Reversible Structural Isomerization of Nature’s Water Oxidation Catalyst Prior to O–O Bond Formation

Abstract

Photosynthetic water oxidation is catalyzed by a manganese–calcium oxide cluster, which experiences five “S-states” during a light-driven reaction cycle. The unique “distorted chair”-like geometry of the Mn₄CaO₅₍₆₎ cluster shows structural flexibility that has been frequently proposed to involve “open” and “closed”-cubane forms from the S₁ to S₃ states. The isomers are interconvertible in the S₁ and S₂ states, while in the S₃ state, the open-cubane structure is observed to dominate inThermosynechococcus elongatus (cyanobacteria) samples. In this work, using density functional theory calculations, we go beyond the S₃⁺Y_z state to the S₃ⁿY_z^• → S₄⁺Y_z step, and report for the first time that the reversible isomerism, which is suppressed in the S₃⁺Y_z state, is fully recovered in the ensuing S₃ⁿY_z^• state due to the proton release from a manganese-bound water ligand. The altered coordination strength of the manganese–ligand facilitates formation of the closed-cubane form, in a dynamic equilibrium with the open-cubane form. This tautomerism immediately preceding dioxygen formation may constitute the rate limiting step for O₂ formation, and exert a significant influence on the water oxidation mechanism in photosystem II.

Introduction

Photosystem II (PSII) is a metalloenzyme that catalyzes water splitting to molecular oxygen in cyanobacteria, algae, and plants. It evolved about 3 billion years ago at the level of ancient cyanobacteria (Figure 1a). The embedded “oxygen-evolving complex (OEC)”, composed of a Mn₄CaO₅ cluster surrounded by water and amino acid ligands (Figure 1b,c), acts as a highly efficient water oxidation catalyst. Due to charge separations in the reaction center of PSII, the OEC is initially stepwise oxidized during the cyclic catalysis, so that it attains four (meta)stable intermediates (S₀, S₁, S₂, and S₃) and one transient S₄ state, the latter of which initiates O₂ formation.¹⁻¹⁰ Accounting also for proton release and charge of the Mn₄CaO₅₍₆₎ complex, the classical five-step “S-state cycle”¹¹ can be refined to instead include nine intermediate states that are separated by kinetically distinguishable proton and electron transfer steps (Figure 1d).^3,12−22

Figure 1.

(a) View of PSII dimer and the OEC location from Thermosynechococcus elongatus (PDB ID: 6W1O)³⁸ (b) Mn₄CaO₅ cluster and its local surroundings in its dark-stable S₁ state. (c) Sketch map of atom labeling and connectivity of the first coordination sphere ligands in the Mn₄CaO₅ cluster. (d) Extended S-state cycle including nine intermediates with sequence of proton and electron transfer and kinetics between transitions;^{3,13−15,17,21,22,70} the red phase is the main focus of this study. (e) Structural flexibility of the OEC cluster in the S₁, S₂, and S₃ states, marked with the reversibility between open (A) and closed (B) cubane structures (for references, see the main text).

Structural polymorphism of the OEC has been proposed and experimentally observed, mainly by electron paramagnetic resonance (EPR) spectroscopy, for some decades.^{1,4,18,23−30} More recently, the first detailed theoretically models were proposed for interpreting these findings.^31,32 However, the proposed alternative structures have thus far eluded verification by structural methods such as protein crystallography.³³⁻⁴⁰ The structural flexibility in the S₂ state is typically attributed to the mobile μ-oxo bridge (O5) between Mn1 and Mn4,^31,41 producing “open” (A) and “closed” cubane (B) forms of the cluster (Figure 1e, see supplementary references in the Supporting Information). As recently discovered by Pantazis and co-workers, orientational Jahn–Teller isomerism in the resting S₁ state⁴¹ generates the precursors for the two interconvertible A and B structures of the S₂ state,³¹ which give rise to the low-spin (S = 1/2) and high-spin (S = 5/2) EPR signals in plant PSII at g = 2 and g ≈ 4.1, respectively, and the latter g ≈ 4.1 (and similar signals around this value) can only be produced by mutations or chemical treatments in cyanobacteria.⁴² These authors also proposed that the closed-cubane form is the entry to the S₃ state,^43,44 in agreement with molecular dynamics studies by Guidoni and co-workers.^32,45 This closed-cubane interpretation for the S₂ high-spin (S = 5/2) state is widely accepted in the field and consistent with the calculations in this report and will hence be employed in this study. However, we note that two competing interpretations exist. First, based on broken-symmetry density functional theory (BS-DFT) calculations with focus on spectroscopic parameter analysis, Corry and O’Malley proposed an isomer in the S₂ state by W1 deprotonation to μ-O4 to rationalize the high-spin (S = 5/2) form⁴⁶ and, on this basis, further identified a high-spin (S = 7/2) deprotonated intermediate with μ-hydroxo O4 during the S₂ → S₃ transition, without invoking a closed-cubane structure.⁴⁷ Second, another model for the high-spin (S = 5/2) S₂ state assumes the early binding of a substrate water to Mn1 as OH^– originating from W3, as suggested by Siegbahn⁴⁸ and later by Pushkar et al.⁴⁹ For a detailed discussion of such models, see Text S6 in the Supporting Information.

