Oxygen Activation and Energy Conservation by Cytochrome c Oxidase.

This review focuses on the type A cytochrome c oxidases (C cO), which are found in all mitochondria and also in several aerobic bacteria. C cO catalyzes the respiratory reduction of dioxygen (O2) to water by an intriguing mechanism, the details of which are fairly well understood today as a result of research for over four decades. Perhaps even more intriguingly, the membrane-bound C cO couples the O2 reduction chemistry to translocation of protons across the membrane, thus contributing to generation of the electrochemical proton gradient that is used to drive the synthesis of ATP as catalyzed by the rotary ATP synthase in the same membrane. After reviewing the structure of the core subunits of C cO, the active site, and the transfer paths of electrons, protons, oxygen, and water, we describe the states of the catalytic cycle and point out the few remaining uncertainties. Finally, we discuss the mechanism of proton translocation and the controversies in that area that still prevail.

Introduction

Dioxygen (O2) is a key constituent of the Earth’s atmosphere, oceans, and lakes and essential for the sustenance of all higher forms of life on the planet. The major source of O2 is the photosynthetic water-splitting reaction of cyanobacteria and green plants. This intriguing reaction is catalyzed by a (Mn)4Ca site, the reactions of which are coupled to a photosynthetic reaction center composed of chlorophyll molecules and accompanying tyrosine and quinone species.[1−3] O2 is thus formed as a side product of oxygenic photosynthesis. Its high oxidizing power led to a fundamental event in the evolution of life on Earth, with development of organisms capable of making use of the thermodynamic oxidation power of O2 in harnessing a maximum amount of energy from the oxidation of the hydrocarbon foodstuffs generated via photosynthetic assimilation of CO2. More than 90% of this O2 consumption by living organisms is catalyzed by a superfamily of enzymes, the so-called heme–copper oxidases, which may be divided into three subclasses, A, B, and C.[4] They are all membrane proteins, and the oxidases of type A include the cytochrome c oxidases (CcO) in mitochondria and in some bacteria such as Paracoccus dentrificans and Rhodobacter sphaeroides (which are the most studied) and which are the subjects of this review. The A-type oxidases further include some quinol oxidases, notably cytochrome bo3 of Escherichia coli, which has also been extensively studied.[5−7] The B- and C-type heme–copper oxidases[4] are found exclusively in bacteria and archaea and will be mentioned here only for comparative purposes. Finally, the NO reductases are also structurally related to the heme–copper oxidases and could have been their evolutionary origin.[8,9]

All heme–copper oxidases have been shown to generate a proton electrochemical gradient (protonmotive force, pmf) across the inner mitochondrial or the bacterial cell membrane and to do so by two different mechanistic means (which are intimately linked to one another). The first is due to the charge and proton separation resulting from the fact that the electrons from cytochrome c on the positively charged side of the membrane are transferred to the active site of O2 reduction within the membrane, whereas the protons required to complete the chemistry of O2 reduction to water derive from the opposite negatively charged side (Figure ). In addition to this vectorial arrangement of the oxygen reduction chemistry, first suggested by Peter Mitchell,[10] all heme–copper oxidases tested to date pump up to one proton per transferred electron across the membrane from the to the side[11] (see Figure ). Whereas the proton pump of the A-type cytochrome c oxidases, covered in this review, appears to be highly efficient with a pumping stoichiometry close to unity even at high values of the opposing protonmotive force,[11,12] the heme–copper oxidases of types B and C seem far less efficient,[13,14] and the distantly related NO-reductases do not pump protons at all.[15,16]

Figure 1

Thirteen-subunit A-type CcO is shown with subunits I, II, and III in blue, red, and green, respectively. Ten nuclear-coded accessory subunits are shown with transparent ribbon representation. Lipid bilayer boundaries (dotted lines) and electron, proton, and oxygen paths (arrows) are also marked. Low-spin heme (yellow), high-spin heme (orange), and copper atoms (purple) are displayed. Figure was prepared with VMD[17] software.

General Thermodynamics

In solution, cytochrome c oxidase catalyzes the overall reaction

The O2/H2O acceptor redox couple has a midpoint redox potential at pH = 7 (Em,7) of 815 mV. Cytochrome c is the donor of the four electrons, and mammalian cytochrome c and many bacterial c-type cytochromes have an Em,7 of ca. 250 mV.[18] Hence, under standard conditions the free energy change (per electron) for reaction is 565 meV (∼13 kcal/mol). However, standard conditions imply an oxygen fugacity of 1 atm, which corresponds to about to 1.2 mM O2 at 25 °C. This is a considerably higher concentration than in air-saturated water, which is 0.258 mM at 25 °C, and very much higher than typical O2 concentrations in mammalian tissues, which vary between 0.005 and 0.025 mM.[19] The actual redox potential (Eh, relative to the hydrogen electrode) of the O2/H2O couple will be ∼15 mV lower than the Em,7 of 815 mV for each 10-fold decrease of [O2] below 1.2 mM, as the Nernst factor is ∼60 mV/4 for a four-electron transfer reaction. A value between 785 and 800 mV thus seems reasonable for physiological conditions, and we choose 800 mV here for simplicity.

The average driving force for reaction is hence ca. 800–250 = 550 meV or ∼12.7 kcal/mol per electron, i.e., ∼51 kcal/mol for the full four-electron reaction.

When incorporated in a membrane, cytochrome oxidase catalyzes the reactionwhich distinguishes “chemical” protons (subscript c; consumed in water formation; see eq ) from pumped protons (subscript p), which are translocated (“pumped”) from the negatively charged side of the membrane (subscript ) to the positively charged side (subscript ). Four protons are taken up on the side of the membrane and transported to the active site within the membrane, where they are consumed in forming water from reduced dioxygen. This is made possible by transfer of four electrons to the active site from cytochrome c on the side of the membrane. In addition, remarkably, four protons are pumped across the membrane for each O2 molecule reduced. Thus, a total of 8 electrical charges are translocated across the membrane per O2 molecule reduced (see Figure ).

The overall protonmotive activity of cytochrome oxidase is equivalent of pumping two protons across the membrane per transferred electron (eq ) even though only one proton per electron is released on the side. The uptake of electrons and protons for the catalytic chemistry of cytochrome oxidase from opposite sides of the membrane is thermodynamically equivalent to pumping of one proton across the membrane per electron, and the proton pump function contributes with a second proton. Protons released on the side do not usually contribute significantly to the ΔpH term of the pmf since the aqueous phase on the side is usually well buffered and/or very large in volume compared to the side. The ΔpH term of the pmf is hence almost entirely due to the consumption of protons on the side of the membrane.

In mitochondria, the proton electrochemical gradient (pmf) is some 200–220 mV in the so-called State 4, where lack of ADP limits the rate of respiration and the pmf is maximal. The pmf decreases to ca. 150 mV during active ATP synthesis when there is an excess of ADP and inorganic phosphate (State 3).[20] At the thermodynamic stoichiometry of 2 H+/e– (see above) cytochrome c oxidase would thus require a minimum redox driving force of 400–440 meV to generate the maximum pmf. The mean driving force of ca. 550 meV from the oxidoreduction chemistry (see above) is hence sufficient and would point at an energy transduction efficiency of ca. 70–80% at maximum load. Careful measurements of the ATP/O (ATP/2e–) ratio in mitochondria respiring only via cytochrome c oxidase, for example, by Hinkle[20] and Chamalaun and Tager,[21] yielded values of 0.98 and 0.94, respectively. From these values and the H+/ATP ratio of 3.67 in animal mitochondria (2.67 H+/ATP based on the structure of the c ring of the ATP synthase,[22] plus 1 H+/ATP due to the coupling of ADP/ATP exchange and transport of inorganic phosphate to translocation of 1 H+),[23] one may deduce the effective H+/2e– ratio of proton translocation by mitochondrial cytochrome c oxidase on the basis ofwhich yields 3.45–3.60, indicating a stoichiometric efficiency of 86–90% under the conditions of ATP synthesis, i.e., at a less than maximum load from the protonmotive force (ca. 150 mV).

Overall Structure of the Core Subunits

Mitochondrial CcO contains three core subunits (named subunits I, II, and III) that are encoded in mitochondrial DNA (mtDNA). The homologues of these three subunits are also found in all bacterial heme–copper oxidases (HCO) of type A, which apart from CcO enzymes also comprise quinol oxidases such as cytochrome bo3 from E. coli.[4,8,9] In addition, mitochondrial CcOs contain up to 10 nuclear encoded “accessory” subunits (Figure ), whose presumably regulatory functions are still poorly understood[24,25] and which are not reviewed here. HCOs of type B and C, found in bacteria and archaea (though no archaeal C-type oxidases are known), lack subunit III, which is thus unique for the type A HCOs. Variations among B and C enzymes also exist with regard to subunit II, which is in many cases replaced by heme c-containing proteins.[4] However, a homologue of subunit I is conserved in all HCOs and even in the distantly related NO reductases.[4,9] This review focuses on the CcO enzymes of type A, which includes mitochondrial cytochrome c oxidase; B- and C-type HCOs (and the NORs) will be mentioned only for comparative purposes. All amino acid numbering corresponds to type A CcO from Bos taurus mitochondria, unless mentioned otherwise.

Subunit I

The cylindrically shaped catalytic subunit I of CcO has a highly conserved characteristic fold found in all HCOs. The 12 transmembrane helices are slightly tilted with respect to the membrane normal and are arranged in three clusters with a quasi-3-fold symmetry[26] (Figure A). Each of the three helical clusters house a hydrophilic pore, shown by a cyan-colored circle in Figure A, which contains polar amino acid residues and water molecules. Two of the pores are known to transport protons required for oxygen reduction and proton pumping at least in the bacterial A-type oxidases, whereas the function of the third pore domain is controversial. It includes the so-called H channel that has been claimed to be the exclusive proton pump pathway in the mammalian mitochondrial oxidase (see section ).

