2.1. S-state cycle and the Mn₄Ca²⁺O _X catalytic centre

The function of the Mn-cluster is to accumulate the four oxidizing equivalents required to split two water molecules and generate a molecule of dioxygen (Barber, Reference Barber2003). The final step leading to O–O bond formation has to be very fast involving two closely located activated oxygen atoms. In this way, the generation of reactive oxygen species is avoided and the triplet state form of O₂ established. Each oxidation step at one catalytic centre is driven by the absorption of one photon, which was elegantly shown by monitoring oxygen release in a series of single turnover flashes of light where the maximum yield of oxygen followed a period of four (Joliot et al. Reference Joliot, Barbieri and Chabaud1969). This finding which gave rise to a scheme known as the S-state cycle (S₀ to S₄) is depicted in Fig. 2 (Kok et al. Reference Kok, Forbush and McGLOIN1970). This cycle is powered by the initial oxidation of a chlorophyll known as P680 generated by light-driven primary charge separation in the reaction centre of PSII, which is coupled to a redox-active tyrosine (Y_Z) serving as an intermediate electron carrier between P680⁺ and the Mn-cluster.

Fig. 2. The S-state cycle showing how the absorption of four photons of light (hv) by the reaction centre primary oxidant P680 drives the splitting of two water molecules and formation of O₂ through a consecutive series of five intermediates (S₀, S₁,S₂, S₃ and S₄). Protons (H ⁺ ) are released during each step of this cycle except for the S₁ to S₂ transition. Electron donation from the Mn₄Ca²⁺ cluster to P680^•+ is aided by the redox-active tyrosine Y_Z. Each step involves a single oxidation of a Mn ion in the cluster, starting at S_O with 3 × Mn^III plus Mn^IV advancing to S₃ with 4 × Mn^IV. The exact oxidation state of S₄ is unknown but could be 3 × Mn^IV plus Mn^V or 3 × Mn^IV plus Mn^IV-oxyl radical (see below). Also shown are half-times for the various steps of the cycle (taken from Barber, Reference Barber2017).

Each photon absorbed provides about 1 V of oxidizing potential per S-state transition since the oxidizing potential of P680⁺ is about 1·25 V. This means that the overpotential for each step of the catalytic cycle is relatively small. As shown in Fig. 2, as the catalytic cycle proceeds from S₀ to S₄, protons are released at each S-state step except for the S₁ to S₂ transition where the metal cluster accumulates one positive charge. The redox-active Y_Z is a neutral tyrosyl radical where its phenolic proton is donated to a nearby base, D1His190, to which it is H-bonded. Therefore, the neutral Y_Z radical is an ideal candidate for facilitating a proton-coupled electron transfer (PCET) from the catalytic centre, as originally suggested by Hoganson & Babcock (Reference Hoganson and Babcock1997). An important study (Haumann et al. Reference Haumann, Liebisch, Muller, Barra, Grabolle and Dau2005), indicated the coupling is sequential, with a strictly alternating removal of electrons and protons rather than hydrogen atom transfer.

A whole range of techniques, particularly magnetic resonance spectroscopy, have shown that each step in this cycle involves an increase in the oxidation of a Mn ion, starting at S₀ with 3 × Mn^III plus Mn^IV advancing to S₃ with 4 × Mn^IV (Cox et al. Reference Cox, Retegan, Neese, Pantais, Boussac and Lubitz2014). The precise details of the final oxidation state, S₄, are not clear because O–O bond formation has to be very fast to avoid side reactions.

Not surprisingly, it has been one of the greatest challenges of photosynthesis research to determine the structure of the Mn-containing cluster of the PSII oxygen-evolving complex (OEC) and thus reveal the molecular mechanism of the water-splitting reaction. Over the years there have been many postulates of its structure mainly derived from X-ray absorption fine-structure (XAFS) spectroscopy (Yachandra, Reference Yachandra2002) and electron paramagnetic resonance (EPR) studies (Britt et al. Reference Britt, Campbell, Peloquin, Gilchrist, Aznar, Cicus, Robble and Messinger2004; Haddy, Reference Haddy2007). However, in 2004 a complete fully refined structure of PSII was determined at 3·5 Å resolution using X-ray crystallography by my colleagues and me at Imperial College London and published in Science (Ferreira et al. Reference Ferreira, Iverson, Maglaoui, Barber and Iwata2004).

We were able to assign, for the first time, over 5000 amino acid side chains of this huge dimeric membrane complex (700 kD) with each monomer consisting of 19 different protein subunits. In so doing we answered many outstanding questions as well as revealing a wide range of important details hitherto unknown. The native electron density map, together with anomalous diffraction data collected at wavelengths of 1·89 and 2·25 Å, provided electron density profiles for Mn and Ca²⁺, respectively. This information was then used to build a model of the metal-cluster. The anomalous electron density attributed to the four Mn ions was ‘pear-shaped’ indicative of the 3 + 1 organization and thus one Mn was assigned to the small domain and three in the large globular domain, whereas the 2·25 Å wavelength map covered one metal ion in the large domain of the native density. From these data the three Mn ions (Mn1, Mn2 and Mn3) and the Ca²⁺, located in the large domain, were modelled with a cubane geometry having bridging oxygens, an organization which was compatible with the native electron density. The fourth Mn ion (Mn4), located in the small domain was modelled so that it was linked to the cubane by one of its bridging oxygens (dangler Mn4) (see Fig. 3a ). Based on values determined by XAFS measurements and comparison with other Mn-containing proteins, the three Mn-di-μ-oxy-Mn bonds of the cubane were spaced at 2·7 Å, while the three Mn-di-μ-oxy-Ca²⁺ bonds were separated by 3·4 Å. The dangler Mn4 was positioned 3·3 Å from the closest Mn ion of the cubane and about 4 Å from the Ca²⁺. The 3 + 1 organization of the Mn-cluster was predicted by Peloquin & Britt (Reference Peloquin and Britt2001) based on the degree of spin coupling detected by resonance spectroscopy and they introduced the term ‘dangler’ for the less coupled Mn ion (Mn4) (Peloquin & Britt, Reference Peloquin and Britt2001).

