Photosynthetic water oxidation: binding and activation of substrate waters for O – O bond formation

Photosynthetic water oxidation occurs at the oxygen-evolving complex (OEC) of Photosystem II (PSII). The OEC, which contains a Mn 4 CaO 5 inorganic cluster ligated by oxides, waters and amino-acid residues, cycles through ﬁ ve redox intermediates known as S i states ( i ¼ 0 – 4). The electronic and structural properties of the transient S 4 intermediate that forms the O – O bond are not well understood. In order to gain insight into how water is activated for O – O bond formation in the S 4 intermediate, we have performed a detailed analysis of S-state dependent substrate water binding kinetics taking into consideration data from Mn coordination complexes. This analysis supports a model in which the substrate waters are both bound as terminal ligands and react via a water-nucleophile attack mechanism.


Introduction
The climate, biology and geology of Earth were transformed by the evolution of oxygenic photosynthesis approximately three billion years ago. 1 All photosynthetic organisms, from the earliest cyanobacteria to modern vascular plants, use solar energy to oxidize water to molecular oxygen (O 2 ), protons and electrons at the oxygen-evolving complex (OEC) of Photosystem II (PSII). 2,3 PSII is a membrane-bound pigment-protein complex that generates a solar lightinduced charge separation in order to oxidize water and reduce plastoquinone. The charge separation is initiated by the primary chlorophyll-a electron donor, P 680 , forming a powerful oxidant, P 680 + , that advances the oxidation state of the Mn 4 CaO 5 cluster in the OEC (Fig. 1) via a redox-active tyrosine, Y Z , that mediates electron transfer from the OEC to P 680 + . The overall four-electron process leads to the formation of O 2 that is released as a byproduct and has accumulated in the biosphere and the release of protons that contribute to the transmembrane proton motive force. 4 The availability of O 2 as an electron sink has since powered oxygenic respiration to give rise to Earth's current biological diversity.
The catalytic cycle of the OEC involves ve metastable redox intermediates known as S i states (i ¼ 0-4), as rst described by Kok and coworkers. 7,8 The darkstable intermediate, S 1 , contains the Mn oxidation state pattern (III,IV,IV,III) 9 (order of oxidation states is based on the numbering of the Mn ions in Fig. 1) and is converted to S 2 upon the loss of one electron. S 2 is present as two spin isomers with oxidation states of (III,IV,IV,IV) (S ¼ 1/2) or (IV,IV,IV,III) (S ¼ 5/2). 10 The next oxidation event results in the loss of one proton and one electron from the OEC to form S 3 , in which all four Mn ions are in the +4 oxidation state (S ¼ 3). 11,12 Oxidation of Y Z following the next charge separation causes release of a proton from S 3 to form a modied Y Z c S 3 0 state, 13 which then forms the S 4 state. The structures of S 3 0 and S 4 are poorly understood. S 4 spontaneously produces O 2 , binds substrate water(s), and releases a proton to form S 0 , thus resetting the catalytic cycle. The general structure of the OEC has been revealed through a series of X-ray diffraction (XRD) studies. 6,[14][15][16][17] However, the atomic structure of any single Kok cycle intermediate has not been determined through these experiments. Early XRD structures of the OEC suffered from X-ray induced reduction of the highvalent Mn centers. 18 Nevertheless, the availability of metal-metal distances from extended X-ray absorption ne structure (EXAFS) spectroscopy 19,20 allowed computational studies (QM/MM and density functional theory (DFT)) to rene XRD models to structures corrected for radiation damage. 5,21 At present, the PSII research community is in general (but not universal 3 ) agreement on the structures of the S 0 , 5 S 1 , 5,10 and S 2 10 intermediates and the Mn oxidation states of all S states. 9 The problem of X-ray induced Mn reduction can be completely avoided by using femtosecond pulses from X-ray free electron laser (XFEL) radiation sources. 22,23 XFEL studies using PSII nanocrystals have produced structures of darkadapted PSII at 6.5-4.9Å resolution. [24][25][26][27] Structures of S 2 , 25,26 S 3 , 26,27 and S 0 26 have also been reported at low resolution, which limits their interpretation. However, QM/MM methods combined with electron density map analysis have shown that the S 1 / S 2 transition involves signicant displacement (and oxidation) of Mn4 (also referred to as the "dangler" Mn). 28 A 1.95Å resolution XFEL structure of dark-adapted PSII was recently reported by Shen and coworkers. 29 However, the resulting XRD model of the OEC does not agree well with EXAFS data of S 1 , 19 which is likely the result of inaccuracy in positioning of the bridging oxo ligands and signicant accumulation of S 0 . 30 The S 4 state decays faster than it is formed and, therefore, cannot be observed as a kinetic intermediate. Consequently, no direct experimental evidence for the nature of O-O bond formation in S 4 has been produced. However, insights from inorganic and computational chemistry have produced two competing mechanisms: water-nucleophile attack and oxo-oxyl radical coupling.
