Serial femtosecond crystallography structure of cytochrome c oxidase at room temperature

Cytochrome c oxidase catalyses the reduction of molecular oxygen to water while the energy released in this process is used to pump protons across a biological membrane. Although an extremely well-studied biological system, the molecular mechanism of proton pumping by cytochrome c oxidase is still not understood. Here we report a method to produce large quantities of highly diffracting microcrystals of ba 3-type cytochrome c oxidase from Thermus thermophilus suitable for serial femtosecond crystallography. The room-temperature structure of cytochrome c oxidase is solved to 2.3 Å resolution from data collected at an X-ray Free Electron Laser. We find overall agreement with earlier X-ray structures solved from diffraction data collected at cryogenic temperature. Previous structures solved from synchrotron radiation data, however, have shown conflicting results regarding the identity of the active-site ligand. Our room-temperature structure, which is free from the effects of radiation damage, reveals that a single-oxygen species in the form of a water molecule or hydroxide ion is bound in the active site. Structural differences between the ba 3-type and aa 3-type cytochrome c oxidases around the proton-loading site are also described.


Results
Crystallization, characterization, LCP injection and data collection. Conditions known to produce lipidic cubic phase (LCP) crystals of ba 3 CcO from Thermus thermophilus sufficiently large for synchrotron based crystallography 24 were modified to yield showers of micrometer sized crystals. A protocol was then developed to prepare microcrystals in glass plates or glass syringes by surrounding the protein-lipid mixture with precipitant solution, similar to a method previously described 25 . Under these conditions a large number of well-diffracting 5-20 µm sized ba 3 CcO crystals appeared after 2-3 days (Fig. 1).
The oxidation state of ba 3 CcO after crystallization was checked by recording an optical absorbance spectrum from microcrystals as prepared in the LCP phase (Supplementary Figure S1). The peak at 415 nm shows that the enzyme is in the oxidised state within the microcrystals. A spectrum was also recorded of microcrystals reduced by the addition of sodium dithionite (Supplementary Figure S1). This caused a shift of the absorption peak to 427 nm with a shoulder at 445 nm, and the appearance of peaks at 528 nm and 560 nm, characteristic of the reduced enzyme.
X-ray diffraction data were collected at SPring-8 Angstrom Compact Free Electron Laser (SACLA) with highly brilliant X-ray pulses at 7.6 keV and approximately 10 fs in duration. Microcrystals were delivered into the X-ray focus at a flow rate of 0.48 µl/min using an LCP injector developed at SACLA 26 . A total of 87,057 frames were collected out of which 11,374 were identified as containing a diffraction pattern using the software Cheetah 27 , of which 8,211 frames were indexed and merged using CrystFEL 28 . Thus a complete diffraction data set was collected from approximately 50 minutes of data collection using 25 µl of sample. The structure was solved by molecular replacement using Phaser 29 with PDB code 3S8F as a search model, and refined to a resolution of 2.3 Å. In the following text we refer to the room-temperature structure described here as the SFX structure (Fig. 2a). 3 CcO structure. A number of structures of ba 3 CcO are available in the protein data bank (PDB), both using vapor diffusion and LCP crystallization methods 11,17,24 . The highest resolution structures at 1.8 Å resolution were obtained by LCP crystallization (PDB codes 3S8F and 3S8G for wild type and A120F mutant enzyme, respectively) 24 . All previously reported structures, however, have been determined using crystals cooled to cryogenic temperatures (~100 K). In the following discussion we use the PDB code 3S8G structure for comparison (referred to as the cryo-LCP structure) as it has the best defined electron density. Overall, the SFX structure ( Fig. 2a) agrees very well with previously solved structures, with a C α root mean-square deviation of 0.32 Å compared to the cryo-LCP structure. The space group, C2, is the same as the cryo-LCP structure with similar unit cell parameters. We can confirm the presence of 13 out of 20 lipid molecules modelled in the cryo-LCP structure, and we also localize two new lipid tails associated with subunit I (Supplementary Figure S2). In the SFX structure we resolve 92 water molecules, which compares with 225 observed in the cryo-LCP structure. This difference in number of modelled water molecules is possibly due to the lower resolution of the SFX structure, or because of increased disorder of water molecules when diffraction data are recorded at room temperature.