The open-cubane S₃ structure contains an extra oxygen ligand to Mn1 due to binding of an additional water molecule. This was proposed first by Siegbahn on the basis of DFT calculations⁵⁰⁻⁵³ that are found on the results from extended X-ray absorption fine structure (EXAFS) experiments,⁵⁴⁻⁵⁶ showing that the S₂–S₃ transition involves the conversion of a five-coordinate Mn(III) to a six-coordinate Mn(IV). Cox et al. confirmed by advanced EPR that all Mn ions in the S₃ state are hexa-coordinate and that the “water-added” open-cubane S₃ structure, S₃^A,W (“W” denotes the extra water binding), is consistent with their experimental data.⁵⁷ Isobe et al. constructed multiple S₃ models^58,59 that vary with regard to total spin and Mn-valence and proposed that the closed-to-open cubane transformation is possible in a stepwise process involving an oxyl–oxo precursor.⁶⁰ By contrast, Capone et al.⁶¹ and Shoji et al.⁶² showed different feasible pathways for a direct closed-to-open cubane conversion. Regardless of the mechanistic details, a consensus has been reached that the OEC cluster in the S₃ state (more precisely the S₃⁺Y_z state, see below) allows for unidirectional conversion from the water-added closed (S₃^B,W) to the open-cubane (S₃^A,W) form, but not vice versa (Figure 1e). S₃^B,W (S = 3) is proposed to be the precursor form of the final S₃^A,W (S = 3) under the pivot/carousel mechanism of water binding during the S₂ → S₃ transition.^43,44,63 Importantly, the dominance of the open-cubane Mn core topology is consistent with the S₃ state structures resolved by serial crystallography using X-ray free electron lasers (XFELs).³⁵⁻³⁸

Nevertheless, alternative S₃ state models that assume early O–O bonding exist.^1,23,64 For example, Corry and O’Malley proposed a chemical equilibrium between “oxo-hydroxo” and “peroxo” for O5–Ox in the S₃ state, based on a comparison of experimental and BS-DFT calculated geometries and magnetic resonance properties.⁶⁵⁻⁶⁷ In higher-plant PSII, a recent combined EPR and DFT study by Zahariou et al. provided evidence, in PSII isolated from spinach for S₃ being a mixed state of S₃^A,W (S = 3) and S₃^B,unbound (S = 6) (“unbound” denotes the unsaturated coordination of Mn4; “S₃^B,unbound” is used throughout to refer to the “S₃^B” in its original publication, and similarly S₄^B,unbound for S₄^B).^68,69 Here, the dominant state (∼80%) has been identified as the S₃^B,unbound state, that is, a closed-cubane S₃ state with penta-coordinate Mn4(IV) without additional bound water. In this view, it should be emphasized that the structural isomerism in the S₃ state introduced here (and discussed later) should apply to that of cyanobacterial PSII, and the less populated S₃ form in higher plants.

Consistent with the abovementioned findings for the S₃⁺Y_z state, it is commonly assumed that the O–O bond formation in the S₄⁺ state also occurs in the open-cubane (S₄^A,W) conformation.^{4,35,37,51,53,57,71−79} However, there have been also several proposals based on a closed-cubane structure (S₄^B,W or S₄^B,unbound),^{4,68,69,73,75,80−87} which is in sharp contrast in terms of geometric configuration. This motivates us to investigate if structural heterogeneity exists just before the S₄⁺ state is formed from the S₃ⁿ state via electron abstraction by Y_z^•. In this paper, we mainly focus on the possibility of structural isomerization in the S₃ⁿY_z^• states (Scheme 1), employing DFT calculations. The correlation of our results with experimental observations and the implications for the mechanism of O–O bond formation are discussed.

Scheme 1. Structural Isomerization between S₃^A,WY_z^• (Left) and S₃^B,WY_z^• (Right) in the S₃ⁿY_z^• (W1 = OH^–) State Explored in the Present Study.

