Structure and Mechanism of Respiratory III-IV Supercomplexes in Bioenergetic Membranes
It is worth noting that supercomplexes composed of cyt. bc1 and CytcO catalyze the same reaction as that catalyzed by quinol oxidases mentioned above, i.e., oxidation of QH2 and reduction of dioxygen to water. However, the energy-conservation efficiency is larger for the supercomplex than for, e.g., the E. coli cyt. bo3 because, in addition to the charge separation and proton pumping by the CytcO part, in the supercomplex there is also a transmembrane charge separation generated by cyt. bc1.
Bacterial heme–copper oxidases consist typically of two to four subunits. The minimum functional unit is composed of subunits I and II, which harbor all four redox-active cofactors that catalyze the reaction in eq 3. Subunits I–III (Cox1–3 in S. cerevisiae, Figure 5A) are often referred to as the “catalytic core” because upon removal of subunit III, many CytcOs lose their activity during turnover, referred to as suicide inactivation (reviewed in ref (131)). The subunit I–III catalytic core is conserved and structurally almost identical in CytcOs from mammals, yeast, and many aerobic bacteria.
On the basis of an analysis of amino acid sequence homology as well as functionally important structural features, e.g., proton pathways (see below and ref (132)), the CytcOs have been classified into three major families named A, B, and C.133,134 Type A includes the mitochondrial as well as the “mitochondrial-like” bacterial cytochromes aa3, e.g., from P. denitrificans, R. sphaeroides, and M. smegmatis. Type B includes e.g. the Thermus (T.) thermophilusba3 CytcO, while type C includes the cbb3 oxidases found, e.g., in R. sphaeroides, R. capsulatus, and P. denitrificans, where a subunit with a diheme cyt. c is the primary electron acceptor instead of CuA.132
The A family CytcOs have two well characterized proton-transfer pathways; the K-pathway named after a conserved Lys (K319 or K340, S. cerevisiae or M. smegmatis numbering, respectively, Figure 5B,C), and the D-pathway named after a conserved Asp at its entrance (D92 or D115 in Figure 5B,C). The A-family is further divided into two subfamilies, A1 and A2. The former is characterized by a subunit I motif “XGHPEVY”, found in, e.g., the mitochondrial CytcOs including that from S. cerevisiae, where “E” is Glu243 in the D proton pathway (Figure 5B), “H” is a ligand of CuB (His241, not shown in Figure 5B), while “Y” is a catalytically active Tyr245 in the catalytic site. The imidazole group of His241 and the phenol group of Tyr245 (Y) are linked by a covalent bond. Similarly, the M. smegmatis CytcO belongs to the A1 subclass. Subclass A2 instead harbors an “YSHPXVY” motif where the Glu is replaced by a Tyr-Ser pair (“YS”) at about the same position in space in the D pathway.
Subunit I in the S. cerevisiae (Cox1) CytcO comprises 12 transmembrane (TM) α-helices. Subunit Cox2 is composed of two TM α-helices and a head domain, which harbors the redox-active CuA site (Figure 5A). Subunit Cox 3 is composed of seven TM α-helices that form a V-shaped cleft, which has been suggested to funnel O2 from the membrane to the catalytic site.129,135 The putative O2 channel in Cox 3 typically harbors three tightly bound lipid molecules, PG, PC, and PE, resolved in crystal structures of CytcO from R. sphaeroides, P. denitrificans, and B. taurus.136 In the S. cerevisiae CytcO, two lipid molecules could be modeled in this cleft.33,34
During turnover of CytcO, electron transfer from cyt. c to the CuA site is followed in time by electron transfer to heme a and the heme a3-CuB catalytic site. Figure 6 illustrates schematically the reaction cycle of the mitochondrial CytcOs. The oxidized state of CytcO is referred to as state O. Electron transfer from reduced cyt. c to the oxidized CytcO results in reduction of first CuB and heme a3, which is associated with uptake of two protons from the membrane n side through the K proton pathway (see Figure 5B) to the catalytic site. Each electron transfer from cyt. c to the catalytic site is associated with proton pumping across the membrane. The two-electron reduced catalytic site binds O2 (state A), which results in breaking the O–O-bond by electron transfer from heme a3 and CuB as well as hydrogen transfer from Tyr245, which forms a radical (state P). In the following reaction steps one electron is transferred to the catalytic site in each of the P → F and F → O transitions. Each of these reduction steps is linked to uptake of two protons from the n side through the D proton pathway, one to the catalytic site and one is pumped across the membrane. The branching point from which the substrate and pumped protons are transferred along different trajectories is located at Glu243.