Figure 2

(A) View of the catalytic subunit I from the side of the membrane. Three helix clusters III–VI, VII–X, and XI, XII, I and II are shown in blue, green, and red, respectively. Two clusters (green and red) hold the redox-active centers. Three pore regions (filled circles) are found in each of the three clusters. (B) Side view of the cylindrically shaped catalytic subunit. TM helices are shown as ribbons, colored according to residue polarity (polar, green; acidic, red; basic, blue; hydrophobic, white). Hemes (yellow) and CuB (orange), buried in the catalytic subunit, are also displayed.

Two of the helical clusters (red and green, Figure A) hold the buried redox centers, the low-spin heme, and the high-spin heme plus CuB, which are all ligated by histidine residues from subunit I (Figure A and 3B). These six histidine ligands are fully conserved in the HCO superfamily, and their spread along the entire sequence of subunit I has played an important role in earlier sequence alignment and homology modeling studies.[4,27]

Figure 3

(A) Top view (from the side of the membrane) showing histidine ligands of heme cofactors. Two TM helices in subunit I (II and X) that carry conserved histidine residues are marked in red. (B) Side view showing the histidine ligands of the CuB center. Two tandem histidines (H290 and H291) originate from helix VII, whereas the His240-Tyr244 cross-link is from helix VI. As before, hemes and CuB are shown in yellow and orange.

The majority of subunit I secondary structure comprises transmembrane (TM) helices that face the lipid bilayer or TM domains of noncatalytic accessory subunits. These interactions are predominantly nonpolar in nature, whereas the loop regions that extend into the positively charged and negatively charged phases on both sides of the membrane comprise charged or polar amino acid residues (Figure B).

Subunits II and III

Subunit II consists of three major structural components, two TM helices and one β-sheet cluster consisting of 10 sheets (Figure A). The two TM helices make extensive contacts to the TM helices VIII and IX of subunit I, whereas the β-sheet cluster on the side of the membrane holds the dimetallic CuA center, which is the first electron acceptor from cytochrome c (Figure A, inset). The role of subunit II in electron transfer to the catalytic subunit is well known. Additionally, due to the presence of a conserved acidic residue (Glu62 in B. taurus) in the second TM helix of subunit II, its role in proton uptake has also been implicated[7,28,29] (see section ). The subunit I/II interface is partly bridged by a magnesium ion, which is ligated by conserved residues from both subunits but which is absent from the type A quinol oxidases[5] as well as from the HCOs of type B and C.

Figure 4

(A) Structure of subunit II. (Inset) CuA center and its ligand sphere. (B) Structure of subunit III. α-Helices and β-sheets are shown in purple and yellow, respectively. Amino acid residues are marked with their one letter codes and numbers. Phosphatidylglycerol (PG) and phosphatidylethanolamine (PE) lipid molecules bound to subunit III are displayed in licorice representation.

Subunit III is highly conserved but found exclusively in the A-type oxidases. It consists of 7 TM helices divided into two clusters of 2 and 5 helices each, forming a V-shaped structure (Figure B), except for the case of the quinol oxidase bo3. In this latter case, two of the TM helices that are “normally” parts of subunit III are linked with subunit I instead but attaining positions in the 3D structure very similar to the case of the cytochrome c oxidases of family A.[5]

Subunit III makes extensive hydrophobic contacts with subunit I. It does not contain any redox-active cofactor and plays no direct role in redox-coupled proton pumping.[30] However, it provides stabilization to the two-subunit enzyme by enhancing the number of turnovers by preventing suicide inactivation.[31−33] As shown by the extensive work of Hosler et al.,[31−33] the suicide inactivation of CcO is enhanced under conditions where the local proton activity near the BNC is low. Under such conditions of limited H+ access CuB may acquire two hydoxide ligands and leave the structure as copper hydroxide. The suicide inhibition has indeed been shown to be caused by depletion of CuB from the enzyme.[33] One role of subunit III is presumably to prevent this effect by enabling fast H+ uptake to the BNC via the D pathway, even at relatively high values of pH of the phase (see below).

The architecture of subunit III is also unique in the sense that it contains three tightly bound lipid molecules, stabilized by polar and nonpolar interactions. The lipid-binding site region has been proposed to play an important role in O2 channeling (see section ),[34] whereas the charged head groups of the tightly bound lipids may enhance the rate of proton transfer by perturbing the pKa of Asp91, the proton uptake site of the proton-conducting D channel.[35] Like subunit III, the D channel is also unique for the A-type HCOs.

Structure of the Active Site

The catalytic subunit I of CcO houses the active site for oxygen reduction, the so-called binuclear center (BNC). The key components of the BNC are two magnetically coupled redox-active metal centers, a high-spin heme (a3) and a copper ion (CuB). The high-spin heme is bound to the protein matrix by means of covalent and noncovalent interactions. On its proximal side, the heme iron is ligated by a highly conserved histidine residue (H376) from helix X (Figure A). On the distal side, between the iron and the CuB, small diatomic ligands such as CO, CN–, and the physiological substrate O2 bind axially to the Fe (see below). The two propionate groups of the high-spin heme interact electrostatically with titratable amino acid residues from subunit I (Figure ), whereas the heme macrocycle is buried between nonpolar residues. The strong electrostatic interactions between the negatively charged propionate groups and the conserved amino acids are known to be important for the proton-pumping function of the enzyme.[36−40] In between the two propionates is a cluster of crystallographically visible water molecules (Figure ).

Figure 5

Proton-pumping elements in HCO of type A. The D channel of proton transfer is displayed, which comprises the proton-uptake site (D91), the asparagine gate (circled with purple dotted line), the serine zone, and a highly conserved acidic residue, Glu-242. Heme propionate region is highly polar and consists of a water cluster (circled with red dotted line), two arginines, and one acidic residue, Asp-364. Water molecules (yellow), hemes (blue, left, heme a; right, heme a3), CuB (orange), Mg ion (brown), and subunit I (green transparent ribbons) are also shown. Nonpolar cavity where water produced at the active site may be released is shown as a pink “cloud”.

The high-spin heme sits next to the redox-active CuB center, with the two metals (Fe and Cu) only ∼5 Å apart (see Table ). The CuB center has a unique structural arrangement in which the copper ion is ligated by two vicinal histidines (H290 and H291 from helix VII of subunit I) and by another histidine (H240) which is covalently linked to a functionally critical tyrosine (Y244). It is important to realize that this tyrosine is part of the unique his-tyr ligand of CuB (Figure B). Both this histidine and the tyrosine are part of a highly conserved sequence motif in helix VI of subunit I of the A-type oxidases, -GHPEVY-. High-resolution crystal structures of bovine oxidase show that the three histidine nitrogens ligating the copper atom are nearly planar forming a T-shaped structure,[26,41,42] an arrangement preferred by reduced, cuprous copper (Cu[I]). In higher oxidation states, the copper center apparently loses planarity, with a water or a hydroxyl ion positioned as the fourth ligand.[43,44] Recent biosynthetic and theoretical work has provided explanations for why the copper is preferred over iron in O2 activation and reduction.[45] The crystal structure data of the fully oxidized enzyme shows electron density between the two metals (Fe and Cu), suggested to be a bound peroxide.[46,47] However, an unambiguous assignment cannot be made, and the exact identity of the ligand structure between Fe and Cu in the fully oxidized state of the enzyme remains under debate.[48]

Table 1

Metal–Metal and Edge–Edge Distances (in Angstroms) between the Redox Centers of Enzymes from Different HCO Subfamilies

subfamily	PDB id	organisma	distanceb	distancec	distanced
A	5B1A	Bt	19.41 (11.82)	13.20 (6.77)	4.83
A	2GSM	Rs	19.36 (11.82)	13.22 (6.95)	4.89
A	3HB3	Pd	19.45 (12.06)	13.27 (6.73)	4.62
A (quinol oxidase)	1FFT	Ec		13.69 (7.79)	5.30
A (caa3 type)	2YEV	Tt	19.47 (11.99)	13.58 (7.11)	4.85
B	3S8F	Tt	19.07 (11.97)	13.70 (7.36)	4.87
C	3MK7	Ps	19.75 (11.65)	13.05 (6.00)	4.58
cNOR	3WFD	Pa	20.45 (13.34)	13.88 (6.49)	4.43
qNOR	3AYF	Gs		13.73 (6.30)	4.58

Bt, Bos taurus; Rs, Rhodobacter sphaeroides; Pd, Paracoccus denitrificans; Ec, Escherichia coli; Tt, Thermus thermophilus; Ps, Pseudomonas stutzeri; Pa, Pseudomonas aeruginosa; Gs, Geobacillus stearothermophilus.

Metal-to-metal (edge-to-edge) distance between CuA and heme a in A- and B-type oxidases, and same distance between heme c and heme b in C-type oxidases and cNORs.

