Research Article

Redox-Coupled Crystal Structural Changes in Bovine Heart Cytochrome c Oxidase

See allHide authors and affiliations

Science  12 Jun 1998:
Vol. 280, Issue 5370, pp. 1723-1729
DOI: 10.1126/science.280.5370.1723


Crystal structures of bovine heart cytochrome c oxidase in the fully oxidized, fully reduced, azide-bound, and carbon monoxide–bound states were determined at 2.30, 2.35, 2.9, and 2.8 angstrom resolution, respectively. An aspartate residue apart from the O2 reduction site exchanges its effective accessibility to the matrix aqueous phase for one to the cytosolic phase concomitantly with a significant decrease in the pK of its carboxyl group, on reduction of the metal sites. The movement indicates the aspartate as the proton pumping site. A tyrosine acidified by a covalently linked imidazole nitrogen is a possible proton donor for the O2 reduction by the enzyme.

Cytochrome c oxidase, a key enzyme in cell respiration, catalyzes the reduction of O2to water at the site involving heme a3 and CuB(heme a3–CuB site hereafter) by means of protons extracted from the matrix side of the inner mitochondrial membrane and electrons from cytochrome c, in a reaction that is coupled with proton pumping (1). The crystal structures of a eukaryotic and a prokaryotic cytochrome c oxidase previously reported at 2.8 Å resolution ushered in a new era for cytochrome c oxidase research (2-4).

For the proton pumping function of cytochrome c oxidase driven by the O2 reduction, an acidic group in the protein must be accessible only to one of the two bulk water phases on both sides of the mitochondrial membrane in a certain oxidation state of the enzyme, and the accessible side must be switched to the other side by a change in the oxidation state, concomitantly with a significant change in the pK of the acidic group. The changes in accessibility and pK can be induced by very small conformational changes in the protein. Thus, careful comparison of the crystal structure in various oxidation and ligand-binding states at high resolution is indispensable for understanding the mechanism of proton pumping by this enzyme.

Crystallization conditions have been improved to provide crystals that diffract x-rays up to 1.9 Å resolution (5). Crystals of the fully oxidized, fully reduced, fully reduced CO-bound, and fully oxidized azide-bound forms were prepared from the crystals of the fully oxidized form (5). Intensity data and phase determinations of these states are summarized in Table1. The crystal structures were refined with the program XPLOR (6). The statistics of the structural refinement of these crystals are given in Table2.

Table 1

Intensity data and statistics of phase determination for the fully oxidized, fully reduced, fully reduced CO-bound, and fully oxidized azidebound forms. Each structure was determined by multiple isomorphous replacement (MIR) (37).

View this table:
Table 2

Statistics of structural refinements of oxidized, reduced, CO-bound, and azide-bound forms. Each structure was refined by simulated annealing followed by positional and Bfactor refinements with XPLOR (6).

View this table:

Structure of the heme a3–CuB site. The electron density for His240 and Tyr244 of subunit I given as an (FoFc) difference Fourier map (7) at 2.3 Å resolution indicates the presence of a covalent bond between Nɛ2 of His240 imidazole group and Cɛ2 of Tyr244 phenol group (Fig.1A). This covalent linkage is also apparent in the electron density distribution of the fully reduced enzyme at 2.35 Å resolution and is consistent with the electron density maps of the fully reduced CO-bound form and the fully oxidized azide-bound form. The covalent linkage has been proposed also for the fully oxidized Paracoccus cytochrome c oxidase at 2.7 Å resolution (8). The covalent linkage suggests that the pK of the TyrOH is significantly lower than that of free TyrOH. Thus, a significant fraction of OH group of Tyr244 may be in the deprotonated form even in the physiological pH.

Figure 1

Crystal structure of Fea3-CuB site of the fully oxidized form at 2.3 Å resolution and a possible role of Tyr244 in the mechanism of O2 reduction. (A) The (F o − F c) difference Fourier map of the oxidized form calculated by omitting His240, Tyr244, and any ligand between Fea3 and CuB from theF c calculation. Contours are drawn at 7σ level (1σ = 0.0456 e3). (B) A possible mechanism for formations of the hydroperoxide intermediate and of the resting oxidized form. Protons (H+) are from the hydrogen bond network and electrons (e) from heme a. Only the O2 reduction site is shown. OH and Odenote Tyr244 and its deprotonated form.