In the oxo-oxyl radical coupling mechanism, 31-33 a Mn(IV)-oxyl radical reacts with a Mn-bridging oxo to generate O 2 ( Fig. 2A). Extensive calculations by Siegbahn suggest that one of the substrate waters rst binds during the S 2 / S 3 transition and then is oxidized to an oxyl radical in S 4 to carry out the reaction. 32 A water-nucleophile attack mechanism for the OEC 34-37 is most consistent with synthetic water oxidation catalysts. 38 As shown in Fig. 2B, such a mechanism would involve a terminal water-nucleophile (preferably bound to calcium, see discussion below) and a formally Mn(V)-oxo electrophile.
These mechanisms can be distinguished by the positions of the substrate waters involved. Kinetics of substrate water exchange throughout the Kok cycle have been determined by membrane inlet mass spectrometry (MIMS) measurements following rapid mixing of PSII with H 2 18 O. This technique, which was pioneered in the Wydrzynski laboratory, allows 34 O 2 and 36 O 2 release by PSII in specic S states (advanced by single-turnover ashes) to be monitored as a function of incubation time with H 2 18 O. 39,40 The resulting rates of substrate water exchange are summarized in Table 1. In S 0 , only one slowly exchanging water (W s ) is resolved with a rate constant of approximately 10 s À1 . The exchange rate of W s is then dramatically slowed to 0.02 s À1 in S 1 . In S 2 and S 3 , two phases of substrate water exchange kinetics are resolved; the slower and faster exchanging waters are referred to as W s and W f , respectively. W s exchanges at a nearly identical rate in S 2 and S 3 , but the rate is much faster than in S 1 . W f exchanges at decreasing rates in S 1 , S 2 , and S 3 .
Determining the molecular mechanism of photosynthetic water oxidation is crucial for applying the principles of Nature's design to synthetic systems for solar  fuel production. In the following analysis, we detail models for substrate binding and O-O bond formation in the OEC in light of experimental evidence and comparisons to well-studied inorganic systems.

Where are the substrate binding sites?
Kinetics and mechanism of terminal water vs. m-oxo ligand exchange from model chemistry Given our current understanding of the structure of the OEC in the S 0 , S 1 , and S 2 states, we can predict that the substrate waters that give rise to O 2 are bound either as terminal ligands to Mn4 and Ca 2+ , or as m-oxo/hydroxo bridges. In order to understand the substrate water exchange kinetics of the OEC, we rst look to inorganic model systems.
The exchange rate of a terminal water ligand on Mn depends greatly on the oxidation state, protonation, ancillary ligands, and geometry. As shown in Table  2, both the water exchange rate and pK a of a terminal aqua ligand decrease dramatically with increasing oxidation state from Mn 2+ to Mn 3+ to Mn 4+ . The hexaaqua Mn(IV) complex has an especially slow water ligand exchange rate (<10 À4 s À1 ), as expected for a high-spin d 3 ion with octahedral geometry for which the ligand eld stabilization energy disfavors ligand dissociation leading to very slow ligand exchange. However, exchange of the terminal water ligands of the [Mn 4 IV,IV,IV,IV (m-O) 5 (terpy) 4 (H 2 O) 2 ] 6+ (terpy ¼ 2,2 0 :6 0 ,2 00 -terpyridine) complex occurs faster than the mixing time for the mass spectrometry measurement ($10 s, k ex > 10 À1 s À1 ), despite the high-spin d 3 conguration of the Mn(IV) ions. This can be explained by the reduced symmetry of the Mn(IV) ions in the tetrameric oxomanganese-terpy complex. Deviations from octahedral geometry of the Mn IV centers resulting from the <90 bite angle of the tridentate terpy ligand and the asymmetry associated with oxo and pyridyl ligands result in loss of degeneracy of the t 2g and e g orbitals, and can lower the barrier for exchange of the terminal water ligands. Indeed, computational studies have found low reaction barriers for the dissociative exchange of terminal waters in m-O bridged Mn(IV) dimers. 42 Generally, ligand exchange for high-valent rst-row transition ions involves a dissociative mechanism, especially for Mn(IV) which is normally six-coordinate and has a small ionic radius that makes an associative mechanism unfavorable. For terminal water ligands, dissociation involves removing a polar but neutral ligand from the cationic metal center, which has a modest barrier. 52 However, dissociative exchange of a m-oxo ligand involves removing an anionic ligand from the cationic metal center, which has a much higher barrier. Therefore, the exchange of m-oxo ligands with bulk water is thermodynamically more challenging than the exchange of terminal water ligands.
