Water oxidation in photosystem II

1More than three billion years ago, the cyanobacteria evolved a light-driven enzyme that was able to split water into molecular oxygen and hydrogen. Biology couples this to the reduction of carbon dioxide (CO2) to carbohydrates and in so doing stores the energy of the sun in energy-rich compounds with water acting as the electron source. We call this whole process photosynthesis and it represents the central metabolic pathway of the biosphere. While there exists a wide variety of photosynthetic organisms from cyanobacteria to algae and higher plants, the cellular machinery responsible for water splitting is unique and uses the same mechanism across all species.

2A byproduct of biological water splitting is dioxygen (O2) (Fig. 1). Its accumulation over the last billion years has led to the formation of the oxygen-rich atmosphere of our planet (Fischer et al.; Hamilton et al. ). It has also caused build-up of the ozone layer in the stratosphere that provides an efficient shield against high-energy UV radiation from the sun, thus protecting and sustaining life on earth. The increase in the concentration of O2 in the atmosphere started about two and a half billion years ago (Bekker et al. ) and led to the ‘great oxygenation event’ (GOE), a catastrophe for most early life forms on earth which had no defense mechanism to cope with the toxic oxygen, leading to mass extinction. On the other hand, the presence of O2 in the atmosphere fostered the development of higher, aerobic life forms, i.e., organisms that evolved to use oxygen for their metabolism. Thus, without the earlier invention of oxygenic photosynthesis, the development of complex life forms with cellular respiration—from simple vertebrates to human beings—would not have taken place. It is, therefore, not only of scientific but also of cultural value for mankind to understand how water oxidation and oxygen release functions in photosynthesis.

3We benefit from photosynthesis in many ways. It is our only source of food and supplies us with much of our energy in the form of fossilized photosynthetic material—oil, coal and natural gas. Furthermore, it provides us with valuable natural materials like wood, paper, cotton, peat and biomass in general. Plants also contain many important bioactive materials that are used as drugs, in cosmetics, as natural dyes etc.

4A basic understanding of the way how Nature stores energy and, in particular, uses sunlight to split water, could open future pathways to harvest and store the sun’s abundant energy, satisfying the energy demands of society in a sustainable fashion i.e., by an “artificial photosynthesis” (acatech et al.; Chabi et al.; Collings and Critchley; Cox and Lubitz; Cox et al.; Faunce et al.; Gratzel; McKone et al.; Messinger et al.; Nocera,; Wydrzynski and Hillier ). In the short review below the current knowledge of the principles of photosynthetic water oxidation in photosystem II of oxygenic organisms are described with special emphasis on the recent X-ray crystallographic data and the results obtained on the electronic structure of the water oxidizing cluster and the water binding and processing derived from advanced EPR techniques in combination with quantum chemical calculations. For a more extensive and detailed description of photosynthetic water oxidation in PS II the reader is referred to some recent reviews (Junge; Pantazis; Shen; Vinyard and Brudvig; Yano and Yachandra ).

5In oxygenic photosynthesis, sunlight is collected by the light-harvesting complexes (LHCs), specialized chlorophyll (and carotenoid) containing proteins, which show a large variation depending on the type of organism (Croce and van Amerongen ). The light energy is funneled to a reaction center (RC) where charge separation across the photosynthetic membrane takes place. Two RCs—photosystem I and II (PS I, PS II)—work in tandem to create the redox potential necessary to drive the coupled reactions (Blankenship ). This process provides electrons with sufficiently negative potential needed to reduce NADP+ to NADPH, which can be considered “bound biological hydrogen”. The electrons for this process originate from water oxidation that takes place in PS II. This process, and subsequent reactions, also generate a proton gradient across the photosynthetic membrane, which is used by the enzyme ATPase to produce ATP from ADP (Junge and Nelson ). In a different cell compartment, ATP and NADPH are then used to reduce CO2 to carbohydrates in the so-called dark reactions (Calvin–Benson cycle) (Calvin ).

6Much of our current understanding of the photosynthetic machinery has come from high-resolution crystal structures of its constituents, particular photosystem I and II. The first crystal structure of a bacterial photosynthetic reaction center (BRC) was reported by Deisenhofer, Michel and Huber in the early eighties (Deisenhofer et al.; Michel, )—who subsequently received the Nobel Prize in 1988 for this work. The structure of the simpler BRC provided a model for the functional core of PS I and PS II, in particular the processes of charge separation and progressive radical pair migration. It did not, however, provide any information on the water oxidizing complex (WOC) which is also called the oxygen-evolving complex (OEC), as this component is not found in the BRC that instead uses another electron source.