The (F oF c) difference Fourier map (7) of the oxidized form (Fig. 1A) shows also a residual density between heme a3 iron (Fea3) and CuB. The refined model for the residual electron density in the heme a3–CuB site of the fully oxidized form (Fig. 1A and Table 3) is a peroxide ligand bridging the two metals (9). The Fea3-O distance is significantly longer than that found in Fe3+-OH compounds and is consistent with the high-spin heme a3 present on the oxidized form (10). The O-O bond length (1.62 Å) is slightly longer than that of H2O2. One hydroxide ion and one water have been proposed for the ligands of the heme a3–CuB site in the fully oxidized bacterial cytochrome c oxidase at 2.7 Å resolution (8). However, this model provides significantly higher residual density in an (F o − F c) difference map than the peroxide model, that is, the bridging peroxide fits the electron density at 2.3 Å resolution significantly better than the hydroxide/water model.

Table 3

Structures of the O2 reduction site in four different states. Positional and B-factor refinement was applied for each structure determination of O2reduction site with the program XPLOR (6,39).

View this table:

The fully oxidized form given here is the enzyme preparation as purified from the tissue under aerobic conditions. In analogy to hemerythrin, in which the peroxide form is the stable oxygenated form (11), the bridging peroxide structure is consistent with the stability of the fully oxidized enzyme (12). The effect of x-ray irradiation under the present experimental conditions is negligible (13). The enzyme form is characterized by much slower reduction of heme a3 (14) and much slower cyanide binding, compared with the enzyme under turnover conditions (15). Thus, the form may not directly participate in the catalytic turnover (16) and is called alternatively the resting oxidized form (17). The fully oxidized form directly involved in the catalytic turnover is called the oxygen pulsed form (18).

No ligand is detectable between Fea3 and CuB in the multiple isomorphous replacement and density modification (MIR/DM) map and the (F oF c) difference map of the fully reduced form at 2.35 Å resolution (not shown). The three histidine imidazoles form a triangle with the CuB on the triangle plane. Such a Cu1+coordination is usually very stable (19). Thus, the CuB 1+ is likely to be a poor ligand acceptor as well as a poor electron donor. This property may inhibit rapid formation of bridging O2 between Fea3 and CuB(Fea3 2+-O=O-CuB 1+), as has been proposed (1).

Inspection of interatomic contacts of the O2 within the O2 reduction site of the fully reduced enzyme shows that the hydroxyl group of Tyr244 is located close enough to form a hydrogen bond to the bound O2 at Fea3 2+ (not shown). This hydrogen bond would stabilize the oxygenated form. Furthermore, Tyr244, acidified by the covalent linkage to the imidazole nitrogen, donates, effectively, protons to the dioxygen reduction site through the hydrogen bond network connecting Tyr244 with the matrix surface (4). The proximity of heme a to heme a3, shown in the crystal structures of bovine heart and bacterial enzymes, suggests that heme a serves as an effective electron donor to heme a3, consistent with a resonance Raman result (20) confirmed by the absorption spectral change (21). Thus, reduction of O2 at heme a3, which is hydrogen-bonded to Tyr244, may be triggered by electron transfer from heme a to heme a3 to produce a hydroperoxo adduct bound at Fea3 3+ (Fig. 1B).

After formation of the hydroperoxo species (Fig. 1B), the third electron from CuB, via His240-Tyr244 linkage, and the second proton from the matrix side, via Tyr244, may be donated effectively to the bound hydroperoxide to form a ferryloxo species (Fe4+=O) and a water. However, the stable CuB 1+, as described above, may not donate the electron equivalent to the hydroperoxo adduct, but to Fe5+=O formed after H2O is released from the hydroperoxo adduct as suggested (22). In either mechanism, the life time of this hydroperoxo intermediate could be too short to detect consistently with the resonance Raman results (22). Addition of another three equivalents of electrons and protons to the hydroperoxo species drives the catalytic process, including the oxygen pulsed state (18), to reproduce the original reduced species (Fig. 1B).

