Crystal structure of CO-bound cytochrome c oxidase determined by serial femtosecond X-ray crystallography at room temperature

Significance A fundamental bioenergetics question is how the redox energy from oxygen reduction in cytochrome c oxidase (CcO) is coupled to proton translocation. Although crystallography has revealed many properties of the enzyme, the coupling mechanism remains elusive, in part due to artifacts caused by X-ray damage from synchrotron radiation. We determined the structure of CO-bound bovine CcO by serial femtosecond X-ray crystallography (SFX) with an X-ray free electron laser, outrunning radiation damage. The SFX structure is compared with the structure obtained with synchrotron radiation in which the Fe–CO bond was cleaved. Our results reveal that ligand binding to heme a3 triggers an allosteric structural transition that may play a critical role in proton translocation in the CcO family of enzymes. Cytochrome c oxidase (CcO), the terminal enzyme in the electron transfer chain, translocates protons across the inner mitochondrial membrane by harnessing the free energy generated by the reduction of oxygen to water. Several redox-coupled proton translocation mechanisms have been proposed, but they lack confirmation, in part from the absence of reliable structural information due to radiation damage artifacts caused by the intense synchrotron radiation. Here we report the room temperature, neutral pH (6.8), damage-free structure of bovine CcO (bCcO) in the carbon monoxide (CO)-bound state at a resolution of 2.3 Å, obtained by serial femtosecond X-ray crystallography (SFX) with an X-ray free electron laser. As a comparison, an equivalent structure was obtained at a resolution of 1.95 Å, from data collected at a synchrotron light source. In the SFX structure, the CO is coordinated to the heme a3 iron atom, with a bent Fe–C–O angle of ∼142°. In contrast, in the synchrotron structure, the Fe–CO bond is cleaved; CO relocates to a new site near CuB, which, in turn, moves closer to the heme a3 iron by ∼0.38 Å. Structural comparison reveals that ligand binding to the heme a3 iron in the SFX structure is associated with an allosteric structural transition, involving partial unwinding of the helix-X between heme a and a3, thereby establishing a communication linkage between the two heme groups, setting the stage for proton translocation during the ensuing redox chemistry.

I n the mammalian respiratory chain, cytochrome c oxidase (CcO), the final electron acceptor, catalyzes the four-electron reduction of oxygen to water and pumps protons across the inner mitochondrial membrane to establish an electrochemical proton gradient that is used to synthesize ATP. CcO contains four redox centers: two heme groups (heme a and heme a 3 ) and two copper centers (Cu A and Cu B ). Cytochrome c delivers electrons to the Cu A center from which they transfer to heme a and then to the heme a 3 -Cu B binuclear center (Fig. 1), where oxygen is reduced to water by using four protons (the substrate protons) from the matrix side of the membrane and four electrons delivered to CcO by cytochrome c. Associated with the oxygen reduction chemistry, four additional protons (the pumped protons) are translocated from the negative side to the positive side of the mitochondrial membrane. Whereas the mechanism of the O 2 reduction is relatively well understood, the mechanism by which the redox energy is coupled to proton translocation remains unresolved in mammalian CcOs (1)(2)(3)(4), although it is quite well-defined in bacterial CcOs (5,6).
The determination of X-ray structures of various derivatives of bovine CcO (bCcO) has been very helpful in understanding many of the features of the enzyme as summarized in a recent review (3). However, the structures obtained with conventional synchrotron light sources were determined from protein crystals exposed to a high X-ray flux and high levels of cryoprotectants at cryogenic temperatures. It has been shown in CcO (7,8) as well as other proteins and enzymes (9,10) that the high-intensity synchrotron X-ray beam has many deleterious effects including reduction, ligand dissociation, and damage (11,12). These deleterious effects cause irreversible changes in the protein structure leading to erroneous structural determinations, especially in protein systems containing electron-rich heavy metals (13). In oxidized forms of CcO, it has been shown that X-ray reduction of the metal centers occurs even with very short exposures to the X-ray beam (7,14). Deleterious radiation-induced artifacts are especially evident in the structure of the CO-bound derivatives (bCcO-CO) obtained at low temperature by synchrotron radiation in which the X-ray beam has been shown to photodissociate the CO from the heme a 3 iron atom (8). Hence, it is a challenging task to obtain damage-free structures of CcO that show its native structure.