7A crystal structure of a functional oxygen-evolving PS II core complex took almost two decades to be obtained. This was first achieved by the groups of Horst T. Witt and Wolfram Saenger in Berlin who reported the first crystallization of PS II from the cyanobacterium Synechococcus (S.) elongatus (later renamed Thermosynechococcus (T.) elongatus) and obtained a structure with 3.8 Å resolution in 2001 (Zouni et al. ). Although this first crystallographic structure provided valuable insight into the arrangement of the protein subunits and the cofactors of PS II it did not allow the positions of the atoms in the OEC, a tetranuclear manganese cluster, to be determined. This is because of radiation damage induced by the intense X-ray beam used for X-ray diffraction (XRD) data collection in modern synchrotrons. X-ray absorption spectroscopy (XAS) experiments with varying beam intensity by Yano and coworkers have shown photoreduction of the oxidized Mn ions, disintegration of the complex and blurring of the electron density associated with the cluster (Yano et al. ). Before the advent of a PS II crystal structure with sufficient resolution attempts have been made to obtain models of the tetranuclear Mn cluster from spectroscopic experiments, in particular from XAS spectroscopy (Yachandra et al.; Yano et al. ). In 2004 the group of James Barber (Imperial College London) presented an improved structure of PS II from S. elongatus at 3.5 Å resolution (Ferreira et al. ). Based on their refined data, including anomalous diffraction to identify the Ca, and also on previous EXAFS data (Robblee et al. ) the authors proposed a heterometallic cubane-type structure of the OEC, which contained a Mn3Ca unit along with a more distant dangler Mn. Such a “dangler model” had been proposed earlier by David Britt’s group (UC Davis) based on constraints from EPR and ENDOR experiments and the distances obtained from EXAFS (Peloquin and Britt; Peloquin et al. ). Finally, in 2011 the groups of Jian-Ren Shen (Okayama, Japan) and Nobuo Kamija (Osaka, Japan), who had reported a low resolution structure of PS II in 2003 (Kamiya and Shen ) published a significantly improved structure from Thermosynechococcus (T.) vulcanus at 1.9 Å resolution, in which the atomic arrangement of the ions in the OEC was finally resolved (see below) (Umena et al. ). Subsequent theoretical studies showed that the actual oxidation state of the OEC in this structure was a mixture of overreduced states, not present in the catalytic cycle (Galstyan et al.; Luber et al. ). The Shen group also succeeded in obtaining a structure with reduced radiation damage using femtosecond (fs) pulsed X-ray crystallography with a free electron laser (XFEL) (Suga et al. ). More recently, time resolved XFEL measurements (serial femtosecond crystallography—SFX) near ambient temperature has emerged as an important tool for the investigation of the various states of the catalytic cycle of the OEC (see below) (Kern et al.,,,; Kupitz et al.; Suga et al., ).

8In Fig. 2 the dimeric PS II protein is shown (Suga et al. ), located in the thylakoid membrane. Each monomer is made up of 20 protein subunits (in cyanobacteria). A few important ones are indicated in Fig. 2A. In Fig. 2B the cofactor arrangement in the D1/D2 subunits is shown—comprising chlorophylls (PD1, PD2, ChlD1, ChlD2, ChlzD1, ChlzD2), carotenoids (CarD1, CarD2), pheophytins (PheoD1, PheoD2), plastoquinones (QA, QB), non-heme iron (Fe) and also the WOC or OEC (Fig. 2C). This cubane-like cluster is bound in the PS II protein and comprises four Mn ions and one Ca ion linked by five oxygen bridges (six in the last stage of the catalytic cycle). Remarkably, it also binds four water molecules, two at the Ca and two at the dangling Mn4. The cluster is connected to the protein surface via specific channels for the efficient uptake of water, and the release of protons and molecular oxygen (Bondar and Dau; Gabdulkhakov et al.; Ho and Styring; Linke and Ho; Murray and Barber; Retegan and Pantazis; Saito et al.; Umena et al.; Vassiliev et al. ). For each PS II monomer about 1300 water molecules have been detected in the crystal structure (Umena et al. ); they are mostly found at the outside membrane region but quite a few are located inside of the PS II protein. The water access channels to the Mn cluster are shown together with a few key amino acid ligands in Fig. 2D. The first channel is via the Asp 61 (shaded blue) the second via the Ca (shaded yellow) and a possible third one via the Cl− binding site (green). In all channels several water molecules have been found in the crystal structure (Ferreira et al.; Gabdulkhakov et al.; Ho and Styring; Murray and Barber; Saito et al.; Umena et al.; Vassiliev et al. ). Cl− depletion inhibits water oxidation; exchange of Cl− with Br− or I− slows the water oxidation kinetics (Damoder et al.; Ono et al. ). The effect of these halogen anions is probably related to charge effects in the protein and proton egress.

9Light excitation of the reaction center of PS II, a multi-pigment assembly of four chlorophyll-a and two pheophytin-a molecules, leads to primary charge separation on a picosecond timescale. This is a single electron transfer process resulting in a charge-separated radical pair (RP) state comprising the radical cation P680·+ and the radical anion Pheo·− (Cardona et al. ). During primary charge separation, the positive charge is initially localized on the accessory chlorophyll ChlD1 and it is stabilized by subsequent transfer on PD1 (see Fig. 2A) (Groot et al.; Holzwarth et al.; Kammel et al. ). The primary acceptor Pheo is the pheophytin pigment bound to the D1 protein (Holzwarth et al. ). The oxidation potential of P680·+ is estimated to + 1.2 to + 1.3 V, the highest known in biology (Rappaport et al. ). The primary RP P680·+Pheo·− is stabilized by subsequent electron/hole transfer steps. This is important as the radical cation can, in principle, oxidize neighboring chlorophylls, and the protein itself. To suppress both side and back reactions, nature has placed secondary donor/acceptors in close proximity to the primary RP. On the donor side of the protein (lumen), it is the redox-active tyrosine residue D1-Tyr161 (known as TyrZ or Yz) next to P680·+ that is able to quickly donate an electron to this species on a nanosecond (20–250 ns) time scale, thereby reducing the radical cation to its initial state P680. The tyrosine radical cation Yz·+ is stabilized by losing a proton to the neighboring D1-His190 in a reversible fashion, forming a neutral tyrosine radical (Yz·) (Berthomieu et al.; Chu et al.; Saito et al. ). Yz· in turn oxidizes the adjacent Mn cluster in 40–1600 µs dependent on the Si transition (Babcock et al.; Brettel et al.; Dekker et al.; Haumann et al.; Karge et al.; Klauss et al.; Noguchi et al.; Rappaport et al.; Razeghifard and Pace ). On the acceptor side of the protein (cytoplasm), the acceptor Pheo·− passes its electron to the bound plastoquinone molecules (to QA and subsequently to QB). Reducing equivalents derived from this process are temporarily stored as reduced plastoquinol (QBH2), a mobile electron carrier, generated after two light absorption/charge separation and protonation events.

