Energetics of proton release on the first oxidation step in the water oxidizing enzyme

Abstract

In Photosystem II (PSII), the Mn₄CaO₅ cluster catalyzes the water splitting reaction. The crystal structure of PSII shows the presence of a hydrogen-bonded water molecule directly linked to O4. Here we show the detailed properties of the H-bonds associated with the Mn₄CaO₅ cluster using a quantum mechanical/molecular mechanical (QM/MM) approach. When O4 is taken as a μ-hydroxo bridge acting as a hydrogen-bond donor to water539 (W539), the S₀ redox state best describes the unusually short O4–O_W539 distance (2.5 Å) seen in the crystal structure. We find that in S₁, O4 easily releases the proton into a chain of eight strongly hydrogen -bonded water molecules. The corresponding hydrogen -bond network is absent for O5 in S₁. The present study suggests that the O4-water chain could facilitate the initial deprotonation event in PSII. This unexpected insight is likely to be of real relevance to mechanistic models for water oxidation.

The core of the Photosystem II (PSII) reaction center is composed of D1/D2, a heterodimer of protein subunits that contains the cofactors involved in photochemical charge separation, quinone reduction, and water oxidation. In PSII, the Mn₄CaO₅ cluster catalyzes the water splitting reaction: 2H₂O → O₂ + 4H⁺ + 4e⁻ (reviewed in ^1,2). The release of protons has been observed in response to changes in the oxidation state (the S_n state, where the subscript represent the number of oxidation steps accumulated) of the oxygen-evolving complex and occurs with a typical stoichiometry of 1:0:1:2 for the S₀ → S₁ → S₂ →S₃ → S₀ transitions, respectively, (e.g. ^3,4). Although the relevant pathway for proton transfer in each S-state transition is not yet clear, proton transfer (PT) may proceed via different pathways in the PSII protein, depending on the S-state transition ^5-7. Candidates for the relevant proton transfer pathways have been reviewed recently ^7-12 and site-directed mutagenesis studies are testing the various possibilities ¹³.

The majority view in the current literature is that the best resolved X-ray crystal structure (1.9 Å) of the Mn₄CaO₅ cluster ¹⁴ represents an over-reduced form due to its reduction by the X-ray beam (e.g., ¹ but see also e.g. ¹⁵ for an alternative explanation). This was suggested to explain the elongation of Mn–Mn and Mn–O distances compared to those obtained from EXAFS (e.g., ^16-21). Detailed oxidation states of the 1.9-Å crystal structure are discussed in recent theoretical studies (see Ref. ²² and references therein). ESEEM and ENDOR studies have suggested that all of the μ-oxo bridges of the Mn₄CaO₅ cluster are deprotonated in the S₂ state, and that the water molecules, W1 and W2, bound to Mn4 are H₂O or OH^{− 23,24}. Because proton release is not observed in the S₁ to S₂ transition, the ESEEM and ENDOR data thus imply that the μ-oxo bridges of the Mn₄CaO₅ cluster are already deprotonated in S₁ ^16,19-21,25. The 1.9-Å structure revealed that two water molecules, W1 and W2, were ligands to the Mn4 atom of the Mn₄CaO₅ cluster, and two waters, W3 and W4, were ligands to the Ca atom ¹⁴. W3 has been proposed as a candidate deprotonation site because it is one of the closest water molecules to the redox active TyrZ ¹⁴ and because the W3-Sr bond in the Mn₄SrO₅ cluster was specifically longer with respect to the Mn₄CaO₅ cluster ²⁶. All of these bound water molecules are candidates not only as potential substrates for water oxidation (e.g., ^14,27-32) but also as ionisable groups that could undergo deprotonation during the enzyme cycle in which positive charge equivalents are accumulated ²¹.

In recent experimental (e.g., ³³) and theoretical studies (e.g., ^34,35) it has been proposed that the proton released on the S₀ to S₁ transition involves hydroxyl form of O5, an oxygen atom that occupies one of the corners of the cubane, linking the Ca, Mn3 and Mn1 and connecting the cubane to Mn4. Recently however, in the radiation-damage-free PSII crystal structure obtained using the X-ray free electron laser, which is assumed to be in the S₁ state, Suga et al. proposed that O5 is a hydroxide ion in S₁ ³⁶. This was based on the observation of significantly long distances between O5 and the adjacent Mn ions ³⁶. This would be conflict with the already deprotonated μ-oxo bridges in S₁ suggested from the ESEEM and ENDOR data ^19-21. Recent report comparing published QM/MM data and EXAFS data ³⁷ with the free electron laser structure ³⁶ suggested that the crystals used for this study could contain significant amount of the reduced S₀ state ³⁸.