The structure and function of the K and D proton pathways have been studied in detail in bacterial A1-type CytcOs,129,130,137−150 and their involvement in proton uptake also confirmed for the S. cerevisiae mitochondrial CytcO.151,152 The K pathway starts near Glu82 in subunit II at the membrane n side (Figure 5B). It is connected via a water molecule to Ser256, which is hydrogen-bonded to the conserved Lys319. Proton transfer from the Lys residue requires a conformational change of the side chain toward the catalytic site.135 From the “up-position” the proton is transferred, via a water molecule and Thr316 to Tyr245 at the catalytic site (see Figure 5B).
Residue Asp92 of the D pathway is positioned at the inside of a cleft at the n-side surface of subunit I. The pathway is composed of polar residues that coordinate ∼10 H2O molecules, which span the distance of ∼20 Å from Asp92 to Glu243 (Figure 5B). The maximum rate of proton uptake to the catalytic site, via the D pathway, is ∼104 s–1 at pH 7, and it drops with increasing pH displaying a pKa of 9.4,153 which is attributed to titration of Glu243153 (but, see ref (154)). Replacement of the Asp or Glu residues by their nonprotonatable analogues, Asn or Gln, respectively, result in impaired activity and a complete block of proton uptake.137,142,155−159
Minor structural changes around the orifice of the D pathway influence the proton-uptake kinetics and proton pumping stoichiometry. For example, one-residue changes at Asp92 or in the vicinity of this residue in bacterial and S. cerevisiae CytcOs result in lower proton-pumping stoichiometry or complete uncoupling of proton pumping from the O2-reduction reaction, often without altering the CytcO turnover or proton-uptake rate.151,153,160 Similarly, changes in the surface-exposed loop of subunit I in the R. sphaeroides CytcO, outside of Asp92, yielded modified pH dependence and uncoupling of proton pumping.161 Also, removal of R. sphaeroides subunit III, which has a loop of residues near Asp92 (S. cerevisiae numbering), resulted in a dramatic shift in the pH dependence of the proton-uptake rate162 and allowed proton uptake via alternative surface protonatable groups, other than Asp92163 (the two subunit I and III loops are found just below D92/D115 in Figure 5B,C, but are not shown in the figure). Collectively, these data indicate that moderate alteration of the D pathway near the entry point modulate proton-pumping stoichiometry and result in changes in the pH dependence of the proton-transfer kinetics through the D pathway.164
Interestingly, in the M. smegmatis and C. glutamicum III2IV2 supercomplexes, in addition to the subunit III (subunits CtaE/F) loop, another loop that extends from cytochrome b (QcrB subunit of complex III) covers the orifice of the D pathway44 and presents an alternative route for proton entry into the D pathway, via protonatable groups of the QcrB loop45 (Figure 5C, the subunit III loop is not shown in the figure, it is positioned between Asp115 and the QcrB loop). As outlined above, the D pathway entrance is highly conserved and the proton-uptake kinetics is controlled by an intricate web of interactions between the pathway residues. A modified architecture as a result from supramolecular interactions between complexes III and IV in the C. glutamicum and M. smegmatis III2IV2 supercomplexes suggests that proton uptake by complex IV could be modulated by structural changes in complex III.65
In the mammalian CytcO, a third proton pathway (H pathway) was suggested based on a structural analysis.121,139 In bacterial CytcOs, the equivalent of this pathway is not involved in proton transfer.165 Structural analyses and data from functional studies of structural variants in which putative residues of the H pathway were modified in the mitochondrial S. cerevisiae CytcO do not support a functional role of this pathway.151,152,166 Furthermore, key residues of the suggested H pathway are not present in CytcO from plant mitochondria,32 which suggest that its involvement in proton pumping would have to be restricted to the mammalian CytcOs.