10Other light-induced processes that occur in PS II are equally dangerous, for which nature has not found a perfect solution to circumvent them. This is the creation of triplet states, e.g., of the chlorophylls in the core antenna via intersystem crossing or in the RC via (triplet) radical pair recombination. While carotenoid molecules are present in PS II to suppress (quench) triplet chlorophyll formation, these states can still react with triplet oxygen (3O2) formed in the water oxidation process yielding singlet oxygen (1O2) (Macpherson et al. ). This very aggressive species can then react with the pigments and/or the protein, resulting in damage of the D1 protein subunit of the PSII supercomplex. The production of such reactive oxygen species (ROS) (Vass ) limits the lifetime of PS II to only 30 min under normal light conditions. Owing to its remarkable efficiency (the turnover time of the OEC is ≈ 2 ms, that of the whole PS II ≈ 10 ms)(Vinyard and Brudvig ), it performs more than 105 reaction cycles before it must be replaced. Fortunately, all organisms performing oxygenic photosynthesis have developed an efficient repair mechanism for the PS II supercomplex by the discrete replacement of the D1 protein (Nixon et al. ).

11PS II thus acts as a water:plastoquinone oxidoreductase (Wydrzynski and Satoh ). The products are molecular (triplet) oxygen (3O2), protons and reduced plastohydroquinone. The light-induced process in PS II is very efficient, with a quantum yield of over 90% and an energy efficiency of about 20%. This value is, however, strongly attenuated when all the subsequent processes are considered, such that the total biomass generated often contains only less than 1% of the original light energy input (Blankenship et al.; Michel ).

12It was discovered by Pierre Joliot almost half a century ago that PS II releases oxygen after four consecutive light flashes (Joliot et al. ) (Fig. 3A). This shows that four light-induced trans-membrane charge separation events are necessary before one O2 molecule is formed and released. This result is in agreement with the fact that the oxidation of two water molecules to produce a single O2 molecule is a four-electron process,

13whereas the charge separation across the membrane is a one-electron process. The oxidizing equivalents for the water oxidation reaction are stored transiently by the tetranuclear manganese cluster, an insight that came from Kok et al.. His so-called Kok cycle (S-state cycle) of water oxidation, shown in Fig. 3B, comprises five distinct states, S0 to S4, which differ by the number of oxidizing equivalents transiently stored in the cofactor, as indicated by their subscript. The cycle is driven by four light absorption events (hν), by which electrons are withdrawn from the metal ions and protons are released to avoid the accumulation of positive charges and thus prevent a “coulombic explosion” of the cluster. O2 is released in a concerted reaction in the last reaction step with an equilibrium constant k > 1.0 × 107 (Nilsson et al. ). Kinetic measurements showed that the reaction times lie in the micro- to millisecond range (40 µs to 1.6 ms) (Klauss et al. ), see Fig. 3B; the complete cycle turns over in about 2 ms. The kinetics of the proton release, which follow a 1:0:1:2 pattern (or 1:0:1:1:1) have also been determined, see (Dau and Haumann; Klauss et al.,; Schlodder and Witt; Suzuki et al. ). The (alternating) electron and proton release led to an extended catalytic cycle for the water oxidation reaction in PS II with nine states that differ in their net electron and proton count (Klauss et al.,; Perez-Navarro et al. ).

14In this sequence, YZ promotes both electron and proton transfer in the catalytic cycle displaying a dual function (Bovi et al.; Klauss et al.; Perez-Navarro et al. ). Between Si and Si+1 states (for i = 0–3) the short-lived SiYZ˙ intermediates exist. Studying these intermediates is critical for understanding the role of YZ˙ in water oxidation and especially in proton removal from the Mn4OxCa cluster (Boussac et al.; Havelius et al.; Nugent et al.; Peloquin et al.; Petrouleas et al.; Retegan et al.; Styring et al. ).

15There are three interlinked reasons why the Mn4OxCa cluster functions in this way: (i) The energetic cost is significantly lowered by first storing oxidizing equivalents and then performing the four-electron chemistry; (ii) this sequence avoids the formation of reactive oxygen species (ROS) as intermediates resulting from partial oxidation of substrate water, which are highly reactive and can destroy the PS II protein complex; and (iii) proton release during charge accumulation leads to charge neutrality so that about the same oxidizing potential (oxidant P680·+) can be used for all Mn oxidation events. This is illustrated by the scheme in Fig. 4 (Messinger and Renger ). Here the Mn ions (and/or their ligands) are oxidized four times by the neighboring tyrosine YZ· triggered by four charge separation events in PS II before the cluster removes four electrons from two bound water molecules in a concerted reaction leading to O2 release and regeneration of the original starting state of the Mn4OxCa cluster. The Mn cluster thus acts as an interface and storage device between the very fast light reaction (ps time scale) and the slow catalytic reaction (ms time scale) of the 4-electron water oxidation chemistry, bridging a kinetic gap of nine orders of magnitude.

16A detailed understanding of the catalytic process of water oxidation in PS II requires knowledge of the electronic structure, i.e., the distribution of the electrons in the cluster, in all consecutive reaction steps. The oxidation and spin states of the Mn ions, representing the total number and configuration of electrons in the Mn valence orbitals, give a basic description thereof. These together with the magnetic interactions between the spin-bearing Mn ions, depending to a large part on the metal ligands, provide a comprehensive picture of the respective electronic state, which governs the chemical and catalytic properties of each S state. Thus, the spin states provide information about how the structure of the cofactor evolves during the S-state cycle, for a recent review see (Krewald et al. ).

17An experimental method to investigate the electronic structure of transition metal complexes is electron paramagnetic resonance (EPR) spectroscopy (Goldfarb and Stoll; Schweiger and Jeschke; Weil and Bolton ). It exploits a fundamental property of matter, which is that unpaired electrons have an intrinsic angular momentum (spin), which can be excited by microwave radiation in a magnetic field. The unpaired electron spin also interacts with other electron and nuclear spins as well as with local electric field gradients, making it a sensitive reporter of its chemical environment. It is thus analogous to other magnetic spectroscopies such as nuclear magnetic resonance (NMR). Since the Mn ions are open-shell species, i.e., exhibit orbitals with single electron occupancy, whereas most of the electrons of the protein and solvent surrounding are paired, the EPR signals of the Mn4OxCa cluster can be detected selectively. It has also been shown that all S states in the Kok cycle can be trapped (except for the elusive S4 state) and that all exhibit paramagnetism (Haddy ).