The fully oxidized resting form may be formed from the enzyme species with Tyr244 in the deprotonated state as described above. The deprotonated tyrosine cannot form a hydrogen bond to the bound O2. In this case, the bound O2 may form a bridging O2 between Fea3 2+ and CuB 1+ to form Fea3 3+-O-O-CuB 2+(Fig. 1B) as the resting oxidized form. A possible physiological role of the fully oxidized resting form is to prevent the potentially very reactive O2 reduction site from producing radical species with accidental contacts of O2 to the site in the absence of sufficient amount of reducing equivalent in the enzyme molecule.

The distance between Fea3 and CuB in the refined crystal structure is 4.9 Å in the fully oxidized state, 5.2 Å in the fully reduced state, 5.3 Å in the fully oxidized azide-bound state, and 5.3 Å in the fully reduced CO-bound state. These four crystal structures indicate that the position of CuBdepends on either the oxidation state of the metal sites or ligand binding to Fea3, whereas Fea3 is fixed. Fea3 2+ in the fully reduced form without any ligand (in a five-coordinated ferrous high-spin state) is significantly closer to the porphyrin plane of heme a3 than in the case of deoxy myoglobins (23).

CO coordinates to Fea3 in a bent end-on manner (Fig.2A and Table 3). The oxygen atom of CO is 2.5 Å away from CuB and forms a distorted trigonal pyramid with the three imidazole nitrogen atoms in the base and the oxygen at the apex. This long bond length as a coordination bond suggests a very weak interaction between CuB and the CO bound at Fea3, which is consistent with the infrared results (24). The refined structure of the azide-bound form at 2.9 Å resolution shows that azide forms a bridge between Fea3and CuB (Fig. 2B). One end of the azide and three histidine imidazoles provide a distorted square planar coordination to CuB in a manner similar to the peroxide binding in the fully oxidized form. The coordination structure of CuB, which includes three imidazole ligands, is in contrast to the reported structure of the azide-bound form of the bacterial enzyme in which one of the histidine imidazoles is missing (3).

Figure 2

Crystal structures of the Fea3-CuB site in the fully reduced CO-bound and fully oxidized azide-bound states and those of the Na+/Ca2+ and Mg2+ sites. (F o − F c) difference Fourier maps for CO bound at Fea3 at 3σ level (A) and azide bridging between Fea3 and CuB at 2.2σ level (B) are given where the bound ligands and fixed waters are not included in theF c calculation. The difference electron density maps at 4σ level (1σ = 0.0436e3) for the Na+/Ca2+ site (C) and for the Mg2+ site (D). Violet, purple, and blue balls are the positions of Na+, Mg2+, and water, respectively. The colors of the electron densities denote the crystals from which the densities are obtained: purple (A), from CO derivative crystal at 2.8 Å resolution; cyan (B), from azide derivative crystal at 2.9 Å resolution; and green (C and D), from the fully oxidized form crystal at 2.3 Å resolution.

Other metal sites. An electron density peak with a trigonal bipyramidal coordination that includes peptide carbonyls of Glu40, Glu45, and Ser441, a water, and the side chain carboxyl of Glu40 (a counterion) is reasonably assigned to a Na+ site (Fig. 2C). The thermal factor for the metal site as a Na+ site supports this assignment (25). However, Ca2+ ion can be accommodated at the site by a small conformational change in the ligands nearby. This finding is consistent with the recent report for competition between Na+ and Ca2+ for binding to a common site (26). The metal site is apparent in all the crystal structures presented here. A corresponding metal site in the bacterial cytochrome c oxidase involves two counter ion ligands out of five ligands (8). The structure difference between bovine and bacterial enzymes could induce the dependency of the Ca2+ effect on the biological species, recently reported (26).

