted for the different heme groups, heme c₁ (554 nm), hemes b_H and b_L (562 nm), and hemes a and a₃ (603 nm), and their concentrations were calculated using the following difference absorption coefficients: Δε₅₅₄ = 21 mM⁻¹cm⁻¹ (58), Δε₅₆₂ = 51.2 mM⁻¹cm⁻¹ (59), and Δε₆₀₃ = 25 mM⁻¹cm⁻¹ (60).

Activity Measurements.

Decylubiquinone (Sigma-Aldrich) was dissolved in 99.8% ethanol and reduced with several crystals of sodium borohydride (NaBH₄; Sigma-Aldrich) to obtain DQH₂. When the solution was clear, 10 to 20 µL of 5 M HCl (depending on the amount of NaBH₄ used for reduction) was added. The sample was then centrifuged for 10 min at 10,000 × g, and the supernatant containing DQH₂ was collected. We found that adding DQH₂ to buffered reaction mixtures did not affect the pH of the mixtures.

The activity of CIII was measured by following absorbance changes as a function of time at 550 nm (reduction of cyt. c) with a Cary 100 UV-VIS spectrophotometer. Baseline absorbance was first recorded with 50 µM oxidized cyt. c from S. cerevisiae (Sigma-Aldrich), CIII₂CIV_1/2 supercomplex (equivalent of 20 nM CIII-CIV), and 2 mM KCN (to block cyt. c oxidation by CIV) in 150 mM KCl, 20 mM KPi buffer pH 7.4, and 0.01% GDN. The reaction was initiated by addition of 100 µM DQH₂ (from a 20 mM solution in ethanol). Control experiments showed that the reduction rate of cyt. c by DQH₂ without supercomplex was negligible.

The activity of CIV was measured by monitoring the oxygen-reduction rate with a Clark-type oxygen electrode (Oxygraph; Hansatech) operated at 25 °C. A baseline oxygen concentration was first recorded with 50 µM oxidized cyt. c from S. cerevisiae, 10 mM ascorbate, 100 µM N,N,N′,N′-tetramethyl-p-phenylenediamine (TMPD) in 150 mM KCl, 20 mM KPi buffer pH 7.4, and 0.01% GDN. The reaction was initiated by addition of 10 nM supercomplex. The activity of CIV (10 nM) was also measured with CIV isolated from the CIV peak shown in SI Appendix, Fig. S1A in either 0.01% GDN or 0.05% DDM.

To measure the activity of dissociated complexes CIII₂ and CIV, DDM was added to a concentration of 0.5% in a sample containing the supercomplex (CIII₂CIV_1/2) in GDN101. After a 20-min incubation, the sample was diluted in the oxygraph chamber to yield a final concentration of 0.05% DDM.

The QH₂:O₂ oxidoreductase activity of the supercomplex was measured by monitoring the oxygen-reduction rate using a Clark-type oxygen electrode operated at 25 °C. A baseline oxygen concentration was first recorded with an equivalent of 20 nM CIII-CIV supercomplex, cyt. c from equine heart or S. cerevisiae (the cyt. c concentration was varied in the range 10 nM to 100 µM) in 150 mM KCl, 20 mM KPi buffer pH 7.4, and 0.01% GDN. The reaction was initiated by the addition of 100 µM DQH₂.

Gel Electrophoresis.

Blue native (BN) polyacrylamide gel electrophoresis (PAGE) was performed with precast gel, NativePAGE 4 to 16% Bis-Tris (Invitrogen), according to the manufacturer´s instructions using Native PAGE 1× running buffer as anode buffer and NativePAGE 1× running buffer supplemented with NativePAGE 1× cathode buffer additive as a cathode buffer (Invitrogen). The gel was run at 4 °C for 60 min at 150 V, before exchanging the cathode buffer to the anode buffer and running for an additional 40 min at 250 V. The gel was then stained with Coomassie Brilliant Blue.

Grid Preparation and Cryo-EM.

Purified supercomplexes (3 µL) at a concentration of 8 mg mL⁻¹, supplemented with S. cerevisiae cyt. c (Sigma-Aldrich; aerobically grown S. cerevisiae contains 95% isoform-1 cyt. c) (61) at a ∼1:12 molar ratio, was applied to homemade nanofabricated holey gold grids (62–64) that had been glow-discharged in air (120 s, 20 mA using PELCO easiGlow). Grids were blotted for 3 s at 4 °C and 100% humidity, followed by rapid freezing in liquid ethane with a Vitrobot Mark IV (Thermo Fisher Scientific). Cryo-EM data were collected using a Titan Krios G3 electron microscope (Thermo Fisher Scientific) operated at 300 kV equipped with a prototype Falcon 4 direct detector device camera. Automated data collection was done with the EPU software package. An initial dataset of 5,690 movies, each consisting of 30 exposure fractions, was collected at a nominal magnification of 75,000×, corresponding to a calibrated pixel size of 1.03 Å. The camera exposure rate and the total exposure of the specimen were 4.6 e⁻/pixel/s and ∼42 e⁻/Ų, respectively (SI Appendix, Table S1).

Cryo-EM Image Processing.

All image analysis was performed with cryoSPARC v2 (65). Movies were aligned with MotionCor2 (66), and contrast transfer function (CTF) parameters were estimated in patches with a 7 × 7 grid. The dataset was manually curated to remove movies with devitrification, large cracks in the ice, or poor CTF fit parameters, which reduced the dataset size to 5,147 movies. Templates for particle selection were generated by 2D classification of manually selected particles, leading to 423,969 selected particles. After particle selection, images were corrected for local motion (67) and extracted in 380 × 380-pixel boxes (SI Appendix, Table S1). Extracted particle images were cleaned with six rounds of ab initio 3D classification and heterogeneous refinement, taking only the classes corresponding to CIII₂CIV₁ and CIII₂CIV₂ after each round. This procedure further reduced the size of the dataset to 73,542 particle images, with 49,273 and 24,269 particle images belonging to the CIII₂CIV₁ and CIII₂CIV₂ classes, respectively. These two classes were combined and used to refine a map of CIII₂CIV₁ with nonuniform refinement (68) and then analyzed with 3D variability analysis (3DVA) (69), with a mask over the region of the map corresponding to cyt. c. Cluster analysis was used to generate three structures that served as initial references in a heterogeneous refinement. The three populations resulting from the heterogeneous refinement corresponded to the CIII-bound, CIV-bound, and in-between populations (SI Appendix, Fig. S7 shows the complete workflow). The maps were locally filtered for presentation in figures. Particle images for the CIII-bound and CIV-bound populations were individually subjected to particle subtraction and local nonuniform refinement with masks over the cyt. c-CIII₂ and cyt. c-CIV regions, respectively. Rigid body fitting of cyt. c was performed with UCSF Chimera (70) based on existing structures (42, 43). Occupancy calculations were performed with custom Python programs (https://github.com/justinditrani). Maps were put on the same grayscale with diffmap (71), and occupancy was estimated by integrating the values of voxels within the mask used for 3DVA (SI Appendix, Fig. S7) for the different cyt. c positions and dividing this value by the number of voxels in the mask.

Data Availability

The cryo-EM data have been deposited in the Electron Microscopy Data Bank (entries EMD-23414 and EMD-23416-23423).