In addition to the three core subunits Cox1–3, the S. cerevisiae CytcO is also composed of nine peripheral subunits called Cox4–9, Cox12, Cox13, and Cox26,33,34 where the latter was identified only recently167,168 (Figure 2A). All of these accessory subunits, except Cox26, have subunit homologues in mammals. Some of these subunits have been suggested to be involved in regulation of the electron transfer and proton pumping activities of the CytcO.169−172 A discussion of the role of all these subunits is beyond the scope of this review, but we briefly discuss those accessory subunits that are relevant in the context of supramolecular interactions with cyt. bc1. A detailed description of all the accessory subunits in S. cerevisiae CytcO is found in ref (173) (see also ref (171)).
Subunit Cox5 is the major interaction partner with cyt. bc1 in the S. cerevisiae supercomplex (Figure 2A). It is homologous to mammalian CoxIV and is expressed as one of two isoforms, called Cox5A or Cox5B, which share 68% sequence identity.173 Expression of the two isoforms depends on the oxygen concentration; the former version is expressed at normoxic conditions (∼200 μM O2), while the latter is expressed at low oxygen concentrations (<0.5 μM).174,175 Early data indicated that the catalytic turnover of CytcO is higher with Cox5B than with Cox5A.176 However, more recent data indicate that the elevated CytcO activity is not simply a result of replacement of Cox5A by Cox5B because a genetic replacement of Cox5A by Cox5B did not yield any differences in the turnover activity nor of the affinity for O2 or cyt. c.177
In the S. cerevisiae CytcO, subunit Cox13 is composed mainly of a single bow-shaped TM α-helix at the periphery of CytcO.34 In the cryo-EM structural model, it interacts with Cox1, Cox3, and Cox12 on the p side and with Cox4 on the n side of the membrane34 (Figure 2A).
The M. smegmatis CytcO core is composed of subunits CtaD (subunit I), CtaC (subunit II), as well as CtaE and CtaF, which together form the equivalent of subunit III. The structure of this subunit I–III core is very similar to that of the canonical CytcO. In addition, the M. smegmatis supercomplex harbors a number of accessory subunits (Figure 2B).43,44 Even though some of these subunits are attached only to the CytcO part of the supercomplex, we consider them being components of the supercomplex rather than of CytcO itself. Furthermore, as already mentioned above, in the M. smegmatis supercomplex subunit QcrB of complex III is extended to interact with complex IV. Figures 2B and 3B show the open and closed positions of the cyt. cc domain (QcrC) in the two halves of the supercomplex.
In addition to the redox-active metal sites, A-type CytcOs harbor a number of nonredox active metal sites (Figure 5A). An Mn2+/Mg2+ (depending on the concentration of the metal in the growth medium) is located near the catalytic site of mammalian and bacterial A-type CytcOs139,178,179 and was also identified in one cryo-EM structure of the S. cerevisiae CytcO.34 In addition, a Ca2+/Na+ site was confirmed in the S. cerevisiae CytcO34 (see also refs (139,179,180)). These metal sites are presumably also present in the actinobacterial supercomplexes.45 Furthermore, a Zn2+ ion is bound in Cox4 of the S. cerevisiae CytcO34 (see also ref (139)). Added Zn2+ also binds near the proton pathways to slow or impair proton uptake.181−185
The bacterial CytcOs are typically monomers. The first crystal structures of the mammalian CytcO revealed a dimer,139 which is consistent with earlier data from functional studies suggesting that formation of the dimer would be functionally relevant.171 As seen in Figure 7, in the mammalian CytcO, the equivalent of subunit Cox12 and Cox13 in S. cerevisiae, i.e., subunits CoxVIb and CoxVIa, are found at the monomer–monomer interface in the crystal structure of the dimeric enzyme (interface subunits are marked in bold text in Figure 7). Here, the CoxVIa subunit adopts a structure different from that of Cox13 in S. cerevisiae.139,186
More recent structural and functional studies showed that the O2-reduction activity of the CytcO monomer was not significantly different from that of the dimer, and only minor structural differences were observed between the monomeric and dimeric forms.187 Furthermore, recent structures of supercomplexes composed of complexes I, III, and IV (sometimes also referred to as respirasomes) from mammals showed that the CytcO bound in these preparations is a monomer38−41 (Figure 7), as also seen for supercomplexes in situ in mammals, yeast, and plants.29
In S. cerevisiae, almost all CytcO is found in supercomplexes.17,72 In variants with only one CytcO (III2IV), the enzyme is obviously a monomer, but also in the III2IV2 variant, the two CytcOs are maximally separated in the supercomplex (Figure 2A). The current data also suggest that the small fraction free CytcO in S. cerevisiae mitochondria is found in monomeric form.17,72,188 Even though a fraction of CytcO dimer was observed upon reconstitution of the S. cerevisiae CytcO in liposomes,189 this observation may be consequence of detergent solubilization of the enzyme prior to reconstitution in a membrane as well as a lipid composition that differs from that of the inner mitochondrial membrane.
