Chapter 8.13. Metalloproteins and systems of biological relevance

The reaction centre of photosystem II (PSII), the water-splitting oxygen-evolving complex (OEC) in photosynthetic organisms, has been subjected to extensive XAS studies from the early days of SR-based X-ray spectroscopy to the present XFEL era. During this period, XAS has proved to be critical to the interpretation and the refinement of the associated crystallo­graphic data and structural modelling of the OEC (Yano et al., 2006; Grundmeier & Dau, 2012), which is now know to consist of an inorganic CaMn₄O₅ cluster with the Mn atoms linked by μ-oxo bridges and coordinated to amino-acid residues (Fig. 1a). Four successive photons are required to oxidize one water molecule as the system cycles round the S₀ → S₁ → S₂ → S₃ → S₄ (→ S₀ with O₂ release) states. These states have all been probed by static and time-resolved XAS experiments at cryogenic and room temperatures to reveal details of the water-splitting mechanism (Dau et al., 2001; Yachandra et al., 1993; Liang et al., 2000; Messinger et al., 2001; Haumann, Liebisch et al., 2005; Haumann, Müller et al., 2005). Mn K-edge EXAFS provided the first structural evidence for the existence of a μ-oxo-bridged metal cluster in the OEC (Kirby et al., 1981), while XANES suggested a mixed metal valence state for the cluster, which is altered during the S₁ → S₂ transition (from Mn³⁺/Mn³⁺ to Mn³⁺/Mn⁴⁺) while the overall coordination is unchanged (Yachandra et al., 1987). The average Mn–Mn distance in the cluster was determined to be ∼2.7 Å, with each Mn coordinated to three O/N atoms at ∼2 Å in addition to the two bridging O/N atoms and having 0.5–2 Mn neighbours. A shell of Mn or Ca atoms at ∼3.5 Å was also detected by XAS and suggested to form part of the cluster (MacLachlan et al., 1992), and was subsequently confirmed by Ca K-edge EXAFS experiments (Cinco et al., 2002) and anomalous diffraction on PSII crystals (Ferreira et al., 2004). The metrical data from XAS, including the geometrical constraints specified by polarized EXAFS studies of oriented chloroplast membranes (George, Prince & Cramer, 1989; Yachandra et al., 1993) and single crystals (Yano et al., 2006), provided crucial information on the composition, shape and size of the cluster prior to high-resolution crystallographic data becoming available. The first high (1.9 Å) resolution crystal structure, ostensibly of the S₁ state of OEC, appeared in 2011 (Umena et al., 2011), and for the first time resolved the positions of the metal atoms, bridging O atoms and protein ligands. Radiation damage to the OEC is a problem for SR-based crystallographic data collection (Yano, Kern et al., 2005) and the metal cluster in this structure is likely to be in a partially reduced state (some Mn³⁺/Mn⁴⁺ converted to Mn²⁺ and longer Mn–Mn bonds than seen in EXAFS) rather than pure S₁. Nevertheless, this structure provided a foundation for exploring the nature of the different S states of the catalytic cycle when used in combination with XAS data (Grundmeier & Dau, 2012) and density-functional theory (DFT) modelling (Siegbahn, 2013). Radiation-damage-free structures of PSII in different states and at varying resolutions (2.2–6.5 Å) have recently been reported using femtosecond XFEL crystallo­graphy (see, for example, Kern et al., 2012, 2018; Kupitz et al., 2014; Suga et al., 2019; Sauter et al., 2016). The quest to finalize the OEC structure and definitively establish the water-splitting mechanism therefore continues into the XFEL era, with XAS (and XES) playing a central role alongside crystallo­graphy.

Figure 1 A selection of metal centres in proteins that have been elucidated by XAS in conjunction with crystallographic studies. (a) The structure of the oxygen-evolving CaMn4O5 complex of photosystem II. (b) The structure of the FeMo cofactor of Mo-nitrogenase. S and Fe atoms are shown as yellow and orange spheres, respectively. (c) The active site of CuZn superoxide dismutase in its oxidized water-bound state. (d) The hydrogenase enzyme bimetallic Ni–Fe active-site cluster.