Fe–Fe distance between low-spin and high-spin heme (edge–edge distance in parentheses),

Distance between the two metals of the BNC

Besides the ligand bridge, heme a and CuB also interact by means of an H bond between the hydroxyl of the hydroxyethylfarnesyl chain of heme a3 and the side chain of Y244. The functional role of this interaction is not known but may be important in modulating the pKa of the tyrosine. The histidine-tyrosine cross-link is a highly conserved motif in the A-, B-, and C-type oxidases, and its role in catalysis is also established (see section ). Interestingly, in the C-type oxidases, the cross-linked tyrosine originates from a different helix (helix VII instead of helix VI) and adopts a configuration slightly different from what is observed in the A- and B-type oxidases[27,49] (see also refs (50−52)). This coincides with the absence of a hydroxyl group on the high-spin heme, which is protoheme or heme B in the C-type oxidases, and may be important in tuning the redox and acid/base properties of the cross-linked tyrosine in C-type oxidases.[53]

The basic BNC structure just described is shared by the distantly related nitric oxide reductases (NORs), consisting of the cytochrome c- and quinol-dependent NORs. However, there are key differences. One is the replacement of the CuB ion with an Fe ion, and the second is the presence of a glutamate ligand to the FeB center in cNORs, which is replaced by water molecules in qNORs.[54−56] Moreover, there is no redox-active tyrosine in the vicinity of the active site in the NORs. Structural and functional studies on NORs remain an active area of research.[57,58]

Synthetic Models

Synthetic heme–copper oxidase adducts have played an important role in shedding light on the oxygen reduction reaction.[59,60] Early on structural and kinetic analysis on heme–copper oxidase model compounds led to the identification of two stable species: μ-oxo (Fe[III]–O–Cu[II]) and μ-hydroxo (Fe[III]–OH–Cu[II]).[61,62] Even though such species have not yet been observed experimentally in the molecular mechanism of CcO, the latter has nevertheless been predicted to occur based on DFT calculations[63−65] in the fully oxidized activated form of the enzyme (OH, see section ). It is also interesting to note that the presence of a true peroxy intermediate (O2–2) has never been observed experimentally in the catalytic reaction of CcO[66] (see also ref (67)) even though quantum chemical calculations on the conversion of state A to state P in the catalytic cycle (section ) supports its presence.[53,68] In contrast, a peroxy species is readily observed in the oxygen reduction reaction of heme–copper oxidase mimics and exists in two structurally distinct forms, side on and end on.[69]

Recent work in synthetic chemistry has shown that artificial chemical constructs that mimic the active site of cytochrome c oxidase are able to catalyze the reaction with a high turnover number.[70] Furthermore, such approaches have highlighted the importance of H bonding in the O–O bond cleavage reaction, such that a phenolic group may assist in further reducing the bound peroxide by supplying an electron and a proton to it.[71,72] Earlier, density functional theory (DFT) calculations performed on A- and C-type oxidases model systems have indeed shown the functional importance of such an H-bonding arrangement between the bound O2 molecule and the cross-linked Tyr, also through a mediating water molecule.[53,73] It is likely that such an H-bonding arrangement enhances the rate of proton-coupled electron transfer necessary for O–O bond cleavage.

Paths of Electrons, Protons, Oxygen, and Water

Reduction of molecular oxygen to water at the active site requires three reactants, molecular oxygen (O2), electrons (e–), and protons (H+), and yields a single product water if one ignores the net translocation of protons. All of the above differ greatly in their mass, charge, and chemical properties. Therefore, it is not surprising that the paths taken to reach and leave the active site are also quite different.

Oxygen Channels and the Escape of Water

Due to its apolar nature the O2 molecule is much more soluble in lipid membranes than in the surrounding aqueous phases and is found to localize at the center of the lipid bilayer.[74−76] In the middle of the membrane two sets of TM helices of subunit III of the A-type oxidases form a V-shaped structure surrounding a hydrophobic channel in subunit I, which leads from the membrane interior to the active site buried in the interior of the enzyme. This path has been identified as an O2 channel, based on crystal structure data[77] and kinetic experiments.[34] Xenon-labeling experiments with cytochrome aa from Rh. sphaeroides(78) and the B-type cytochrome ba from T. thermophilus(79,80) have confirmed and extended this conclusion.

Two water molecules are produced at the BNC in each catalytic cycle. By labeling O2 with 17O Ferguson-Miller et al.[81] showed by paramagnetic resonance spectroscopy that the labeled product water reached the Mg site (see Figure ), which was substituted by a manganese ion in the experiment, within milliseconds of the reaction with O2. This finding was supported by computational work[37] that suggested how water produced at the BNC would reach the Mg domain above the hemes via dissociation of the delta-propionate (heme a3)-arginine ion pair (see purple dotted lines between R438 and heme-propionate in Figure ). Facile water access between the BNC and the Mg domain may be expected to lead to uncoupling pathways of protons, so such a pathway of water transport must be carefully controlled (gated). The computational studies[37] indeed indicated that dissociation of the propionate-arginine ion pair was strongly dependent on the redox state. In addition, redox-dependent water dynamics in a tight nonpolar confinement (Figure ) and hydration/dehydration transitions may also be expected to prevent such an uncoupling (see also ref (82)).

Electron Transfer Paths

Electrons as elementary particles are often described by the wave–particle duality. In the context of electron transfer in proteins, both the wave (dynamic electron tunneling) and the particle (equilibrium redox potential) nature of the electrons are at play. Long-range electron transfer in proteins is accomplished through buried metal centers, which are bound to the protein matrix. A wide variety of metal centers are found in redox-active proteins, such as hemes, coppers, iron sulfur clusters, manganese clusters, etc. A survey of a large number of crystal structures of redox-active proteins has revealed that electron transfer at biologically relevant rates takes place through redox centers that are placed not more than ca. 14 Å apart when measured edge-to-edge.[83] The process of electron transfer in proteins through fixed redox centers is relatively well understood.[83,84] It seems that the nature of the intervening proteinaceous medium in electron transfer between the electron donor and the electron acceptor groups is of relatively little importance in practice.[85] Although electron transfer rates do depend on the medium structure[86] and electron transfer calculations may reveal preferred pathways, for example, in CcO,[87] it seems that the number of relevant pathways between the donor and the acceptor becomes very high in proteins, especially at longer distances. Under such circumstances it is the distance not the medium composition that is the main determinant of the rate of electron transfer.

Electron transfer in CcO is initiated with the formation of a cytochrome c–CcO (donor–acceptor) complex. Reduced cytochrome c binds on the surface of subunit II on the side of the membrane and donates an electron to the CuA center of the enzyme. The detailed nature of this step was recently elucidated from the crystal structure of a complex between cytochrome c and CcO.[88] The interface between the two proteins is found to be stabilized by weak electrostatic interactions, enabling rapid association and dissociation of the donor–acceptor complex, in agreement with earlier kinetic experiments and modeling approaches.[89−91] The crystal structure of the complex identified a new class of protein–protein interaction termed “soft and specific”, which had not been previously predicted.[92]

The intraprotein electron transfer processes in CcO are relatively well understood. The primary electron acceptor is a cupredoxin-type center, CuA, which consists of two closely spaced (∼2.5 Å) copper atoms, each ligated by a cysteine and a histidine residue from subunit II. CuA undergoes a one-electron redox transition that can formally be written as CuII/CuII + e– ↔ CuI/CuII, while the CuI/CuI structure is too high in energy to be occupied. The edge-to-edge distances from the CuA center to heme a, heme a3, and CuB are ca. 12, 16, and 20 Å, respectively. Thus, the bis-imidazole-ligated low-spin heme a located approximately 1/3 into the membrane will be the acceptor of the electron from CuA in a pure electron tunneling event with a time constant (τ = inverse rate constant, 1/k) of ca. 10–50 μs depending on the source of enzyme.[93] The rate is insensitive to variations in pH and to heavy water substitution and fits very well with the rate of 8.7 × 104 s–1 (τ ≈ 11 μs) predicted by the Moser–Dutton ruler.[85] In equilibrium titrations, however, the redox potential of heme a has been found to be weakly linked to nonstoichiometric proton uptake especially when heme a and CuB are oxidized (ca. 0.4 H+).[94]

Pure electron transfer, not coupled to any proton transfer or conformational change, from the electron-queueing heme a to the BNC occurs by nanosecond electron tunneling.[95,96] This is due to the fact that the hemes a and a3 are very closely spaced with only a 7 Å (edge-to-edge) distance between them. During enzymatic turnover, it is this electron transfer step that has been found to drive the internal proton transfer step that initiates the proton-pumping function (see section ).[97]

Proton Transfer Pathways

The proton is 1837 times heavier than the electron and cannot tunnel across distances longer than 1–2 Å on biologically relevant time scales (see ref (93)). However, long-range proton transfer (tens of Angstroms) is typical of biological systems such as respiratory and photosynthetic complexes.[98] Therefore, some sort of “medium” is necessary to accomplish proton transfer across longer distances. It is observed in crystal structures of several enzymes involved in redox-coupled proton transfer that there are cavities of varying sizes and volume formed by surrounding charged or polar amino acids that can hold bound water molecules.[99,100] Combined with site-directed mutagenesis studies, some of these hydrated cavities have been identified to function as proton transfer pathways in A-type oxidases,[6,101−105] but they may also have other roles such as “dielectric wells”.[106,107] Proton transfer in such pathways occurs by means of the classical Grotthuss mechanism described in detail elsewhere.[108]

In HCOs, proton channels are responsible for the transfer of protons from the side of the membrane to the site of O2 reduction and for pumping from the to the side of the membrane (Figure ). The proton pump mechanism requires a proton-loading site (PLS, also called pump site or proton trap). In its simplest form the PLS accepts a proton from the side under conditions where it is isolated protonically from the side and donates the proton to the side under conditions where it is isolated from the side. Thus two separate proton channels are required for the proton-pumping process: one from the side of the membrane to the PLS and another from the latter to the side. The former is relatively well defined (section ), whereas the latter is still poorly understood.