To overcome these problems, we have used serial femtosecond X-ray crystallography (SFX), to determine the structure Significance A fundamental bioenergetics question is how the redox energy from oxygen reduction in cytochrome c oxidase (CcO) is coupled to proton translocation. Although crystallography has revealed many properties of the enzyme, the coupling mechanism remains elusive, in part due to artifacts caused by X-ray damage from synchrotron radiation. We determined the structure of CObound bovine CcO by serial femtosecond X-ray crystallography (SFX) with an X-ray free electron laser, outrunning radiation damage. The SFX structure is compared with the structure obtained with synchrotron radiation in which the Fe-CO bond was cleaved. Our results reveal that ligand binding to heme a 3 triggers an allosteric structural transition that may play a critical role in proton translocation in the CcO family of enzymes.
of bCcO-CO at room temperature using an X-ray free electron laser (XFEL). SFX has been shown to be a powerful way to outrun conventional radiation damage (15,16). In SFX, a continuously flowing liquid jet containing a suspension of microcrystals intersects a pulsed X-ray beam from an XFEL before their destruction (i.e., diffraction before destruction) (15). The diffraction patterns are all still snapshots. They are sorted by hit finding with the program Cheetah (17), indexed, and merged into a complete dataset with the program CrystFEL (18,19), to obtain damage-free room temperature structures of the protein (20). As a comparison with the SFX structure, we also determined the structure of bCcO-CO with synchrotron radiation. The comparison of these two structures reveals an allosteric transition involving changes in a critical region of helix-X.

Results
Cytochrome c oxidase was isolated from bovine hearts by standard procedures (21,22). For the synchrotron measurements, large crystals (∼300 × 300 × 60 μm) were grown at near physiological pH (pH 6.8). For the SFX measurements, modified crystallization conditions at the same pH were developed to yield microcrystals with a size of ∼20 × 20 × 4 μm (SI Appendix, Fig. S1). The freshly prepared crystals of oxidized bCcO were reduced anaerobically by minimal amounts of dithionite and exposed to CO.
The SFX Structure of bCcO-CO. The SFX data were obtained by injecting a stream of a bCcO microcrystal slurry into the vacuum chamber using a gas dynamic virtual nozzle at the Coherent X-ray Imaging (CXI) experimental station of the Linac Coherent Light Source (LCLS) at the Stanford Linear Accelerator Center (SLAC). A typical high-resolution SFX diffraction pattern from CO-bound bCcO is shown in SI Appendix, Fig. S2. In a total accumulation time of 225 min, 32,651 crystal hits were obtained, from which 26,264 were indexed and merged. The structure was resolved to a resolution of 2.3 Å. The crystals are in space group P2 1 2 1 2 1 with unit cell parameters of a, b, c = 178.2, 189.8, 209.6 Å, respectively, and α = β = γ = 90°. (See SI Appendix, Table S3, for the data analysis and refinement parameters.) The calculated unbiased F o -F c difference map, based on the refinement performed without the CO ligand, shows strong positive density the Cu B -heme a 3 binuclear center ( Fig. 2A). In the refined structure with CO modeled in (Fig. 2B), the Fe-CO and Fe-His bond lengths are 1.89 and 1.98 Å, respectively, and the Fe-C-O moiety adopts a bent conformation with an angle of ∼142° (Fig.  2C). Electronically, the Fe-C-O favors a linear geometry that is perpendicular to the heme plane. However, the close proximity of Cu B to the heme a 3 iron atom in bCcO forces the iron-bound CO to move away from Cu B leading to the bent Fe-C-O conformation, possibly associated with an Fe-CO bond tilted from the heme normal, which cannot be differentiated by the current data.
Structure of bCcO-CO Obtained by Synchrotron Radiation. As a comparison with the room temperature SFX structure, we obtained the bCcO-CO structure with synchrotron radiation at 100 K, at the Advanced Light Source (ALS) of the Lawrence Berkeley National Laboratory (beamline 8.2.1). The data collection strategy and analysis are described in SI Appendix,  (8). They are also comparable to those determined by SFX discussed above, with the exception that the b unit cell axis is 182.55 versus 189.8 Å, which possibly reflects the difference in the data acquisition temperatures.