The difference maps at 2.3 Å resolution show a structure of Mg2+ coordinated by Glu198 of subunit II, His368 and Asp369 of subunit I, and three water molecules (Fig. 2D). Two out of three water molecules were not shown in the crystal structures previously reported at 2.8 Å resolution (2).

Neither the ligand binding nor oxidation states induced any detectable conformational change in the Na+/Ca2+ and Mg2+ sites.

Redox-coupled conformational change in the protein moiety. A segment from Gly49 to Asn55moves toward the cytosolic surface by about 4.5 Å at the position of the carboxyl group of Asp51 on reduction of the fully oxidized enzyme (Fig. 3). One of the amino acid residues in the segment, Asp51, in the fully oxidized state is completely buried inside the protein (Fig. 3). One of the oxygen atoms of the carboxyl group of Asp51, which is completely buried inside the protein in the fully oxidized state and does not contribute to the accessible surface calculated with a 1.4 Å probe, contributes to the accessible surface in the fully reduced state (Fig. 3) (27). No other significant conformational change is detectable in the electron density distribution of the protein moiety.

Figure 3

Redox-coupled conformational change in the segment from Gly49 to Asn55. The conformation of the segment in the fully oxidized form is stereoscopically shown in red, and that in the fully reduced form in green. The yellow structure denotes subunit II with no redox-coupled conformational change. The accessible surface for the fully oxidized state (35) is indicated by dots.

Asp51, in the fully oxidized state, is connected with the matrix surface by a network that includes a peptide unit, hydrogen bonds, a cavity, and a water path (Fig.4A). The peptide bond connecting Asp51 and Tyr54 with hydrogen bonds on both ends can act as a unidirectional proton transfer path from Tyr54 to Asp51, because it is in an equilibrium state between the two tautomers, -CO-NH- and -C(OH)=N-, where the former is much more abundant than the latter. Tyr54, a propionate group of heme a, Tyr371, a fixed water, and Arg38 are connected by hydrogen bonds. Arg38 is on the boundary of a large cavity that lacks any detectable electron density but is large enough to contain randomly oriented or mobile water molecules. The cavity is connected with the matrix surface by a water path, which includes some hydrophilic amino acid side chains of helices XI and XII of subunit I (4) and the hydrophobic farnesyl group of heme a. The diameter of the tube at the narrowest point is approximately three-quarters the size of water. However, molecular dynamics calculations have shown that small molecules can travel through spaces much smaller than their row radii in static models (28). Thus, Asp51 in the fully oxidized state can take up protons only from the matrix side but is inaccessible to the cytosolic side. The above hydrogen bond network was not described in (2), because Asp51 at the cytosolic end of the network is inaccessible to the bulk water without movement of the peptide backbone (29). This network interacts with heme a at several points (Fig. 4A), suggesting a function of the network under redox control.

Figure 4

Schematic representations of the structure of the proton pumping system and a possible function of Asp51 in the proton pumping by cytochrome c oxidase. (A) The hydrogen bond network from Asp51 to the matrix surface. Dotted lines denote hydrogen bonds. The small, dark circles indicate fixed waters. A thick stick denotes a side view of the porphyrin plane of heme a, and thin sticks (from top to bottom) are the side chains, propionates, formyl, and hydroxyl farnesylethyl. The hydroxyl farnesylethyl group is in the water path with its hydroxyl group hydrogen bonded to Ser461 and Thr424 via a water fixed on the boundary of the cavity. The porphyrin plane is a part of the boundary of the cavity. The formyl group is hydrogen-bonded to Arg38, and one of the propionate groups is hydrogen-bonded to Tyr371 and Tyr54. The Na+ site is also shown. (B) A possible mechanism for a unidirectional proton transfer by Asp51. Dotted lines denote hydrogen bonds. Direction of the movement of Asp51 side chain is shown by arrows. Deprotonated carboxyl groups are shown by broken lines.