XAS has been used to characterize the metal centres in several molybdenum-containing proteins or enzymes, including the nitrogenases, which catalyse a central reaction of the global nitrogen cycle: the fixing and reduction of N₂ gas to NH₃ with the associated release of H₂. Nitrogenases are formed of two component metalloproteins, an Fe protein and an MoFe protein, which contains two FeMo cofactors (`FeMoco') and two FeS (`P') clusters. Alternative nitrogenases contain vanadium or iron in place of molybdenum. Nitrogenase was the first enzyme to be studied by XAS (Cramer et al., 1976), and Mo K-edge XAS of the Clostridium pasteurianum and Azotobacter vinelandii nitrogenases provided the first structural evidence for the presence of a common but unique Mo–Fe–S cluster in the MoFe protein, with 3–4 S atoms at ∼2.3 Å, 2–3 Fe atoms at ∼2.7 Å and one S atom at ∼2.5 Å (Cramer, Gillum et al., 1978). V and Fe K-edge studies of the VFe protein of A. chroococcum nitrogenase demonstrated a structure analogous to FeMoco and suggested that 3–4 O/N atoms were coordinated to vanadium at ∼2.1 Å, along with two S atoms at ∼2.3 Å and three Fe atoms at ∼2.7 Å (Arber et al., 1987). XAS was also used to propose the existence of an FeFe cofactor in the Fe-only nitrogenase from Rhodobacter capsulatus (Krahn et al., 2002). While many of the structural models of FeMoco and the P-cluster proposed on the basis of XAS studies of the enzymes and synthetic Mo–Fe–S complexes were discarded following the publication of the first 2.7 Å resolution crystal structure of A. vinelandii nitrogenase (Kirn & Rees, 1992), XAS proved its utility in providing structural constraints and in aiding interpretation of the crystallo­graphic data for the metal centres of the MoFe protein. The crystal structures showed [MoFe₃S₃] and [Fe₄S₃] clusters linked by three S²⁻ bridges, with the molybdenum in the cluster directly coordinated to three S atoms, two O atoms from homocitrate and one N atom from a His ligand (Fig. 1b). In the presence of the heavy atoms the homocitrate and histidine ligands in the cluster were less well determined in the initial crystal structures. A 3D EXAFS refinement (Cheung et al., 2000) using 1.6 Å resolution crystallographic coordinates of the Klebsiella pneumoniae MoFe protein showed that EXAFS-derived distance restraints can be used in the early stages of crystallographic refinement to substantially improve the final crystallographic model (Strange et al., 2003). The accuracy of this approach was confirmed by a 1.16 Å resolution crystal structure of the MoFe protein (Einsle et al., 2002), which also revealed the presence of an interstitial light atom, later confirmed to be a C atom, coordinated by Fe at the centre of FeMoco. An analogous carbide at the centre of the FeV cofactor was confirmed by DFT modelling and XES data (Rees et al., 2015). XAS and spectroscopic studies of nitrogenases have also focused on substrate binding and the probing of associated conformational changes in FeMoco. Conformational changes were observed to a long Fe—Fe bond length in FeMoco upon the binding of propargyl alcohol to an Fe atom adjacent to the Mo atom (George et al., 2012), while EXAFS of the resting enzyme showed that the interaction between the Mo atom and the shell of Fe atoms originally at 5.08 Å was altered by reaction with CO (Scott et al., 2014). Recently, XAS and crystallography were combined to assign individual Fe-atom oxidation states to the resting state of FeMoco (Spatzal et al., 2016), while high-energy resolution fluorescence-detected XAS and EXAFS were used to probe the local electronic and geometric structure of selenium-substituted FeMoco to reveal antiferromagnetic coupling in the cluster (Henthorn et al., 2019). The MoFe protein crystal structure has now been resolved to 1 Å resolution (Spatzal et al., 2011), yet these recent experiments demonstrate that nearly 45 years after its first application and when used in combination with crystallography or other methods, for example vibrational spectroscopy and DFT, XAS still has an important role in discovering the important functional states of FeMoco in nitrogenase proteins.