D Channel

One of the best characterized proton transfer pathways in the A-type oxidases is the D channel, which is named after the highly conserved residue D91 at the side surface of subunit I (Figures and 6A). Already before the availability of the crystal structures, biochemical experiments revealed the importance of this aspartate in proton transfer linked to the proton-pumping activity.[6] The D channel begins at the side of the membrane with the solvent-exposed D91 forming the proton uptake site. Lined with a number of water molecules and polar amino acid residues, the channel continues inside the protein up to about halfway into the membrane (Figure A). Above D91, two conserved asparagine residues appear to form a gate.[109,110] Site-directed mutagenesis of several conserved amino acid residues of the D channel has led to interesting phenotypes.[111,112] For example, mutation of N98 (N139 in enzyme from Rh. sphaeroides) to an aspartate leads to decoupling of the oxygen reduction reaction from proton pumping so that the latter is completely abolished. The molecular reasons for this unusual effect are being actively discussed.[110,113]

Figure 6

Proton transfer pathways in type A (A), B (B), and C (C) HCOs.

The D channel ends at a highly conserved acidic residue, E242, from where it is unclear from the crystal structures how protons are transferred to the active site (substrate protons) and for pumping to the side. Between E242 and the heme groups “above” the glutamic acid there is an apolar cavity that appears empty in all crystal structures to date (Figure ). Computational approaches provided key insights into how protons could be transferred further, viz. by side chain conformational transitions of E242,[114,115] and dynamic water molecules in the cavity above it.[82,116,117] Several research groups have proposed a functional role of water molecules transiently present in this cavity.[118−122] Overall, the kinetics of proton transfer through the D channel and its role in supply of at least two substrate and all four pumped protons is well established from functional site-directed mutagenesis experiments on A-type cytochrome c and quinol oxidases from bacteria[123] and from yeast mitochondria.[124] By contrast, Yoshikawa and colleagues claim that the translocation of the pumped protons in mammalian mitochondrial CcO, which are also of type A, does not occur via the D pathway but via another pathway—the H channel (see section ).

Crystal structures,[5,78,125−128] site-directed mutagenesis,[129,130] and molecular dynamics simulation studies[131] reveal that a proton channel analogous to the D channel is unique for the A-type heme–copper oxidases and is entirely lacking from the B- and C-type oxidases. It seems that this “deficiency”—and the lack of subunit III—may be related to the relatively poor proton-pumping efficiency of the latter oxidases[14] compared to the high efficiency of the A-type enzymes. We argued[13,132] that this suggests that the B- and C-type enzymes are more ancient in evolution compared to the enzymes of type A, but this view is not fully accepted.[133]

K Channel

The K channel in the A-type CcOs is based on a highly conserved lysine residue (K319), which when mutated to a nonpolar residue completely abolishes catalytic turnover.[106] Similar to the D channel, an acidic residue (from subunit II) forms the putative proton uptake site at the side of the membrane[28,134] (see Figure ). The K channel terminates close to the BNC near the cross-linked tyrosine, which forms an H bond with the hydroxyl of the hydroxyethyl farnesyl side chain of the high-spin heme a. It has been proposed that the K channel supplies one or two substrate protons in the reductive phase of the catalytic cycle (see section ). In comparison to the highly hydrated D channel, the K channel comprises fewer predicted water molecules and is also scarce in polar amino acid residues[26,126] (Figure A). In all resolved crystal structures of A-type oxidases, lysine (K319) adopts a conformation in which its side chain is ca. 18 Å from the BNC. However, experiments and computer simulation studies have led to the proposal that the lysine side chain may undergo an “up-flip” that brings it closer to the BNC by ca. 5 Å and which may be important for the proton-pumping function of the enzyme[126,135−137] (section ).

A wide variety of ligands are known to bind to the region next to the K channel, as observed crystallographically and also characterized biochemically.[43] This has led to the suggestion that the proton transfer through the K channel is regulated.[43] Despite a vast amount of literature on K channel structure and function, its precise role remains unclear. In other words, it is not known why this channel is needed in addition to the D channel and why it supplies only some of the substrate protons (see the computational work by Blomberg and Siegbahn[138] and Sharma and Wikström[136]). It is also unclear how it operates together with the D channel. Nevertheless, the pore of the channel, formed by the polar residues from helices VIII and IX of subunit I (Figure ), is conserved in all three subfamilies (type A, B, and C; see also panels A, B, and C in Figure ), suggesting that this proton transfer route may have existed in the earliest forms of HCOs.[27]

In the B- and C-type oxidases, which lack the D channel altogether, the K channel analogue is responsible for the transfer of both the substrate and the pumped protons[129,130] (Figure B and 6C). Site-directed mutagenesis as well as molecular dynamics simulation studies on C-type oxidases have provided significant insights into the putative proton transfer paths for substrate and pumped proton transfer.[131,139] The situation is less clear in the B-type oxidases, which are also bona fide proton pumps. While the route of proton uptake is biochemically tested[129] and also analyzed through molecular simulations,[140] the way the protons are gated toward the side of the membrane remains unclear.

Catalytic Cycle

R and A States

The states of the catalytic cycle as discussed here (Figure ) as well as elsewhere refer primarily to the structure of the binuclear iron–copper center, including the adjacent tyrosine residue that is covalently linked to one of the histidine ligands of CuB. In the R (reduced) state the ferrous heme a is five-coordinated high spin and the CuB is cuprous (CuB[I]). In this state the heme readily binds ligands, such as O2, CO, and NO, which results in a high- to low-spin transition that is revealed by MCD, magnetic susceptibility, and optical experiments.[141,142] Interestingly, the binding of ligands such as O2 to ferrous heme a seems to occur only under conditions where the nearby CuB is also reduced.[143] A possible reason for this is that the fourth OH– ligand of CuB[II] prevents access to the iron of heme a and is only removed by protonation to water that readily dissociates upon reduction of the copper.

Figure 7

Catalytic cycle. Square encompasses the binuclear site with the heme a, CuB, and the covalently linked tyrosine (HO-tyr). Distal histidine ligand of heme a and the three histidine ligands of CuB are not shown for simplicity. Uptake of protons to complete the chemistry of water formation is shown, but proton pumping is not shown. One proton is pumped across the membrane in each of the one-electron reactions, but for the A → F reaction the situation is more complicated: formation of state PR is linked to loading the PLS from the side of the membrane; its release to the side is driven by uptake of the chemical proton in formation of state F (see text). Structures of intermediates R, A, PM, PR, and F are well established (see text), whereas those of states OH and EH are still more hypothetical.

Binding of the physiological ligand O2 to the originally five-coordinated iron of heme a results in the so-called A state, originally named Compound A by Britton Chance et al.,[144] who first demonstrated this state experimentally in low-temperature optical experiments.

The resonance Raman (RR) band of state A at 571 cm–1 (Table ; see Ferguson-Miller and Babcock[142]) is indicative of binding O2 “end on” to the iron by comparison to the spectra and crystal structures of HbO2, and this binding geometry was proven by the elegant RR work of Kitagawa and Ogura[145] using 16O–18O-labeled O2, by which a “side on” geometry could be excluded. On the basis of quantum-chemical calculations the A intermediate is an S = 0 open-shell singlet biradical state (Figure ). The O–O bond length is intermediate between free dioxygen and peroxide and may best be regarded as a superoxide coordinated to a ferric iron, even though the state is often abbreviated as ferrous–oxy, Fe[II]–O2. The spin distribution shows that the iron is low-spin ferric antiferromagnetically coupled to the unpaired electron located on the two oxygens of superoxide.[146,147]

Table 2

Spectroscopic Data on Key Intermediates of the Catalytic Cycle

intermediate	resonance Raman Fe–O vibration (cm–1)	light absorption maxima (nm)a
A (Fe[II]–O2)	571	590, 430
P (Fe[IV]=O)	804	607, 442
F (Fe[IV]=O)	785	580, 442
OH (Fe[III]–OH)	450	NA, NA

In a difference spectrum vs the O state. References to this data are found in the main text.

Figure 8

Oxygen-splitting A → PM transition. Density functional theory-based geometry optimizations (def2-SVP/TZVP/BP86/disp3/MARIJ)[148−154] and energy calculations (def2-TZVP/B3LYP/disp3/eps4)[148,149,151,152,155−157] were performed with Turbomole[158] software on large model systems of the BNC. Spin density (α, green; β, pink) are plotted at an isosurface value of 0.01 e/Å3.

At room temperature and at 1 mM O2, state A is formed with a time constant of ca. 10 μs.[159] Prior to this O2 diffuses to the active site from the lipid membrane, where it is dissolved at a concentration at least ca. 5-fold higher than in the surrounding aqueous phases.[74−76] The diffusion occurs via channels in the structure of subunit I that have been clearly identified from the X-ray diffraction data as well as from mutagenesis experiments (see section ). The second-order rate constant for O2 binding is 1.38 × 108 M–1 s–1, and the dissociation constant (KD) is as high as 0.28 mM, which means that the off rate is ca. 40 000 s–1. At an O2 concentration of 100 μM ([O2] in tissues is usually in the range 5–25 μM (see ref (19))) the free energy change of dioxygen binding is ca. +0.6 kcal/mol. These apparently unfavorable numbers are traded off kinetically by fast “trapping” of the O2 once bound by electron transfer,[160] which renders the apparent Michaelis constant to be 3 orders of magnitude lower than KD. Blomberg et al. reported a calculated free energy change of −1 kcal/mol for the R → A reaction step, but this was based on a 1 mM activity of O2.[161]

P and F States

There are two kinds of P states which were named such because the structure of the BNC was originally thought to be ferric–cupric–peroxide[162] but which was later shown to have a ferryl heme–cupric structure with scission of the original O–O bond.[145] The PM state is formed spontaneously by decay of the A state under conditions where both the low-spin heme and the CuA center are oxidized (the so-called mixed-valence state). This is the state originally described by Chance in low-temperature experiments and called Compound C.[144] The reducing equivalents required to split the O–O bond are taken from the ferrous to ferryl transition of the heme, from the cuprous to cupric transition of CuB, and from oxidizing the covalently bonded tyrosine residue to a neutral tyrosyl radical. EPR data (see discussion of the EPR-active PR state below) using labeled O2 have identified a fourth OH– ligand on CuB[II]. FTIR data support the notion that the proton of this hydroxide ligand stems from the tyrosine residue, which thus donates both an electron and a proton in the process.[163] The free energy change of the conversion of state A to state PM is from −3[146] to −4.3[164] kcal/mol based on DFT calculations (see also Figure ), which is in agreement with the experimental observations that the reaction appears for all practical purposes to go to completion.