Room-temperature ba
Active site ligand. The active site of CcOs is made up of heme a 3 , Cu B and surrounding residues. This site is where molecular oxygen binds in to the reduced heme a 3 iron and the O=O double bond is cleaved through the uptake of four electrons and four protons, producing two water molecules. In several structures of oxidized CcO the electron density at the active site has been interpreted to indicate the presence of peroxide (O 2 2− ) bound between the heme a 3 iron and Cu B 24, 30, 31 . This interpretation has been questioned by Kaila and co-workers who claim that a stable peroxide in the active site of oxidized CcO is unlikely 32 . Their arguments were based upon quantum chemical calculations in combination with spectroscopic data and re-refinement of two X-ray structures. For the ba 3 CcO, as many as 34 structures have been deposited in the PDB from crystals grown using the method of vapor diffusion 11 and from lipidic cubic phase crystallization 24 . These structures are summarized in Supplementary Table S1. Structures resulting from vapor diffusion crystallization show conflicting results regarding the ligand bound in the active site, where the omit map electron density appears spherical in some cases (best fitted with a single-oxygen species such as water) and elongated in other cases (best fitted with a peroxide species) (Supplementary Figure S3). The authors attribute this uncertainty to effects of temperature as well as time of exposure to X-rays 24,33 . All fourteen structures of ba 3 CcO resulting from LCP crystallization display somewhat elongated omit map electron density in the active site, which has repeatedly been interpreted to indicate a bound peroxide molecule (Supplementary Table S1). However, this elongated density is potentially due to X-ray induced reduction 24 .
Our SFX structure displays residual electron density for a ligand in the active site. Since this residual density in the unbiased F o -F c omit map is spherical in shape it strongly suggests that a single-oxygen species, presumably a water molecule or hydroxide ion, is bound between the heme a 3 iron and Cu B (Figs 3 and 4a). Moreover, the peak height of this active site feature (0.99 e − /Å 3 ) is very similar to that of a strongly bound water molecule located between Ser309 and the heme a 3 secondary alcohol (0.86 e − /Å 3 ), further supporting our interpretation that the active site residual density corresponds to a single tightly bound water molecule or hydroxide ion. In contrast, a peroxide molecule was modelled as the active site ligand in the cryo-LCP structure 24 , and the F o -F c omit map calculated from the deposited cryo-LCP data clearly displays an elongated electron density peak in the active site (Fig. 4b). When a water molecule is modelled into the active site of the SFX structure, some residual positive F o -F c electron density remains (Fig. 4d). However, if a water molecule is modelled into the cryo-LCP structure in place of the peroxide, it results in a poor fit with a much stronger residual density (Fig. 4e). Similarly, the structure of bovine heart aa 3 -type CcO derived from XFEL diffraction data using large cryo-protected crystals (PDB code 3WG7) 31 also shows an elongated shape of the active site peak in the F o -F c omit map (Fig. 4c). Moreover, if a single water molecule is built into this electron density in place of the peroxide ligand, the fit is poor with a strong residual F o -F c density similar to in the cryo-LCP structure (Fig. 4f). The SFX structure is shown in green/blue and the bovine heart aa 3 -type CcO (PDB code 3WG7) in grey. The magnesium ion of the bovine heart structure is shown in brown. Water molecules are shown in red (SFX) and grey (bovine heart). Heme a 3 is shown in black (SFX) and grey (bovine heart). The amino acid numbering refers to the ba 3 CcO.
