Electron Transfer Pathways in Cytochrome c Oxidase

Abstract

Mixed quantum mechanical/molecular mechanics calculations, were used to explore the electron pathway of the terminal electron transfer enzyme, Cytochrome c Oxidase. This enzyme catalyzes the reduction of molecular oxygen to water in a multiple step process. Density functional calculations on the three redox centers allowed for the characterization of the electron transfer mechanism, following the sequence Cu_A → heme a → heme a₃. This process is largely affected by the presence of positive charges, confirming the possibility of a proton coupled electron transfer. An extensive mapping of all residues involved in the electron transfer, between the Cu_A center (donor) and the O₂ reduction site heme a₃-Cu_B (receptor), was obtained by selectively activating/deactivating different quantum regions. The method employed, called QM/MM e-pathway, allowed the identification of key residues along the possible electron transfer paths, consistent with experimental data. In particular, the role of arginines 481 and 482 appears crucial in the Cu_A → heme a and in the heme a → heme a₃ electron transfer processes.

1. Introduction

Cytochrome c Oxidases (ferrocytochrome-c:oxygen oxidoreductase, EC 1.9.3.1, hereafter known as CcO) act as the terminal enzymes of aerobic respiration in eukaryotic organisms and many bacteria. In the terminal reaction, CcO takes electrons from a soluble iron-containing electron transfer (ET) protein, cytochrome c, and passes them on to dioxygen (O₂) which is reduced to water following the reaction: O₂ + 4H⁺ + 4e⁻ → 2H₂O [1]. The free energy obtained from the reduction of O₂ is converted to an electrochemical proton gradient, across mitochondrial or bacterial membranes, that is subsequently used for the adenosine triphosphate (ATP) synthesis by the membrane bound H⁺-ATP synthase.

CcO are members of a large super-family of heme/copper containing proteins composed from 3 to 5 subunits in bacteria and up to 13 in mammalian mitochondria. However, only subunits I and II are essential for the function of the enzyme, which is consistent with the high degree of sequence conservation in these subunits for the various terminal oxidases. Subunit I contains two heme centers. The first of these, low-spin heme a, is thought to act as an electron input device to the second heme a₃. The high-spin heme a₃ is part of the binuclear center, with a copper atom (Cu_B), and is the site of dioxygen reduction. Subunit II, which contains another copper center (Cu_A) with two Cu atoms, has been shown to be the primary acceptor of electrons from reduced cytochrome c [2, 3]. Due to its location, within the membrane, the Cu_A site has constituted a challenge to full characterization because of the spectral interference from the heme groups and the inherent difficulties of studying a membrane bound enzyme. [4, 5] Only with the determination of the crystal structure of CcO, as well as of an engineered Cu_A domain, was it possible to ascertain the dinuclear nature of this center [6–8]. EPR spectral analyses have shown that Cu_A is a mixed-valence [Cu(II)/Cu(I)] complex in which an unpaired electron is equally distributed over the two Cu ions[9] (confirming the hypothesis suggested in 1962[10]). A high degree of delocalization onto two bridging cysteine sulfur atoms (Cys252 and Cys256 – Rhodobacter sphaeroides CcO numbering, used throughout and unless mentioned otherwise the residues are present in subunit I) is observed and some spin density on a nitrogen atom of a bonded imidazole group (His_II260). In addition to these bonds, a weak axial interaction with a sulfur atom from a methionine (Met_II263) and a carbonyl oxygen from a glutamate (Glu254) residue can also be found with the two copper atoms in the Cu_A center [11].

Electron transfer from cytochrome c occurs by electrostatic binding to subunit II, which should be followed by electron transfer through the Cu_A center to heme a or heme a₃, see Figure 1. Heme a is coordinated by two axial histidine ligands (His102 and His421) provided from different helices. On the other hand, heme a₃ has one axial histidine ligand (His419). Neither of the hemes is covalently bound to the enzyme (besides the coordination chemistry to the axial histidines) and although the generally accepted view that heme a is an intermediary between Cu_A and heme a₃[12, 13] the possibility for a direct electron transfer from Cu_A to the heme a₃-Cu_B site has not been excluded [14].

Figure 1.

Overall structure of Cytochrome c Oxidase from Rhodobacter sphaeroides. A - Possible electron transfer path shown. The redox active and inactive cofactors of CcO are displayed as well as the D- and K-pathways used for proton uptake. B - Relative position of all residues proposed to participate in ET process.