The mechanism of m-oxo exchange has been investigated for di-m-oxo di-Mn(III,IV) complexes. 53 These studies have shown that m-oxo exchange requires protonation of the oxo before dissociation and exchange; therefore, m-oxo pK a s are good predictors of their rates of exchange. The m-oxo pK a decreases dramatically with increasing Mn oxidation state in Mn model complexes ( Table 3 53 Because the pK a of the water ligand bound to a high-valent Mn center is lower than solvent water (Table 2), a m-oxo can be more readily protonated by ligated water, which opens up the m-oxo bridge and lowers the energetic barrier for dissociation. 53 In this mechanism, the terminal water and m-oxo are in dynamic equilibrium with bulk water. For Mn dimers without terminal water ligands, the exchange is much slower, such as in [Mn 2 III,IV (m-O) 2 (bpy) 4 ] 3+ , because intramolecular proton transfer between terminal-bound water and m-oxo bridges is eliminated. 53 There have also been measurements of terminal water and m-oxo ligand exchange of the di-Mn center in Mn-catalase. 54 The active site of Mn-catalase contains a redox-active di-Mn center for which ligand exchange has been studied in the di-m-oxo di-Mn(III,IV) state. It was found that the terminal water ligand exchanges within the mixing time of one minute, but the m-oxo ligands require over one hour for exchange. These results parallel those obtained for Mn model complexes in solution, although the exchange rate of the m-oxo ligands is much lower in the protein active site.  In summary, electrostatics play a dominant role in the dissociative ligand exchange reactions of high-valent oxo-manganese complexes. There is a large barrier for m-oxo exchange due to the high negative charge of the oxo ligand. Thus, protonation of the m-oxo ligand is required to promote ligand exchange. Because of the large pK a shi of the m-oxo ligands between di-m-oxo di-Mn(III,IV) and di-moxo di-Mn(IV,IV) complexes, m-oxo ligands exchange at a modest rate for di-m-oxo di-Mn(III,IV) complexes but do not exchange for di-m-oxo di-Mn(IV,IV) centers. On the other hand, terminal water ligands exhibit fast rates of ligand exchange for both Mn(III) and Mn(IV) centers. Moreover, the rate of terminal water ligand exchange is similar for Mn(III) and Mn(IV) because both have similar electrostatic potential (ESP) charges. 52 Ammonia does not compete with substrate water The water analog ammonia inhibits the OEC by binding to two sites in PSII. The rst is in the outer coordination sphere of the OEC and upon binding of ammonia, alters the hydrogen-bonding networks in such a way that the S 2 state S ¼ 5/2 isomer is favored. 56 The binding of ammonia to this site is competitive with chloride 57,58 and can be accessed by larger amines, 56,58 but its specic location has not been resolved.
Ammonia also binds directly to Mn in the OEC in the S 2 state resulting in the formation of an altered multiline g ¼ 2 EPR signal. 59 However, the substrate water exchange kinetics of S 2 do not change when ammonia is bound. 60 Therefore, the binding site of ammonia cannot be the binding site of a substrate water.