Amino acid ligands are omitted for clarity.

Results and Discussion

Unidirectional Structural Isomerization in the S₃⁺Y_z State

Unlike the structural interconversion simply caused by O5 shuttling between Mn1 and Mn4 in the S₂ state (the most likely mechanism accounting for the EPR isomers, see Text S6 in the Supporting Information), isomerization in the S₃⁺Y_z state results from Mn3 ligand exchange between O5 and Ox. We revisited this process by our quantum chemical model (Figure S1 in the Supporting Information) and determined a direct conversion pathway connecting the open and closed-cubane structures. A notable phenomenon is that the incidental proton transfer is directed toward the moiety becoming a terminal ligand; protonation of the μ-oxo ligand is impossible in either isomer (O5H in S₃^A,W or OxH in S₃^B,W), which is justified by the relaxed potential energy scan for proton translocation between Ox and O5 (Figure 2d, Text S1 in the Supporting Information).

Figure 2.

(a) Relative Gibbs free energy profiles for the conversion between S₃^A,WY_z(W1=H₂O) and S₃^B,WY_z(W1=H₂O) in all the possible spin states of the S₃⁺Y_z state. Because of the close similarity to the other spin states, more information regarding the changes of (b) geometric structures, (c) electronic configurations along the MEP, and (d) relaxed PES scan curve of proton transfer between Ox and O5 are exemplified in the highest 13-et/αααα spin state. Spin populations are displayed in yellow contours and key interatomic distances are given in Å.

In contrast to Capone et al.,⁶¹ where the formal oxidation state of Mn3 is lowered from (IV) to (III), while Mn1 acquires a partial radical character in the proximity of the transition state (TS), our results show that the electronic configuration of the OEC cluster essentially remains constant along the minimum energy paths (MEPs). That means all Mn keep valence (IV) throughout as reflected by the Mulliken spin populations (Figure 2c, Text S2 and Table S2 in the Supporting Information). The reason may be attributed to the exclusion of structural and thermal fluctuations along the MEPs, which are instead present during the molecular dynamics simulations. Anyhow, consistent with Isobe et al.,⁶⁰ our calculated S₃^A,W → S₃^B,W barriers of 24.3–25.6 kcal/mol and stabilization energies of 13.4–14.8 kcal/mol for the open-cubane form (S₃^A,W) on all the spin surfaces (Figure 2a,b, Table S1 in the Supporting Information) show that the same conclusion can be drawn, that is, conformational change in the S₃⁺Y_z state is essentially confined to the unidirectional closed-to-open cubane conversion and forbidden reversely. Consequently, the bidirectional structural flexibility prevalent in both S₁ and S₂ states has disappeared in the following S₃⁺Y_z state, with an overwhelming preference for the open-cubane structure.

It is worth mentioning that the abovementioned conclusion is strictly only valid for the S₃⁺Y_z state and does not apply for the S₂⁺Y_z^• state, which can be formed from the S₃⁺Y_z state by electron back donation from Y_z to Mn under certain conditions, as shown in some experimental findings.⁸⁸⁻⁹⁰ The structural equilibration in the S₂ⁿY_z^• state was suggested as a requirement for water exchange in the S₃⁺Y_z state,⁹¹ in which the open-to-closed conversion is involved and readily reversible, but necessitates a Mn(III) center within the cluster.

Reversible Structural Isomerization in the S₃ⁿY_z^• State

By various experimental approaches, it has been established that O₂ formation and release begins after the flash-induced formation of the S₃⁺Y_z^• state only after a lag phase of about 200 μs.^12,70−92 This lag phase has been assigned to a deprotonation reaction in the S₃⁺ to S₃ⁿ transition, as shown in Figure 1d. Since the initial deprotonation site has been widely acknowledged as W1(H₂O) (via the egress gate Asp61 to the lumen) during the S₃ → S₄ transition,^{52,53,78,93−96} this ligand was formulated as a hydroxide (OH^–) in our S₃ⁿY_z^• model, in agreement with a series of previous computational work.^52,53,94,96 In analogy to the abovementioned case of S₃⁺Y_z, redox-related events were not observed at any of the Mn centers along the whole MEP (Figure 3c and Table S4 in the Supporting Information) and various spin couplings do not significantly affect the energetics even after Y_z^• addition. The redox-irrelevance and spin-insensitivity for such a ligand exchange are understandable because the octahedral coordination geometry of Mn3(IV) basically maintains during the simultaneous movements of O5 and Ox in opposite directions, and the two oxygens never approach a bonding distance to cause Mn reduction.

Figure 3.