Considering the ET to begin with an electron transfer from Cu_A to heme a, several proposals exist in the literature with different initiating residues. After leaving the copper atom some suggest that the electron should proceed through His260 [6, 15, 16] via, possibly, arginines 481 and/or 482 of subunit I to the propionate ring A of heme a, B panel in Figure 1. Alternatively, the electron transfer can start at Cys252 [15, 17] or even Cys256 [16] and favoring these two paths is the large degree of electron delocalization on the sulfur atoms. Likewise, from heme a to heme a₃ several hypotheses exist for ET. Most start at His421 then following through the only residue between the two ligand histidines, Phe420 or by means of Val417, Arg418 and Tyr415 as possible intermediates [6, 14]. Another possibility considers a direct jump from the methyl group of the D ring of heme a to its counterpart on the heme a₃ [18] or even a through-space jump between two neighboring CH groups in either hemes [19]. There is also experimental work suggesting the presence of tryptophan and/or tyrosine radicals in the vicinity of the heme groups, associated with the formation of reactive intermediates [20–24]. Finally, it has also been considered that additionally to the path from Cu_A to Cu_B via heme a, a direct pathway to heme a₃ can not be ruled out [14]. In spite of all the research done in this area, the exact path for electron transfer remains unknown.

The most common method to study this kind of process consists of mutational experiments. However, even though this technique, in principle, points out the importance of some residues within a protein its results can also be misleading since mutations can produce changes in the original structure that may mask the real cause for the loss/gain of activity or, in this particular case, electron transfer. Alternatively, computational tools can provide a quick and cheap alternative in mapping electron transfer pathways with the key advantage of not introducing significant perturbations on the system [25–27]. Much effort has also been dedicated to the assessment of the time evolution of an electron transfer process by diverse approaches as described in the following work and references within [28, 29]. In the present work, however, we will explore the electron transfer pathway, in Cytochrome c Oxidase, by means of a new computation technique [30], that has already shown to give good results [31]. This method, QM/MM e-pathway, consists basically in activating-deactivating different regions in the quantum region, by means of a QM/MM scheme, which allows us to follow the progress of an electron and so describe its path between a donor and acceptor. The theoretical analysis indicates a mechanism for the electron transfer that follows the sequence Cu_A → heme a → heme a₃, which is significantly affected by electrostatic perturbations of the active site - i.e. to proton translocation. We also identify the key residues along the electron transfer pathway between these redox centers.

2. Computational Methods

2.1 QM/MM

The subjacent concept to mixed Quantum Mechanics/Molecular Mechanics (QM/MM) is to use different levels of theory to model diverse parts of the system. The main advantage of these methods consists of the ability of adequately describing large systems such as proteins [32–38]. The reactive part (for example the active site) is usually treated with a robust high level method and so providing the possibility of a full quantum treatment, essential for bond breaking/forming processes or electron transfer. The remaining part of the system is included by means of a lower level as Molecular Mechanics offering the adequate structural constraints as well as the electrostatics and the van der Waals interactions with the core reacting region. In this work, the QM/MM calculations have been carried out with the QSite 4.5 program which is part of the Schrödinger Suite [39]. The B3LYP hybrid density functional method has been used for the QM part of the system [40–44]. A mixed basis set was used with the double ζ basis set 6–31G* [45] for all atoms except the metals where the lacvp* pseudopotential [46] was employed. The OPLS-AA [47, 48] force field was used for the treatment of the Molecular Mechanics part.

2.2 QM-MM e-pathway

The QM/MM scheme is fundamental for the electron transfer search path algorithm that will be described shortly here. The methodology is based on the idea that regions contained in MM part can not allocate explicit electron density given that the smallest unit is the atom. So, if an unpaired electron is present in the system it must necessarily occupy the quantum region. This said, the strategy consist in including in the QM part all residues that are considered to be in the “transfer region” (TR), that is the amino-acids contained between the redox centers. The donor and acceptor must be in the oxidized from, meaning that the electron has already left the donor but still hasn’t reached the acceptor. The first step consists of building the MM parameters for both donor and acceptor in the oxidized form, that is to say each redox center is fully (QM) optimized separately in the correct state and placed in the MM region, keeping their QM geometries and ESP charges. Thus, these groups are removed from the quantum section and only the transfer region is used where an extra electron is injected and a doublet state specified. A single point calculation (at the HF/6-31G* level to improve spin localization) is performed and the residue that contains the extra spin density is identified. This residue is then removed from the QM region and a second iteration is performed. The spin density must now find another host and this procedure is repeated until a complete path between the donor and acceptor is found. This method was previously tested and proved to produce electron transfer pathways consistent with previous experimental and theoretical results at an inexpensive computational cost [30, 31].