The PR state (Figure ) is formed from state A under conditions where the enzyme is initially fully reduced, i.e., both CuA and heme a are reduced upon O2 binding to heme a3 (contrast with the PM state described above). The PR state has the same optical absorption spectrum as state PM (Table ), but the fourth electron required to split the O–O bond now stems from oxidation of heme a rather than from oxidation of the tyrosine, which however supplies the necessary proton as indicated by FTIR studies.[163] PR has an unusual EPR signal (as opposed to PM which is EPR silent), which arises from CuB[II] interacting magnetically with the ferryl heme a3(66) but which is quenched in PM presumably due to the additional unpaired spin of the tyrosine radical. The structure of the PR state is thus established as Fe[IV]=O2– Cu[II]–OH– Tyr-O–. PR is a unique state of the catalytic cycle also for the reason that it represents the only relatively stable state of the binuclear site that is apparently not charge neutralized according to the principle outlined by Rich.[165] The charge of the porphyrinate dianion of heme a3 is compensated by the two positive charges of the ferryl iron (Fe4+=O2–); the CuB[II] cation is compensated by the hydroxide ligand and the tyrosinate anion. Propionate D of heme a3 is charge compensated by making an ion pair to a conserved arginine, but propionate A remains uncompensated so that the overall charge is −1. Calculated this way, all other known intermediates of the catalytic cycle have an overall charge of zero. However, as shown by time-resolved experiments by Belevich et al.[97] (see also refs (93 and 98)), the electroneutrality principle of the binuclear center is not violated in the PR state but the excess negative charge provides the driving force for uptake of a proton to the proton-loading site of the pump mechanism (PLS, see section ).

The F state is formed upon protonation of state PR (Figure ). The dramatic difference in the optical absorption spectrum between PM (or PR) and F must be due to a change in the absorption spectrum of the heme itself. The OH ligand of CuB identified in the P states (see above) is directed toward the axial oxo group of the ferryl heme, and therefore, its protonation to water may be expected to change the axial heme symmetry with a change in the α-band as a result, either due to H bonding from the cupric aquo ligand to the oxo group or, perhaps more likely, due to loss of the resulting water molecule from the coordination sphere of CuB. On the other hand, protonation of the tyrosinate close to the heme edge and the heme’s hydroxyethyl farnesyl side chain would not be expected to cause such a dramatic change in the optical spectrum. Moreover, several studies have shown that protonation of a state equivalent to PM (with ferryl heme, cupric copper, and the neutral tyrosine radical) yields a state (called F(dot) or F′) with optical characteristics similar to the F state but without the additional electron.[166] Since protonation of the tyrosine radical is unlikely,[167] these observations independently suggest that it is protonation of the OH– ligand of CuB that changes the P-like heme spectrum to become F like (Table ). Furthermore, Verkhovsky et al.[168] found that in cytochrome ba3 from Th. thermophilus, which is a type B member of the heme–copper enzyme family,[4] an F-like species is entirely absent from the catalytic reaction sequence. Instead, uptake of the first proton to the binuclear center after the reaction with O2 (analogous to formation of F) yields a species with optical P-like character. This may hence be a natural example of a case where the proton is taken up into the active site without a spectral shift (from P like to F like), suggesting that the proton acceptor in the catalytic sequence of the ba3 enzyme is the covalently bonded tyrosinate (known to be present in the Th. thermophilus enzyme), in contrast to the OH– ligand of CuB as in the A-type oxidases (however, see ref (169) for a different interpretation).

O and E States

The BNC in the “as-isolated” cytochrome c oxidase is in the so-called O state (“oxidized”) with ferric high-spin heme a3 and cupric CuB magnetically coupled so that the EPR characteristics of neither is observed.[141,142] There has been much discussion with regard to the more detailed structure of this state, especially concerning the nature of the bridging ligand between the two metals. “Resting” vs “pulsed” forms as well as “slow” vs fast” have been described (see ref (93) and citations therein). Recent crystallographic work has claimed that the bridge comprises a peroxide molecule,[46,126] which is very difficult to understand in terms of the otherwise quite well understood structures of catalytic cycle intermediates (see ref (93)). This is in part augmented by the recent radiation-damage-free structure of B-type oxidase in which the electron density between the two metals has been ascribed to a hydroxide or a water molecule[170] (see also ref (171)). At any rate, it has become clear that state O of the BNC is not a catalytically active state but a relaxed form (or a combination of such forms) generated spontaneously in the absence of electron donors. The bridging ligand in that state could possibly be dioxygen itself or superoxide.[48] Light-triggered electron injection into CcO in the O state results in fast reduction of CuA and the low-spin heme a but no further electron transfer into the binuclear site within several milliseconds.[172] In contrast, when electron injection is performed immediately after reoxidation of CcO with O2, there is electron transfer into the BNC in the submillisecond time domain, and the reaction is accompanied by proton pumping.[172,173] On the basis of such data a catalytically active form of the O state, called OH, was postulated,[173] different from both pulsed and fast forms described earlier.

OH was early on suggested to have a hydroxide ligand on heme a3 (hence the suffix H) based on the resonance Raman work by Rousseau et al.,[174] who identified such a high-spin ferric–OH intermediate with an Fe–O stretching vibration at 450 cm–1 (Table ), which is a very low frequency compared to other ferric heme hydroxide compounds that are usually low spin due to the strong-field hydroxide ligand. Ishigami et al.[175] attribute the low-frequency and the high-spin state to a strong hydrogen bond to the oxygen of the ferric OH– ligand. Alternatively, it is due to strong bonding of CuB to the oxygen atom (see below). It was also predicted[176] that the Em of the OH state would have to be much more positive than the redox potentials of ca. 350–400 mV found in anaerobic equilibrium redox titrations of CcO.[177] This prediction was fulfilled in the electron injection experiments by Belevich et al.,[178] which showed that injection of a single electron into the OH state led to exclusive reduction of CuB (based on the optical spectral change) with an Em value at least 120 mV higher than the Em of the low-spin heme a, i.e., >500 mV. It may be of interest to note that on the basis of recent density functional theory (DFT) studies the OH state is an electronic equilibrium between ferric–OH cupric tyrosinate and cuprous tyrosine radical states.[63,179,180] In both cases the addition of an electron would yield a ferric–OH cuprous tyrosinate state (EH; Figure ). Moreover, recent DFT calculations[180] support the prediction[176] that the Em values of the OH/EH and EH/R redox couples (see Figure ) are much higher than those obtained in equilibrium redox titrations.

Earlier we proposed based on DFT calculations that one of the key differences between the relaxed O and activated OH states may be the trigonal planar geometry of the CuB center in the latter state.[63] This imparts CuB a high operational redox potential and cross-linked Tyr a radical character, whereas in the O state the CuB is ligated by three histidine residues, a fourth oxygenous ligand (−OH–), and the tyrosine is anionic. Overall, the conclusion from the study was that the strained BNC structure and a neutral radical on Tyr makes the OH state kinetically competent in redox-coupled proton pumping in comparison to the O state. In order to model the activated OH and FH states, we optimized the geometries by displacing the water near CuB to a position such that it does not ligate the copper (see Figure ). In subsequent studies, Blomberg reached very similar mechanistic conclusions based on DFT calculations performed on larger model systems.[64,180,181] In order to further look into this, we compared the optimized structures of the activated states OH from these two independent studies.[63,64,180,181] The data in Figure clearly shows that the structure of the active site is very similar in the two cases. Also, in both studies it was concluded that the structure has the tendency of favoring the cuprous–tyrosine radical electron distribution. Furthermore, using the reported coordinates[180] we calculated the energies of the two structures, which are within 1 kcal/mol from one another.

Figure 9

Comparison of structures of the FH (A) and OH (B) states from studies by Sharma et al.[63] (orange) and Blomberg[180] (green).

Redox Potentials

Figure summarizes the main features of the catalytic cycle and the structures of its intermediates. Originally, the only known redox potentials were those measured by anaerobic redox titrations in which the fully oxidized enzyme was reduced and reoxidized in the presence of suitable redox mediators.[18,141] Briefly, in those titrations the Em,7 of CuA was found to be ca. 245 mV (n = 1) and almost independent of pH. By contrast, the Em,7 values of hemes a and a3 were found to be interdependent presumably due to an electrostatic redox interaction. Thus, the Em,7 values of both hemes a and a3 were found to be near 380 mV when the companion heme was oxidized but ca. 230 mV when the companion heme was reduced. The Em,7 of CuB was estimated to be ca. 390 mV on the basis of appearance of the high-spin ferric EPR signal of heme a3 at g = 6 upon CuB reduction that abolishes the magnetic coupling between them (see ref (93)).