Fig. 3. (a) Comparison of the Mn₄Ca²⁺O₄ cubane model from coordinates (PDB 1S5L) deposited by Ferreira et al. in Reference Ferreira, Iverson, Maglaoui, Barber and Iwata2004 with (b) the 1·9 Å structure of the Mn₄Ca²⁺O₅ cluster from Umena et al. (Reference Umena, Kawakami, Shen and Kamiya2011) (PDB 3WU2) and (c) the chemically synthesized Mn₄Ca²⁺O₄-cubane by Zhang et al. (Reference Zhang, Chen, Dong, Shen, Dau and Zhao2015). Note that unfortunately, the numbering of the Mn and O are different in (b) and (c) to that originally given in (a) (taken from Barber, Reference Barber2016b).

At that time there was no similar chemical structure known in biochemistry or chemistry and therefore the model was greeted with some uncertainty, especially since it had been conceived from relatively low-resolution data. But the main opposition to this model was the argument that the metal cluster, and surrounding amino acid ligands, would be seriously disturbed by radiation damage during data collection (Yano et al. Reference Yano, Kern, Irrgang, Latimar, Bergmann, Glatzel, Pushkar, Loll, Sauer, Messinger and Yachandra2005). Indeed the cubane structure did not seem to be consistent with polarized XAFS measurements made at much lower radiation levels (Yano et al. Reference Yano, Kern, Sauer, Latimer, Pushkar, Biesiadka, Loll, Saenger, Messinger, Zouni and Yachandra2006). Unfortunately, the claim that the Ferreira et al. cubane model was invalid because of significant radiation damage caused a substantial level of confusion for several years, both for specialists and non-specialists. However, it has now become clear that the 2004 cubane model of Ferreira et al. was valid within the limitation of its resolution and therefore the extensive radiation damage argument and associated alternative postulated XAFS structures for the Mn-cluster (Yano et al. Reference Yano, Kern, Sauer, Latimer, Pushkar, Biesiadka, Loll, Saenger, Messinger, Zouni and Yachandra2006) turned out to be wrong. However, doubts in the interpretation of the polarized XAFS data were presented by Sproviero et al. (Reference Sproviero, Gascon, McEVOY, Brudvig and Batista2008) who concluded that the polarized XAFS spectra were consistent with the cubane model of Ferreira et al., which they went on to refine using QM/MM calculations.

In 2011, 7 years after the coordinates of the cubane model by Ferreira et al. were deposited, the uncertainty of the cubane model was finally lifted with the report of 1·9 Å PSII crystal structure from Umena et al. (Reference Umena, Kawakami, Shen and Kamiya2011). At this resolution, the electron densities of individual metal ions could be resolved and bridging oxygens inferred. As shown in Fig. 3b , the resulting model was overall similar to that of Ferreira et al. (Reference Ferreira, Iverson, Maglaoui, Barber and Iwata2004) except an additional oxo bridge was proposed to link the dangler Mn outside the Mn₃Ca²⁺ cubane (Mn4) to a Mn ion of the cubane (Mn3). Importantly this additional oxo-bridge had already been predicted by Siegbahn (Reference Siegbahn2008) and by Dau et al. (Reference Dau, Grundmeier, Loja and Haumann2008) based on a modification of the original 2004 cubane model using Density Function Theory (DFT) and other quantum mechanical methods. Nevertheless, the 1·9 Å structure of Umena et al. (Reference Umena, Kawakami, Shen and Kamiya2011), together with high-resolution diffraction data obtained by femtosecond X-ray free-electron laser (XFEL) pulses (Suga et al. Reference Suga, Akita, Hirata, Ueno, Murakami, Nakajima, Shimizu, Yamashita, Yamamoto, Ago and Shen2015, Reference Suga, Akita, Sugahara, Kubo, Nakajima, Nakane, Yamashita, Umena, Nakabayashi, Yamane, Nakano, Suzuki, Masuda, Inoue, Kimura, Nomura, Yonekura, Yu, Sakamoto, Motomura, Chen, Kato, Noguchi, Tono, Joti, Kameshima, Hatsui, Nango, Tanaka, Naitow, Matsuura, Yamashita, Yamamoto, Nureki, Yabashi, Ishikawa, Iwata and Shen2017; Young et al. Reference Young, Ibrahim, Chatterjee, Gul, Fuller, Koroidov, Brewster, Tran, Alonso-Mori, Kroll, Michels-Clark, Laksmono, Sierra, Stan, Hussein, Zhang, Douthit, Kubin, de Lichtenberg, Long Vo, Nilsson, Cheah, Shevela, Saracini, Bean, Seuffert, Sokaras, Weng, Pastor, Weninger, Fransson, Lassalle, Bräuer, Aller, Docker, Andi, Orville, Glownia, Nelson, Sikorski, Zhu, Hunter, Lane, Aquila, Koglin, Robinson, Liang, Boutet, Lyubimov, Uervirojnangkoorn, Moriarty, Liebschner, Afonine, Waterman, Evans, Wernet, Dobbek, Weis, Brunger, Zwart, Adams, Zouni, Messinger, Bergmann, Sauter, Kern, Yachandra and Yano2016) have provided reliable atomic structures of the OEC at cryo-and room-temperatures, particularly in its dark-stable S₁ state. Therefore this has given validity to the detailed quantum mechanical calculations for the OEC and its functioning (Cox et al. Reference Cox, Retegan, Neese, Pantais, Boussac and Lubitz2014; Lohmiller et al. Reference Lohmiller, Krewald, Navarro, Retegan, Rapatskiy, Nowaczyk, Boussac, Neese, Lubitz, Pantazis and Cox2014; Luber et al. Reference Luber, Rivalta, Umena, Kawakami, Shen, Kamiya, Brudvig and Batista2011; Yamanaka et al. Reference Yamanaka, Isobe, Kanda, Saito, Umena, Kawakami, Shen, Kamiya, Okumur, Nakamura and Yamaguchi2011) and more rigorous interpretation of experimental results (Ames et al. Reference Ames, Pantazis, Krewald, Cox, Messinger, Lubitz and Neese2011; Bovi et al. Reference Bovi, Narzi and Guidoni2013; McConnell et al. Reference McConnell, Grigoryants, Scholes, Myers, Chen, Whittaker and Brudvig2012; Rapatskiy et al. Reference Rapatskiy, Cox, Savitsky, Ames, Sander, Nowaczyk, Rogner, Boussac, Neese, Messinger and Lubitz2012).