18The first EPR signal observed by Charles Dismukes in 1981 (Dismukes and Siderer ) came from the trapped S2 state. Centered at g ≈ 2 it included over 20 resolved lines and was thus called “multiline signal” (see Fig. 5). The structure is due to the coupling of the electron spin with the four 55Mn nuclei (I = 5/2) via the electron nuclear hyperfine interaction. The EPR signal represents an effective low-spin ground state of Seff = ½ (Bencini and Gatteschi ). The spin state depends on the oxidation state of the Mn ions, their geometry and in particular on the (bridging) ligands which connect the metal ions. These mediate antiferromagnetic or ferromagnetic exchange interactions between the Mn ions leading to either a low-spin state, minimizing the number of unpaired electrons, or a high-spin state, maximizing the number of unpaired electrons. The studies of the groups of Ono and Kusunoki on oriented PS II membranes and theoretical investigations (Hasegawa et al., ) and subsequent advanced EPR studies from the Britt (UC Davis) (Britt et al.,; Peloquin et al. ) and Lubitz (MPI Mülheim) laboratories (Cox et al.; Kulik et al.,,; Lohmiller et al.,; Su et al. ) further constrained all four 55Mn hyperfine tensors in the S2 state and allowed the spin coupling in the tetranuclear manganese cluster to be interrogated. These data together with results collected on the S0 state, which also resolves a multiline signal (Ahrling et al.; Kulik et al.,; Lohmiller et al.; Messinger et al. ), and density functional theory calculations allowed the local oxidation states and set of coupling pathways to be determined (Ames et al.; Krewald et al.,,; Pantazis et al., ). The oxidation state assignment for the S2 state, which comes from this analysis, is shown in Fig. 5 (top), i.e., Mn4(III, IV, IV, IV). According to the low oxidation scheme S0 includes a MnII (Kolling et al.; Pace et al.; Terrett et al. ), but this was excluded by the 55Mn ENDOR experiments and analysis of the S0 state (Kulik et al.,; Lohmiller et al. ). The reader interested in details on how the electronic configuration of a polynuclear Mn cluster, especially the Mn4CaO5, can be probed by EPR spectroscopy and double resonance techniques in relation with calculations of magnetic properties by quantum mechanical methods is referred to the following review (Lohmiller et al. ). Additional support for the high oxidation state scheme has come from the detection of 14N (I = 1), 2H (I = 1) and 13C (I = ½) hyperfine couplings of the S2 state using ESEEM and ENDOR techniques performed by the group of R. D. Britt (Britt et al.; Marchiori et al.; Perez-Navarro et al.; Oyala et al.; Stich et al.; Stull et al. ); see also (Lohmiller et al.; Perez-Navarro et al. ). X-ray spectroscopies also agree with this assignment (Chatterjee et al.; Zaharieva et al., ). These studies rule out the alternative lower oxidation state models for the manganese cluster (Kolling et al.; Pace et al. ) that are, however, still discussed in the literature (Chen et al.; Terrett et al. ).

19In the S2 state, in addition to the low-spin Seff = ½ form showing the characteristic multiline signal around g ≈ 2, the cluster can also be found in a high-spin Seff = 5/2 state under certain conditions. This is evident from an EPR signal around g = 4.1 (see Fig. 5), which has been observed earlier by several research groups (Boussac et al.; Casey and Sauer; Haddy et al.; Zimmermann and Rutherford ). Pantazis et al. could show computationally that the two electronic structures are a direct consequence of two different spatial conformations of the manganese cluster, namely a “closed cubane” (Seff = 5/2) and an “open cubane” (Seff = ½) form, which have almost the same energy (Bovi et al.; Isobe et al.; Pantazis et al. ). In these two structures one of the oxygen bridges, O5, is occupying different positions. In turn this also led to a change of the MnIII position (Jahn–Teller ion) and thus of the open coordination site of the cluster resulting in a change of the precise electronic properties of the cofactor. EXAFS data also support this model (Chatterjee et al. ). This shows that the effective spin Seff is a crucial parameter for describing a particular state of the cluster and assigning it a spatial structure. It is postulated that O5 has not a fixed position but toggles between two structures in the dynamic S2 state, which is important for the water binding and the catalytic mechanism (see below).

20Modifications of the Mn4OxCa cluster, such as exchange of Ca2+ with Sr2+ (Boussac and Rutherford; Boussac et al.; Chu et al.; Cox et al.; Koua et al.; Yachandra and Yano ), its complete removal (Boussac et al.; Lohmiller et al.; Styring et al.; Vrettos et al. ) and the binding of small molecules (Beck et al.; Perez-Navarro et al.; Oyala et al.; Oyala et al.; Retegan and Pantazis; Su et al. ), provided further insight into the conformation and number of ligands of individual Mn ions (coordination geometry) and substrate access in the S2 state. Upon Ca depletion water oxidation functionality is lost, but the S2 state (S2′) is still formed (Boussac et al., ). Interestingly, the EPR and 55Mn ENDOR data of S2′ show that the calcium has no substantial effect on the magnetic parameters of the S2 state; this ion is thus not crucial for maintaining the electronic structure of the tetranuclear Mn cluster (Lohmiller et al. ). Instead, it might serve as stage for the delivery of water molecules to the reaction site (Nakamura et al.; Service et al.; Ugur et al. ), affect the function of YZ, (Miqyass et al.; Retegan et al. ) and introduce some structural flexibility allowing the cofactor to toggle between the different motifs of the open and closed cubane structures (see Fig. 5). The interactions of small molecules which mimic the substrate water (methanol, ammonia) also support this potential role for Ca2+. These molecules associate with the Ca2+ and Asp61 water channels that lead to the Mn4OxCa cofactor (Marchiori et al.; Navarro et al.; Oyala et al.; Retegan and Pantazis; Schraut and Kaupp ) (see Fig. 2D). In addition, both of these molecules reduce turnover efficiency; these results implicate that at least one (or possibly both) of these channels are involved in substrate delivery.