On reduction of this enzyme, the hydrogen bond between the peptide amide of Ser441 and the carboxyl group of Asp51is broken, and Asp51 loses its accessibility to the matrix side through the hydrogen bond network. The conformation of the segment from Gly49 to Asn55 seems to be determined only by the oxidation state of the metal sites, because the fully oxidized azide-bound form shows the conformation identical to that of the fully oxidized form, and the fully reduced CO-bound form provides the conformation of the fully reduced form. These properties strongly suggest that CuB does not control the conformation of Asp51 and that the conformation of the Asp51segment of the resting oxidized form, which is not involved in the catalytic turnover (16, 17), is identical to that of the oxygen pulsed state, which is active under turnover conditions (18).

The above results suggest the following redox-coupled proton pumping mechanism (Fig. 4B). Asp51, in the fully oxidized state, is hydrogen-bonded with Ser205 of subunit II and Ser441 of subunit I each at both its hydroxyl group and its peptide amide. The protonated carboxyl group (–COOH) of Asp51 in the fully oxidized state become accessible to the bulk water to release protons on reduction. Meanwhile, the carboxyl group breaks three hydrogen bonds out of four, leaving the one to the hydroxyl of Ser441. A water molecule is fixed with hydrogen bonds to Ser205 at both its hydroxyl and peptide amide and to the peptide carbonyl of Ser202. On reoxidation of the enzyme, the fixed water is replaced by the deprotonated carboxyl group (-COO) of Asp51. The -COOextracts a proton of the peptide amide and creates a deprotonated peptide. The peptide is reprotonated readily by Tyr54, which is hydrogen bonded to the carbonyl group of the peptide providing the -C(OH)=N-form. Transition to the stable tautomer of the peptide and reprotonation of Tyr54 by means of a proton from the matrix side give the original structure of the fully oxidized state (Fig. 4B).

The redox transition of heme a, which interacts with the hydrogen bond network between Asp51 and the matrix surface as described above, is likely to contribute to driving the active transport of protons. The mechanism by which protons are released on oxidation of the enzyme, in contrast to the one presented in Fig. 4B, is also possible (30), consistent with biochemical results (31).

The present results suggest a proton pumping site apart from the heme a3–CuB site, in contrast to the direct coupling mechanism for the proton pumping in which one of the histidines liganded to CuB serves as a proton carrier (32). The latter mechanism is supported by the crystal structure of the azide-inhibited bacterial enzyme that lacks the electron density of one of the histidines coordinated to CuB (3). Iwata and colleagues proposed a possible proton transfer path to the proton pumping site, His290 (in the numbering of bovine heart enzyme), from Asp91 via Glu242 (3). On the other hand, the crystal structure of bovine heart enzyme shows the electron density distribution for His290 even in the azide-bound fully oxidized form as described above. In any forms of bovine heart enzyme, the hydrogen bond network from Asp91 ends at Glu242. No possible proton transfer path is detectable from Glu242 to His290 in the crystal structures of bovine heart enzyme. Furthermore, no fixed water connecting the two amino acids is detectable even in the crystal structures of bovine heart cytochrome c oxidase at 2.30 to 2.35 Å resolution. Thus, any crystal structure of bovine heart cytochrome c oxidase available at present does not support the proposal (33) for the proton transfer from Glu242 to His290 or to His291.

Glu242 is hydrogen bonded to the sulfur atoms of Met71, Sγ-O distance of 3.2 Å, in any crystal structure given here, indicating that the carbonyl group is in COOH form, regardless of the oxidation and ligand-binding states, which is consistent with the recent infrared investigation (33).

If the proton pumping pathway includes the O2 reduction site (the heme a3–CuB site), protons to be pumped must be sorted effectively from those for making water as suggested (32, 34). Otherwise, the protons to be pumped are used for making H2O at the O2reduction site. However, the crystal structure of bovine enzyme shows no structure suggesting such a function.

All the amino acids in the network connected to Asp51 in bovine heart enzyme are conserved in animals from human to sea urchin. However, Asp51 and Tyr54 are not conserved in plant and bacterial enzymes. These differences suggest an evolution in the proton pumping pathway.


View Abstract

Navigate This Article