2.3 System setup

We have started by selecting an adequate crystal structure, the wild type 2.3 Å resolution Cytochrome c Oxidases from Rhodobacter sphaeroides [49]. From the crystal, we modeled the oxidized ferric/cupric state (O), an intermediate known to accept electrons from cytochrome c, a reduction that is expected to be accompanied by the uptake of protons by the enzyme. The O state was modeled with an hydroxyl group as the sixth ligand of heme a₃ and a second hydroxyl group added to the Cu_B center . Preparation of the system included solvation of the protein with a water box of at least 12 Å layer of waters (~97×92×108 ų) and neutralization using a 0.150 mM NaCl salt. Due to its proximity to the heme propionates, Asp407 was neutralized (all other ionic groups were kept charged). Molecular dynamics simulations were run using NAMD v2.6 [50]. The protein was described with the CHARMM22 force field [51] while the parameters for the redox centers have been taken from the following references [52–54]. The system was initially minimized to remove possible bad contacts, then heated to 300 K and equilibrated during 300 ps. A 2 fs time step was used with lengths for all bonds containing hydrogen fixed using the SHAKE algorithm [55]. Throughout all metal centers and ligands were maintained frozen.

For the QM/MM calculations the system was reduced, maintaining only an 8 Å water layer around the protein resulting in a total of 29088 atoms. Each individual redox center was fully optimized separately, at both reduced and oxidized states. The Cu_A site contains a total of 104 atoms including the sequence Cys252, Ser253, Glu254, Leu255 and Cys256. The side chains of Met263, His217 and His260 were also added and, naturally, the two copper atoms. This leads to a total charge of 0 and doublet for the oxidized state and total charge −1 and singlet state for the reduced Cu_A site. The QM section of the heme a center is comprised of 96 atoms with the two side chains of histidines 102 and 421 and the heme, including the propionates but excluding hydroxyethylfarnesyl group beyond C11 atom. The total charge for this system is −2, for the reduced state, with two negative charges coming from the propionates, two from the porphyrin nitrogen atoms and +2 from the iron. The reduced low spin iron is in a singlet state. The oxidized state exhibits a −1 global charge with a +3 iron atom in doublet state. The heme a₃-Cu_B site contains 136 atoms in the QM section, including the heme truncated in a similar manner as heme a, the histidine 419 side chain and the iron atom for the heme center. The Cu_B site includes the side chains of the three histidines 284, 333 and 334 as well as the covalently linked (to His284) tyrosine 288. Additionally, two OH⁻ groups are coordinated to the iron and copper atoms. In oxidized state the high spin iron has 5 unpaired electrons (sextet state) and the copper is +2 with an extra uncoupled electron leading to a total charge of the QM partition of −1. The reduced states with a total charge of −2 have been explored considering either the iron or the copper reduced giving way to a total sextet state with the charges of either Fe³⁺/Cu⁺ or Fe²⁺/Cu²⁺. These centers were all fully optimized individually at the DFT/6-31G* level and convergence criteria were set to default values. Once the structures were fully optimized, the point charges were localized at each atomic center and were derived from molecular electrostatic potential (ESP). These structures and charges were used in all QM/MM calculations always maintaining the redox centers frozen when present in the MM partition and non-bounded cutoff of 100 Å were used to ensure a correct electrostatic description. Energy calculations were done by combining all these sites into one quantum region and single point calculation performed.

When studying the electron transfer pathway, the initial transfer region (between the redox centers) that contained 329 atoms was allowed to relax through 7 cycles of QM optimization in neutral state (24 hours of 32 processors), previous to the addition of the extra electron (a full optimization was computationally limited given the size of the system). This was necessary as we noticed that without this step the spin density was extremely delocalized, probably due to a poor force field structure. This situation changed upon only 3–4 cycles of optimization leading to highly localized spin densities. These were, in most cases, very well limited to a certain part of the residues, either the side chain or the backbone.

The boundary regions between the quantum and classical atoms were treated using a capping (link) atom, namely a hydrogen atom to satisfy the valence of the quantum chemical system [56]. Attention was paid to ensure that any relevant part of the protein was included in the QM partition. For example, in the case of the transfer regions it was essential that the spin would not localize on any of the frontier residues so avoiding any incorrect description of these limiting residues. Additionally, special care was place in not breaking any π-conjugated system (peptide bond, etc,) with a hydrogen atom link. Although the DFT wave function is strictly speaking not an eigenfunction of the operator S², the evaluation of <S²> still provides a useful control of the quality of wave function. Minor spin contamination occurs however never exceeding 0.37 % in the case where all metal centres are included. When the transfer region is considered the value of <S²> never exceeds 0.762 (ideal value of a pure doublet spin eigenstate is 0.750).