21The EPR measurements have been extended to the S3 state. These results are of particular importance since this is the last metastable state prior to O–O bond formation and O2 release. The EPR signal of the S3 state has first been reported to originate from a ground state with an integer spin of Seff = 3 by the groups of Petrouleas (Sanakis et al. ) and Boussac (Boussac et al. ). Recent pulse high field (W-band) EPR experiments by Cox et al. (see Fig. 6A) corroborated this assignment and directly proved that the effective spin state is indeed Seff = 3 by measuring the Rabi oscillations via a spin nutation experiment (Cox et al. ). In addition 55Mn ELDOR-detected-NMR (EDNMR) experiments at W-band could successfully be performed showing that in the S3 state all Mn ions are in the MnIV state Mn4 (IV,IV,IV,IV) and octahedrally coordinated. Thus, no ligand oxidation takes place as previously proposed (Kawashima et al.; Messinger et al. ). The results also show that a sixth ligand is binding to the open coordination site of the MnIII present in the S2 state when it is oxidized. This ligand is most probably the second substrate water molecule which also loses a proton in the S2 → S3 transition. Data from XAS, XES and FTIR further support these finding (Noguchi; Sakamoto et al.; Zaharieva et al., ). Mn oxidation in the S2 to S3 transition has been proposed by Dekker (Dekker et al., ) and Ono (Ono et al. ) and substantiated by Dau and coworkers (Haumann et al., ). It has been included in mechanistic models (Pecoraro et al.; Vrettos et al. ) and has also theoretically been supported by Siegbahn (Siegbahn ). Very recently, SFX data have shown additional electron density acquired during the S2 to S3 transition, consistent with the binding of a light atom (e.g., an oxo or hydroxyl ligand) in the manganese cluster (Kern et al.; Suga et al. ).

22The magnetic resonance experiments described above together with theoretical calculations allowed a reliable characterization of the S0, S2 and S3 states with respect to oxidation and spin states of individual ions and their spin coupling in the tetranuclear Mn cluster summarized by Krewald et al.. Together with information about binding of the two substrate water molecules described below spatial models of the S states could be obtained that form the basis for developing a catalytic mechanism for the OEC (Fig. 8).

23Knowledge of the binding and dynamics of the substrate water molecules is crucial for formulating the reaction mechanism of photosynthetic water oxidation. Candidates for substrate waters, besides the four H2O/OH− molecules (Fig. 2C) directly attached to the cluster (Ames et al.; Suga et al.; Umena et al. ), are oxygen bridges within the cluster as well as possible water molecules binding at a later point in the catalytic cycle. Apart from X-ray diffraction spectroscopic techniques are used to investigate water binding, in particular EPR (Cox et al.; Fiege et al.; Kawamori et al.; McConnell et al.; Nagashima and Mino; Rapatskiy et al. ) and vibrational (infrared) spectroscopy (Debus,; Kim et al.; Kim and Debus; Nakamura et al.; Noguchi; Sakamoto et al.; Suzuki et al. ). These are used in conjunction with time-resolved mass spectrometry (MIMS) which detects the uptake of H218O labeled water into the product O2; for a comprehensive review on the identification of possible water substrates by MIMS see (Cox and Messinger ). MIMS experiments showed that the two substrate water molecules exchange at different rates in all of the S states (Hillier and Wydrzynski,, ). The more slowly exchanging water (Ws) has an exchange rate of the order of seconds (Messinger et al. ) while the faster exchanging water (Wf) has an exchange rate of milliseconds (Hillier et al. ) (see Fig. 3B). The observation of two rates implies that the two substrates bind at two chemically distinct sites. They also demonstrate that the O–O bond is not formed until the (S4) state is reached. Interestingly, it was recently observed that during the lifetime of the S3YZ˙ state, that lies between S3 and the elusive S4 state while the YZ˙ radical exists, the water exchange is slowing down, which may suggest that the two oxygen species are “arrested” in a bonding formation (Nilsson et al. ).

24For the detection of substrate water molecules, EPR spectroscopy makes use of interactions with nuclear spins, e.g. 2H (in 2H2O) (Fiege et al.; Kawamori et al.; Nagashima and Mino ) or preferentially 17O (nuclear spin I = 5/2) in PS II samples with isotope-labeled H217O as the solvent (Lohmiller et al.; McConnell et al.; Rapatskiy et al. ). Methodological and instrumental developments in our laboratory (Cox et al., ) allowed us to characterize the 17O signals in the S2 and S0 states using electron–electron double resonance (ELDOR)-detected NMR (EDNMR) at high frequency (94 GHz) (Lohmiller et al.; Rapatskiy et al. ), which yields significantly better sensitivity compared with 17O ENDOR (McConnell et al.; Rapatskiy et al. ). Through spectral simulations, we could identify the µ-oxo bridge O5 to be an exchangeable ligand in the S2 state (Fig. 7A, B) (Lohmiller et al.; Perez-Navarro et al.; Rapatskiy et al. ). O5 represents the most probable candidate for the two-fold deprotonated substrate Ws for the following reasons:It is the only exchangeable 17O nucleus bound both to Mn and Ca, a requirement imposed by MIMS experiments (Cox and Messinger; Hendry and Wydrzynski; Messinger ).Shortening of an Mn–Mn bond observed by EXAFS has been explained by a deprotonation of an O-bridge and has been assigned to O5 (Robblee et al.; Messinger ).Similar timescales for the water exchange were found both for Ws in the MIMS, see ref. (Hillier and Wydrzynski ), and for O5 in the EPR experiments, i.e., complete exchange in < 10 s (Rapatskiy et al. ).Orientation dependent EDNMR studies showed that the measured hyperfine coupling can only be assigned to O5 or O4) (Rapatskiy et al. ).Both Ca/Sr exchange experiments and ammonia inhibition studies demonstrate O5 and only O5 to be an exchangeable oxygen bridge (Cox et al.; Perez-Navarro et al. ).