Table summarizes the midpoint redox potentials (Em,7) of the catalytic cycle intermediates as well as the corresponding standard free energy changes (−ΔG0′) for each reaction step. Experimental estimates are available only for the PM/F and F/OH redox couples; their uncertainty is due to the assumed protonmotive force of 220 mV in well-coupled rat-liver mitochondria supplied with an excess of ATP[176] but is hardly larger than some 20–30 mV. The discrepancy to the free energies calculated by DFT methods (Table ) is significant especially for the PM → F transition. This transition has a calculated driving force much higher than that of the subsequent F → OH transition, while the experimental estimates for these two reactions are comparable. It should be emphasized that the sum of standard free energies in Table should amount to the −ΔG0′ of the overall reaction , which is 4 × 550 or 2200 meV (see section ). On this basis and from the experimental estimates of −ΔG0′ for the PM → F and F → OH steps, the mean values for the OH → EH and EH → R steps may be estimated (in parentheses, Table ). Figure shows a plot comparing the Em,7 values of the oxidase reaction steps (in red) with the corresponding values of the four one-electron reactions of O2 reduction to water in aqueous solution[93] (in blue). The figure shows how binding to the enzyme’s active site has a remarkable smoothening or leveling effect on the four otherwise very different redox potentials. This leveling is, in fact, much more prominent than thought originally[141] for two reasons. First, at that time, the experimental Em,7 values of the PM → F and F → OH steps were overestimated due to the erroneous belief that each of these steps would be coupled to pumping of two protons instead of one.[182] Second, at that time it was not realized that the redox potentials measured in anaerobic equilibrium experiments were not relevant for the potentials of the active forms OH and EH. At optimum, the leveling of the Em,7 values for each of the four one-electron reaction steps should occur at the potential of the O2/2H2O redox couple, which is 800 mV in the conditions used here (section ). As shown in Figure , estimates based on currently available experimental data do not deviate much from that optimum. It is also important to realize that the bound dioxygen in state A receives four electrons essentially simultaneously (no experimentally discernible intermediate), so that no “reactive oxygen species” are produced in the reaction.

Table 3

Em,7 Values (mV) and Driving Forces (−ΔG0′ in meV) in the Catalytic Cyclea

	Em (exp)	Em,7 (exp)	Em,7 (exp) (pmf = 0)	–ΔG0′ (exp)	–ΔG0′ (Kaukonen)	–ΔG0′ (Blomberg)
PM → F	375 (pH 7.7)	417	857	607	798	871
F → OH	350 (pH 7.2)	362	802	552	332	477
OH → EH			(660)	(410)	342	277
EH → R			(660)	(410)	518	316
R → Ab				0	–41	43
A → PMc				220	124	225

Em cyt c = 250 mV; Em O2/H2O= 800 mV; pmf assumed at 220 mV. Experimental observations from ref (162). Conversion to pH = 7 and pmf = 0 (Wikström and Verkhovsky[176]). Computational data from Kaukonen[146] and Blomberg[180] converted with Em (cyt c) = 250 mV and Em (O2) = 800 mV. Note that the Em,7 values for the PM/F, F/OH, OH/EH, and EH/R redox transitions are related to the corresponding values of −ΔG0′ by adding 250 mV (the Em,7 of cyt c). The values in parentheses for the OH → EH and EH → R redox reactions were obtained from the known sum of all of the −ΔG0′ (exp) values (i.e., 2200 meV), from which the experimental and calculated values of the other transitions were subtracted. The obtained sum of −ΔG0′ (exp) values for the OH → EH and EH → R transitions was assumed to be distributed equally between them.

These values are based on the equilibrium constant of O2 binding and differ slightly due to different assumed O2 activities.

This value is obtained from the equilibrium constant as estimated by DFT calculations

Figure 10

Redox potentials. Redox potentials for the four one-electron reactions reducing O2 to water are shown in blue for the reaction in aqueous solution[93] and in red for the reactions catalyzed by cytochrome c oxidase (Table ). Note that the ordinate may also be the standard change in Gibbs free energy (−ΔG0′, in meV), which is obtained by subtracting 250 mV from the Em,7 values plotted. Red point at 2.5 reaction equivalents is not an Em,7 value but the combined −ΔG0′ value for the reactions R → A and A → PM to which 250 mV was added (cf. Table ).

Proton Translocation

General Principles

The most entertained mechanism of proton translocation by the A-type heme–copper oxidases follows a set of common principles that will be outlined first. The mechanistic stoichiometry has been found to be up to one pumped proton per electron transferred from cytochrome c to dioxygen, the electrical charge translocation being equivalent to two charges/e– due to the uptake of the electrons from the side of the membrane and the protons for formation of water (“chemical” protons) from the side. Today, these stoichiometric features are well documented in experiments with isolated mitochondria, bacterial cells, and isolated CcO from different sources incorporated into phospholipid vesicles.

Wikström and Krab[183] proposed a directly coupled proton pump (DCP) principle in which there would be direct electrostatic coupling between the translocated proton and the transferred electron. Such a mechanism would not require large conformational changes or long-range structural interactions. According to this principle, electron transfer would initially be partially compensated for electrostatically by the transfer of a proton originating from the side of the membrane to a “pump site” (other subsequent descriptions are “proton trap” and “proton-loading site”) located at a short distance from the electron in the BNC in order to enable sufficient Coulombic stabilization by the electron. In the next step, another proton originating from the side would arrive at the BNC, completing the oxygen reduction chemistry and annihilating the electron charge. This event would therefore destabilize the proton in the “pump site”, lowering its pKa and thus causing its ejection to the side of the membrane. Morgan et al.[184] described this by “binding of scalar protons to oxygen drives the ejection of the vectorial protons from the pump”. Rich[165] independently arrived at the same principle, based on the electroneutrality concept of the BNC that he had previously established experimentally.[94]

This basic principle of the functioning of the proton pump can be traced back to early proposals where reduction of heme a would be coupled to protonation of the PLS (cf. ref (185)). The competing view is that electron transfer from heme a to the BNC is required to drive the first proton-pumping event, viz. loading of the PLS with a proton. This latter idea was implicit in the mechanistic proposals by Morgan et al.[184] and Rich[165] as well as in the “water-gate” mechanism[116] and the mechanisms proposed by Popovich and Stuchebruchov[186] and Siletsky et al.[187] More recent experimental observations have supported this suggestion,[176] because electrometric measurements of charge translocation have shown that the proton movements associated with the proton pump are linked to oxidation of heme a by the BNC, not to its reduction by CuA (see also Belevich et al.[178]).

All protons to be pumped are taken up from the side of the membrane via the so-called D pathway (or D channel) characterized by a conserved aspartate at its opening, an “asparagine gate” domain, a “serine domain”, and finally a conserved glutamic acid situated roughly halfway into the membrane dielectric at ca. 10 Å from both heme groups (Figure ; section ). Yoshikawa et al.[47] maintain that in bovine CcO all pumped protons are transferred via another pathway, the so-called H channel, and by an indirectly coupled mechanism entirely different from the one discussed in this section (see section ).

Several candidates for the “pump site” (PLS) have been proposed. The first one, His-284, one of the ligands of CuB, was suggested before the crystal structure was known as part of the so-called histidine cycle mechanism.[184,188] A detail of that mechanism, viz. that proton pumping would be linked only to the reactions of the oxidative phase of the catalytic cycle, was later shown to be erroneous by kinetic experiments, which demonstrated proton pumping coupled to all four one-electron transfers into the BNC.[173] Subsequently, another histidine ligand of CuB has been proposed as the PLS,[189] along with the D and A propionates of heme a3(190−192) and water molecules in their vicinity.[193] Common to all these proposals is the vicinity to the BNC, allowing strong Coulombic proton–electron coupling as required by the DCP principle. A final decision has not been reached, and it is possible that the PLS is not a single binding site but a more poorly defined domain that the proton is occupying, possibly including the propionate A of heme a3 and water molecules in the vicinity[194] (Figure ).

Proton-Pumping Mechanism in Type A Cytochrome c Oxidases

Time-resolved dynamic and static equilibrium experiments on type A heme–copper oxidases from both mitochondria and bacteria over a period of nearly four decades have led to the proposal of a general mechanism that follows the directly coupled proton pump principle described above and which conforms to the notion that uptake of the chemical protons is a crucial element of the proton pump. In both these respects it differs fundamentally from alternative mechanisms which will be summarized separately below (section ). The mechanism is proposed to be essentially the same in all four one-electron reaction steps of the catalytic cycle (Figure ), albeit with some small variations that will be discussed. The mechanistic time frame is largely based on two kinds of time-resolved experiments, either flow-flash data where the reaction is started by photolysis of heme a3[II]-CO in the presence of O2 or experiments where a single electron is photoinjected to CcO in a defined state, usually OH. In both cases the progress of electron transfer is usually measured by optical spectroscopy, in some particular studies by resonance Raman and EPR spectroscopy; proton transfer has been measured by electrometry and proton uptake/release by using pH indicators in phospholipid vesicles inlaid with CcO. Figure shows this mechanism as applied to the OH → EH reaction.

Figure 11

Proton pump mechanism. Mechanism is depicted as revealed from time-resolved electron injection experiments of the OH → EH transition[172,178] but revised from ref (178) to account for the effect of mutating the K channel lysine, which blocks the 150 μs phase.[98,172] The 150 μs phase includes protonation and movement of lysine-319 closer to the BNC (blue circle below heme a3), which is necessary to allow electron transfer from heme a to heme a3 and the linked uptake to the PLS of the proton to be pumped (blue circle above heme a3) in states marked III and IV, respectively. Red color indicates the position of the injected electron.