2.2. Synthesized cubanes mimicking the Mn₄Ca²⁺O _X cluster

Of considerable importance is the report that the PSII cubane cluster can be synthesized as a Mn₄Ca²⁺O₄ molecule (Zhang et al. Reference Zhang, Chen, Dong, Shen, Dau and Zhao2015) whose structure is very similar to that of the OEC (compare Figs 3c with 3d ) and remarkably similar to the original model of Ferriera et al. (compare Figs 3c with 3a ). As with the Ferreira et al. structure, the synthetic Mn₄Ca²⁺O₄ molecule was also missing the additional oxo-bridge to Mn4. The chemical synthesis of the Mn₄Ca²⁺O₄ structure in organic solvent shows that this cluster does not require protein to assemble and as long as there is an appropriate source of ligands, the structure is stable in an oxidized state. Moreover, cyclic voltammetry conducted in 1,2-dichloroethane showed that it can advance through the S-state cycle to S₃ (Zhang et al. Reference Zhang, Chen, Dong, Shen, Dau and Zhao2015). However, no oxygen evolution was detected when a small amount of water was present. Nevertheless, this extraordinary finding may have implications for the evolutionary origin of the OEC and provides some explanation for the robustness of the in vivo cluster to site-directed mutations (see below) and X-ray radiation damage.

Before the synthesis of the Mn₄Ca²⁺O₄ cluster by Zhang et al. (Reference Zhang, Chen, Dong, Shen, Dau and Zhao2015), Christou and colleagues (Mukherjee et al. Reference Mukherjee, Stull, Yano, Stamatato, Pringouri, Stich, Abbroud, Britt, Yachandra and Christou2012) reported the synthesis of a Mn₃Ca₂O₄ molecule which is structurally similar to that of Mn₄Ca²⁺O₄ (see Fig. 4c ). Another important study was the synthesis of a Mn₃CaO₄ cubane by Agapie and colleagues (Kanady et al. Reference Kanady, Tsui, Day and Agapie2011) (see Fig. 4d ). However, neither showed redox properties like that of the Mn₄CaO₄ synthesized by Zhang et al. (Reference Zhang, Chen, Dong, Shen, Dau and Zhao2015).

Fig. 4. Diagrammatic comparison of synthesized cubane structures (b), (c) and (d) with the Ferreira et al. X-ray model (a). Mn purple, Ca²⁺ green and red are oxo bonds (taken from Barber, Reference Barber2016b).

The fact that a Mn₄Ca²⁺O₄ structure like the one proposed by Ferreira et al. can also exist in the lattice of a macromolecular Mn–Ca molecule was shown by Misra et al. (Reference Misra, Wernsdorfer, Abboud and Christou2005). They synthesized a [Mn₁₃Ca₂O₁₀(OH)₂(OMe)₂(O₂CPh)₁₈(H₂O)₄] molecule in a highly oxidized state (Fig. 5a ) which consisted of two Mn₄Ca building blocks. As can be seen in Fig. 5, one form is the well-known Mn₄O₄ cube (Ruettinger et al. Reference Ruettinger, Ho and Dismukes1999) with a Ca attached to the cube by one of its bridging oxos (Fig. 5c ), while the other building block (Fig. 5b ) has a similar cubane structure to that of Mn₄CaO₄ of PSII. This study was the first to show that Mn and Ca can form discrete molecular complexes and also gave credibility to the Mn₄CaO₄ cubane model of Ferreira et al. (Reference Ferreira, Iverson, Maglaoui, Barber and Iwata2004) well before the publication of the 1·9 Å structure of Umena et al. (Reference Umena, Kawakami, Shen and Kamiya2011).

Fig. 5. (a) [Mn₁₃Ca₂O₁₀(OH)₂(OMe)₅(O₂CPh)₁₈(H₂O)₄] macromolecule synthesized by Misra et al. (Reference Misra, Wernsdorfer, Abboud and Christou2005) composed of two building blocks shown in (b) and (c). Ca is turquoise, Mn purple and yellow and oxygen in red.