25It is the only exchangeable 17O nucleus bound both to Mn and Ca, a requirement imposed by MIMS experiments (Cox and Messinger; Hendry and Wydrzynski; Messinger ).

26Shortening of an Mn–Mn bond observed by EXAFS has been explained by a deprotonation of an O-bridge and has been assigned to O5 (Robblee et al.; Messinger ).

27Similar timescales for the water exchange were found both for Ws in the MIMS, see ref. (Hillier and Wydrzynski ), and for O5 in the EPR experiments, i.e., complete exchange in < 10 s (Rapatskiy et al. ).

28Orientation dependent EDNMR studies showed that the measured hyperfine coupling can only be assigned to O5 or O4) (Rapatskiy et al. ).

29Both Ca/Sr exchange experiments and ammonia inhibition studies demonstrate O5 and only O5 to be an exchangeable oxygen bridge (Cox et al.; Perez-Navarro et al. ).

30This is in agreement with earlier low frequency FTIR studies of the OEC that showed that a µ2-oxo or µ3-oxo bridge is exchangeable (Chu et al. ). It is remarkable that in the OEC a µ-oxo bridge is readily exchangeable. For structurally similar model compounds an exchange has only been observed on much longer timescales (Rapatskiy et al.; Tagore et al., ). These results further demonstrate the dynamic nature of this O5 ligand in the OEC (see above, Fig. 5).

31Recent 17O EDNMR investigations of H217O binding to the S0 state of the OEC show that an OH− is introduced in the initial step of the cycle (the 1st substrate water that has lost a proton) after O2 release in the S4 to S0 transition (Lohmiller et al. ). In the S0 and S1 state the O5 is thought to exist in an open cubane-type structure (in S0 with S = ½, and in S1 with S = 0 ground state) as shown in Fig. 8.

32As described above, recent spectroscopic results, including high field EPR indicate that a new water molecule (O6) is inserted into the manganese cluster in the S3 state as an OH− at the open coordination site of Mn1 (Cox et al. ). Its close proximity to O5 suggests it is the second substrate Wf (Fig. 6B). How it is exactly inserted remains an open question. Recent SFX data (Kern et al.; Suga et al. ) indicate that the Glu189, which bridges the Ca2+ ion and Mn1 in the S1 state detaches from the Ca2+ ion upon formation of the S3 state. This removes steric hindrance to direct water binding via the Ca2+ channel, i.e., it is originally one of the Ca2+ waters. Another possibility is that this OH− (O6) is introduced in an indirect fashion by toggling between the open and closed cubane structures and is thus derived from W2, one of the terminal ligands of Mn4 (compare structure S3′ in Fig. 8). Within this sequence the water that binds during the S2 to S3 transition is not one of the substrates of the current cycle, but a substrate of the next cycle (Cox and Messinger ). It also implies that O5 (the existing μ-oxo bridge) and O6 (the new water-derived oxygen ligand) may switch identities. There is significant theoretical support for this sequence, which explains how redox tuning of the cofactor can precede water binding (Retegan et al. ). This hypothesis also better agrees with MIMS data which shows both substrates are bound to the cofactor (most likely to Mn ions) in all S-states and that the rate of exchange does not dramatically change upon moving from S2 to S3 (Cox and Messinger; Hendry and Wydrzynski, ).

33As a consequence of the above results, it is assumed in the following discussion of the catalytic cycle of the OEC that the O5 bridge derives from one of the substrate waters (Ws). A possible cycle is shown in Fig. 8 (Cox et al.; Krewald et al. ). The first water Ws binds as an OH− (O5) in the S0 state, being deprotonated during the subsequent oxidation step to S1. In the S3 state the second substrate water enters the reaction as OH− (O6) to Mn4 (or Mn1) as suggested by the results obtained from EPR and XFEL experiments on the S3 state (Cox et al.; Kern et al.; Suga et al. ). A further oxidation and deprotonation step leads to S4. Experimental evidence shows that formation of S4 is triggered by proton release (Haumann et al.; Klauss et al. ). Current theoretical models suggest that a proton is transferred from W1 (H2O) to Asp61, as in the preceding S2 to S3 transition (Siegbahn ). Mutants of PSII lacking Asp61 show a pH dependency of the lag phase during the S3–S0 transition which has been attributed to a deprotonation of the Mn4CaO5 (Bao and Burnap; Dilbeck et al. ) in support of this model.

34In the current literature, there are several proposed mechanisms that suggest the early onset of O–O bond formation in the S3 state (Corry and O’Malley ). This is mainly based on the SFX crystal structure of Suga et al. which modeled a short distance between the O5 and O6 bridges in the S3 state (Suga et al. ). We note that the more recent higher resolution SFX structure of the S3 state (Kern et al. ), does not exhibit this short O5–O6 distance, and is instead more consistent with the high field EPR structure described above. Furthermore, earlier biophysical measurements categorically rule out O–O bond formation in the S3 state. The MIMS data described above shows that both substrates still rapidly exchange with bulk water in the S3 state (Messinger et al. ). It is only upon oxidation and proton release to form the S3Yz· state directly preceding S4 that substrate exchange is ‘arrested’ indicating the onset of O–O bond formation, i.e., formation of a peroxo type intermediate (Nilsson et al. ).

35How the O–O bond is precisely formed in the S4 state is still unknown, owing to a lack of experimental data. There are two popular chemical mechanisms for this transition that differ with regard to what component of the cofactor is oxidized to form S4: (i) a ligand (one of the substrate waters); or (ii) a metal (one of the Mn ions). In the case of a ligand oxidation event, forming a MnIV-oxyl moiety, O–O bond formation is proposed to proceed by a radical coupling mechanism described below. In the case of a metal oxidation event, forming a MnV-oxo type moiety, O–O bond formation is instead proposed to proceed by nucleophilic attack of the electrophilic oxygen (O5) bound to the MnV by a nearby water, e.g., W3 bound to the Ca2+. Unfortunately the nature of the last oxidation (ligand vs. metal centered) does not resolve this question. XAS data which monitors an oxidation state change of the Mn ions during the S3 to S0 transition appears to rule out a MnV intermediate (Haumann et al. ), favoring instead a ligand (oxygen) oxidation. We note that while the nucleophilic attack mechanism initially envisaged the involvement of a MnV intermediate, it was later suggested to have substantial MnIV oxyl character in the S4 state based on theoretical studies (Vinyard and Brudvig ).