3. Results

In the first part, results from the study of the redox centers will be presented. By combining all active redox sites in the quantum region we were able to map the electron progress. The occupied orbitals are described and the energetics presented. The second part focuses on the transfer region between these centers by applying the QM/MM e-pathway procedure. In this case, all metals are removed from the quantum section and kept as point charges in the molecular mechanics section leaving only the connecting amino-acids in the so called transfer region. The highest affinity residues are identified and a complete path between Cu_A to heme a₃-Cu_B is offered.

3.1 Redox centers

In order to obtain an estimate of the relative energies for the different redox centers we performed geometry optimizations. These calculations were done separately for each site (Cu_A, heme a and heme a₃-Cu_B) given that a full optimization with all centers would be prohibitive. The Cu_A has been modeled to include in the quantum region both metals and the residues 217, 252, 253, 254, 255, 256, 260 and 263. Both hemes include the iron atom and the porphyrin ring with the hydroxyethyl farnesyl group truncated between the C11 and C12 atoms. In heme a the two binding histidines were added. For the heme a₃-Cu_B site all residues bounded to both the iron and copper atoms were included, namely the histidines 419, 284 (and the covalently linked tyrosine 288 to this histidine) 333 and 334 as well as the two OH⁻ groups, as the heme a₃ and Cu_B ligands used to model the O state. Finally, single point calculations were performed combining all sites in the quantum region and an electron added or removed depending on the desired final state.

Once cytochrome c electrostatically binds to subunit II, an electron is expected to move to the Cu_A center. This will be our starting point. We are interested in depicting the moment when the electron leaves from Cu_A to the next site and for this reason the Cu_A site will be in its reduced geometry while both heme sites are in their oxidized states. The structures were optimized at the DFT level and the charges calculated. The calculations were performed considering the enzyme in the oxidized state (O) but the model applies to other parts of the cycle as well. We have optimized the Cu_A center in both the oxidized and reduced states and observed, as would be expected [57, 58], minor differences between the two structures. The Cu-Cu distance varies from 2.80 Å to 2.73 corresponding to the oxidized (delocalized mixed valence state) and reduced state respectively. Minor changes take place in the Cu-S (Cys) distances where we observe a slight increase (inferior to 0.1 Å) in bond length from oxidized to reduced state (all distances can be viewed in supporting information, SI). We find the ESP partial charges on copper to be 0.43 and 0.53 for the oxidized structure while the reduced exhibits 0.39 and 0.43. The atomic spin densities were determined to be 0.13 and 0.28 for the two copper atoms in the mixed valence state with spin density distributed also on the ligand sulfur atoms (0.13 Cys256 and 0.33 Cys252) and residual density on the histidine nitrogens (0.020 and 0.043). Only minor spin contamination was detected (0.755). Naturally, no spin density is found in the reduced state. The next center to be optimized was heme a which presents, in the oxidized state, an Fe³⁺ low spin center with one unpaired electron. Concerning the heme a₃, three situations have been explored: low spin (equivalent to the heme a) and high spin, both coupled and uncoupled to the copper Cu_B atom. According to experimental evidence [59, 60] the iron atom should exhibit a high spin state but our calculations show that the lowest energy state to be the low spin. However, care must be taken as the theory level seems to play an important role in accessing the correct spin state. In the case of a cobalt system, for instance, it has been recently reported [61] that the lowest energy spin state obtained with DFT was not necessarily the correct one. Given this discrepancy between experimental and computational results we have decided that it would be important to perform all studies in the high and low spin states. The results show that the changes in geometry driven by the spin state do not introduce any differences in the electron transfer processes and that only minor energy changes are observed between the redox centers. For this reason, and given the experimental evidence, all calculations shown within this paper, were performed in the high spin uncoupled state (the low spin state results are shown in SI).

The distance between Fe_a3 (high spin uncoupled state) and the Cu_B atom is 4.75 Å for the oxidized state, in agreement with experimental data [16, 49, 62]. Considering a QM region containing both hemes a and a₃ as well as the two copper centers (Cu_A and Cu_B) we have carried out single point calculations on the previously optimized structures. We intended to compare the relative energies when the electron travels from center to center maintaining the initial structure. For this, we have constrained the initial guess to the spin population of interest in each fragment by assigning the desired charge and spin multiplicity. That is to say, for example, if we wish to observe the spin density localized on the Cu_A center, the first point of the electron path, we must assign a +1 charge to both copper atoms and a spin multiplicity of 1. This guides Jaguar (37) to generate a high quality initial guess wavefunction biased to a reduced state on the Cu_A center. By repeating this procedure on all metals we are now able to compare the relative energies and the affinity of each center. In Figure 2 we can observe the spin-density plot for CcO when the electron is present in the Cu_A site. We do not detect any spin density in the Cu_A site given that the extra electron actually “fills” a hole present in the oxidized state where both copper atoms have a formal charge of +1 with complete d¹⁰ orbitals.