Lubitz and coworkers have proposed that ammonia binds to Mn4 trans to O5 based on 17 O-ELDOR-detected NMR (EDNMR) measurements. 60 In these experiments, dark-adapted PSII (poised in S 1 ) is incubated with H 2 17 O for tens of minutes and perturbations in the 17 O-EDNMR signal are detected when ammonia (as NH 4 Cl at pH 7.6) is present following advancement to the S 2 state by 200 K illumination and subsequent annealing at 260 K. In accompanying 1 H-ENDOR measurements, no change in the proton environment around the OEC was observed. 60 Other lines of evidence have suggested that ammonia binds as a bridging ligand between two Mn ions. This motif was rst proposed by Britt and coworkers based on ESEEM measurements of the ammonia-bound S 2 state in higher plant PSII. 61 The quadrupole coupling between the S ¼ 1/2 S 2 state of the OEC and bound 14    (amino) ligands display h # 0.3. Therefore, the measured quadrupole coupling of ammonia to S 2 is highly deviant from axial symmetry and could represent a deprotonated bridging ligand such as an imido (NH 2À ) or nitrido (N 3À ). However, Britt and coworkers have very recently revealed the origin of this anisotropy. In wild type cyanobacterial PSII, they measured the 14 NH 3 coupling as 1.62 MHz with h ¼ 0.40, but in a D1-D61A mutant, the interaction was nearly completely axial (1.54 MHz, h ¼ 0.04). 62 This work provides convincing evidence that bound ammonia has a strong hydrogen bond to Asp61 and is, therefore, a terminal ligand to Mn4. In a recent study using QM/MM methods and EXAFS simulations, we have proposed that instead of replacing a terminal water or m-oxo group, ammonia binds as a sixth ligand to Mn4 resulting in a complete octahedral coordination sphere. 63 In this model (Fig. 3), ammonia binds to the S ¼ 5/2 S 2 state but then induces a redox switch stabilizing the S ¼ 1/2 S 2 state by protonating O5 and deprotonating W2. This additional ligand would lower the reduction potential of ammonia-bound S 2 as has been observed by both thermoluminescence 64 and ash O 2 measurements. 65 While both substrate waters are already bound in S 2 , 41 an additional water may bind during the S 2 /S 3 transition as computationally predicted. 32 We have proposed that ammonia, a "harder" Lewis base than water, binds in S 2 to the dangler Mn4 in an analogous site as water binds in S 3 to the dangler Mn4. 63 The addition of a ligand to Mn4 in either S 2 or S 3 trans to O5 would cause the other terminal water ligands (W1 and W2) to shi their positions towards the cuboidal core of the OEC. In this model of ammonia binding, both water-nucleophile attack and oxo-oxyl radical coupling O-O bond formation mechanisms are feasible.
A model for substrate water exchange in the OEC W2 exchanges rapidly. In agreement with previous analyses, 66,67 we assign the kinetic features of W f primarily to the terminal water W2. This water is already bound in S 2 as resolved in measurements at reduced temperatures 41 and in PSII cores from the red alga Cyanidioschyzon merolae. 68 In S 0 and S 1 , W2 is present as a Mn 3+ -aqua species and would be rapidly exchanged (>120 s À1 ), consistent with the predicted exchange kinetics of [Mn(H 2 O) 6 ] 3+ (Fig. 4, Table 2).
For the water-nucleophile attack mechanism, the substrate waters are W2 and W3. W3, which is bound to Ca 2+ , is also expected to exchange rapidly (specically in S 0 and S 1 , see discussion below). Therefore, the W f kinetic (measured by monitoring 34 O 2 ) may reect contributions from both W2 and W3.
The exchange rate of W2 is rst resolved aer the S 1 / S 2 transition with a value in the S 2 state of k f $120 s À1 . As discussed above, S 2 is present as two spin isomers. Given the changes seen in XFEL experiments at the Mn4 position measured at physiological temperature, we have asserted that the S ¼ 1/2 isomer is dominant. 28 Therefore, W f would be bound as a terminal Mn 4+ -aqua species in the S 2 state and exchanges at a rate comparable to those measured in model complexes ( Table 2).