36Only a nucleophilic attack mechanism has been observed for first row molecular transition metal water oxidation catalysts, for which there is mechanistic data, (Codola et al. ) with radical coupling to be thus far only described for second row transition metals, e.g., ruthenium (Cox and Lubitz; Romain et al. ). Nucleophilic attack mechanisms have thus been proposed for the OEC by a number of groups (Ferreira et al.; Pecoraro et al. ); see (Barber; Vinyard and Brudvig; Vinyard et al. ) for recent reviews. In this sequence, a nearby water (e.g., W3) attacks a manganese bound oxygen, most likely O5. Upon re-reduction of the four Mn ion and loss of the product O2, the newly inserted O6 bridge presumably fills the site vacated by O5 leading to rapid recovery of the S0 state.

37In contrast the oxo-oxyl radical coupling mechanism involving two Mn bound oxygens, is most compatible with the assignment of the two substrates sites described in the previous section (O5 and O6). Such a mechanism was first proposed ten years ago by Per Siegbahn (Siegbahn ). Importantly high field S3 state EPR data has demonstrated that there is the requisite spin alignment of the two putative substrate oxygens and the Mn ions to which they are bound (Mn4-O5-O6-Mn1 = βαβα) on which this mechanism is based.

38To illustrate the current situation, in Fig. 9 three different structures are shown preceding the O–O bond formation in S4. They are derived from the S3 open and closed cubane structures, S3A and S3B, and from S3′ in which a second water has not (yet) been bound (Fig. 8). In panel (A) and (B) the open and closed cubane structures are shown with all Mn(IV) and O5 and O6 in close contact that could react to form dioxygen via an oxyl-oxo coupling mechanism. In panel C) the dangling Mn4 is 5-coordinate; in the shown geometry the formation of a Mn(V) = O species is preferred in which the electrophilic oxygen could be attacked by a nucleophilic water (e.g., bound to Ca) or a neighboring oxo group as pointed out by Pantazis. In this context it should be mentioned that in the recent SFX structure (Kern et al. ) only the open cubane structure has been observed for S2 and S3 which led the authors to propose O5 to react with either O6 (here called Ox) bound to Mn1 or with W3 (at Ca) or W2 (at Mn4).

39There are also other alternative mechanisms. A very recent example, proposed by the Yamaguchi group, involves an O–O˙ radical intermediate, formed following electron transfer to YZ (Shoji et al. ). Similarly, there are proposed mechanisms that do not fall into either of the two categories, such as a recent proposal from the Sun group. In this novel mechanism, the dangler Mn acts as the site of catalysis, as in a nucleophilic attack like mechanism, forming a MnVII-dioxo intermediate following charge rearrangement of the Mn cluster in the S4 state (Zhang and Sun ). As above for the nucleophilic attack mechanism, it is unclear if this mechanism is compatible with XAS data (Haumann et al. ). For a more detailed comparison of alternative mechanisms the reader is directed to a recent in-depth review (Pantazis ).

40Owing to the increasing experimental support for the oxo-oxyl coupling mechanism first proposed by Siegbahn, a brief description of its key steps is given below (Fig. 10). The first stage of the mechanism involves oxidation of O6 to form an oxyl radical, in concert with deprotonation. In the Siegbahn mechanism, the proton associated with O6 in S3 migrates to W1, which is already deprotonated during S3 formation. The generated oxyl radical carries an unpaired β spin electron, because O6 is antiferromagnetically coupled to Mn1. The oxyl radical then attacks the O5 bridge which carries a small excess of α spin. This results in formation of an σ O–O bond between O5 and O6—with the new σ bond consisting of the α spin electron of O5 and the β spin electron of O6 (Fig. 10 [S4 oxyl]). The α spin electron comes from the cleavage of the existing Mn–O bond between O5 and Mn4. The β spin electron of the cleaved Mn–O bond is transferred back to Mn4, reducing its oxidation state to III+. Note that the β electron of the cleaved Mn–O bond has the same spin as the three existing d-electrons of Mn4, and thus this transfer can occur without energy penalty due to spin conversion.

41Subsequently, a peroxo product (a bound O2 with a single σ bond) is formed bridging Mn3 and Mn1 (Fig. 10 [S4 peroxo]). Mn4 is now five coordinate with the product peroxide sitting along its Jahn–Teller (JT) axis (Li and Siegbahn; Siegbahn ). Cleavage of the two remaining Mn–O bonds leads to formation of the final O2 product. Importantly the cleavage of each Mn–O bond results in an α spin electron being transferred to each of the two Mn ions that were attached to the peroxide and a β spin electron being transferred to O5 and O6. As before, this is an energetically favorable situation as the electron gained by each metal is of the same spin (α) as their existing d-electrons. The two new β spin electrons occupy two π* orbitals giving rise to the triplet O2 product. The displacement of O2, via binding of the first substrate water of the next S cycle (coupled to its deprotonation), leads to formation of the observed (metastable) S0 state (Fig. 10, S0). The O2 may detach in a stepwise fashion, during its concerted replacement with OH−, passing through a transient superoxo intermediate (Li and Siegbahn ).