3.1. Base-catalysed nucleophilic attack mechanism

Prior to the 2004 crystal structure, there had been many postulates of how PSII splits water to form dioxygen. One idea suggested by Messinger et al. (Reference Messinger, Badger and Wydrzynski1995) and Pecararo et al. (Reference Pecararo, Baldwin, Caudl, Hsieh and Law1998) was that O–O bond formation in PSII resulting from a based catalysed nucleophilic attack of a hydroxyl group onto an electrophilic terminal oxo derived from the deprotonation of the second substrate water molecule. In particular, Pecararo et al. (Reference Pecararo, Baldwin, Caudl, Hsieh and Law1998) conducted studies on synthesized Mn complexes and despite there being no crystal structure of the Mn₄Ca²⁺ cluster at that time they proposed that a ‘nucleophilic attack by a calcium-ligated hydroxide on an electrophilic oxo group ligated to a high-valent manganese to achieve the critical O–O bond formation’. The high oxidation state was taken to be Mn^V in line with the suggestion of Limburg et al. (Reference Limburg, Brudvig and Crabtree1997), but could also be a Mn^IV-oxyl radical (Brudvig, Reference Brudvig2008). The Ca²⁺ would act as a weak Lewis acid and align a specific H₂O molecule for an efficient nucleophilic attack of an OH on the highly electrophylic terminal oxo of Mn4. This mechanism was also favoured by Britt et al. (Reference Britt, Campbell, Peloquin, Gilchrist, Aznar, Cicus, Robble and Messinger2004) based on EPR studies.

The cubane structure of the Mn₄Ca²⁺O₄ cluster of PSII derived from X-ray crystallography in 2014 by my colleagues and I was consistent with this proposed mechanism and discussed in our papers at that time (Barber et al. Reference Barber, Ferreira, Maghlaoui and Iwata2004; Ferreira et al. Reference Ferreira, Iverson, Maglaoui, Barber and Iwata2004). Importantly, the dangler Mn4 was found to be immediately adjacent to the Ca²⁺ and close to side chains of several key amino acids, including the redox-active tyrosine, Y_Z , and therefore we suggested that Mn4 and Ca²⁺ form a ‘‘catalytic’ surface outside the cubane for binding the two substrate water molecules and their subsequent oxidation. The progression through the S-state cycle would lead to deprotonation of these water molecules and the formation of a highly electrophilic Mn^V-oxo or Mn^IV-oxyl radical, and to a nucleophilic hydroxyl on the Ca²⁺ in the former case and an oxo–oxyl coupling mechanism in the latter so as to generate a dioxygen molecule (not shown in Fig. 6).

Fig. 6. Diagrammatic representation of a mechanistic scheme for water splitting and dioxygen formation in PSII reproduced from Barber (Reference Barber2017). The substrate water molecules and products of the oxidation reactions for each S-state are shown in red. Intermediates that may exist between the S-state transitions (Dau & Haumann, Reference Dau and Haumann2007) are not depicted nor is the possible peroxide intermediate just prior to O–O bond formation. The essence is that a proton (H⁺) and an electron (e⁻) are removed by PECT for the flash-induced S0 to S1, S2 to S3 and S3 to S4 light-driven transitions but only an electron is removed during the S1 to S2 transition resulting in the accumulation of a positive charge in the metal cluster as shown. The final S3 to S4 flash-induced transition progresses to a highly electrophilic terminal oxo (electron deficient) ideally poised for a nucleophilic attack by a hydroxyl coordinated to the nearby Ca ion (electron rich) (taken from Barber, Reference Barber2017).

Thus as shown in Fig. 6, the formation of the O–O bond of dioxygen would result from the based catalysed nucleophilic attack of hydroxyl onto the electrophilic oxo of Mn4 with a likely peroxide intermediate. The electrophilicity of Mn^V-oxo would be enhanced by the high oxidation level of the cubane (3 × Mn^IV) plus a net positive charge (formed during the S₁ to S₂ transition) acting together as an ‘oxidizing battery’. The deprotonation of the substrate waters would be aided by nearby bases such as CP43Arg357 and by the weak Lewis acidity of Ca²⁺. Indeed, there is an extensive H-bonding network leading from Yz to the lumenal side of OEC (Ferreira et al. Reference Ferreira, Iverson, Maglaoui, Barber and Iwata2004; Murray & Barber, Reference Murray and Barber2007; Umena et al. Reference Umena, Kawakami, Shen and Kamiya2011). However, as stated above, the dangler Mn4 and Ca²⁺ each bind two water molecules according to the high-resolution structure of PSII (Umena et al. Reference Umena, Kawakami, Shen and Kamiya2011) (see Fig. 6). This may indicate a ‘carousel mechanism’ providing efficient delivery of substrate water molecules to the active sites and allow maximum turnover rates for the S-state cycle rather than being restricted by the rate of water diffusion to the catalytic site just prior to the formation of the S₀-state. This can explain the action of ammonia, often considered to be an analogue of water. It has been shown that NH₃ binds strongly to Mn4 (Oyala et al. Reference Oyala, Stich, Debus and Britt2015) and does not inhibit production of O₂ but rather slows the turnover of the S-state cycle (Boussac e t al. Reference Boussac, Rutherford and Styring1990; Navarro et al. Reference Navarro, Ames, Nilsson, Lohmiller, Pantazis, Rapatskiy, Nowaczyk, Neese, Boussac, Messinger and Lubitz2013). This would be consistent with it inhibiting the carousel process but not the chemistry of O–O bond formation. The idea of a carousel mechanism was first proposed by Batista and colleagues but for a different mechanism for the S₂ to S₃ transition (Askerka et al. Reference Askerka, Wang, Vinyard, Brudvig and Batista2016; Vinyard & Brudvig, Reference Vinyard and Brudvig2017) and supporting in part the proposals described below in Section 4.4.