The redox cycling and structure of the copper and zinc centres in CuZn superoxide dismutase (SOD), which catalyses the conversion of the radical to O₂ and H₂O₂, has been explored by a number of XAS studies (Blackburn et al., 1987; Murphy et al., 1997); the first was by Blumberg et al. (1978), who used Cu and Zn K-edge XANES to probe the electronic states of oxidized and reduced protein. The crystal structure of oxidized (bovine) SOD showed that four histidine ligands are coordinated to the solvent-exposed copper, with one of these histidine ligands being coordinated to zinc (Fig. 1c). The loss of this Cu–His–Zn bridge and proton donation from the bridging histidine were proposed to be crucial steps in the catalytic mechanism (Tainer et al., 1983). The first direct structural evidence for a three-coordinate copper site in reduced SOD was provided by Cu K-edge EXAFS (Blackburn et al., 1984), with Zn K-edge EXAFS showing the zinc site to be unperturbed by the cycling of the redox state of copper. The first crystal structures of reduced SOD were ambiguous, suggesting an intact Cu⁺–His–Zn²⁺ bridge in reduced bovine SOD1 protein (Rypniewski et al., 1995). A subsequent 1.15 Å resolution structure of fully reduced bovine SOD was shown to be consistent with the XAS data and the accepted catalytic mechanism (Hough & Hasnain, 2003).

Type I (`blue') copper centres mediate electron-transfer reactions. Ligand K-edge XAS was used as a direct probe of the covalency of metal–ligand bonds (Glaser et al., 2000). Experiments performed at the S K edge showed that the high covalency of the unusually short Cu—S(Cys) bond is responsible for the small hyperfine splitting observed in the type I proteins (Shadle et al., 1993). The first direct structural information on the type I copper proteins was provided by Cu K-edge XAS of oxidized Pseudomonas aeruginosa azurin in 1978 (Tullius et al., 1978). This study revealed the presence of a remarkably short (2.1 Å) Cu—S bond together with two or three N atoms coordinated at 1.97 Å. The first crystal structures were reported in 1978 for P. aeruginosa azurin (at 3 Å resolution) and Populus nigra plastocyanin (at 2.7 Å resolution), and confirmed that the active site consisted of two N(His) ligands, one S(Cys) ligand and one S(Met) ligand. EXAFS and other spectroscopic data showed that the hydrophobicity of the axial ligand is the dominant factor in determining the redox properties of type I proteins (Berry et al., 2003). The function of the methionine ligand (or an alternative fourth ligand) in modulating the redox potential was also addressed by variable-pH spectroscopy, ligand binding and XAS of azurin mutants (Murphy et al., 1993; Strange et al., 1996) to show that the methionine ligand in the native protein constrains the active site to adopt a preferred Cu²⁺ orientation that facilitates conversion to Cu⁺. XAS studies of Rhus vernicefera stellacyanin conducted at different pH values showed that the fourth ligand was either N- or O-donating, at 2.7 Å, and not a sulfur ligand (Strange et al., 1995). These findings from XAS were confirmed by the subsequent crystal structures, which showed that glutamine is the fourth ligand, with a Cu–O(Gln) distance of 2.2–2.5 Å (Hart et al., 1996; Koch et al., 2005).

Iron–sulfur electron-transport proteins were early subjects of XAS studies. EXAFS was used to correct the initial crystal structure of rubredoxin, where an unusually short Fe—S bond in the single FeS(Cys)₄ cluster had been proposed (Bunker & Stern, 1977). Improved resolution (1.2 Å) crystal structures subsequently confirmed the EXAFS structural model. A systematic EXAFS study was also made of 2Fe and 4Fe iron–sulfur clusters in proteins and model compounds (Teo et al., 1979), while the presence of an Fe₃S₄ site in the 3Fe iron–sulfur clusters in ferredoxin II (Antonio et al., 1982) and the inactive form of aconitase (Beinert et al., 1983) was established by EXAFS and was subsequently confirmed by crystallo­graphy (Kissinger et al., 1989; Robbins & Stout, 1989).

