Unexpected Electron Spin Density on the Axial Methionine Ligand in CuA Suggests Its Involvement in Electron Pathways

The Cu A center is a paradigm for the study of long-range biological electron transfer. This metal center is an essential cofactor for terminal oxidases like Cytochrome c oxidase, the enzymatic complex responsible for cellular respiration in eukaryotes and in most bacteria. Cu A acts as an electron hub by transferring electrons from reduced cytochrome c to the catalytic site of the enzyme where dioxygen reduction takes place. Different electron transfer pathways have been proposed involving a weak axial methionine ligand residue, conserved in all Cu A sites. This hypothesis has been challenged by theoretical calculations indicating the lack of electron spin density in this ligand. Here we report an NMR study with selectively labeled methionine in a native Cu A . NMR spectroscopy discloses the presence of net electron spin density in the methionine axial ligand in the two alternative ground states of this metal center. Similar spin delocalization observed on two second sphere mutants further supports this evidence. These data provide a novel view of the electronic structure of Cu A centers and support previously neglected electron transfer pathways. ABSTRACT : The Cu A center is a paradigm for the study of long-range biological electron transfer. This metal center is an essential cofactor for terminal oxidases like Cytochrome c oxidase, the enzymatic complex responsible for cellular respiration in eukaryotes and in most bacteria. Cu A acts as an electron hub by transferring electrons from reduced cytochrome c to the catalytic site of the enzyme where dioxygen reduction takes place. Different electron transfer pathways have been proposed involving a weak axial methionine ligand residue, conserved in all Cu A sites. This hypothesis has been challenged by theoretical calculations indicating the lack of electron spin density in this ligand. Here we report an NMR study with selectively labeled methionine in a native Cu A . NMR spectroscopy discloses the presence of net electron spin density in the methionine axial ligand in the two alternative ground states of this metal center. Similar spin delocalization observed on two second sphere mutants further supports this evidence. These data provide a novel view of the electronic structure of Cu A centers and support previously neglected electron transfer pathways.

cellular respiration in eukaryotes and in most bacteria. Cu A acts as an electron hub by transferring electrons from reduced cytochrome c to the catalytic site of the enzyme where dioxygen reduction takes place. Different electron transfer pathways have been proposed involving a weak axial methionine ligand residue, conserved in all Cu A sites. This hypothesis has been challenged by theoretical calculations indicating the lack of electron spin density in this ligand. Here we report an NMR study with selectively labeled methionine in a native Cu A .
NMR spectroscopy discloses the presence of net electron spin density in the methionine axial ligand in the two alternative ground states of this metal center. Similar spin delocalization observed on two second sphere mutants further supports this evidence. These data provide a novel view of the electronic structure of Cu A centers and support previously neglected electron transfer pathways.