Resonance Raman spectroscopy has shown the existence of a peroxide species bridging the heme a 3 iron and Cu B in the so-called "resting oxidized state" of aa 3 -type CcO from bovine heart 34 . This has not yet been explicitly investigated for the ba 3 CcO, but electrochemically induced UV-vis difference absorbance spectra suggest that there is no peroxide molecule at the active site of ba 3 CcO in the oxidized (as prepared) state of the enzyme 35 . On the other hand, it has been demonstrated that the redox sites of ba 3 CcO crystals become reduced during X-ray data collection at cryogenic temperatures 33 . Thus the apparent peroxide species in the cryo-LCP structures   Table S1) may result from the recombination of two hydroxyl radicals produced from X-ray damage to water molecules near the active site, as the authors also suggest 24 . In contrast, our SFX structure represents a radiation-damage free structure of ba 3 CcO for which a water molecule or hydroxide ion is likely the active site ligand.
Proton translocation pathway in ba 3 CcO. The ba 3 CcO utilizes a single proton uptake pathway, the so-called K-pathway analogue, in order to provide protons for the oxygen chemistry at the active site, as well as to transport protons that are pumped across the membrane. In the SFX structure we can track a proton pathway that stretches from Glu15 II ("II" denotes subunit II), via Thr315, Wat721, Tyr248, Thr312, Tyr244, Ser309, Wat738 and the secondary alcohol of heme a 3 to Tyr237 in close proximity of the active site (Fig. 2b). The positions of the two water molecules as well as the side chain conformations of key residues are conserved when comparing with the cryo-LCP structure 24 . Most importantly, the room-temperature SFX structure confirms that there is a gap in the hydrogen bond network between Thr312 and Tyr244, with a distance of 4.45 Å between the two residues. It is possible that transient ordering of a water molecule bridges this gap such that a functional proton pathway is formed when a proton is taken up from the negative side of the membrane.
Structural differences around the proton-loading site. Transmembrane proton pumping against an electrochemical gradient requires a protonatable group with alternating access to the positive and negative sides of the membrane during the catalytic cycle. There is an ongoing debate as to the identity of this group, referred to as the proton-loading site, where it has been suggested to be localized in the vicinity of the heme a 3 propionates; the surrounding water cluster; or a near-by aspartic acid residue (Asp372 in ba 3 CcO) 36,37 . Several of the residues in this region are conserved throughout the CcO superfamily including a tryptophan residue (Trp229 in ba 3 CcO). It has been shown that in the aa 3 -type CcO from Paracoccus denitrificans the corresponding residue forms a tryptophan radical during enzymatic turnover 38 . This, however, is not the case for ba 3 CcO. It was recently proposed that the transient formation of the tryptophan radical is involved in proton pumping, and consequently the lack of such a radical in ba 3 CcO could be related to the lower proton pumping stoichiometry displayed by this enzyme 39 . In light of this debate, it is interesting to note that the SFX structure and all previously published ba 3 CcO structures differ from the aa 3 -type CcO structures in the vicinity of this conserved tryptophan residue (Fig. 2c). Specifically, the tyrosine residue (Tyr136 in ba 3 CcO) that hydrogen bonds to the tryptophan is in close proximity of the heme a 3 propionates in ba 3 CcO, whereas it is pointing more towards the associated water cluster in the aa 3 -type CcOs. It may also be of relevance for the difference in proton pumping efficiency that all aa 3 -type CcOs contain a magnesium ion as part of the above mentioned water cluster. There is no magnesium ion present in our SFX structure or any of the previously deposited ba 3 CcO structures. Instead, the position of the magnesium ion is occupied by the carboxyl group of a glutamic acid residue (Glu126 II ) in the ba 3 CcO structures (Fig. 2c). Calculations have shown that the magnesium ion affects the proton affinity of the heme a 3 propionate A in the aa 3 -type CcOs 40 .

Discussion
We describe a method to prepare large quantities of well-diffracting microcrystals of ba 3 CcO in lipidic phase and present the first room-temperature structure of a CcO free of radiation damage. The omit electron density map shows that the ligand bound at the active site in the oxidized form of ba 3 CcO is most likely a single-oxygen species corresponding to a water molecule or hydroxide ion bound between the heme a 3 iron and Cu B . Considering the fact that it is localized between two positively charged groups, the heme a 3 iron and Cu B, a hydroxide ion may be the more likely alternative. Our conclusion is based upon the spherical shape of the electron density peak in the unbiased F o -F c omit map and comparison of residual densities when modelling a single water molecule at this site in other deposited CcO structures (Fig. 4). It is also apparent that the proton transfer pathway and the water cluster associated with the presumed proton-loading site, described in previous cryo-temperature structures 11,24 , are conserved at room temperature.