Electron transfer from CuA to heme a is fast (section ) and due to pure quantum-mechanical electron tunneling without linkage to proton movements. However, the redistribution of charge will change the pKa values of nearby acidic groups, and depending on their location and protonic connectivity and on the pH of the aqueous and phases, secondary proton movements may occur as seen as a weak pH dependence of the equilibrium Em,7 of heme a.[94] Reduction of heme a may raise the pKa of the proton-loading site (PLS), but this is insufficient to load the site with a proton from the side at pH 7. Loading of the PLS is observed separately, coupled to the transfer of the electron from heme a to the BNC, in time-resolved experiments where the fully reduced enzyme (state R) reacts with O2.[97] However, such experiments do not distinguish the time sequence of subsequent events of the pump mechanism, which have been addressed by time-resolved electron injection data.

In CcO from P. denitrificans, single electron injection in the OH state first leads to ca. 10 μs electron transfer from CuA to heme a and then to a 150 μs phase of proton uptake from the side, which can be attributed to protonation of the PLS.[178] This phase is totally blocked by mutating the key lysine residue in the K pathway to methionine,[172] which supports the notion that electron transfer to heme a3 in the BNC is actually required for loading the PLS. This observation may, in fact, be an example of the K pathway functioning as a “dielectric well” as proposed by Rich.[106,165] Apparently, the Em,7 of heme a3 is not raised sufficiently by protonation of the PLS relative to the Em,7 of heme a, so that movement of positive charge closer to heme a3 by K channel activity is necessary to make the electron transfer to the BNC thermodynamically feasible. An “upswing” of the protonated lysine side chain toward the side may be sufficient to achieve this, and such dynamics have been observed in MD simulations.[136] After the 150 μs phase, the injected electron is distributed roughly 60/40 between heme a3 and heme a, which means that even with the PLS protonated and the assistance of the K pathway, the Em,7 of heme a3 is only slightly higher than that of heme a. The relatively low Em,7 of heme a3 may be a feature unique for the OH state and may explain why the electron transfer in this case requires “dielectric assistance” from K channel activity. In contrast, in the A → PR → F and subsequent F → OH transitions of the fully reduced enzyme, blockade of the K channel has only a slight effect on the rate.[195] This is possibly attributable to a much higher Em,7 of the electron acceptor than is the case in the OH → EH transition but may nevertheless also be due to an “upflip” of the lysine in this case also.[196]

The 150 μs phase is followed by an 800 μs phase, which is rate limited by net protonation of the BNC via the K pathway.[172,178] This amounts to protonation of the tyrosine and accompanying electron transfer from heme a3 to CuB as verified by optical spectroscopy. Optical spectra reveal that the final destination of the injected electron is indeed CuB alone. The lack of any discernible reduction of either heme a or a3 in the final EH state indicates that the Em,7 of CuB is at least 120 mV higher than that of the two hemes, implying an Em,7 well above 520 mV (assuming that 1% heme a reduction would have been detected). This experimental estimate is in good agreement with the Em,7 values of the OH/EH redox couple of 603 and 528 mV, respectively, obtained from DFT calculations (Table ).

After the 800 μs phase there is finally a reaction phase with a time constant of ∼2.6 ms, which is discernible by the electrometric technique (but not observed by optical spectroscopy), and was interpreted to represent ejection of the proton in the PLS into the aqueous medium on the side of the membrane.[178] The driving force for this process is suggested to be the lowered pKa of the PLS due to neutralization of the BNC by the uptake of the substrate proton. The relatively low rate (compared to a maximum turnover of the Paracoccus enzyme of some 1000 s–1) may be due to the special conditions in which only one electron is injected.

Earlier electrometric experiments[172,173] had shown that each of the four one-electron reduction steps in the catalytic cycle (Figure ) is associated with pumping of one proton across the membrane and with additional uptake of a chemical proton into the binuclear site. In agreement, time-resolved pH indicator data with CcO vesicles have demonstrated that one proton is pumped to the side in each of the P → F and F → OH transitions and that each of these transitions is coupled to uptake of two protons from the aqueous side.[197]

Prevention of Leaks

The prevention of leaks is a fundamental property of a redox-linked proton pump mechanism such as the one briefly outlined above (see refs (190) and (198)). For example, electron transfer to the BNC must not be allowed without loading the PLS with a proton. Most of all, primary protonation of the BNC (“chemical proton”) linked to the electron transfer must be prevented since, if allowed to occur, the exergonic chemistry driving the endergonic proton pump would be completed without linkage to proton pumping. We note here that Brzezinski et al.[199] proposed an alternative mechanism in which the oxygen reduction chemistry does occur primarily but results in a high-energy conformation of the structure that relaxes secondarily by linkage to proton pumping. This possibility is hard to exclude definitively, but it is made unlikely by both experimental data and conceptual difficulties.[93]

Stimulated by the prediction of water molecules in the hydrophobic cavity between Glu-242 and the hemes, an area that is empty according to the crystal structures, Wikström et al.[116] proposed that such water molecules in a restricted hydrophobic pocket could arrange themselves to catalyze controlled primary proton transfer to the PLS rather than to the BNC, thus securing the first requirement of a functioning redox-linked proton pump. This so-called water-gated mechanism has subsequently been criticized[200] but has also gained further support from more sophisticated multiscale approaches[118] as well as from comprehensive molecular dynamics simulations that have included the lipid membrane.[82] Yet, it is evident that this mechanism would not alone be sufficient to prevent proton leaks in the pump mechanism.

Another acute possibility for a leak concerns the situation later on in the mechanistic pump sequence, where the pumped proton has been deposited in the PLS and the chemical proton has been taken up at the BNC, causing a substantial destabilization of the PLS proton (see above). The issue now is how the system prevents the PLS proton from leaking back into the aqueous phase, a reaction that is far more favorable thermodynamically than the ejection into the phase required for a functioning pump. It is obvious that the gating required here must be of kinetic nature. Blomberg and Siegbahn[198] suggested a positively charged transition state for proton transfer from the side to the PLS, which would be stabilized (lowered in energy) by Coulombic interaction with the electron in heme a but would be raised in the case where the electron has moved to the BNC, thus preventing back-leak of the proton from the PLS. Another possibility is gating by the position of the side chain of glutamic acid 242 at the end of the D pathway.[122,201] One promising kinetic gate may be provided by the conserved asparagine domain in the D pathway.[109,110] The importance of this latter gate is highlighted by the fact that several mutations in this area cause decoupling of the proton pump without affecting electron transfer.[112,202−205] These gating possibilities have not been quantified in a precise manner, and it remains possible that they all contribute to the high efficiency of proton pumping by the A-type oxidases, which is not shared by those of type B and C, which lack the D pathway.

Proposed Alternative Mechanisms

Role of the so-Called H-Channel

Yoshikawa et al.[47,99,206,207] emphasized the importance of the structure that they have called the H channel as the sole transfer path of the pumped protons across the membrane. The H channel is located in subunit I and passes the low-spin heme a but is very far away from the binuclear site. Hence, any redox-linked proton coupling involving the H channel would have to be via long-distance interactions, in contrast to the DCP mechanisms described above but in agreement with proposals of indirect conformational coupling by Papa et al.[208] and Rousseau et al.[209]

The H channel structure is not unique for mammalian CcO, but most of it is shared by the A-type bacterial heme–copper oxidases (but see below). Extensive mutations in this structure of bacterial CcO[104,210] have shown no effects whatsoever on proton pumping and very small perturbations of overall redox activity. This is in complete contrast to results with the mammalian enzyme, published by Yoshikawa et al.,[47] where bovine CcO genes were expressed in HeLa cells and three mutations in the H channel structure were reported to abolish proton pumping without effect on electron transfer. Unfortunately, to our knowledge, these results have not been reproduced by other laboratories, and such confirmatory experiments should have very high priority. One reason for this is that if the Yoshikawa mechanism is correct it would have profound evolutionary implications, indicating that the mechanism of proton pumping would be fundamentally different in mammalian (mitochondrial) CcO, not only from that of A-type CcO in aerobic bacteria despite their very close structural and functional homology but also from yeast mitochondrial CcO where mutations in the D and K channels have comparable effects to those in bacterial A-type oxidases and mutations in the H channel are ineffective.[211]

We may analyze the Yoshikawa mechanism by two approaches: one structural and one functional. As already mentioned, the main H channel structures are found also in the bacterial cytochrome c oxidases of type A with the exception of the aspartate residue near the side of the membrane unique to the vertebral CcOs, close to the bimetallic CuA center, and a Tyr440-Ser441 motif. In most other respects key hydrophilic residues of the H channel in mammalian CcO are conserved in the bacterial A-type CcOs. However, all of those hydrophilic H channel residues that were conserved between mammalian and bacterial CcOs have been replaced by apolar amino acids in the bacterial quinol oxidases, which are also of type A.[98] Yet, the bacterial A-type quinol oxidases are proton pumps just like their “cousins” in the CcO group. Hence, the presence of a H channel structure correlates with the function of the oxidase as a cytochrome c oxidase (instead of being a quinol oxidase) but not with the function as a proton pump.

Linkage of the Proton Pump to the Catalytic Cycle

From a functional point of view, Yoshikawa’s mechanism defines the steps in the catalytic cycle that are coupled to various aspects of proton transfer, viz. the uptake of the four “chemical” protons to complete the redox chemistry at the BNC, the uptake of protons to be pumped from the side into a storage mode near the Mg site, and the release of the stored protons to the aqueous side of the membrane. According to the original mechanism by Yoshikawa and colleagues,[212] all four protons to be pumped are taken up from the side of the membrane after formation of the R state (see Figure ) and before this state reacts with O2 and are deposited in a hydrogen-bonded network around and above heme a. In each of the subsequent one-electron transfer reactions of the catalytic cycle (see Figure ) one proton was proposed to be released to the side from this network and one proton taken up to complete the oxygen reduction chemistry at the BNC. However, this scheme is at variance with the time-resolved measurements of proton release and uptake on bacterial enzyme by Brzezinski et al.,[197] who showed that the P → F and F → O transitions are each associated with uptake of two protons from the side. In a subsequent paper[213] this disparity was realized, the proton-accepting hydrogen-bonded network was revised, and it was proposed instead that each of these two transitions may be associated with uptake of two protons for the O2 reduction chemistry so that the P → F and F → O reaction steps would each be linked to release of 1 proton from the “store” to the P side and to uptake of two chemical protons to the BNC. Since each step is also linked to transfer of one electron from cytochrome c on the side to the BNC, each reaction step is coupled to translocation of two electrical charges across the membrane. This charge translocation stoichiometry is consistent with electrometric data for those reaction steps[98] and also with measurements of the number of protons taken up on the side and released on the side[197] as well as with the observations of net consumption of about two protons from the medium in the oxidative phase of the catalytic cycle.[94] However, the consumption of all four chemical protons in the oxidative phase is very difficult to fit to the structures of the P, F, and O states as they were determined by other data (see section and Figure ).