Figure 2.

Unpaired electrons spin-density plot for a system where the Cu_A site is reduced and both heme a and heme a₃-Cu_B are oxidized.

Heme a and a₃ contain an Fe³⁺ d⁵ ion. The strong octahedral field induced by the heme a ligands gives way to low spin complexes and so the unpaired electron shall occupy one of the T_2g orbitals. The highest occupied orbital is located at the bisection of the axis in the porphyrin plane (nitrogen atoms), corresponding to the d_xy orbital. The high spin iron atom from heme a₃, with an unpaired electron in each d orbitals presents a spherical spin density, together with residual spin density in each of the six octahedral ligands. The atomic spin densities for both these atoms are 1.12 for low spin heme a-Fe³⁺ and 4.05 for high spin heme a₃-Fe³⁺. In the Cu_B center, the Cu d_xz orbital interacts with one of the oxygen p orbital lobes of the coordinated OH⁻ group and the σ* nitrogen orbitals from the three ligated histidines.

We are now interested in observing the changes in spin density and energies as the electron flows from the Cu_A center to the acceptors. With Cu_A relaxed to the reduced state but heme a and a₃ still in the oxidized geometries we are able to identify the moment when the electron leaves the donor. Figure 3A shows the system where the electron has moved to heme a leaving a hole in Cu_A. The spin density of Cu_A center, in the mixed-valence [Cu(II)/Cu(I)] state, shows a spin density equally distributed over the two d_xz Cu orbitals, as would be expected [9]. Spin density is also observed on both sulfur p orbitals from Cys252 and Cys256 and the nitrogen atom from His217 and His260.

Figure 3.

Unpaired electron spin-density plots for a system where: the Cu_A and Cu_B sites are oxidized and A) heme a reduced and heme a₃ are oxidized. B) heme aoxidized and heme a₃ reduced.

The reduced heme a presents an Fe²⁺ metal center which does not show any spin density (no unpaired electrons); the heme a₃ site displays an identical spin plot as the previous case, left A panel in Figure 3. The same procedure was repeated for heme a₃ and, in this case, we can observe that the spin density corresponding to the 4 unpaired electrons remaining on the reduced state exhibit a modified spin plot. The change in the shape of the spin density to a less spherical contour reflects the uneven occupation of the d orbitals (Fig. 3B). On the other hand, it is very clear that the electron has left the heme a, now in the Fe³⁺ state, confirmed by the spin density present on this center. The spin population on the Cu_A site is identical in both situations.

Finally, we have constrained the initial guess so that the electron should be localized on the Cu_B atom. Figure 4 depicts this situation corresponding to the final state in the electron transfer chain. Little spin density is observed on the Cu_B site as a consequence of the extra electron “filling” the hole from the oxidized state. Also the OH⁻ group bounded to copper contains less evidence of spin density.

Figure 4.

Unpaired electron spin-density plots for a system where the Cu_A and both heme a and heme a₃ are oxidized and Cu_B is reduced.

The summation of all partial charges, for each section of the quantum region, is shown in Table 1; the total partial charges agree with the above orbital analysis. The labels: Red Cu_A, Red heme a, Red heme a₃ and Red Cu_B refer to the calculation where initial spin population guess has been constrained to the reduced Cu_A, heme a, heme a₃ and Cu_B, respectively. For the initially constrained Cu_A reduced state, we observe a −1 charge. The same fragment in the other situations exhibits a total charge of zero, in agreement with an oxidized Cu_A center which contains a Cu²⁺ and Cu⁺ atoms and negative charges from the two cysteines and the Glu_II254 residue adding up to 0 formal charge. From Table 1 we can also see that for the oxidized heme a center the total charge is about −1 while in the reduced state it is −2. This charge is again consistent with the iron +2/+3 oxidation state, corresponding to the oxidized/reduced states, plus the −4 charge from the porphyrin group (2 from the propionates and 2 from the deprotonation of the porphyrin nitrogens). In the heme a₃-Cu_B site it is also patent from the charge distribution when the electron is present in the heme and when it moves to the copper center. Thus, summing up the ESP charges for the different redox centers confirms the localization of the electrons observed in the previous orbital analysis. In the heme a₃-Cu_B site case there is a larger degree of delocalization, as it had been observed by the spin density plots, due to the close proximity of the redox centers. In agreement with our findings the perturbed heme a₃ spectra observed by Ji et al. [60] could be explained by the delocalization of this single electron between the heme a₃ and Cu_B centers. This one electron reduced state, termed the E state, has been reported to have an electron distribution among the redox centers [63–65].