(a) Relative Gibbs free energy profiles for the conversion between S₃^A,WY_z^•(W1=OH^–) and S₃^B,WY_z^•(W1=OH^–) in all the possible spin states of the S₃ⁿY_z^• state. Because of the close similarity to the other spin states, the highest 14-et/ααααα spin state was selected for visualizing more information regarding the changes of (b) geometric structures and (c) electronic configurations along the MEP, and (d) relaxed PES scan curve of proton transfer between Ox and O5.

Interestingly, the obtained reaction landscapes of the structural isomerization in the S₃ⁿY_z^• state is fundamentally changed with regard to both thermodynamics and kinetics (Figure 3a,b, Table S3 in the Supporting Information) as compared to that of the S₃⁺Y_z state (Tables S1 in the Supporting Information), allowing for a dynamically reversible isomerization S₃^A,WY_z^• ⇌ S₃^B,WY_z^• in chemical equilibrium. First, the relative thermodynamic stability of S₃^B,WY_z^• is greatly enhanced to only 2.4–3.9 kcal/mol higher in free energy than S₃^A,WY_z^• (vs 13.4–14.8 kcal/mol in the S₃⁺Y_z state). Superficially, according to the relationship between ΔG° and equilibrium constant K, this energy difference would still correspond to a major population of S₃^A,WY_z^• in the equilibrium at room temperature; however, overemphasis on the precise quantitative population of the isomers in the S₃ⁿY_z^• state would be undesirable because of the calculated small energy gap and the well-known intrinsic limitations in the accuracy of DFT methodology,⁹⁷⁻¹⁰⁰ and the ambiguous direction of the equilibrium shifting given the consumption of S₃^A,WY_z^• and/or S₃^B,WY_z^• when proceeding to the S₄ state. Thus, the isomerism suggested here in S₃ⁿY_z^• resembles the situations in the S₁ and S₂ states,^31,32,41 where the closed-cubane structures are also deemed important for the catalytic progression despite the calculated slight energetic disadvantages compared with the open-cubane forms, that is, +3.2 kcal/mol for S₁^B and +(1–2) kcal/mol for S₂^B (see Text S3 in the Supporting Information for the detailed analysis).^{26,31,32,41,48,101−104} As a consequence of the markedly closer energies of the isomers in the S₃ⁿY_z^• state, the predominance of the open-cubane structure is undermined and the significance of S₃^B,WY_z^• should be highlighted in addition to S₃^A,WY_z^•. Strictly speaking, one should not overlook the importance of either isomer in the S₃ⁿY_z^• state, considering the aforementioned uncertain factors that would lead to an indefinite identification of a dominant or most reactive component.

Besides the thermodynamics, the free energy barriers from S₃^A,WY_z^• to S₃^B,WY_z^• for all the possible spin states are dramatically reduced to 14.8–16.2 kcal/mol (vs 24.3–25.6 kcal/mol in the S₃⁺Y_z state), which allows for smooth production of S₃^B,WY_z^• at a level of milliseconds kinetics (see Text S4 in the Supporting Information for more details). It is noteworthy that the direct reactant for the isomerization turns out to involve the protonated O5H, which is reachable by facile deprotonation from Ox and vice versa (Figure 3d); this remarkably contrasts the situation in the S₃⁺Y_z state, where O5H is not achievable for S₃^A,W (Figure 2d). These results show a feasible pathway from S₃^A,WY_z^• to S₃^B,WY_z^• preceded by Ox deprotonation and demonstrate that the structural heterogeneity lost in the S₃⁺Y_z state becomes available again in the S₃ⁿY_z^• state. This leads to a more balanced constituent of the isomers, as compared to the dominance of the S₃^A,W conformation and the high energetic barrier for isomerization in the S₃⁺Y_z state.

W1 Deprotonation Facilitates the Open-to-Closed Isomerization

As shown above, a magnitude of ca. 10 kcal/mol decrease in both barrier heights and relative energies from S₃⁺Y_z to S₃ⁿY_z^• has largely changed the equilibrium distribution of the isomers. This is mainly manifested in the feasibility of S₃^A,WY_z^• converting to S₃^B,WY_z^•, since B to A is attainable in both the S₃⁺Y_z and S₃ⁿY_z^• states. Quite evidently, the S₃ⁿY_z^•(W1=OH^–) state is differentiated from S₃⁺Y_z(W1=H₂O) by its oxidized Y_z^• unit and deprotonated W1 ligand, that is, the asynchronous departure of an electron and a proton from two separated sites. Thus, two virtual states S₃⁺Y_z^•(W1=H₂O)* and S₃ⁿY_z(W1=OH^–)* characterizing the single effect were artificially fabricated in order to clarify the ultimate reason for the observed difference. Since the spin state selectivity is expected to bring little impact on the isomerization, only the highest spin states were studied for a comparison, as shown in Figure 4.