Although the electrophilic oxo in the S₄ state is shown attached to Mn^V it could equally be an energy equivalent terminal Mn^IV-oxyl, which would be available for an oxo–oxyl coupling mechanism.

Fig. 6 is a mechanistic scheme for water splitting and dioxygen formation in PSII based on the acid–base, hydroxyl/water nucleophilic attack mechanism for O–O bond formation discussed above and taken from Barber (Reference Barber2017).

4.1. Comparison with Fe–Ni carbon monoxide dehydrogenase (CODH)

There are anaerobic prokaryotic organisms which use the oxidation of carbon monoxide as an energy source to split water. The enzyme that catalyses this reaction is CODH. There are some interesting similarities between PSII and the Fe–Ni CODH, which supports the nucleophilic attack mechanism for photosynthetic O–O bond formation (Barber, Reference Barber2017). In CODH, the energy of the oxidation of a CO molecule to CO₂ is used to split one water molecule to produce carbon dioxide and two reducing equivalents.

(2) $$\hbox{CO} + 2\hbox{H}_{2} \hbox{O} \to \hbox{CO}_{2} + 2\hbox{e}^{\rm -} + 2\hbox{H}^{+} $$

This chemistry is the well-known ‘water-gas shift reaction’ discovered in 1780 by the Italian physicist, Fontana (Reference Fontana1780), and today adopted as a large-scale industrial process to make pure hydrogen gas from carbon monoxide. This commercial process requires shifts in reaction temperature from high (approximately 400 °C with an iron-chromium oxide catalyst) to low (approximately 225 °C with copper-based catalysts) (Smith et al. Reference Smith, Loganathan and Shantha2010).

In contrast, in the case of Fe–Ni CODH, the generated reducing equivalents are produced at ambient temperatures and are used to drive the reductive chemistry of the organism rather than producing hydrogen gas, usually by electron transport involving nearby Fe₄–S₄ centres (Gong et al. Reference Gong, Hao, Wei, Ferguson, Tallant, Krzycki and Chan2008; Svetlitchnyi et al. Reference Svetlitchnyi, Dobbek, Meyer-Klaucke, Meins, Thiele, Romer, Huber and Meyer2004). Remarkably, the catalytic centre of Fe–Ni CODH has a similar geometry to that of Mn₄CaO₅ of PSII (Fig. 7). X-ray crystallography at high resolutions (Dobbek et al. Reference Dobbek, Svetlitchnyi, Gremer, Huber and Meyer2001; Jeoung & Dobbek, Reference Jeoung and Dobbek2007) has shown that Fe–Ni CODH consists of a Fe₃NiS₄ cubane with a fourth ‘dangler’ Fe attached to the cubane via a bridging S of the cubane and an additional S bridge to make a Fe₄NiS₅ cluster as shown in Fig. 7b .

Fig. 7. Comparison of (a) the Mn₄Ca²⁺O₅ cluster of PSII at 1·9 Å using PDB 3WU2 (Umena et al. Reference Umena, Kawakami, Shen and Kamiya2011) with (b) Fe₄Ni²⁺S₅ cluster of CO dehydrogenase at 1·03 Å using PDB 4UDX (Fesseler et al. Reference Fesseler, Jeoung and Dobbek2015). (c) Overlay of the two structures with a RMSD of 0·65 between the metal atoms. Oxo bonds in red, sulphur bonds in yellow, manganese ions in purple, calcium and nickel ions in green as labelled and iron ions in orange (taken from Barber, Reference Barber2017).

Despite the two catalytic centres being composed of different elements, the overlay of their structures, shown in Fig. 7c , is remarkably good with a root mean square deviation (RMSD) of 0·65 between the metal atoms. In fact, they are the only known examples of metallo-catalytic centres having heterocubanes with an additional metal in exo.

It is known that in CODH, the CO is activated by binding to the Ni²⁺ of the cubane, which causes a shift in the coordination of the Ni atom from square-planar to a square-pyramidal geometry (Jeoung & Dobbek, Reference Jeoung and Dobbek2007) and increases the electrophilicity of CO. Similarly, water is activated by forming a complex with the dangler Fe4²⁺. This induces a movement of the cysteine Ni ligand leading to the possible formation of a carboxyl bridge between the Ni²⁺and Fe4. Thus, the mechanism is a base-catalysed nucleophilic attack by the Fe4 bound water/hydroxyl group onto CO (see Fig. 8). A decarboxylation of the intermediate leads to CO₂ release and the reduction of the metal cluster (Jeoung & Dobbek, Reference Jeoung and Dobbek2007). Oxidation back to the active state involves electron transfer to nearby Fe₄–S₄ centres including ferredoxin or to a closely bound acetyl-CoA synthase depending on species (Gong et al. Reference Gong, Hao, Wei, Ferguson, Tallant, Krzycki and Chan2008).