Table 1.

Total ESP charges for each fragment in the quantum region. The values in bold indicate the presence of an extra electron in the initial guess for that site. All sections with grey background specify that all sites in these calculations are in their oxidized geometries except for Cu_A which is in its reduced geometry. Using these geometries we display the energy differences between the acceptor sites and Cu_B with and without an extra potassium ion added near to the heme a₃ propionates. The last part of the table contains data using reduced optimized geometries. The energy differences (in kcal/mol) between the heme a₃-Cu_B site and the heme a were obtained for: 1) both hemes in their reduced geometries and Cu_B oxidized; 2) heme a and Cu_B in reduced geometries and heme a₃ oxidized.

		Red CuA	Red heme a	Red heme a3	Red CuB
Charges	CuA	−1	0	0	0
	Heme a	−1	−2	−1	−1
	Heme a3	−2	−2	−3	−2.4
	CuB	1	1	1	0.4
Energy diff. (kcal/mol)	without K+	35.5	−44.1	1.1	0
	with K+	66.5	−17.2	7.8	0
	1	-	−24.9	-	0
	2	-	−1.0	-	0

In the second part of table 1 (upper section), the differences in energy between the all sites and the Cu_B site are shown. In the first row (without K+) we can observe that the reduced heme a state is found ca. 80 kcal/mol lower than the reduced Cu_A site indicative of a higher electron affinity for the heme. Likewise, the reduced heme a₃ and Cu_B sites lie about 35 kcal/mol below the reduced Cu_A. The highest energy lying fragment is the Cu_A, which indicates a lower affinity, consistent with the relay function attributed to this center. According to these values the electron is more stable in the heme a, so that we would not expect the electron to proceed to heme a₃.

Based on Belevich et al. [66], the electron transfer from heme a to the binuclear center should follow a proton transfer to an unidentified protonatable site above the hemes. For this reason, we have perturbed the system in order to model the presence of an extra proton in the vicinity of the a₃ propionates. A hydrogen bonded water molecule, present in the middle of the carboxylate groups, was altered to a potassium atom (included in the classic part) and the energies recalculated without any further optimization. These differences between the energies of each reduced fragment and Cu_A are present in the last row of Table 1 (with K+). We can see that (despite the simplicity of the calculation) the presence of a positive charge in the proximity of the a₃ propionates has an accentuated stabilizing effect on the heme a₃ and Cu_B sites. The heme a now lies ~84 kcal/mol below the Cu_A only 4 kcal/mol more stable while the heme a₃ presents a drop in energy grater than to 24 kcal/mol. The drop is even more enhanced for the Cu_B center, where another 31 kcal/mol are observed.

These modeling efforts aimed to address the electron affinity at the initial stage of the electron transfer, when Cu_A is reduced and the other redox centers are oxidized; the calculations were performed keeping the oxidized geometries for the hemes and Cu_B centers (and no additional K⁺ ion). The final electron location however, will depend on the affinity of the centers in the reduced optimized geometries. Thus, additional calculations using reduced geometries were necessary. We began by optimizing heme a and heme a₃ to their reduced geometries. In these conditions heme a now lies 25 kcal/mol below heme a₃-Cu_B (~44 kcal/mol when both sites are in their oxidized geometries). Contrary to the first case, when both heme groups were present in their oxidized geometries, we could not localize the electron in heme a₃ (heme a always recover it, indicating its high electronic affinity) but only in Cu_B. Thus, in the last two rows in table one we could only fill the heme a and Cu_B fields. Next, we optimized the system considering the Cu_B site also in its reduced state. Here we observe a chemical change with a spontaneous proton transfer from the OH⁻ group bounded to the iron atom to the hydroxyl coordinated to the Cu atom, forming a water molecule at the copper center (a movie for this process can be found in SI). This electron-proton coupled transfer leads to differences in the electronic affinity of each center where now Cu_B and heme a have degenerate affinities (last row in table 1). These results seem to point that the electron and proton transfer might be coupled in order for the electron to reach the final Cu_B metal center, an extended topic of discussion in cytochrome c oxidase chemistry [67].