The interface surface between cyt. bc1 and CytcO within the S. cerevisiae supercomplex is surprisingly small,24 with a main part of the cyt. bc1–CytcO interactions on the matrix (n) side of the supercomplex where the N-terminal domain of Cox5 binds to Cor134 (Figure 2). In addition, the C-terminal domain of Cox5 on the p side of the membrane interacts with the C terminus of Qcr6 and with the cyt. c1 domain (Figure 2). The first supercomplex structures34,35 were determined with the Cox5A isoform (in ref (34), Cox5B was removed genetically). Many of the residues of Cox5A that are involved in binding to cyt. bc1 in the supercomplex are the same in the two isoforms of Cox5. Accordingly, a recent structural study of supercomplexes composed of CytcO with either Cox5A or Cox5B did not show any isoform-dependent interactions.33
Only minor structural changes result from formation of the supercomplex. The data suggest that the N terminus of the TM α-helix of the Rieske iron–sulfur protein (Rip1) in cyt. bc1 undergoes a conformational change upon interactions with a cardiolipin molecule within the supercomplex.34 However, the authors also noted that this change would not impact the FeS-containing head domain of the iron–sulfur protein,34 i.e., the function of cyt. bc1 is unlikely to be altered as a result of supercomplex formation. Furthermore, the structural comparison of the N terminus of the iron–sulfur protein was made to the crystal structure of cyt. bc1, i.e., any differences in interactions with cardiolipin may also reflect differences in the organization of cyt. bc1 in crystals and in the cryo-EM sample, respectively. The conformation of the other cyt. bc1 subunits that interact with CytcO (mainly Cor1, but also cyt. c1 and Qcr8, see Figure 2A) are not altered by the supramolecular interactions.33,34 Another difference in structure possibly caused by the supramolecular interaction is the configuration of the N-terminal domain of Cox5A. This protein segment may bind ATP, which has been suggested to allosterically regulate the CytcO activity.190 Because upon forming a supercomplex this domain is shifted toward cyt. bc1, the structural difference may be a consequence of binding of CytcO to cyt. bc1 within the supercomplex.34,35 However, because a structure of the S. cerevisiae CytcO alone (i.e., not part of a supercomplex) is not available, the structural comparison of subunit Cox5A was made for the equivalent subunit of the isolated mammalian (bovine heart) CytcO and the S. cerevisiae CytcO in a supercomplex.34 Therefore, the structural difference may reflect that of the equivalent subunits in the different CytcOs.
We also note that the turnover activity of free CytcO is the same as that of CytcO in a supercomplex with cyt. bc1,37 which suggests that the putative structural changes seen upon supercomplex formation are not functionally relevant. In conclusion, because the supramolecular interaction surface is small and any structural differences that may occur upon supercomplex formation are minor,33−35 the activities of cyt. bc1 and CytcO are unlikely to be “regulated” upon formation of the supercomplex.