Figure 4.

Relative Gibbs free energy profiles for the conversions between the virtual states S₃^A,WY_z^•(W1=H₂O)* and S₃^B,WY_z^•(W1=H₂O)* (left) and between the virtual states S₃^A,WY_z(W1=OH^–)* and S₃^B,WY_z(W1=OH^–)* (right) in their respective highest spin states; “*” denotes a virtual state.

The situation for the virtual S₃⁺Y_z^•(W1=H₂O)* and S₃ⁿY_z(W1=OH^–)* states fairly coincides with that of the S₃⁺Y_z(W1=H₂O) and S₃ⁿY_z^•(W1=OH^–) states, respectively, in terms of the reaction energetics and geometric parameters (Tables S5–S8 in the Supporting Information), as well as the proton mobility between Ox and O5 (Figures S2 and S3 in the Supporting Information). The comparison clearly reveals that it is the occurrence of W1(H₂O) deprotonation, rather than appearance of the Y_z^• radical, that substantially promotes the A to B isomerization in the S₃ⁿY_z^• state. This is reasonable because the covalent bonding interactions within the Mn₄CaO₆ cluster should be much more powerful than the electrostatic effect brought by the distal Y_z^• group. Specifically, we expect that the strong σ donation from W1=OH^– reinforces its coordination to Mn4 but considerably weakens the O5–Mn4 bonding, due to the “structural trans effect” in octahedral transition metal complexes.^105−108 The diminished overlap between the Mn4 3d and O5 2p orbitals can in turn stabilize the O5 2p–H s covalency, increasing the basicity of O5 and explaining the accessibility of O5 protonation in both S₃^A,WY_z^•(W1=OH^–) and S₃^A,WY_z(W1=OH^–)*. Furthermore, the Mn3–O5 bond is weakened by the protonated O5H, which therefore becomes easier to be substituted by Ox (oxo). The altered bond strengths can be seen from variations of the key bond lengths and Wiberg bond orders (Table S9 in the Supporting Information). To sum up, the feasibility of the open-to-closed isomerization in the S₃ⁿY_z^• state is directly attributable to W1(H₂O) deprotonation, which causes a series of subtle changes in Mn–ligand interactions.

Although Y_z^• itself produces little chemical effect on the isomerization, its formation is necessary for the subsequent W1(H₂O) deprotonation. After the third flash given to dark-adapted PSII, the S₃⁺Y_z^• state forms accompanied by deprotonation of the phenolic oxygen of Y_z to the ε-nitrogen of His190, and thus an extra positive charge accumulates within the vicinity of the Mn₄CaO₆ cluster. Thereafter, the Mn4-bound W1 serves as an ideal deprotonation site for charge compensation because it is in strong hydrogen-bonding interaction with the negatively charged D1-Asp61, which connects further to the proton exit channel and the lumen.^{39,52,56,81,95,109−114} Thereby, the occurrence of Y_z oxidation is an essential prerequisite for the reversible structural isomerization in the S₃Y_z^• state, from a perspective of the causal relationship.

It is noted that up to date, there is still no unambiguous/conclusive assignment for the protonation states of the titrable groups (especially W2 and Ox) of the OEC cluster in the S₃⁺Y_z and S₃ⁿY_z^• states; however, our models adopt the protonation states suggested by Cox et al.,⁵⁷ which reproduced the experimental EPR and electron–nuclear double resonance (ENDOR) and electron–electron double resonance-detected nuclear magnetic resonance (EDNMR) of the S₃ state, and are in good agreement with most computational studies.^{51,60,61,91,94,115−118} Still, we have also performed extensive additional computations and found that our conclusion still holds even if different protonation state distributions were considered (see Text S5 and Tables S10–S14 in the Supporting Information for the details).