This version of the mechanism further suggests that all four protons to be pumped are taken up from the phase in a single step following formation of state R and before oxygen binds to form state A. This means altogether that the reaction steps OH → E and E → R are each associated only with release of one proton from the Mg site to the phase, which is in accordance with the observations of release of pumped protons, but incompatible with the observed charge translocation during these two reaction steps. The only charge translocation of each of the OH → E and E → R steps in the mechanism is electron transfer from cyt c to the BNC and release of a proton to the side from the PLS, which together amount to less than one electrical charge equivalent across the membrane. In contrast, electrometric experiments on bacterial enzymes have shown that two electrical charges cross the membrane in each of these reaction steps, which is consistent with the mechanism discussed earlier (section ). For these reasons, Yoshikawa’s mechanism was recently further revised.[207] In the revised version, each electron transfer to the binuclear center is coupled to uptake of one proton from the side of the membrane to the BNC for the chemistry of O2 reduction, to uptake of one proton from the side into a hypothetical “proton pool”, and to ejection of one proton from the H-bond network in the Mg domain to the side of the membrane. When the “proton pool” is filled with four protons in the R state, they are all transferred to the H-bond network in the Mg domain prior to binding of O2 to the BNC. As noted by the authors, the distance between the “proton pool” and the H-bond network in the Mg domain must be very small to be consistent with measurements of electrical charge displacement during individual reactions of the catalytic cycle.

With these amendments the new mechanism proposed by Yoshikawa et al.[207] is consistent with time-resolved measurements of electron and charge transfer summarized above (section ). Nevertheless, we find the requirement of two adjacent domains capable of taking up and releasing four protons quite problematic. Moreover, the Mg domain that is particularly important in this alternative model is entirely lacking from the quinol oxidase cytochrome bo3, which nevertheless belongs to the A class of heme–copper oxidases, lacks the H channel, but is an efficient proton pump.

The four-proton collection mechanism is based on an observed conformational change in helix X of subunit I between the two heme groups that appears to allow protonic contact between the H-bonded network in the Mg domain and the H channel only when the binuclear center is fully reduced and unliganded but which is closed in all other cases.[214] However, these are relatively small changes in a static crystal structure and not unexpected as a result of the fairly large change in valence of heme iron and copper in the BNC. In the absence of dynamic data the link between this change to the mechanism of proton translocation is speculative despite the wide attention that it has received due to the spectacular structural X-ray work of the Yoshikawa group. We also note here that recent atomistic simulations on the entire 13-subunit bovine enzyme and on a smaller core (subunits I–III) do not support a role of the H channel in proton translocation but suggests that it may function as a “dielectric well”.[215]

Linkage of the Proton Pump to Oxidoreduction of Heme a

Very early on, Artzatbanov et al.[216] proposed a mechanism of proton pumping by cytochrome c oxidase where uptake of the pumped proton from the side of the membrane is linked to reduction of the low-spin heme a and subsequent release of the proton to the aqueous side linked to reoxidation of heme a. This theme is still maintained in the mechanistic proposals of Papa et al.[208] and Rousseau et al.,[209] but it has been experimentally excluded. As summarized by Siletsky and Konstantinov,[185] the time-resolved measurements of membrane potential show that the main phases of charge separation are linked to oxidation of heme a by the BNC, not to its reduction. In fact, in electron injection experiments the relatively small membrane potential generation linked to reduction of heme a can be unequivocally attributed to the electrogenic electron transfer from CuA to heme a.

In two papers published in 2000, Capitanio et al. investigated the release and uptake of redox Bohr protons coupled to oxido–reduction of cytochrome a and CuA in CO-liganded cytochrome c oxidase.[217,218] In the first paper they studied this in the isolated enzyme;[218] in the second paper they used both the isolated enzyme and the oxidase reconstituted in liposomes.[217] They found that a complete redox transition (2 electrons) of cytochrome a and CuA was linked to uptake or release of 0.4–0.7 protons, as recalculated from Figures 1 and 3 in ref (217). With the single exception of the experiment where cytochrome a is initially completely oxidized, H+/2e values around an average of 0.54 are found (note that this H+/2e is the sum of the H+/e of cytochrome a and the H+/e for CuA as both participate equally in the redox transitions). This number is intermediate between the values this group had reported earlier (0.65–1.0 H+/2e for CO-liganded oxidase)[219] and those obtained by Verkhovsky et al. (0.20–0.43).[220]

The present values for the Bohr H+/2e would imply a 24–41 mV/pH unit dependence of the apparent midpoint potentials, shared between cytochrome a and CuA. This is not very different from the known slight pH dependences of the midpoint potentials of cytochrome a (determined in CO-liganded cytochrome c oxidase, a decrease of ca. 10–26 mV/pH unit[221]) and CuA (a decrease of 9 mV/pH unit[222]). However, this demonstration of “redox Bohr protons” is certainly not sufficient for postulating that one or both of these two redox centers (cytochrome a and CuA) are involved in a proton pump. Also, it should be noted that a putative role of a redox reaction in a proton pump does not imply that the stoichiometry of the pump should be reflected in the stoichiometry of Bohr protons observed upon incomplete turnover of the pump mechanism.[183]

In addition to the demonstration of Bohr protons with the isolated enzyme, Capitanio et al. concluded from their experiments with CO-liganded cytochrome c oxidase reconstituted in liposomes that upon oxidation of cytochrome a and CuA these Bohr protons are released at the outside of the vesicles and upon reduction they are taken up from the inside.[217] They conclude this from experiments with these proteoliposomes in the presence of cytochrome c and a trace amount of mitochondria providing a way to slowly rereduce the enzyme (with succinate as the source of reducing equivalents) after oxidation with ferricyanide. What they observe measuring pH changes in the medium is (i) fast oxidation of cytochrome a with concomitant release of Bohr protons, (ii) slow rereduction with release of scalar succinate protons, and (iii) a pH drift interpreted as equilibration of protons putatively taken up from the inside of the vesicles during phase ii.

This interpretation, if correct, would have a very interesting consequence. Oxidation of cytochrome a and CuA by ferricyanide probably occurs via cytochrome c (see the comparable reduction of cytochrome a by ferrocyanide in the absence and presence of cytochrome c(223)). Rereduction by succinate catalyzed by the added trace amount of mitochondria will also be mediated by cytochrome c. Because electron transfer to the binuclear center is blocked by CO, the output of electrons (to ferricyanide) followed by their return (from succinate) result in proton translocation in this view. Reduction and reoxidation of cytochrome c would thus be sufficient for generation of ΔμH+, a genuine perpetuum mobile.

Conclusions in Brief

The mechanism of dioxygen activation in cellular respiration is today fairly well established compared to the situation 25 years ago.[141] Major advances include the role of the unique tyrosine residue in the active site that is covalently linked to one of the histidine ligands of CuB and which is conserved throughout the superfamily of respiratory heme–copper oxidases. The most recent results suggest that the catalytic importance of this residue considerably exceeds the original idea of assistance in the breaking of the O–O bond[63,180,181] and that it is of key importance in modulating the redox potentials of the catalytic site intermediates. These redox potentials are now established with reasonable accuracy from both experiment and calculations, and the remarkable “leveling effect” by the active site, relative to the potentials of O2 reduction in solution, is even more dramatic than originally estimated.[141]

The conservation of energy by means of redox-coupled proton pumping in CcO is also relatively well understood, at least in the biochemically and biophysically characterized bacterial oxidases of type A but, as we suggest, also in the mitochondrial type A oxidases. In particular, the time-resolved spectroscopic and electrometric approaches have been central in identifying the order of the electron and proton transfer events during a cycle of the proton pump. The current dominant view is that the proton to be pumped to the side of the membrane is transferred to a pump site near the heme a3 propionates in linkage to each electron transfer from the low-spin heme to the binuclear site and prior to the transfer of the proton to be consumed at that site. The protonation of the binuclear site finally lowers the pKa of the pump site with expulsion of its proton to the side. The latter is not trivial because especially in the presence of a protonmotive force this proton must be prevented from returning to the side by kinetic gating processes, some of which are beginning to be revealed not least by computational means. The proton-pumping mechanism seems to be very similar in the four one-electron reactions of the catalytic cycle but with at least one exception. Electron transfer from heme a to the binuclear site coupled to proton transfer to the pump site requires charge compensation from partial proton transfer in the K pathway for the case of the OH → EH transition (and possibly the EH → R step). This is not absolutely mandatory for the A → F and F → OH transitions possibly because of the considerably higher redox potentials of the binuclear site (and hence higher driving forces) in these cases (Table ).

Recent multiscale computational approaches have been beneficial in the modeling and simulation of transient high-energy states, which are difficult to characterize experimentally, thereby providing molecular insights into the available experimental data as well as making new predictions.