3.2 Electron transfer pathways

In this section, the potential electronic paths between Cu_A and Cu_B are investigated using the QM-MM e-pathway scheme. The goal consists of identifying all possible active residues between the donor and acceptor. QM/MM calculations were performed to optimize the geometry of all the residues around the metal atoms in their oxidized form. The ESP charges were calculated at the DFT level of theory and were used on these atoms when included in the MM region. The system was then ready for application of the QM-MM e-pathway scheme.

3.2.1 Cu_A to heme a electron transfer

With both Cu_A and heme a in their oxidized states and geometries included in the MM region, 18 residues have been selected to be introduced in the QM area, as well as the propionate groups from both heme groups, counting a total of over 300 atoms. The residues: Gln251, Cys252, Ser253, Glu254, Leu255, Cys256, Gly257, Ile258, Ser259, His260, Ala261, Tyr262, Met263, Pro480, Arg481, Arg482, Tyr483 and Ile484 were elected in order to include all amino-acids in the intermediary area, between Cu_A and heme a, and taking into account information from the literature about residues known to be important for this electron transfer pathway as depicted in Figure 5.

Figure 5.

All residues included in the QM section are shown. The heme propionates have also been included but the remaining part of the heme (part shown in grey) was included in the MM region as well as the rest of the protein. Each residue’s backbone is shown in “think” colored representation with the respective sidechain in “thin” grey.

The first residue to be identified, with the highest electron affinity, is Arg482 and smaller amounts spread through propionate D from both heme a and heme a₃, pale blue in Figure 6. When we remove the side chain of Arg482 from the quantum region, the spin density is now observed in the side chain of Arg481 and some small amount on the propionate groups, purple in Figure 6. The next residue to be identified is in the vicinity of the Cu_A center, Ser253 (and also some on Glu254, displayed in orange in Figure 6) which lies immediately next to Cys252. Once this residue is removed, the spin density moves to the backbone of both Arg481 and Arg482 (green in Figure 6). Finally, the full removal of these two residues leads to spin density on the backbone of Pro480 (yellow in Figure 6), completing a possible electron transfer pathway between Cu_A and heme a. We should note that in these calculations the electron is extremely localized, normally found in π conjugated molecular orbitals of each residue.

Figure 6.

Electron transfer pathway for Cytochrome c Oxidase between Cu_A and heme a.

Interestingly, not including the heme a₃ propionate groups in the quantum region at the beginning shifts the first spin density to Arg481. Arg482 appears as the second residue. Proceeding with this study in the same manner as the previous situation, the presence/absence of the a₃ propionates does not induce any more changes. Not including the propionates into the quantum region involves a fix charge description at the propionate oxygens, resulting into a less flexible modeling of the system polarization. The shift of spin density from Arg481 to Arg482 indicates the plasticity of the system in diverting the electron flow as a response to the electrostatic environment, in agreement with the previous results where we perturbed the system by adding a positive charge (the potassium addition shown in table 1).

3.2.2 Heme a to heme a₃ electron transfer

Once the electron transfer path between Cu_A and heme a was established, we proceeded by changing the transfer region to be included in the QM part as to incorporate the adequate residues between the two heme groups (see figure 6 SI). All residues bridging the porphyrin rings were added, as well as the two arginines 481 and 482 and the two propionate groups. The first residue to be highlighted is Arg481 side chain as well as Arg482 and the D propionate groups of the two hemes (red color in Fig. 7). Removing the side chain of Arg481 highlights the bridging Phe420 and the two heme histidines (419 and 421) (blue). We have also investigated if the insertion of the two tryptophans (172 and 280)[20–23] in the vicinity of the two heme D propionates had any effect on the spin density of the residues within the transfer region. When including both tryptophans we observe very small amounts of spin density on Trp172, but most of the spin is still located on Arg481, as was the case when these residues were not included. Furthermore, we have repeated the calculations with the inclusion of a potassium ion halfway between the heme a₃ propionates (as done for the energy calculations previously) which leads to no changes in the spin populations with the arginines still showing the highest affinity for the electron. The complete electron transfer pathway can be seen in Figure 7.

Figure 7.

Electron transfer pathway for CcO between heme a and heme a₃.