The exposure of crystals at 100 K to synchrotron radiation, at a total exposure of less than 25 MGy, led to CO dissociation from the heme a 3 iron, as well as its movement toward Cu B (Fig. 2 D-F). We define this photoproduct as bCcO-CO*. The CO occupies a position in which the Cu-C and Cu-O distances are similar, as shown in Fig. 2F. The Cu B -Fe distance decreased from 5.30 Å, when the CO was coordinated to the heme a 3 iron atom, to 4.92 Å. This change in distance involves primarily movement of Cu B ; the position of the iron atom changes less than that of Cu B , which changes by ∼0.28 Å. This change is a result of the Cu moving from its nearly planar position, defined by the three histidine ligands in the CO-bound structure, to an out-of-plane position in the bCcO-CO* structure.
Structural Changes in Helix-X. Apart from the differences in the CO position and the Cu B -heme a 3 distances, the overall structure of the CO-bound and CO-photodissociated proteins were quite similar, except for a section of helix-X. Helix-X is a critical structural element of bCcO as it lies between heme a and heme a 3 and contains their axial ligands (H378 and H376, respectively) ( Fig. 1). In bCcO-CO*, the helical structure of this part of the sequence is stabilized by a series of H-bonding interactions between backbone N-H groups and C=O groups located three or four residues earlier along the sequence. In the SFX-based undamaged bCcO-CO structure, the [380-384] amino acid fragment is partially unwound due to the disruption of the H-bonding interactions as illustrated in Fig. 3 3A). We define the synchrotron structure with the intact H-bonding interactions as a closed structure and the SFX structure with the disrupted H-bonding interactions as an open structure. How does ligand binding to heme a 3 trigger the partial unwinding of the helix-X? One might expect that this modulation is a consequence of a change in the position of the helix-X histidine, H376, which is coordinated to heme a 3 , where the exogenous ligand binds. Despite the large differences in the backbone between residues 380 and 384, there are only minimal differences (Fig. 3A) between residues 376-379, the region that contains the heme axial His ligands (376 and 378). This demonstrates that it is not the histidine movement due to the presence or absence of a ligand on heme a 3 that causes the structural differences. Instead, the differences originate in the interaction of the C pyrrole ring of heme a 3 with residue V380, one of the most common heme face contact residues (23). V380 is the only residue that is in contact with heme a 3 face (Fig. 4) and is part of the region of the helix-X backbone that changes. Upon release of the CO, a heme a 3 distortion brings about the change in position of V380 (SI Appendix, Fig. S5).
To quantitate the heme a 3 distortion, we used normal-coordinate structure decomposition (NSD) (24) to analyze the heme structures. In the SFX structure of bCcO-CO the composite out-of-plane distortion is 0.603 Å, whereas in the bCcO-CO* structure it is 0.866 Å, with the largest changes occurring in the doming (from 0.428 to 0.670 Å) and saddling (from 0.256 to 0.517 Å) coordinates. A full description of the NSD results and descriptions of the doming and saddling structural distortions are presented in SI Appendix. The heme distortion results in a 0.5 Å movement of the C 2 and C 3 atoms of the C pyrrole ring of heme a 3 , as well as a change in the orientation of the vinyl group attached to it (see the expanded view in Fig. 4 and SI Appendix, Fig. S5). Together these conformational changes induce the movement of V380, triggering the partial unwinding of the helix-X structure. Thus, V380 initiates the change in the helix-X structure and transmits the heme a 3 state to the peptide backbone, regulating the open and closed structures.

Discussion
The SFX method was used to determine the structure of bCcO-CO, a large membrane-bound protein, without radiation damage and/or photodissociation, at room temperature near physiological pH (pH 6.8). The CO, which serves as a surrogate for the physiological ligand, O 2 , binds to heme a 3 in a bent conformation (142°), although a contribution of a Fe-CO bond tilted away from the normal to the heme plane cannot be excluded. A perpendicular orientation of CO with respect to the heme plane is the electronically preferred structure of a heme CO complex due to π-backbonding (25). However, it has been shown in model complexes that a bent conformation occurs when steric factors are present that restrict the perpendicular orientation (26). In the bCcO-CO complex, Cu B lies over the heme plane causing the CO to be bent to 142°, and the Fe-Cu B distance is elongated to 5.30 Å due to the steric interactions.