42New experimental data is needed to probe these key stages in O–O bond formation and release. SFX crystallography (Kern et al.,; Kupitz et al.; Suga et al.,; Young et al. ) is poised to address structural changes during the S3 to S0 transition, and may suggest an alternative proton transfer sequence. Specifically, as described above, the Siegbahn mechanism starts with internal proton transfer from O6 to W1. We note, however, that recent SFX measurements indicate that Glu189 detaches from the Ca2+ ion during S-state progression. Glu189 is proximal to Mn1 and could act as an internal base, accepting a proton from O6, similar to old ideas put forward by Gerry Babcock (Hoganson and Babcock ). Within this revised scheme W1 does not need to undergo deprotonation—with proton transfer to Asp61 switched off due to a conformational change (gating) (Kern et al. ).

43In biology, there is only one catalytic site that is able to split water—the OEC—and its structure and function is identical (Su et al. ) in cyanobacteria, algae and all higher plants; however, differences in the second coordination sphere between higher plants and cyanobacteria have been reported (Retegan and Pantazis ). Many other enzymes have found different ways to perform their task—but not the unique, highly optimized OEC in PS II. The work compiled above shows that this unique metal cluster requires the presence and concerted action of all four manganese ions in the Mn4OxCa complex. The oxidation states of the Mn ions are always III or IV—there exists no MnII in the whole cycle. The possible presence of a MnV in the last step (S4) has, however, not yet been finally clarified. The importance of the total electron spin state of the cluster has been highlighted—and is understandable considering the necessary restriction to form and release triplet dioxygen 3O2. The Ca2+ is probably important for efficient water delivery and the Cl− for maintaining the correct charge balance of the OEC. Water delivery is optimized through highly efficient water channels leading to the Mn cluster. In the favored model the two substrate waters bind to neighboring redox-active Mn ions, and deprotonation of these waters is coupled to the oxidation events of the cluster. Close proximity and proper alignment of the two active oxygens are assured to efficiently form the O–O bond and finally release the 3O2 molecule with high efficiency.

44In spite of the great progress in the understanding of the OEC in PS II there are still several challenges to be tackled. One pertinent open question is the structure of the elusive S4 state. A possible stabilization and characterization of this state would finally complete our knowledge of the S states of the Kok cycle. A missing key experiment is further the detection of the 6th oxygen in the manganese cluster by EPR (using 17O labeling) to support the recent crystallographic XFEL (SFX) data on the S3 state (Kern et al.; Suga et al. ). Furthermore the determination of the water binding kinetics should be followed for all state transitions by rapid freeze quench (RFQ) techniques combined with advanced EPR techniques. First experiments along these lines have successfully been performed in our laboratory using time resolved RFQ 17O EDNMR measurements (Rapatskiy et al., unpublished data). The exchange of certain amino acids in the vicinity of the OEC could shed light on the fine-tuning of its electronic structure in the S state cycle and would be important for explaining its high turnover frequency. Furthermore, a great challenge for chemists working in the field of water oxidation is the understanding and modeling of the sophisticated mechanism by which PS II repairs photo-damage of its protein, of the pigments and of the manganese cluster.

45The properties of the OEC in PS II described above represent valuable design features for bioinspired light-driven molecular catalysts for water oxidation. The following points are considered essential for such catalysts: The material of the catalyst must be abundant, easily accessible, inexpensive (i.e., no precious metals), non-toxic, sufficiently stable under working conditions and should be scalable.The required four oxidizing equivalents must be stored in the catalyst to couple the fast 1-electron photochemical reaction with the slow 4-electron chemical water oxidation process.The binding of the two substrate water molecules should preferentially take place at two well-defined neighboring redox-active metal centers, accompanied by successive deprotonation and activation of the water molecules.The redox steps of the catalytic metal centers should be of similar magnitude and in the range of 1 eV, and should finally lead to a concerted water oxidation, O2 formation and O2 release, to avoid reactive oxygen intermediates. This also requires a sequential release of the 4 protons.A functional matrix mimicking the protein (in PS II) is needed for efficient transport of water to the reaction zone, release of dioxygen and for the correct proton management.In the photochemical act, a light-induced species must be generated with sufficient oxidative power to oxidize water (+ 1.23 eV).Effective coupling of the charge separation unit with the catalytic center (metal cluster) is necessary to diminish the overpotential (comparable to the function of the tyrosine YZ in PS II).The catalytic unit should be stable; in case of deterioration/destruction a mechanism of self-repair or healing should be in place to avoid deactivation of the catalytic process.

46The material of the catalyst must be abundant, easily accessible, inexpensive (i.e., no precious metals), non-toxic, sufficiently stable under working conditions and should be scalable.

47The required four oxidizing equivalents must be stored in the catalyst to couple the fast 1-electron photochemical reaction with the slow 4-electron chemical water oxidation process.

48The binding of the two substrate water molecules should preferentially take place at two well-defined neighboring redox-active metal centers, accompanied by successive deprotonation and activation of the water molecules.

49The redox steps of the catalytic metal centers should be of similar magnitude and in the range of 1 eV, and should finally lead to a concerted water oxidation, O2 formation and O2 release, to avoid reactive oxygen intermediates. This also requires a sequential release of the 4 protons.

50A functional matrix mimicking the protein (in PS II) is needed for efficient transport of water to the reaction zone, release of dioxygen and for the correct proton management.

51In the photochemical act, a light-induced species must be generated with sufficient oxidative power to oxidize water (+ 1.23 eV).

52Effective coupling of the charge separation unit with the catalytic center (metal cluster) is necessary to diminish the overpotential (comparable to the function of the tyrosine YZ in PS II).

53The catalytic unit should be stable; in case of deterioration/destruction a mechanism of self-repair or healing should be in place to avoid deactivation of the catalytic process.

54At present, it is still a great challenge to create a catalyst that fulfills the above criteria and could thus compete with the Mn cluster in PS II. The future will show if at least part of the above points can be fulfilled. Attempts to synthesize water splitting chemical catalysts and devices are too numerous in the literature to be discussed here—but show the great importance of this research field for a future society using (only) renewable, sustainable energy based on sunlight driven processes (Andreiadis et al.; Blakemore et al.; Kanan and Nocera; Kurz; Najafpour et al.; Tran et al.; Zhang et al.; Matheu et al. ).