As modeled in Fig. 4, the formation of S 3 involves the release of a proton from W2 and a 3-fold slowing of its exchange rate (Table 1). This change is qualitatively consistent with the differences expected between Mn 4+ -aqua and Mn 4+ -hydroxo species. W s in S 0 and S 1 involves O5/W2 exchange. Identifying the origin of the substrate water exchange rates for the W s kinetic phase is more complex. The water exchanges fastest in S 0 , dramatically slows down in S 1 , and then dramatically speeds up again in S 2 and S 3 . These observations are difficult to explain based on exchange at a single site. Herein, we propose a new model for W s exchange that involves contributions from both terminal and bridging ligands (W2 and O5, respectively) whose interconversion is possible only in the S 0 and S 1 states owing to the decreasing pK a of the m-oxo species, O5, with increasing Mn oxidation states.
It has been determined that only one of the m-oxo bridges denoted O5 slowly exchanges with H 2 17 O in the dark-stable S 1 state over tens of seconds. 54,69 We assume that O5 also exchanges with bulk water in S 0 (oxidation states III,IV,III,III). QM/MM models of the S 0 state 5 predict that O5 is protonated, as expected if its pK a is compared to Mn(III,III) dimers (Table 3). Consequently, O5 exchange would be expected to occur relatively rapidly in S 0 (10 s À1 ) because it is already protonated at physiological pH. During exchange, protonated O5 may dissociate from Mn3 becoming a terminal water ligand to Mn4. Thus, in the S 0 state, W2 (purple, Fig. 4) and O5 (blue, Fig. 4) are expected to be in rapid equilibrium with each other. In S 1 (oxidation states III,IV,IV,III), O5 is present in a chemical environment analogous to Mn(III,IV) dimer complexes and the pK a of O5 in S 1 is predicted to be much lower than in S 0 (Table 3). Based on QM/MM models, O5 is not protonated in the S 1 state 5,21 and, thus, would have a higher energetic barrier for exchange. This change is proposed to be the reason for the 500-fold decrease in the exchange rate of W s observed in S 1 as compared to S 0 . W f is not resolved in S 0 and S 1 and reects the rapid exchange of W2 (and W3, discussed below). Because m-oxo exchange involves subsequent dissociation, binding, and association steps, 53 isotopically labeled substrate will appear in both the O5 and W2 positions. Therefore, W2 is labeled with two rates. The rst is rapid (>120 s À1 ) and reects direct exchange with bulk water. The second kinetic is slow (0.02-10 s À1 ) and reects the interconversion of labeled O5 with W2. Crucially, this mechanism implies that the rate of W s exchange is controlled by the rate of O5 exchange in S 0 and S 1 , but that O5 is not necessarily a substrate water.
For the dominant S 2 state S ¼ 1/2 spin isomer (oxidation states IV,IV,IV,III), the O5 chemical environment is analogous to Mn(IV,IV) dimer complexes in which the m-oxo pK a is very low (Table 3). Given the large energetic barrier for m-oxo exchange, the measured rate of W s in S 2 (2 s À1 ) is too fast to be assigned to O5. W s in S 2 and S 3 involves a terminal water bound to Ca 2+ . In S 0 and S 1 , W3, which is a substrate water in the proposed water-nucleophile attack mechanism ( Fig. 2B and 4), exchanges at a rate >120 s À1 and its kinetic features are merged with those of W2. However, the S 1 / S 2 transition does not involve the release of a proton to the bulk, but does involve an oxidation. The result is the build-up of a positive charge and a signicant contraction of the OEC as recently observed in photothermal beam deection experiments. 13,70 This accumulated charge (and contraction) is not released until the S 0 state is reformed aer O 2 release. 70 As a result, we hypothesize that the Ca 2+ -W3 electrostatic interaction, as well as the surrounding hydrogen-bonding network, is stronger in S 2 and S 3 compared to S 0 and S 1 . This effect of the increased charge would have to be dramatic to explain the 10 6 -fold decrease in W3 exchange compared to the Ca 2+ -aqua ion in bulk water ( Table 2).