4. Discussion

The energetic analysis suggest that the reaction from Cu_A to the binuclear center should occur in two sequential steps via heme a. Our calculations show that for the oxidized O state, the most favorable site for the electron is the intermediate heme a site. As seen in table 1 adding one electron to the heme a center results in 45 kcal/mol lower energy than the addition to the heme a₃-Cu_B. When considering the reduced geometries however, this energy gap is reduced to degeneracy between both centers. This degeneracy, however, is only accomplished when a proton is coupled to the electron transfer. When reducing the Cu_B site we observe a spontaneous proton transfer from the Fe-OH group to the Cu-OH one, indicating that the electron transfer from heme a to the heme a₃-Cu_B site is coupled to a proton transfer. We should emphasize here that our models do not include an extra proton in the system. Even in absence of the extra proton, the reduced Cu_B abstracts one proton from the iron hydroxyl group. While this might not be a real chemical step in the process (where protons are pumped to the binuclear center), it clearly indicates the necessity of a proton to stabilize the electron transfer. It might also indicate that the hydroxyl group bound at the iron might be part of the proton transfer pathway to the copper center. Furthermore, the importance of a proton transfer along the Cu_B reduction is further confirmed by the strong stabilizing effect of a positively charged environment in the proximity of the propionate a₃ (table 1), a result in agreement with the principle of electroneutrality introduced by Mitchell and Rich [67].

The results show the importance of Cys252, Ser253, Glu254, Pro480, R482 and the D propionate group of heme a in the first electron transfer between Cu_A and heme a. Most likely, not all these residues should be involved in the electron transfer by physically hosting the electron. Some (or even all of them) might simply act as bridge mediators between a longer through space jump [68]. Electron tunneling in biological systems has been the subject of much research through the years [69]. According to some experimental studies, His260 seemed to be the preferential residue for the beginning of the pathway which does not concur with our findings [70]. Nevertheless, experimentally observed spin density on both sulfur cysteine atoms support Cys252 as a possible precursor for this ET process. The highly conserved subunit I Arg482 is expected to be part of the electron transfer to heme a propionates [71]. For example, a reduction from 93000s⁻¹ to 50s⁻¹ in the electron transfer rate from Cu_A and heme a has been observed in the R482P mutant [72]. The same study concludes that the Arg482 side chain and the peptide backbone largely influence the rate of electron transfer to heme a. Similar observations on the role of Arg482 have been made by other studies mapping tunneling ET pathways by means of semiempirical methods [18]. According to our results Arg482 should have an active part in the Cu_A to heme a electronic migration and the backbone is expected to play an important part as well. This point demonstrates the capabilities of our new technique, QM/MM e-pathway, to map important residues along electron transfer pathways. It has also been seen, by adding/deleting the propionate groups to the quantum region, that the affinity of the arginine pair for spin density is dependent on the environment. A small change in the electrostatic description switches the propensity of the process between the two arginines, which might open a possibility to alter the final destination of the electron flow to either heme a or heme a₃. The active role for the propionate groups seen here, in the ET, acting as a vehicle of electrons and not only as an electrostatic anchor for binding, has also been underlined previously in several heme systems [30, 31, 73–75].

The second step involves the electron transfer from heme a to heme a₃. In the first QM/MM iteration, we find Arg482, Arg481 and the two propionate groups. Thus, one could imagine a direct propionate-propionate ET assisted (bridge mediated) by the arginine. Arg481 could also act as an intermediate, a vehicle for the electrons in their passage between propionates. However, mutational studies seem to favor the first mechanism. Mutating the arginine into a lysine, R481K, shows normal activity while in most of the other cases the activity is reduced to less than 10% [72]. Lysine side chain does not present a conjugate system, as opposed to arginine, and it is not a good electronic conductor. The distance between the two heme centers is within that of a direct electron transfer between the metal centers. Phe420, a phenylalanine in between both irons, appears in the second QM/MM e-pathway iteration. Mutation of Phe420 leads to a partially active enzyme which is consistent with this residue acting only as bridge mediated direct metal-metal ET [76]. Theoretical kinetic studies should confirm the different rates for these two ET mechanisms [77, 78].

Finally, it has been shown that the ion pair D-propionate (from heme a)-Arg481 can be affected by the proton pumping from the D-channel [79]. Our results supports that electron transfer through heme a might be associated with proton/electron cooperative coupling at this center, which could explain the presence of heme a in oxidases as a possible intermediate towards the binuclear site [80].

5. Conclusions

QM/MM methods, together with a novel algorithm (QM/MM e-pathway) were used to map the electron transfer process between the three redox centers in Cytochrome c Oxidase. Our calculations confirmed a two step ET process following the Cu_A → heme a → heme a₃ sequence. On the first step Arg482 delivers the electrons to the metal center by means of their propionate groups. For the heme to heme electron transfer we support the existence of two different pathways[18]. One would involve directly the heme a propionate-heme a₃ propionate contact, mediated by Arg481. The second pathway would involve directly the metal centers, mediated by Phe420. The overall process as seen through the energetic analysis is very sensitive to the additional presence of positive charges in the vicinity of the propionate groups, supporting a proton-electron coupled mechanism.