Hirata and co-workers recently presented a structure of bovine heart aa 3 -type CcO from XFEL diffraction data collected using large cryo-protected crystals where the crystals were physically translated between each X-ray exposure 31 . We propose that the use of microcrystals in order to collect room-temperature SFX data has major advantages over the use of large, cryo-cooled crystals at an XFEL. Specifically, it is clearly beneficial to determine the structure of a protein at room temperature, as this is the more physiologically relevant temperature. Further, each microcrystal is only exposed once to an XFEL pulse when using SFX. This completely avoids the risk of radiation damage accumulating due to X-ray induced free radicals propagating through the crystal. In this study an average X-ray dose of 13 MGy per crystal was used, well below the limit where radiation damage has been detected in metal clusters using XFEL radiation 22 . Microcrystals are also favoured over large crystals if the aim is to trigger a reaction within the crystal. For microcrystals the penetration of the reaction trigger, may it be light or a chemical, is more likely to reach and activate a large population of the protein molecules within the crystal. This idea has been very successfully demonstrated in several recent examples 26,41,42 . It is therefore exciting to consider future opportunities for time-resolved structural studies now that we have demonstrated the possibility to collect high resolution SFX data on CcO crystals. This opens up a unique opportunity to record structural snapshots of CcO as oxygen is reduced to water while protons are simultaneously pumped across a biological membrane.

Expression and purification.
Recombinant ba 3 CcO from Thermus thermophilus was produced and purified as described 43  Absorbance spectroscopy. Microcrystals of ba 3 CcO were prepared as described above. 10 μl crystal-containing LCP was added to a CaF 2 glass cell with a path length of 0.2 mm and the absorbance spectrum recorded between 200 and 700 nm in a Hitachi U-2910 spectrophotometer. The absorption spectrum of the monoolein mixed with buffer only was subtracted from the spectrum of the samples. To record a spectrum of the reduced sample, sodium dithionite was mixed with monoolein that was subsequently added to the crystal-containing LCP, giving a final dithionite concentration of 4.8 mM. The final protein concentration in the LCP was approximately 0.1 mM for both samples. However, the local crystal concentration in the cell may vary between different measurements.
Data collection and processing. 15  a 75 μm diameter nozzle at a flow rate of 0.48 μl/min. The sample-to-detector distance was 53 mm and data was recorded using an eight sensor module CCD detector 45 . Diffraction images were processed using the software package Cheetah 27 through a pipeline adapted for SACLA 46 . Images containing more than 20 diffraction peaks were identified as hits and further processed with CrystFEL 28 where DirAx 47 and Mosflm 48 were used for indexing. Intensities from all images were merged and scaled using Partialator without partiality correction. A per-crystal resolution cut-off was employed that rejected data beyond 1.2 Å past the crystal resolution as calculated by the peak finding in CrystFEL. The crystal hit rate was about 12% and the indexing rate was 72%. The average radiation dose per crystal was calculated using the program RADDOSE-3D 49 . This can be considered as an upper estimate of the actual dose due to the possibility that photoelectrons escape from the microcrystals, which is not taken into account in the calculation.
Structure determination and refinement. The structure was solved by molecular replacement with Phaser 29 using the high-resolution cryo-LCP wild type ba 3 CcO structure (PDB code 3S8F) as a template. The model was built in Coot 50 and refinement at a resolution of 2.3 Å was carried out in REFMAC5 51 using anisotropic displacements with three rigid-body motion groups and an X-ray weighting term of 0.05. Final R work and R free were 16.2% and 19.8% respectively with 99.6% of the side chains within accepted Ramachandran regions. Data collection and refinement statistics are summarized in Table 1.