Fig. 8. A structurally based diagrammatic comparison of a base nucleophilic attack of a hydroxyl onto an electrophile leading to oxygen atom transfer for (a) O–O formation in the S₄ state of PSII and (b) CO conversion to CO₂ in CODH. Note that the numbering of Fe ions in (b) are different to those in PDB 4UDX (Fesseler et al. Reference Fesseler, Jeoung and Dobbek2015) so as to compare with the numbering of the Mn ions in (a) taken from PDB 3WU2 (Umena et al. Reference Umena, Kawakami, Shen and Kamiya2011). Unfortunately, the numbering of the Mn₄Ca²⁺O₅ in PDB 3WU2 (Umena et al. Reference Umena, Kawakami, Shen and Kamiya2011) differs from that of the earlier Mn₄Ca²⁺O₄ in PDB 1S5L (Ferreira et al. Reference Ferreira, Iverson, Maglaoui, Barber and Iwata2004). Note that in the S₄ state there is the storage of a net positive charge generated during the S₁ to S₂ transition (see Fig. 6) (taken from Barber, Reference Barber2017).

Given the striking geometrical similarities between the catalytic centres of CODH and PSII, together with the fact they both catalyse the removal of reducing equivalents from water, it would be reasonable to conclude that there may be common features in their catalytic mechanisms. Of course PSII water splitting involves high-valent oxidative chemistry, while CODH is low-valent organometallic chemistry. Moreover, the water-splitting reaction of PSII is non-reversible while that of CODH is. For these reasons, I wish to emphasize that the comparison of PSII with Fe–Ni CODH does not imply any direct evolutionary link but a similarity in their function to extract reducing equivalents from the water.

4.2. Ligands to the Mn₄Ca²⁺O₅ cluster

The Ferreira et al. structure (Reference Ferreira, Iverson, Maglaoui, Barber and Iwata2004) identified seven amino acids as ligands to the Mn₄Ca-cluster, six from the D1 reaction centre protein: Asp170, Glu189, His332, Ala344, Glu333 and Asp342 and Glu354 of the inner antenna chlorophyll-binding protein, CP43. That these seven amino acids are ligated to the Mn₄Ca-cluster has been confirmed in the 1·9 Å structure of Umena et al. (Reference Umena, Kawakami, Shen and Kamiya2011) and the precise details of the ligation pattern revealed (see Fig. 9). The coordination properties of the three Mn ions of the cubane are totally satisfied by amino acid ligands. However, as stated above, Umena et al. (Reference Umena, Kawakami, Shen and Kamiya2011) assigned two water ligands for the non-cubane dangler Mn4 and also for the Ca ion as shown in Fig. 9. Ferreira et al. (Reference Ferreira, Iverson, Maglaoui, Barber and Iwata2004) had previously suggested that water molecules could act as ligands to Mn4 and Ca²⁺ and that this may be the site for dioxygen formation.

Fig. 9. Diagrammatic representation of the amino acid side chains acting as ligands to the Mn₄CaO₅ cluster of the OEC together with four water molecules providing ligands to Ca and Mn4 (constructed from Umena et al. crystal structure (2011)). D1 residues shown in orange and green depict the CP43 residue Glu354 (taken from Barber, Reference Barber2016a).

The amino acids which serve as ligands to the metal cluster are highly conserved in PSII over the whole cyanobacterial, plant and algal kingdoms (Murray, Reference Murray2012). Yet, there are a large number of reports that their replacement by other residues using site-directed mutagenesis does not inhibit oxygen evolution completely and sometimes not at all (Debus, Reference Debus2008). One example is the replacement of the negatively charged D1-Glu189 by positively charged Lys, which made no difference to the rate of oxygen evolution (Clausen et al. Reference Clausen, Winkler, Hays, Hundelt, Debus and Junge2001). This was also true for other D1-Glu189 mutants leading to the logical conclusion that it cannot be a ligand to the Mn₄Ca-cluster. Another example is the replacement of acidic CP43-Glu354 by neutral glutamine where O₂ evolution persisted at a reduced rate of 10–18% compared with the wild-type (Service et al. Reference Service, Yano, McCONNELL, Hwang, Niks, Hille, Wydrzynski, Burnap, Hillier and Debus2010). It is also possible to totally remove CP43 and yet assemble the OEC (Buchel et al. Reference Buchel, Barber, Ananyev, Eshaghi, Watts and Dismukes1999). Thus compared with many other metallo-enzymes, the assembled Mn₄CaO₅ cluster is rather robust to changes in its first coordination sphere. There are exceptions since, for example, D1-Asp170 also acts as a high-affinity site for Mn²⁺ binding (Diner et al. Reference Diner, Nixon and Farchaus1991) so that the photo-oxidation of P680 to P680⁺ can facilitate its conversion to Mn³⁺. In this way, as long as CaCl₂ is present, the Mn₄CaO₅ is assembled step-by-step to form the OEC, a process known as photoactivation (Ananyev & Dismukes, Reference Ananyev and Dismukes1996).

4.3. Synthesized chemical model systems

Ruthenium model chemistry was the first to achieve homogeneous water oxidation and has subsequently become the best characterized. However, functional manganese model complexes have also been synthesized and have the added advantage of being more biologically relevant to the OEC.