ABSTRACT:
The Cu A center is a paradigm for the study of long-range biological electron transfer. This metal center is an essential cofactor for terminal oxidases like Cytochrome c oxidase, the enzymatic complex responsible for cellular respiration in eukaryotes and in most bacteria. Cu A acts as an electron hub by transferring electrons from reduced cytochrome c to the catalytic site of the enzyme where dioxygen reduction takes place. Different electron transfer pathways have been proposed involving a weak axial methionine ligand residue, conserved in all Cu A sites. This hypothesis has been challenged by theoretical calculations indicating the lack of electron spin density in this ligand. Here we report an NMR study with selectively labeled methionine in a native Cu A . NMR spectroscopy discloses the presence of net electron spin density in the methionine axial ligand in the two alternative ground states of this metal center. Similar spin delocalization observed on two second sphere mutants further supports this evidence. These data provide a novel view of the electronic structure of Cu A centers and support previously neglected electron transfer pathways.
Copper ions are essential for electron transfer (ET) in all living organisms. 1 This biological function is performed by copper centers in proteins with unique electronic features: the mononuclear Type 1 (blue) sites and the binuclear Cu A (purple) centers. 2,3 These sites share some common features: a rigid structure that minimizes the reorganization energy, highly covalent copper-thiolate bonds with cysteine ligands (one or two, respectively in these sites) and two equatorial histidine ligands that maintain the geometry of the metal sites ( Figure 1). [4][5][6] Type 1 and Cu A centers also share the presence of weak axial ligands, whose role in fine tuning the electronic structure and functionality of these sites has been intensely discussed, particularly in the blue copper sites. [7][8][9][10][11][12][13] The most common axial ligand in blue copper sites is a methionine residue, with variable Cu-S(thioether) distances ranging from 2.6 to 3.0 Å, that have also been shown to tune the reduction potential and the electronic structure of the metal site. 14 The nature of the axial ligand in Type 1 copper sites is variable, and a glutamine residue or no axial ligand can be found in these centers. [15][16][17] In the case of Cu A centers, a Met residue and a peptide bond are the conserved weak axial ligands in all known sites, except for the recently characterized PmoD-Cu A , with two Met axial ligands. 18,19 The role of the Met axial ligand in the Cu A site has also been matter of debate. 13 There is a general consensus supporting that the main role of this ligand is to tune the reduction potential 11 while preserving the reorganization energy, as shown by Solomon and coworkers. 12,20 The involvement of this residue in ET pathways, however, is more controversial. 12,13 The electronic structure of Cu A sites can be described by a double-well potential with two partially populated electronic states in thermal equilibrium: σ u * and π u (separated by a small energy gap of 600 cm -1 ). 12,20,21 A stronger interaction with the axial ligand decreases this gap, making the π u ground state more accessible. It has been proposed that this phenomenon could increase the superexchange coupling and favor an ET pathway through this position. 12 C signal at 31.3 ppm coupled with two protons at 20.7 and 7.8 ppm (Figure 2A). The former 1 H resonance has been previously assigned to a Hε1 of His114. 21,33 The current data show that these resonances correspond to a CH 2 moiety from the axial Met ligand. To further corroborate this assignment, we performed 1D and 2D NOE experiments, that revealed a strong dipolar coupling between these two proton resonances, confirming that they correspond to geminal protons ( Figure 2B). We assign these resonances to the ϒ-CH 2 moiety from Met160, since these protons are three bonds away from one of the copper ions. The signal located at 20.7 ppm had less intense NOEs with other resonances ( Figure 2B). One of them shows a correlation in the HMQC spectrum with a 13 C located at 32.9 ppm that we assign to a β-CH 2 moiety from the Met ligand. The geminal proton of this signal could not be clearly identified in the diamagnetic region. The chemical shifts of these protons are less distant from the diamagnetic region. To complete the assignment, we then prepared a sample with the ε- of electron spin density in the Met160 sidechain. Since most of these signals fall within the diamagnetic region in uniformly labeled samples, none of these assignments could have been identified without resorting to this selective labeling strategy.