In contrast to CO, O 2 is expected to bind to the heme a 3 iron in the same orientation as the CO but without the imposition of steric constraints owing to its sp 2 electronic configuration, which favors a 120°Fe-O-O angle. The absence of steric crowding would also suggest a shorter Fe-Cu B distance for O 2 compared with CO. In accord with this, in the reduced bCcO-NO complex, in which Fe-N-O adopts an electronically favored bent geometry, the Fe-Cu B distance (4.92 Å) is short (8). On the other hand, our synchrotron data revealed that CO dissociation from the heme iron leads to the movement of Cu B toward the heme a 3 iron. The apparent facile change in position of Cu B , as a function of the ligand coordination on the heme a 3 iron atom, suggests that during the catalytic cycle, Cu B is able to move in response to the change in the ligation states of the heme a 3 , thereby modulating the reactivity of the binuclear center. As suggested recently, a shortening of the Fe-Cu B distance could play an important role in ligand stabilization (27). A structure of the CO-bound derivative of bCcO was reported by Muramoto et al. (8) in which the crystals were placed in a capillary and measured at room temperature (280 K) by synchrotron radiation. In that structure, CO remained bonded to heme a 3 , but the orientation of the CO was reported (8) to be essentially perpendicular with an Fe-C-O angle of 164-168°, in sharp contrast to the 142°we obtained in the SFX measurements reported here. Because the measurements of Muramoto et al. (8) were made with synchrotron radiation, radiation damage may have affected the final structure. In addition, CO photodissociation by the synchrotron X-ray beam would be expected, so the structure may represent a photostationary state, in which the CO was continually being dissociated and rebound to the heme a 3 iron atom.
Allosteric Structural Transition. The data reported here show that ligand binding to the heme a 3 iron is associated with partial unwinding of the helical structure in helix-X generating an open conformation. In both the SFX bCcO-CO structure (open) and the synchrotron-based bCcO-CO* structure (closed), all four of the metal redox centers are reduced. However, in the former structure, CO is coordinated to heme a 3 whereas in the latter, heme a 3 is ligand-free in a five-coordinate high spin configuration, i.e., À In Eq. 1, φ represents photodissociation by the synchrotron radiation beam, and the ⋯ represents the observation that the CO is close to, but not coordinated to, Cu B . Thus, the removal of ligand coordination of the iron atom on heme a 3 is sufficient to induce the change from the open to the closed conformation and establish the α-helical H-bonding. Structural differences in helix-X, in various derivatives of bCcO (28), and in its bacterial analogs (29), have been reported previously (SI Appendix, Table S2). Recently, new high-resolution structures of reduced and oxidized bCcO obtained with synchrotron radiation at 50 K were reported (30). To reduce the effects of radiation damage each frame was taken with a 10-s exposure and the crystals were translated by 100 μ after each round of 10 shots. For the measurements, 16 oxidized and 8 reduced crystals were used. In their reduced and oxidized structures, helix-X adopted closed and open conformations, respectively, similar to those defined in this work (SI Appendix, Figs. S6 and S7), although in bCcO-CO* the helix does not relax all of the way back to the same position as in the fully reduced enzyme, likely due to restrictions imposed by the low-temperature measurements. Inspection of other reported structures of bCcO, revealed that when the ferrous heme a 3 iron is in an exogenous ligand-free state, helix-X adopts a closed conformation; conversely, when the heme a 3 iron is coordinated by an exogenous ligand, in either the oxidized or reduced state, helix-X adopts an open conformation. It is notable that the reduced bCcO-NO complex has alternate structures (PDB: 3AG3) (8), with helix-X in both the open and the closed structures, possibly due to partial photodissociation of the NO by the synchrotron radiation. (See SI Appendix, SI Text and  Table S2, for a discussion of the helix-X conformations in other derivatives.) The series of structural changes in the helix-X region induced by ligand binding to heme a 3 (red arrows) are summarized in Fig. 5. Ligand binding to heme a 3 (indicated by a in Fig. 5) leads to a change in the heme distortion (indicated by b in Fig. 5). The change in the heme structure triggers the movement of V380 (indicated by c in Fig. 5), which induces a partial unwinding in helix-X (indicated by d in Fig. 5), causing a reorientation of S382 (indicated by e in Fig. 5). This poises S382 to interact with the OH of the farnesyl side chain, but it is over 8 Å away when heme a is reduced. However, during the catalytic reaction, heme a alternates between an oxidized and reduced state (gray arrows). When heme a becomes oxidized (indicated by f in Fig. 5) the OH of the farnesyl side chain rotates by 160°(indicated by g in Fig. 5) to be within ∼3 Å of S382 in the ligand-bound (open) state (SI Appendix, Fig. S8). The allosteric structural transition establishes a dynamic communication pathway between the two hemes, offering structural insights into previously proposed proton translocation mechanisms in bCcO (1,4,28,31).