4.3.1. Mechanism of O₂ production from organo-Ru complexes

The blue dimer, cis,cis-[(bpy)₂(H₂O)Ru^IIIORu^III(H₂O)(bpy)₂]⁴⁺ (where bpy = 2,2′-bipyridine) was the first designed, well-defined molecule known to function as a catalyst for water oxidation (Gersten et al. Reference Gersten, Samuels and Meyer1982). As in the case of PSII, it met the stoichiometric requirements for water oxidation by utilizing PCET reactions in which both electrons and protons are transferred. This avoided charge build-up, allowing for the accumulation of multiple oxidative equivalents at the Ru–O–Ru core. Since then a wide range of both dinuclear Ru and mononuclear Ru organo-complexes have been synthesized, which show high rates of O₂ evolution using Ce⁴⁺ as an oxidant. A significant number of them, particularly the mononuclear Ru molecules, generate the O–O bond by nucleophilic attack of an electrophilic oxo of Ru^V = O (Duan et al. Reference Duan, Wang, Li, Li and Sun2015; Murakami et al. Reference Murakami, Hong, Suenobu, Yamaguchi, Ogura and Fukuzumi2011; Sala et al. Reference Sala, Maji, Bofill, Garcia-Anton, Escriche and Llobet2013; Tong et al. Reference Tong, Duan, Xu, Privalov and Sun2011). This mechanism and the rate of turnover are dependent on the design of the ligands (Concepcian et al. Reference Concepcian, Jurss, Brennaman, Hoertz, Patrocinio, Murakami Iha, Templeton and Meyer2009). One example is [Ru^II(bpc)(bpy)(OH₂)]⁺ (where bpc = 2,2′-bipyridine-6-carboxylate and bpy = 2,2′-bipyridine). A proposed mechanism for these types of mononuclear Ru catalysts at low pH is shown in Fig. 10. The [Ru^II-OH₂]⁺ is first oxidized to [Ru^III-H₂]²⁺, followed by a 2H⁺/1e PCET oxidation step generating [Ru^IV = O]⁺. [Ru^IV = O]⁺ is not reactive towards the water and therefore has to be further oxidized to generate the active species [Ru^V = O]². Thereafter, water nucleophilic attack on the Ru^V oxo occurs, forming the hydroperoxo intermediate [Ru^III-OOH]⁺, and oxygen evolves from [Ru^IV-OO]²⁺. The carbonate ligand (bpc) enhances electron transfer steps without obviously influencing the PCET (Duan et al. Reference Duan, Wang, Li, Li and Sun2015).

Fig. 10. (a) Mechanistic scheme for the generation of dioxygen by treating [Ru^II(bpc)(bpy)(OH₂)]⁺ (where bpy = 2,2-bipyridine and bpc = 2,2-bipyridine-6-carboxylate) with Ce^IV. A water nucleophilic attack onto Ru^V = O is proposed for the rate determining step (RDS) (taken from Duan et al. Reference Duan, Wang, Li, Li and Sun2015 with permission).

This type of rational design of Ru-complexes with a general bipyridine–dicarboxylate ligand motif has produced many molecular catalysts having exceptionally high turnover frequencies (TOF). For example, the mononuclear ruthenium complex [Ru(bda)(isoq)²] (where bda = 2,2′-bipyridine-6,6-dicarboxylic acid and isoq = isoquinoline) has a high catalytic activity for water oxidation when using Ce^IV as the oxidant under acidic conditions (Duan et al. Reference Duan, Wang, Li, Li and Sun2015). However, higher rates of O₂ formation are achieved which are much faster than PSII (1000 s⁻¹), with turnover rates of 8000 s⁻¹ at pH 7·0 and 50 000 s⁻¹ at pH 10 and is combined with a high chemical stability (Matheu et al. Reference Matheu, Ertem, Benet-Buchholz, Coronado, Batista, Sala and Llobet2015). These very impressive turnover rates are obtained with Ru complexes having the general formula [Ru ⁿ (tda)(py)] ^m+ with tda being [[2,2 : 6′,2″-terpyridine]-6,6″-dicarboxylate]. Again the mechanism is thought to involve a nucleophilic attack by water onto the oxo of Ru^V as in Fig. 10. The high rates are attributed to: (i) easy access to the high oxidation states, provided by the tda⁻² ligand through the anionic nature of the carboxylates and the capability of stabilizing a coordination of seven to Ru in high oxidation states and (ii) the ability of the dangling carboxylate to be a proton acceptor in an intramolecular fashion, while the incoming substrate water molecule undergoes the nucleophilic attack.

4.3.2. Mechanism of O₂ production from organo-Mn complexes

In the context of the water/hydroxide nucleophyllic attack mechanism for PSII O–O bond formation, a number of efforts have been made to prepare Mn^V oxo complexes, which relate to OEC of PSII. The groups of both Crabtree and Brudvig (Cady et al. Reference Cady, Crabtree and Brudvig2008; Limburg et al. Reference Limburg, Brudvig and Crabtree1997; Reference Limburg, Vrettos, Liable-Sands, Rheingold, Crabtree and Brudvig1999; Reference Limburg, Vrettos, Chen, De Paula, Crabtree and Brudvig2001; Vrettos et al. Reference Vrettos, Limberg and Brudvig2001) and Pecararo et al. (Reference Pecararo, Baldwin and Gelasco1994, Reference Pecararo, Baldwin, Caudl, Hsieh and Law1998) have provided evidence for Mn^V oxo intermediates related to the OEC that were very short lived. One catalyst, [Mn₂ ^III/IV(-O)₂(terpy)₂(OH₂)₂](NO₃)₃ (where terpy = 2,2′:6′2″-terpyridine) has provided a model system for the OEC of PSII (Limburg, et al. Reference Limburg, Vrettos, Liable-Sands, Rheingold, Crabtree and Brudvig1999; Reference Limburg, Vrettos, Chen, De Paula, Crabtree and Brudvig2001). These studies used a range of added oxidants and often involved the radical coupling of two adjacent oxos.