When Ca 2+ is substituted with Sr 2+ , only W s is signicantly affected: k s is threeto four-fold faster in S 1 , S 2 , and S 3 when Sr 2+ is present. 71 Both O5 and W3 are adjacent to Ca 2+ /Sr 2+ and are, therefore, reasonable candidates for W s . Because Ca 2+ and Sr 2+ have very similar Lewis acidities (similar pK a s of the aqua ion, Table  2), W3 is expected to be fully protonated throughout the Kok cycle providing an adequate nucleophile in the water-nucleophile attack mechanism. 36 However, Sr 2+ is considerably larger than Ca 2+ (ionic radii are 1.12Å and 0.99Å, respectively 72 ) resulting in a weaker electrostatic interaction between Sr 2+ and its ligands. This difference is reected in the faster exchange rate of the water ligands bound to [Sr(H 2 O) 6 ] 2+ compared to [Ca(H 2 O) 6 ] 2+ (Table 2), and is also consistent with the faster exchange of W s when Sr 2+ is substituted for Ca 2+ in the OEC.
In our model for OEC turnover and substrate water exchange shown in Fig. 4, the observed W s kinetics reect O5 exchanging with W2 in S 0 and S 1 , and W3 exchange in S 2 and S 3 . The W f kinetic reects both W2 and W3 exchange in S 0 and S 1 , and only W2 in S 2 and S 3 . This new substrate exchange model provides a complete description of all of the kinetic phases of substrate exchange throughout the S-state cycle and is consistent with data from model high-valent oxomanganese complexes. It, thus, provides a framework from which to design additional experimental measurements to test and validate this hypothesis.

Mechanism of O-O bond formation
The structure of the S 4 intermediate will determine if the mechanism of photosynthetic water oxidation involves water-nucleophile attack or oxo-oxyl radical coupling. Unfortunately, few experimental clues about S 4 are available. Given that the Mn oxidation states in S 3 are (IV,IV,IV,IV), the S 4 state is predicted to contain either a Mn IV -oxyl radical or Mn V -oxo species. A Mn IV -oxyl radical species has been suggested through computational studies by Siegbahn and supports an oxo-oxyl radical coupling mechanism. On the other hand, a high-spin Mn V -oxo is predicted to serve as the electrophile in a waternucleophile attack mechanism. Borovik and coworkers have recently reported experimental support for the latter option through studies of [Mn V H 3 buea(O)] (where buea ¼ tris[(N 0 -tert-butylureaylato)-N-ethylene]aminato). 73,74 The buea ligand imposes C 3 symmetry around Mn and stabilizes the Mn-oxo group through intramolecular hydrogen bonding. [Mn V H 3 buea(O)] is a high-spin S ¼ 1 system as conrmed by EPR spectroscopy and supported by DFT calculations. 73 Therefore, the d xz and d yz orbitals in [Mn V H 3 buea(O)] are degenerate and the Mnoxo bond is likely to be weaker (and more reactive) than corresponding Mn V -oxo species with tetragonal symmetry (S ¼ 0). By quantifying the hyperne contributions from 17 O-labeled [Mn V H 3 buea(O)] using EPR spectroscopy, Borovik and coworkers determined that 0.45 spins reside on the oxo group. Therefore, the Mn V -oxo group is strongly covalent, but does not contain an oxyl radical. For [Mn III H 3 buea(O)] 2À , only 0.30 spins reside on the oxo group, suggesting that the Mn-O bond becomes more strongly covalent (and the unit more electrophilic) with increasing Mn oxidation state. 74 To date, no experimental evidence of a Mn IV -oxyl radical species has been found, calling into question the accuracy of the oxo-oxyl radical coupling mechanism. However, if a reactive high-spin Mn V ]O species is present in the S 4 state, analogous to that found in [Mn V H 3 buea(O)], a water-nucleophile attack mechanism would lead to O-O bond formation.

Conclusions
Determining the molecular mechanism of O-O bond formation in photosynthetic water oxidation remains one of the great challenges in bioinorganic chemistry. In the absence of direct experimental evidence of the S 4 intermediate, we can predict its structure and, therefore, the mechanism of O-O bond formation by comparing the OEC to model Mn complexes and by determining where the substrate waters bind throughout the catalytic cycle.
In our proposed model of OEC turnover shown in Fig. 4, the substrate waters are W2 (terminal water on Mn4) and W3 (terminal water on Ca 2+ ). The substrate water exchange kinetics in S 0 and S 1 are complicated by the dynamic equilibrium between W2 and the slowly exchanging m-oxo/hydroxo bridge, O5. This assignment of substrates dictates that O-O bond formation occurs via a water-nucleophilic attack of W3 (as water) on W2 (as a high-spin Mn V ]O species).