As indicated above, the monomer–monomer interface in the mammalian dimer139 involves subunits CoxVIa and CoxVIb139 (Figure 7). The equivalent subunits in the S. cerevisiae CytcO, Cox13, and Cox12, respectively, were suggested to define a monomer–monomer interface also in a putative dimer of the S. cerevisiae CytcO.173 Because in the S. cerevisiae supercomplex the cyt. bc1–CytcO interface involves subunit Cox5, subunits Cox12 and Cox13 are exposed on the opposite side of the CytcO (see Figures 2 and 7). Therefore, if a CytcO dimer would be formed in S. cerevisiae by interactions through Cox12 and Cox13, a chain of supercomplexes would form in the membrane. Indeed, such a multisupercomplex structure was suggested by Schägger for yeast and mammalian mitochondria.18 However, to our knowledge, there is no published data in support of such a scenario. Furthermore, Hartley et al. noted that the bow-shaped topology of Cox13 would hinder dimerization of CytcO.34 In addition, the suggested binding of the respiratory supercomplex factor 2 (Rcf2, see below) at Cox13 would probably also prevent CytcO dimerization through interactions via Cox13.33
Figure 7 shows known structures of supercomplexes in which complexes III and IV are in direct contact (see also Table 1), as well as the mammalian complex IV dimer. The orientation of the mitochondrial respiratory complexes in relation to complex IV is shown in the main left-hand side panel, with CytcO subunits that interact with the other complexes indicated in different colors (bold text is used to indicate interactions for each supercomplex). To the right are shown bacterial complex III-IV supercomplexes with known structures. The inset on lower right shows an overlay of all supercomplexes but instead aligned to the complex III2 dimer.
As seen in Figure 7, there is a great variability in the relative orientation of complexes III2 and IV, i.e., the interaction surfaces of these complexes in supercomplexes varies between different organisms. In the mammalian I1III2IV1 supercomplex,38 the surface of the homologous subunits of complex III that interact with complex IV in the S. cerevisiae supercomplex, instead bind to complex I. In this mammalian supercomplex, main interactions with cyt. bc1 occur via CytcO subunit CoxVIIa (Cox7 in S. cerevisiae). The details of the cyt. bc1–CytcO interactions in S. cerevisiae as well as interactions within the mammalian CytcO dimer are discussed in the previous sections.
In the plant supercomplex from V. radiata mitochondria the approximate relative orientation of complexes III2 and IV is similar to that of S. cerevisiae. However, the protein–protein interaction sites differ and the orientation angle differs by 18° (defined by heme bHs in complex III2, and hemes a and a3 in complex IV, Figure 7).32 As with the S. cerevisiae supercomplex, subunit Cox5 (Cox4 in V. radiata mitochondria) faces toward complex III. However, on the matrix side the interactions between Cox5 and Cor1, observed in S. cerevisiae, are absent in V. radiata because the equivalent of Cox5 in the latter is shorter by ∼100 amino acid residues at the N terminus. Instead, the main interactions are found on the cytosolic side between V. radiata Cox4 and Qcr6, which are more extensive in the V. radiata than in the S. cerevisiae mitochondrial supercomplex.32
In the M. smegmatis III2IV2 supercomplex the main III2–IV interactions are mediated via complex IV subunits CtaE and CtaF, which together form the equivalent of CytcO subunit III, and QcrB (cytochrome b) of complex III2, which is also bound to complex IV via the extended QcrB loop on the periplasmic (n) side (Figure 7).43,44
In the structure of the F. johnsoniae supercomplex composed of an alternative complex III and CytcO, interactions are mediated via the CytcO subunit III.47 The authors noted that this subunit III lacks TM α-helices 1 and 2, i.e., consists of five TM α-helices. These five TM α-helices are equivalent to subunit CtaE of the M. smegmatis CytcO, which also interact with complex III2 in this supercomplex. As noted above, in M. smegmatis, the equivalents of TM α-helices 1 and 2 are present and formed by the CtaF subunit. This observation shows that subunit III of CytcO displays a structural variability that may be adopted to accommodate different interaction partners.47
The variability in the interaction surfaces of complexes III and IV most likely excludes a universal structure–function modulation that would be a consequence of III2–IV supercomplex formation in mitochondria. The situation is different for actinobacterial supercomplexes where formation of the III2IV2 supercomplex introduces new architecture to otherwise conserved structural elements, for example, those involved in proton uptake and pumping in complex IV.
Cardiolipin is typically found in membranes that are involved in energy conversion, i.e., that maintain an electrochemical proton gradient.191−193 The phospholipid is unique in having a dimeric structure consisting of two phosphatidyl moieties linked to glycerol and four acyl chains. The pKa values of the two phosphate groups were reported to be different with one pKa being above 8.0, i.e., the cardiolipin headgroup would carry only one negative charge at neutral pH.194 The high-pKa headgroup was suggested to act as a proton trap near enzymes that maintain or utilize electrochemical proton gradients.194 However, results from more recent studies indicate that the two pKas are similar (≤ ∼3) and that cardiolipin carries two negative charges at neutral pH.195,196