However, both porphyrins and corroles and some related complexes have been shown to produce dioxygen by a nucleophilic mechanism by generating a Mn^V oxo as the electrophilic species (Gao et al. Reference Gao, Liu, Wang, Na, Åkermark and Sun2007, Reference Gao, Åkermark, Liu, Sun and Åkermark2009) (see Fig. 11). A hydroxyl nucleophilic attack of water on the Mn^V oxo yielded dioxygen (steps 2 to 3 in Fig. 11) as deduced from DFT calculation, spectroscopy and isotopic O¹⁸ exchange experiments.

Fig. 11. Mechanism for a base-induced hydroxyl nucleophilic attack on a Mn^V = O of nitrophenylcorrole (structure is shown and depicted as Mn^III in the scheme) using t-BuOOH as the oxidant and n-Bu₄NOH as the nucleophilic base (taken from Gao et al. Reference Gao, Åkermark, Liu, Sun and Åkermark2009 with permission).

In another study, it was shown that both mononuclear and dinuclear ‘Pacman’ Mn^IV corrole complexes could act as a catalyst for water splitting (Gao et al. Reference Gao, Liu, Wang, Na, Åkermark and Sun2007). In principle, the dinuclear Mn complexes could react via oxygen–oxygen radical coupling, but in these systems this route was found to be energetically less favourable than the nucleophilic attack mechanism.

4.4. Alternative mechanisms are less compelling

An alternative to the base-catalysed nucleophilic mechanism for O–O bond formation in PSII, is radical pair coupling, which was proposed by Babcock and colleagues (Hoganson & Babcock, Reference Hoganson and Babcock1997; Tommas & Babcock, Reference Tommas and Babcock1998) prior to the elucidation of the OEC structure. DFT calculations and X-ray diffraction data have proved to be remarkably good for studying the OEC as emphasized by Siegbahn who predicted, using the Ferreira et al. cubane model as the starting point, the existence and structure of the Mn₄Ca²⁺O₅ cluster three years before the 1·9 Å crystal structure of Umena et al. was published (Siegbahn, Reference Siegbahn2008). This powerful DFT approach has also been used extensively by Siegbahn (Reference Siegbahn2006, Reference Siegbahn2009, Reference Siegbahn2013) and Li and Siegbahn (Reference Li and Siegbahn2015) to explore the mechanism of water oxidation and particularly the nature of S₄ in the S-state cycle.

Initially, Siegbahn favoured a surface radical coupling mechanism involving a terminal Mn^IV-oxyl radical (Siegbahn & Crabtree, Reference Siegbahn and Crabtree1999) but according to his calculation, this mechanism has a higher energy barrier than an oxyl–oxo coupling mechanism. He therefore concluded that the most likely mechanism is the formation of a terminal oxyl radical within the cubane at S₄, which attacks a nearby bridging oxo, O5 to form the O–O bond (see Fig. 12). This proposed mechanism requires significant conformational changes due to the insertion of a substrate water molecule into the cubane during the S₂ to S₃ transition and reconstruction of the cubane during the S₄ to S₀ transition. However, a recent XFEL study (Young et al. Reference Young, Ibrahim, Chatterjee, Gul, Fuller, Koroidov, Brewster, Tran, Alonso-Mori, Kroll, Michels-Clark, Laksmono, Sierra, Stan, Hussein, Zhang, Douthit, Kubin, de Lichtenberg, Long Vo, Nilsson, Cheah, Shevela, Saracini, Bean, Seuffert, Sokaras, Weng, Pastor, Weninger, Fransson, Lassalle, Bräuer, Aller, Docker, Andi, Orville, Glownia, Nelson, Sikorski, Zhu, Hunter, Lane, Aquila, Koglin, Robinson, Liang, Boutet, Lyubimov, Uervirojnangkoorn, Moriarty, Liebschner, Afonine, Waterman, Evans, Wernet, Dobbek, Weis, Brunger, Zwart, Adams, Zouni, Messinger, Bergmann, Sauter, Kern, Yachandra and Yano2016) obtained room temperature diffraction data, which did not provide any evidence of structural changes in the S₃ state as required by the Siegbahn oxyl–oxo coupling mechanism. Thus, in my opinion, this alternative mechanism is unlikely although very recently a XFEL study did detect some structural changes close to O5, which was suggested to be involved in O–O bond formation (Suga et al. Reference Suga, Akita, Sugahara, Kubo, Nakajima, Nakane, Yamashita, Umena, Nakabayashi, Yamane, Nakano, Suzuki, Masuda, Inoue, Kimura, Nomura, Yonekura, Yu, Sakamoto, Motomura, Chen, Kato, Noguchi, Tono, Joti, Kameshima, Hatsui, Nango, Tanaka, Naitow, Matsuura, Yamashita, Yamamoto, Nureki, Yabashi, Ishikawa, Iwata and Shen2017) and also see Vinyard & Brudvig (Reference Vinyard and Brudvig2017) for a balanced review of this mechanism.

Fig. 12. A mechanism for O–O bond formation during the S₄ to S₀ transition postulated by Siegbahn (Reference Siegbahn2006; Reference Siegbahn2008; Reference Siegbahn2009; Reference Siegbahn2013) whereby formation of a terminal oxyl radical originating from a second substrate water incorporated into the cubane during the S₂ to S₃ transition leads to a radical attack on an adjacent oxo ligand (O5) incorporated within the Mn₃CaO₄-cubane from the first substrate water during the S₄ to S₀ transition.