The observed hyperfine shifts (δ obs ) in paramagnetic compounds have different contributions: the diamagnetic contribution to the chemical shift of the nucleus without the effect of the unpaired electron (in this case, δ dia corresponds to reduced Cu A ), the contribution to the chemical shift due to the dipolar coupling between the nuclei and the magnetic moment of the unpaired electron (δ pc ) and the contribution from the Fermi contact shift due to non-null electron spin density in the nucleus (δ con ). 38 δ con can be calculated from (See details in supplementary material): In general, the low magnetic anisotropy of Cu A sites leads to the assumption that the δ pc term is negligible. 39 However, since we are dealing with nuclei with small hyperfine shifts and close to the copper site, the pseudocontact component that depends on the distance of the nuclei to the paramagnet may not be disregarded. Based on the structure available for this protein, 4 we calculated the pseudocontact contribution to the isotropic shifts and, therefore the contact contributions for all nuclei. As shown in Table 1, all 13 C and 1 H nuclei from Met160 (except for the Cβ) have non-negligible contact shifts in Tt Cu A . The largest contact contributions can be located to the ε-CH 3 and ϒ-CH 2 moieties, that are next to the copper binding S δ (Met) atom. The δ con of three out of the five resonances from the Met residue showed a non-Curie temperature dependence, suggesting that this shift reflects an averaged contribution of both the σ u * and π u states ( Figure Table S1) and allowed a stereospecific assignment of the geminal H ϒ 's.
Comparison of the δ con with the calculated spin densities allows us to confirm that the proton at 20.7 ppm corresponds to the Hϒ2 and the resonance at 7.8 ppm to the geminal Hϒ3. (Figure 4 and Table S1).  To confirm these findings, we studied two second-shell mutants of Tt Cu A in which three loops from the bacterial protein were replaced by the homologous ones from the human oxidase (Tt3Lh) 40,41 and a plant oxidase (Tt3LAt) from Arabidopsis thaliana 42 (Figures S1 and S2). These mutants are good mimics of the Cu A -containing eukaryotic oxidases 40-43 since they include changes only in second sphere residues, conserving all metal ligands. The NMR spectra of Tt3LAt were assigned and resembled those of the previously reported Tt3Lh (Figure S1-2 and Table S3). Both variants preserve the identity of all Cu A ligands but display a smaller energy gap (ca. 240 cm -1 ) between the two alternative ground states. 40 HMQC experiments in 13 C labeled Met and ε-13 CH 3 Met samples of this chimeric proteins revealed hyperfine shifted resonances with a pattern resembling that observed for Tt Cu A . 1D and 2D NOE of these signals also showed similar correlations, enabling the assignment of all 1 H and 13 C signals corresponding to the Met ligand in these Cu A variants (Table S3 and Figures S3 -5). To calculate the contact shifts on these two variants, we solved the crystal structures of the two Cu A mutants. Both proteins display Cu A sites structurally similar to Tt Cu A ( Figure S6 and Table S4). The Cu-S(Met) distances range between 2.5 and 2.6 Å, i.e., within a much shorter range than in blue copper sites, reflecting minor structural distortions induced by the mutations. 4,14 As in the wild-type protein, the nuclei from the Met axial ligand have net contact shifts in both chimeric proteins (Table 1).
These data indicate the presence of unpaired electron spin density in the axial Met residue in all three Cu A variants.
Here we report the presence of electron spin density both in the methionine ε-CH 3 and ϒ-CH 2 nuclei in the three Cu A sites here studied (Tt Cu A , Tt3Lh and Tt3LAt). The contact shifts on the ϒ-CH 2 are larger than those observed in the homologous nuclei of T1 sites, despite there is only one electron delocalized between two copper ions in the Cu A centers. Blue copper proteins display net display net contact shifts in the ε-CH 3 (Met) and in the ϒ-CH 2 protons. In general, normal T1 sites show shifts in the ε-CH 3 while distorted T1 sites are characterized by non-null shifts in the ϒ-CH 2 nuclei (Table 1). In azurin, displaying a long Cu-S(Met) distance (3.0 Å), there is no contact contribution to any Met nucleus. We propose that the shorter Cu-S(Met) distance in native Cu A sites compared with T1 centers could account for this electron spin density.
The current results differ with calculations performed for the biosynthetic Cu A -azurin. 13 This apparent discrepancy can be attributed to the longer Cu-S(Met) bond (2.9 Å) in this model protein, compared to the short Cu-S distance in native Tt Cu A and the two loop mutants here analyzed. 44 This differential interaction with the axial ligand is also reflected in the redox properties of this site: while the reduction potential of Cu A -azurin is rather insensitive to axial ligand mutations, changes in the axial Met in Tt Cu A give rise to changes in the reduction potentials resembling those reported for Type 1 copper centers. 9,11,[44][45][46] The current experimental data support the notion that the axial Met ligand of native Cu A sites has net electron spin density in both σ u * and π u states. This scenario validates the feasibility of ET pathways involving the axial Met ligand in heme-copper oxidases and claims for the reinterpretation of the current picture of ET in terminal oxidases.   download file view on ChemRxiv MorgadaLlases_MainArticle.pdf (1.20 MiB)