Alternative Computational S₃ State Models

For the S₃⁺Y_z and S₃ⁿY_z^• states, Corry and O’Malley proposed “oxo–hydroxo ⇌ oxo–oxo ⇌ peroxo” and “oxo–oxo ⇌ peroxo” equilibria to describe the chemical nature of “O5–Ox” in S₃⁺Y_z and S₃ⁿY_z^•, respectively.^65,67 Early O–O bond formation in the S₃ state was also explored by Pushkar et al.⁶⁴ and Isobe et al.^58,59 We note that the “oxo-hydroxo” model with all octahedral Mn(IV) employed in this study adequately fits the vast majority of results from EXAFS,^{5,116,119,120} EPR/ENDOR/EDNMR, and X-ray absorption and emission spectroscopies in the S₃ state.^57,118,121 In contrast, while it still remains unclear if the “oxo–oxo” model is consistent with these spectroscopic data, the “peroxo” model would produce two anisotropic Mn(III) and thus is clearly inconsistent with the experimental observations. It has been also ruled out by all the latest XFEL experiments with updated essential details (e.g., O5–O6/Ox distance),³⁶⁻³⁹ despite support from one initial study.³⁵ The calculated S = 4 ground state of the peroxo model⁶⁵ does not agree with the S = 3 signal observed experimentally.⁵⁷ On the basis of substrate-water exchange,^24,122 although we cannot fully exclude the “peroxo” model given the option of suitable structural/redox equilibria, obviously a stable peroxide can be ruled out. From the aspect of computational modeling, calculations by coupled cluster theory, which is beyond traditional DFT, also strongly disfavors the scenario based on an early-onset O–O bond formation in the S₃ state.¹²³

However, it remains possible that the “peroxo” form could constitute part of the redox equilibria in the S₃ state and it might be catalytically relevant, but it should not represent the dominant form in the S₃ state. For the S₃ⁿY_z^• state, the “peroxo” model was indeed considered as one possible option because water exchange dramatically slows down as compared to S₃,¹²⁴ but the model was also ruled out by the authors in that report because of the inconsistency with the results from time-resolved X-ray experiments.^3,70,124 Still, we have in detail considered the “oxo–oxo” model in both the S₃⁺Y_z and S₃ⁿY_z^• states (Text S5 and Tables S12–S14 in the Supporting Information), and we can conclude that it does not change the basic conclusion of this study. Finally, we emphasize that the “oxo-hydroxo” model should be adopted (for cyanobacterial PSII) because of its representation of the most stable form of the ground state in the dominant population of the S₃⁺Y_z and its derived S₃ⁿY_z^• states; for high-plants, the “oxo-hydroxo” model is also valid in the novel closed-cubane S₃ structure according to Zahariou et al.,⁶⁸ but the circumstances of the structural isomerization, if exist in the water-unbound form, would need further investigations.

Comparison to Experimental Observations

Since experimental techniques probing into the S₃ → S₄ transition remain difficult, so far there is very limited information regarding the morphological changes of the Mn₄CaO₅₍₆₎ cluster upon formation of the S₃ⁿY_z^• state. Therefore, the structural isomerization found in this study should be seen as a theoretical prediction pending experimental verification. However, some suggestive evidence still exists in support of our proposal.

Nilsson et al. discovered that substrate-water exchange is arrested in the S₃ⁿY_z^• state because of the observed dramatically slowed kinetics as compared to earlier S states.¹²⁴ As discussed therein, the possible reasons include the impossibility to generate a Mn(III) center that is required for water exchange,⁹¹ H⁺ release that leads to much stronger binding of the deprotonated group, and reconstruction of the H-bonding network after proton-coupled electron transfer upon Y_z oxidation. Our proposed reversible isomerization is compatible with the observation because, in contrast to the S₃^A,WY_z ⇌ S₂Y_z^• equilibrium that supports water exchange in the S₃⁺Y_z state, the S₃^A,WY_z^• ⇌ S₃^B,WY_z^• equilibrium does not facilitate water exchange due to the lack of Mn(III) formation. In fact, the chemical equilibrium would cause extensive rearrangement of the locations and H-bonding orientations of water molecules and may thus even contribute to slowing down the rate of substrate water exchange in the S₃ⁿY_z^• state. Such changes in the H-bonding network have also been suggested to affect the distribution of the conformational microstates of water molecules and to thereby affect the rate of the S₃ⁿY_z^• → S₀Y_z transition.¹²⁵

Bao and Burnap studied the O₂ release kinetics by site-directed mutagenesis and found that both the lag and slow phases during the S₃⁺Y_z → S₀ⁿY_z transition are retarded. On that basis, they suggested that “proton tautomerization” and/or “structural isomerization” precede(s) dioxygen formation.⁹³ Specifically, they suggested two interpretations. For proton tautomerization, they proposed that it would be followed by O–O bond formation via W3–W2 nucleophilic attack,^72,81,82,85 while in case of open-closed structural isomerization, O–O bond formation by oxo–oxyl radical coupling between W2 and O5 may occur, in line with previous suggestions.^80,83,84 We note that O–O bond formation via water nucleophilic attack (WNA) appears less favorable on the basis of recent experiments^36,37,124 and theoretical calculations.^71,126 Indeed, our present results provide further support for the variant radical coupling mechanism using a closed-cubane S₄ structure for O₂ evolution because our theoretical finding shows that the S₄^B,W structure could be obtainable via the open-to-closed rearrangement in the S₃ⁿY_z^• state (rather than in the S₃⁺Y_z state as assumed in ref (84)).