In proton translocation studies of bovine and bacterial CcOs, three proton channels have been postulated (SI Appendix, Fig.  S9). The K and D channels connect the negative side of the membrane to the vicinity of the heme a 3 -Cu B binuclear center (32)(33)(34)(35), whereas the H channel passes from the negative side of the membrane via the heme a center to the positive side (31,36,37). Because the catalytic sites of the heme-copper oxidases from eukaryotic and prokaryotic species share the same structural architecture, it has been proposed that the proton pumping mechanism is conserved across species (38,39). However, this general concept has been questioned by the lack of agreement between those mechanisms shown to be operative in bacterial CcOs (5, 6) and those proposed for bCcO (1,4,28,31). Illustrative of the distinct properties of the different species is the finding that although a structural change in helix-X, similar to that in bCcO, is seen in Rhodobacter sphaeriodes CcO (RsCcO) (29), the OH group of the heme a farnesyl side chain does not undergo the 160°rotation upon the change in the redox state of heme a as it does in bCcO. Thus, what functional role, if any, the conformation change in helix-X plays in bacterial CcOs remains to be determined.
Based on structural and mutagenesis studies of bCcO (31), it has been proposed that the H channel is the major pathway for proton translocation in bCcO, and the translocation is gated by D51, heme a, and the S382 residue/farnesyl side chain OH pair (3). Thus, repositioning of S382 during the allosteric transition plays a critical role in proposed H channel mechanisms (1,3,4). Contrary to the H channel mechanisms in the bovine enzyme, extensive mutagenesis, theoretical, and structural studies in bacterial CcOs have led to the conclusion that the major proton translocation pathway is the D channel (5,40,41). (See SI Appendix, SI Text, for additional information on the H and D channel mechanisms.) To interrogate the effect of S425 in RsCcO (equivalent to S382 in bCcO) on proton translocation in the D pathway, Brzezinski and coworkers (42) mutated S425 to alanine and measured its proton translocation efficiency during the O 2 reaction cycle as a function of pH. They found that structural changes in S425 propagate to the D pathway, thereby modulating proton transfer rates (42). These data suggest that the S382-linked allosteric structural transition may play a role in proton translocation in both mammalian and bacterial oxidases.
Ligand Reactivity. It is well established that in solution, following CO photodissociation from the heme a 3 iron atom, the CO first binds to Cu B ; it subsequently dissociates from Cu B and exits the protein matrix without geminately rebinding to the heme a 3 iron atom (43). This has been postulated as evidence for conformational changes at the binuclear center thereby hindering the CO from rebinding to the heme a 3 iron atom (43). However, our current data do not support this hypothesis, because no structural changes in the binuclear center that can limit the rebinding rate are identifiable in the structure of bCcO-CO*.