Materials and Methods
Chemicals were purchased from Sigma-Aldrich otherwise stated. 13 C-labeled Met and ε-13 CH 3 labeled Met were purchased from Cortecnet.

Protein expression and purification
Selectively labeled samples on the Met residues were obtain by overexpression in E. coli cells by addition of the corresponding labeled Met after induction with IPTG and expression for 4 h at 37°C or overnight at 30°C. Cells then were harvested and the proteins were purified following previously optimized protocols. 1,2 For crystallization, Tt3Lh variant was purified with an N-terminal His 6 tag using a His-Trap column (GE Healthcare). The tag was then cleaved with Thrombin protease (Sigma) for 12 h followed by size exclusion chromatography to remove the protease and the Tag.

Protein NMR spectroscopy
NMR experiments were carried out on a Bruker Avance II Spectrometer operating at 600.13 MHz (1H frequency). 1 H and 1 H-13 C HMQC/HSQC spectra were acquired with a triple-resonance (TXI) probehead and direct 13 C acquisition NMR spectra were acquired with a broadband observe (BBO) probehead tuned at the proper frequency.
1D 1 H spectra were acquired with spectral windows of 48 to 360 kHz using presaturation, PASE and SuperWEFT schemes (total recycle times around 100 ms and 10 ms for H 2 O and D 2 O samples, respectively). 2D 1 H-1 H NOESY spectrum were acquired using a phase-sensitive NOESY sequence with solvent presaturation in 99.9% D 2 O samples. The spectral window was set to 36 kHz and mixing times ranged between 4 to 20 ms. 1 H, 13 C HMQC experiments were acquired on spectral widths ranging from 48 to 360 kHz in the 1 H dimension (1024 points) and 36 to 100 kHz in the indirect dimension (32 points). The delay for coherence transfer was set to 2 ms, and the relaxation delay was set to 30 ms. 1 H, 13 C HMQC experiments on ε-13 CH 3 Met samples were acquired using a spectral window of 6 kHz in the direct dimension and 12 kHz in the indirect dimension. The delay for magnetization coherence transfer was set to 1.67 ms, optimized for the detection of the paramagnetic signal.
1D 13 C NMR spectra were acquired using an excitation pulse of 6.9 μs at 12.6 W using inverse-gated decoupling. The carrier frequency was set to 800, 200, or -450 ppm depending on the group of signals being studied. Spectral window and delays also varied in each experiment, which resulted in total recycle times of approximately 20 ms for the acquisition of signals around 1000 and -450 ppm, and approximately 100 ms for those near the diamagnetic region. No further signals were detected when the carrier was moved to different frequency offsets.
Signal assignment of the spectra of Tt3LAt was done based on a set of different homonuclear and heteronuclear spectra. Intra-residual signal connectivities were made by 1 H-1 H NOESY experiments in 10% and 99.9% D 2 O with mixing times between 4 and 20 ms. 13 C signals were assigned by means of 1 H, 13 C HMQC spectra tailored to the paramagnetic nature of the resonances, as already reported.
Analyses of the chemical shifts and their temperature dependences were performed as described previously. 2-5

Determination of contact shifts
The contact contributions (δ con ) to the chemical shifts (Tables 1 and S3) were calculated from: Diamagnetic chemical shifts (δ dia ) were obtained by saturation transfer difference experiments in samples containing ~10% reduced protein, by addition of substoichiometric amounts of sodium ascorbate to the fully oxidized sample. Signals were irradiated for ~50 ms at a power of ~5 mW.

( )
where µ 0 is the vacuum permeability, µ B is the Bohr magneton, S is the electronic spin (1/2 for the oxidized Cu A center), k B is Boltzmann's constant, T is the temperature in Kelvin, r is the module of a vector r connecting the nucleus and the averaged coordinates of both copper ions, θ is the angle between the z component of the g tensor and the r vector, and g | and g⊥ are the parallel and perpendicular values of the axial g tensor. Distances and angles were obtained from the corresponding crystal structures. g | and g⊥ were obtained from the EPR spectra of each protein in X band (frequency = 9.4 MHz), 4K (data not shown). This approach is valid for systems with null or small spin-orbit coupling, such as the present one. Magnetic axes were defined as described by Neese and co-workers. 6 The pseudocontact shift proved to be small in all cases (Table S3), as expected.