Thus, the open-closed isomerization in the S₃ⁿY_z^• state may correspond to the proposed “structural isomerization” preceding dioxygen formation⁹³ and to thereby constitute the rate limiting 1–2 ms phase (slow phase) that follows a 200 μs lag phase (Figure 1d) and precedes the much more rapid O₂ formation.^{3,12−22,92,127} According to the Eyring–Polanyi equation of TS theory assuming a standard pre-exponential factor,^128−130 the 1–2 ms kinetics is calculated to correspond to an activation free energy ∼14 kcal/mol at room temperature. Given the limited errors from DFT methodology and possibly experimental measurement, a safer quantity for the barrier should be around 13–15 kcal/mol for a process that occurs on a timescale of milliseconds.⁷¹ Siegbahn ascribed the slow phase of O₂ formation to an intramolecular proton transfer step with 10.2 kcal/mol barrier,⁵² but as he pointed out, considering a typical accuracy within 3–5 kcal/mol, which normally overestimates barriers for a DFT hybrid functional,^51,97,98,131 it is far below the required limit for a millisecond process.⁵² By comparison, our calculated barriers of 14.8–16.2 kcal/mol for all the possible spin states are in much better agreement with the experimental kinetics (see Text S4 in the Supporting Information for more details). If the open-to-closed transition in the S₃ⁿY_z^• state was indeed the main rate limiting step for O₂ formation, this would mean that O₂ formation would start exclusively from the S₃^B,WY_z^• state, and radical coupling from the S₃^A,WY_z^• state would not be possible for reasons still to be determined. Thus, at the current stage of knowledge, we do not emphasize the proposed isomerization as the only possibility taking place in S₃ⁿY_z^• → S₄⁺Y_z before compelling experimental evidence emerges; however, the reversible open-to-closed structural rearrangement should be regarded as a viable mechanism or at least as part of processes responsible for the slow phase.

Implications for the Mechanism of O–O Bond Formation

The proposed reversible open-closed interconversion in the S₃ⁿY_z^• state has important implications for the mechanism of O–O bond formation in the S₄ state. This is illustrated in Scheme 2, which starts from the two structural architectures S₃^A,W and S₃^B,unbound observed by XFEL³⁵⁻³⁸ and EPR experiments,^44,68 respectively. The first route (a) → (b) from S₃^A,W to S₄^A,W-1 expresses Siegbahn’s oxo(O5)–oxyl(Ox) coupling mechanism that he found to be energetically most favorable.^51,53,71,76 Here, the Mn1(IV)-bound Ox radical couples with μ-O5 in an open-cubane structure. Alternatively, the radical could be localized at W2 if it is deprotonated instead of Ox, and its coupling with μ-O5 in S₄^A,W-2 might be an option, which, however, has not gained support from recent DFT calculations.^79,84

Scheme 2. Possible mechanisms of the S₃ → S₄ transition and O–O Bond formation in the S₄ State.

S₃^A,W and S₃^B,unbound in gray stand for the two potential starting configurations in the S₃ state, resolved in cyanobacteria and higher-plant PSII by XFEL and EPR experiments, respectively.^{35−38,44,68} The process focused in this work is highlighted in the green dashed box. Candidate substrates are encircled in red (favored) or blue (possible alternatives). Mn formal oxidation states (IV)(V) are displayed in different colors. The superscript “W”/“unbound” means hexa/penta-coordinate Mn4 with a water bound/unbound water trans to O5. The annotations for sequence numbers: (a,g) Y_z oxidation followed by proton release; (b,d,h) intramolecular proton transfer followed by Ox/W2/O5/Mn4 oxidation; (c) reversible open-closed rearrangement in the S₃ⁿY_z^● state, as proposed in this study; (e,i) reversible water binding to the five-coordinate Mn4(IV) in the closed-cubane structure; and (f) irreversible closed-to-open conversion in the S₃ state. Other proposed mechanisms are discussed in the text. It is noted that the oxygen labeling for S₃^B,W and S₃^B,unbound (and their derivatives) is chosen for consistency with that established by serial crystallography for the S₃^A,W36,38,39 and for convenience to describe all the transitions uniformly. These labels do not reflect the origin of the oxygens with regard to the S₁ and S₂ states because several options for water insertion during the S₂ → S₃ transition are still discussed;^{6,7,32,38,43,48,52,62,82,101,120,132−139} an alternative nomenclature based on S₃^B,unbound and the pivot/carousel water insertion is shown in Figure S4 in the Supporting Information.