XAS studies of FeFe and NiFe hydrogenases, enzymes that perform the reversible oxidation of H₂, are greatly complicated by the presence of different iron sites, i.e. in the multiple Fe–S clusters and in the active sites. Nevertheless, Ni K-edge EXAFS was used more than 30 years ago to demonstrate the presence of a novel Ni–S cluster in the active sites of the NiFe hydrogenases from Methanobacterium thermoautotrophicum (Lindahl et al., 1984) and Desulfovibrio gigas (Scott et al., 1984). In the FeFe hydrogenase from Clostridium pasteurianum, EXAFS showed the active site to contain an Fe–Fe interaction in the oxidized state at a distance of ∼2.8 Å, which increased by ∼0.5 Å upon reduction (George, Prince, Stockley et al., 1989). While incomplete [for example, the ∼2.7 Å Ni–Fe interaction in the first crystal structure of oxidized D. gigas hydrogenase (Volbeda et al., 1996) was not detected by EXAFS, possibly due to interference between Ni–Fe and Ni–S scattering (Gu et al., 1996)], these EXAFS-derived models suggested new kinds of metal-centred biological structures. The Fe–Fe enzymes are now known to consist of a combination of 4Fe–4S and Fe–Fe domains, with distinctive nonprotein CO and CN⁻ ligands of iron in the latter domain being essential for biological activity (the NiFe enzyme active site is similar; Fig. 1d). XAS has shown that the irreversible O₂-induced degradation of the active site, which is a problem facing the development of robust `hydrogenase-based' biofuel cells or biomimetics, results from specific damage to the 4Fe–4S (but not the Fe–Fe) domains (Stripp et al., 2009). Using more refined methods of analysis and improved data quality, current XAS investigations have shown good agreement with, and have aided in the interpretation of, crystallographic data for the hydrogenases (Gu et al., 2003; Hiromoto et al., 2009).

Currently (mid-2020) around 5400 iron haem structures have been deposited in the PDB. Of these entries, only ∼40 unique proteins (∼170 non-unique) are solved to a resolution (<1.2 Å) where the accuracy and precision of XAS at the metal site is matched by crystallography. Important questions addressed by XAS include the nature of the subtle structural changes that occur upon the binding of ligands to the Fe atom in the ferric and ferrous forms, the small (∼0.5 Å) displacement of Fe from the haem plane in deoxyhaemoglobin (Eisenberger et al., 1978; Perutz et al., 1982), which is seen as a key factor in ligand-binding affinity and cooperativity, and intermediates in peroxide chemistry in P450 and analogous enzymes (see, for example, Stone et al., 2005; Krest et al., 2015). Many of the earlier studies were performed before crystal structures became available and, more recently, when used in combination with MX data, XAS continues to probe transient states that are inaccessible to crystallography. Multiple-scattering XANES has been used to probe the structures of the deoxy forms of myoglobin and haemoglobin, while polarized single-protein crystal experiments have been used to characterize carboxymyoglobin (Della Longa et al., 1999), cyanomyoglobin (Arcovito et al., 2007) and the photoreduction of the iron in aquomet-myoglobin (Della Longa et al., 2003). Other steady-state XAS studies have included those in combination with laser flash photolysis to examine the geminate rebinding of carbon monoxide to myoglobin at cryogenic temperatures (Chance et al., 1983; Della Longa et al., 2001). Thirty years ago a groundbreaking submillisecond time-resolved XANES study was made of carboxymyoglobin, finding a −3 eV shift in the absorption energy of the transient photolysed product compared with the ligand-bound protein (Mills et al., 1984). X-ray absorption experiments with <100 ps resolution on third-generation synchrotrons have recently allowed structural interpretations of the Fe K-edge spectra of transient carboxymyoglobin and nitrosomyoglobin complexes, showing ligand dissociation and doming of the haem (Stickrath et al., 2013; Silatani et al., 2015). Time-resolved XAS of carboxymyoglobin performed using an XFEL has suggested that a rearrangement of the haem occurs within the first 70 fs of photolysis occurring, with iron moving out of the haem plane within 400 fs (Levantino et al., 2015). Structural information from photoactivated transition-metal complexes has also been demonstrated at XFELs using femtosecond-resolution EXAFS (Britz et al., 2020).

XAS remains a powerful synchrotron-based X-ray method. It yields detailed local structural and electronic information on biological samples in different states of matter and does not require, although it can use, samples with long-range order or single crystals. The metal environments of `uncrystallizable' metalloproteins can easily be studied in their chemically relevant states. When used in combination with crystallo­graphic information, XAS can also provide important information to help interpret or improve crystallographic models and identify metal redox states. Techniques for rapid data collection for both MX and XAS experiments at synchrotrons and at XFELs will lead to an expansion of time-resolved applications to study catalysis in situ using crystals or other states of matter.