To better understand the ligand binding/dissociation reactions of bCcO, it is necessary to consider the differences between CO and O 2 binding based on the binding reaction scheme shown in Fig. 6, Inset, and the corresponding free energy curves. This reaction coordinate is analogous to those derived for the myoglobin reaction with ligands (44). In bCcO, for a ligand to bind the heme a 3 iron, it has to first migrate into the protein matrix by overcoming the outer barrier (OB) to form the encounter complex (labeled in gray). Subsequently, it binds to Cu B by crossing inner barrier 1 (IB-1) and then moves from Cu B to the heme a 3 iron by crossing inner barrier 2 (IB-2). The initial migration into the protein is expected to be similar for CO (red curve) and O 2 (gray curve), represented by their overlap in the outer barrier. Because the bimolecular binding of O 2 to heme a 3 is over 10 3 times faster than that of CO (45), the inner barriers must be lower for O 2 than for CO. For the CO reaction, IB-2, and possibly also IB-1, is higher compared with the O 2 reaction because in essentially all heme proteins the barrier for binding to the ferrous iron atom is much higher for CO than for O 2 (44,46). For example, owing to this barrier, in myoglobin the binding rate is 35 times slower for CO than for O 2 . Frauenfelder and Wolynes (47) have shown that this difference in the CO and O 2 binding rates in heme proteins is a result of the spin transitions required for the binding of CO to a heme iron atom, which does not occur for O 2 binding.
Consistent with the hypothesized energy curves shown in Fig.  6, Woodruff and coworkers were able to determine that the on rate of CO from Cu B to the heme a 3 iron is 1,030 s −1 (43), at least 35 times slower than the on rate of O 2 from Cu B (48). In any case, the energy curve associated with the CO reaction indicates that once CO is photodissociated from the heme a 3 iron in bCcO, it could either rebind to the heme iron by crossing the IB-2 or escape out of the protein by crossing IB-1 and OB. Which route it takes depends on the branching ratio R (R = k IB-2 /k IB-1 ), in which k IB-1 and k IB-2 represent the rate constants for crossing inner barrier 1 and inner barrier 2, respectively. The low branching ratio for CO, due to the higher IB-2 than IB-1 barrier, accounts for the observed low geminate CO recombination yield.
The Inverse Correlation of the Fe-CO and C-O Vibrational Modes. In CO-bound heme proteins, there is a well-established inverse correlation between the frequencies of the Fe-CO and the C-O stretching modes. However, the frequencies of the modes in CcO-CO fail to fall on the inverse correlation line; instead, they lie above the line as shown in SI Appendix, Fig. S10 (49). It was shown by Yu and coworkers (26) that distortion of the CO, from its typical linear and perpendicular orientation, by distal pocket constrictions, can result in deviations from the correlation curve, in which the data points lie above the curve, possibly due to a reduction in the effective mass of the Fe-C-O moiety due to its bent conformation, thereby modifying the frequencies of the stretching vibrations. The structural data we report here reveal that the Fe-C-O moiety of bCcO-CO is bent due to the steric interaction with Cu B ; hence, they offer a molecular explanation for the deviation of the CcO data from the inverse correlation line. It is notable that in some studies of CcO-CO, a second set of data falling on the inverse correlation line were also observed (49), indicating that in a certain population of the enzyme molecules, Cu B has moved away from the heme a 3 iron atom, allowing the CO to bind in a linear fashion.

Materials and Methods
Isolation and purification of bCcO from bovine hearts was carried out by standard procedures (21,22). To generate highly purified bCcO from the isolated enzyme, crystals were formed and harvested. These crystals were redissolved and used as the source for subsequent crystal growth. Large crystals were grown and then crushed to form seed stock to grow microcrystals (∼20 × 20 × 4 μm) for the SFX measurements and large crystals (∼300 × 300 × 60 μm) for the synchrotron radiation measurements. All crystals were grown in phosphate buffer at pH 6.8 with variable amounts of detergent (decylmaltoside) and PEG-4000, which were optimized for each preparation. For the SFX measurements the crystals were reduced with a minimal amount of dithionite and exposed to CO to form the CO derivative, which was monitored by microscopic spectroscopy. The crystal suspensions were kept at 4°C throughout and used within a few hours of their preparation. For the synchrotron measurements, the crystals were soaked in ethylene glycol by a stepwise procedure to reach 45% in a CO-saturated buffer. The crystals were immediately frozen and stored under liquid nitrogen for the X-ray measurements.