Protein crystallization, data collection and structure determination
Crystallization screenings were carried out using the sitting-drop vapor diffusion method and a Gryphon (Art Robbins Instruments) or a Mosquito (TTP Labtech) nanoliter-dispensing crystallization robot. Crystals of hTt3L grew after 1 day from an 80 mg/ml protein solution, by mixing equal volumes of protein solution and mother liquor (100 mM Hepes, 1.6 M (NH 4 ) 2 SO 4 , 0.1 mM NaCl, pH 7.5), at 18 ºC. Similarly, crystals of Tt3LAt grew after 3 days from a 75 mg/ml protein solution, using 100 mM Hepes, 1.5 M LiSO 4 , pH 7.5, as mother liquor. In both cases, single crystals were cryoprotected in mother liquor containing 30% glycerol as cryoprotectant and flash-frozen in liquid nitrogen. X-ray diffraction data were collected at the synchrotron beamline ID23-1 (ESRF, Grenoble, France), at 100 K, using wavelengths of 1.367875 Å and 1.382671 Å. Diffraction data were processed using XDS and scaled with Aimless 7 from the CCP4 program suite. 8 The crystal structures of Tt3Lh and Tt3LAt were solved by molecular replacement using the program Phaser 9 and the atomic coordinates of the soluble subunit II from Cytochrome c oxidase from Thermus thermophilus (PDB ID code 2CUA, chain A) as search probe 10 . The structures were refined by iterative cycles of manual model building with Coot 11 and refinement with Phenix.refine 12,13 . Copper atoms were manually placed in mF o -DF c sigma-A-weighted electron density maps employing COOT and the resulting models were refined as described above. The final structures were validated through the Molprobity server (http://molprobity.biochem.duke.edu) 14 . They contained more than 97% of residues within favored regions of the Ramachandran plot, with no outliers. Figures were generated and rendered with Pymol version 2.0 (Schrödinger, LLC). Atomic coordinates and structure factors were deposited in the Protein Data Bank under the accession codes 6PTY (Tt3Lh) and 6PTT (Tt3LAt).

Computational methods
The initial structure of oxidized Tt Cu A was obtained from PDB ID code 2CUA and was relaxed through an equilibration process which consisted of an energy minimization followed by a slow heating from 0 K to 300K (400 ps). Starting from these equilibrated structures, 20 ns long production MD simulations in explicit water were performed at 1 atm and 300 K using the Berendsen barostat and thermostat, respectively. Periodic boundary conditions and Ewald sums were used to treat long-range electrostatic interactions and a 12 Å cut-off was used for computing direct interactions. In order to to keep bonds involving hydrogen atoms at their equilibrium length, the SHAKE algorithm was used. All simulations were performed with the PMEMD module of the AMBER16 package 15-17 . The Amber ff14SB force field was used for all residues but the Cu site, whose parameters where developed using the MCPB.py model in AmberTools17 18 . Snapshots of each system were slowly cooled to 0 K (200 ps) in order to obtain the initial structures for the QM/MM simulations. These were performed at the DFT level using the SIESTA 19 code with the QM/MM implementation Hybrid 20 . For all atoms, basis sets of double zeta plus polarization quality were employed with cut-off and energy shift values of 150 Ry and 25 meV. All calculations were performed under the spin-unrestricted approximation using the generalized gradient approximation functional proposed by Perdew, Burke, and Ernzerhof (PBE) 21 . The scaled position link atom method was used to treat the interface between the QM and MM sections. Due to the fact that the carbonyl ligand is involved in a backbone amide bond, a large QM subsystem was required in order to assure stability. This QM section included both copper atoms and the aminoacids shown in Table S1. The rest of the protein and water molecules were treated classically using the Amber force field. All atoms included in the MD simulation were included in the QM/MM system and geometry optimization was performed at the QM/MM level for both proteins. Finally, single point QM calculations at the DFT level were performed on the optimized QM section using Gaussian09 22 in order to obtain spin densities. The mixed triple-zeta/double zeta (TZVP) basis set was used for Cu and S atoms, while the 6-31G* basis set were used on all the other atoms. Atom contributions to molecular orbitals were computed with the software Chemissian. Figure S1. 1 H NMR spectra of non-labeled wild-type TtCu A and the variants Tt3Lh and Tt3LAt. All spectra were recorded in phosphate buffer, 50mM, pH 6.0. Figure S2. 13 C NMR spectra of wild-type TtCu A and the mutant proteins Tt3Lh and Tt3LAt. All spectra were recorded in phosphate buffer, 50mM, pH 6.0. Figure S3. 1 H, 13 C HMQC spectra of 13 C-Met labeled samples of the different Cu A variants: TtCu A (blue), Tt3Lh (red) and Tt3LAt (green). All spectra were recorded in phosphate buffer, 50mM, pH 6.0. Figure S4. 1 H, 13 C HMQC spectra of ε-13 CH 3 labeled samples of the different Cu A variants: TtCu A (blue), Tt3Lh (red) and Tt3LAt (green). The purple and orange circles indicate the position of the ε-CH 3 -Met on TtCu A and the chimera variants, respectively. All spectra were recorded in phosphate buffer, 50mM, pH 6.0. Figure S5. 1