For the intermediate S₃^A,WY_z^• state prior to S₄, our present results provide the first theoretical basis for the (reversible) conversion to S₃^B,WY_z^• by route (c), thereby diversifying alternative pathways leading to O–O bond formation in the S₄ state with a closed-cubane type. Specifically, S₃^B,WY_z^• with the octahedral Mn4(IV) coordination may proceed to S₄ by O5 or Mn4 oxidation through route (d), producing the O5 radical in S₄^B,W-1, or alternatively, Mn4(V) in S₄^B,W-2. Both options, W2–O5 coupling (blue) on Mn4 and O5–Ox coupling involving multiple metal (Mn and Ca) centers (red), are worth considering. It is noted that O5–Ox coupling in the S₄^B,W-1 state resembles the variant oxyl–oxo mechanism by Li and Siegbahn,⁸⁴ which was based on previous experimental proposals.^4,80,124

The S₃^B,unbound state may either evolve to S₃^A,W via S₃^B,W after (reversible) water binding by the “pivot“ or “carousel” mechanism^{43,44,120,132−134} through the route (e) → (f) and then advance to the O₂ formation routes described above, or, as suggested by Pantazis and collaborators, directly proceed to S₄^B,unbound without water binding by the route (g) → (h).^68,69,87 The latter pathway involves a penta-coordinate Mn4(IV) center and would lead to Mn4(V) where nucleophilic Ox-O5 coupling^68,69,87 or hydroxo-oxo coupling between W2 and O5 might be possible. We note that the finding in the present study can provide an additional route (a) → (c) → (i) → (h) from S₃^A,W to S₄^B,unbound (other than from S₃^B,unbound).

Alternative mechanisms proposed in the literature include WNA from Ca-bound W3 onto the electron-deficient Mn4(V)=O (W2)^6,72,81,85 and oxyl-oxo coupling between W1 and μ-oxo O4.⁷⁸ Both appear inconsistent with mass spectrometric and EPR-based substrate water exchange data, which show that both substrates are bound to Mn(IV) in the S₃⁺Y_z and S₃Y_z^• states (excluding WNA),^122,124 and are best consistent with O5 as the slow exchanging substrate water.^{4,24,80,83,122,124,140−143} Nevertheless, these suggestions will also be further scrutinized in future DFT calculations.

Since S₃^A,WY_z^• and S₃^B,WY_z^• are nearly isoenergetic and for both states, low-energy routes for O–O bond formation via radical coupling have been determined,^51,84 the intriguing possibility arises from the results of this study that O–O bond formation may occur via two routes, or even more, if also the S₃^B,unboundY_z^• → S₄^B,unboundY_z → S₀Y_z path in “water-deficient” catalytic sites is taken into account.^68,69 While recent water exchange experiments in the S₂ state have reported first evidence for two possible fast exchanging water substrates,^24,122 the current water exchange data in the S₃ state are best consistent with only one set of substrate waters. This would favor that either S₄^B,W/S₄^B,unbound (substrates: W2 and O5) or S₄^A,W (substrates: Ox and O5) would be involved. However, the present study suggests that the energetic and kinetic differences between these possible routes are so small that minor differences between species or experimental conditions could favor one or the other pathway.

Conclusions

In summary, we have identified a reversible open-to-closed isomerization for the S₃ⁿY_z^• state, in contrast to the unidirectional conversion in the S₃⁺Y_z state. This isomerization immediately before O₂ formation is activated by deprotonation of a Mn-bound water (W1) after tyrosine Y_z oxidation. The structural rearrangement may constitute or contribute to the slow kinetic phase that prepares the Mn₄CaO₆ cluster for O₂ formation. Thus, the restored structural heterogeneity prior to the S₄ state diversifies the viable options for O–O bond formation in PSII. In this way, the availability of both open and closed-cubane structures in the S₄ state may reflect a “two-pronged” arrangement of the OEC, allowing for efficient and robust water oxidation, and may have contributed to its evolutionary development. The elegant structural reversibility triggered by proton release in the natural enzyme may provide a useful reference for designs of artificial catalysts.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.2c03528.

• Model construction, spin state definition, computational details, supplementary texts, tables, figures, and references, and optimized Cartesian coordinates (PDF)

The authors declare no competing financial interest.