The 1.9-Å structure shows the presence of a chain of strongly H-bonded 8 water molecules (O4-water chain) directly linked to O4 (linking Mn4 and Mn3 in the Mn₃CaO₄-cubane), whereas O5 has no direct H-bond partner ¹⁴. Notably, the activation energy is specifically low in the S₀ to S₁ transition, and can be rationalized as a rate-limiting electron transfer followed by a deprotonation step ²¹. As far as we are aware, none of the QM/MM studies have considered the O4-water chain explicitly (i.e., QM region) irrespective of the H-bond network around the Mn₄CaO₅. In the 1.9-Å structure, the water molecule W539 (B-factor: Monomer A=23.9, Monomer B=25.2) is situated near the O4 atom of the Mn₄CaO₅ cluster ¹⁴. Remarkably, the H-bond between O4 and W539 (O4–O_W539) is unusually short (~2.5 Å) in comparison with typical O–O distances of ~2.8 Å for standard (asymmetric) H-bonds in H₂O ^39,40, even if we consider the uncertainty of ~0.16 Å in the measured distances within the 1.9-Å structure ¹⁴. Notably, “single-well H-bonds” are very short and typically have O–O distances of 2.4 to 2.5 Å ^41,42. This results in an essentially barrier-less potential between the H-bond donor and acceptor moieties. The appearance of single-well H-bonds in protein environments is often associated with PT events ^43-45.

Here we look for short H-bonds in the environment of the Mn₄CaO₅ complex and investigate their significance in energy terms by adopting a large-scale QM/MM approach based on the crystal structure analyzed at 1.9 Å resolution ¹⁴.

RESULTS

A short H-bond between O4 and W539

The O4–O_W539 bond is unusually short in the original 1.9-Å structure (2.50 Å, Figure 1) ¹⁴. When a H-bond is assumed for the O4–O_W539 bond, two H-bond patterns are possible. Case 1: the H atom is from O4, an OH⁻ moiety, so that the bond can be written O4–H…O_W539. Case 2: the H atom is from the water, W539, so that O4 is O²⁻, and the bond can be written as O4…H–O_W539 (Supplementary Figure 1). Using these two cases, we investigated whether formation of the short H-bond between O4 and W539 is energetically possible in the S₋₁ to S₁ states of the high oxidation state model. In the low oxidation state model these Mn valence states corresponds to the S₁ to S₃ states ¹⁵ and this option was also investigated (see Table 1), however in the present study, if not otherwise specified, the S-states given refer to the high oxidation state model. For discussions of the Mn oxidation state of the low oxidation state model ⁴⁶, see Supplementary Discussion.

Figure 1. H-bond networks near the Mn₄CaO₅ cluster.

(a) Overview. The O4-water chain is the only H-bond network directly linked to the Mn₄CaO₅ cluster (O4–O_W539 = 2.50 Å). The O5 path is not directly H-bonded to the Mn₄CaO₅ cluster (O5–O_W2 = 3.08 Å). (b) The water chain linked to O4 of the Mn4CaO5 cluster in the 1.9-Å structure ¹⁴. Water molecules (red), Cl⁻ 2 (green), and Mn₄CaO₅ (purple, orange, and red for Mn, Ca, and O atoms, respectively) are depicted as balls. H-bonds or ionic interactions are represented by dotted arrows. (c) H-bond distances in the water chain (Å).

Table 1.

H-bond geometries (in Ångstroms) of the O4 site. O5 = O²⁻; O4 =OH⁻ in pre-PT and O²⁻ in post-PT. RMSD = root-mean-square deviation of the optimized heavy atoms with respect to those of the 1.9-Å structure. Short H-bond distances (< ~2.5 Å) are in bold.

	X-ray	pre-PT (O4 = OH−)			post-PT (O4 = O2−)
High (Low)a		S1 (S3)	S0 (S2)	S−1 (S1)	S1 (S3)	S0 (S2)	S−1 (S1)
Mn1-4		III,IV,IV,III	III,IV,III,III	III,III,IV,II	III,IV,IV,III	III,IV,III,III	III,III,IV,II
O4–539	2.50	2.45	2.55	2.64	2.58	2.57	2.49
539–538	2.77	2.55	2.59	2.61	2.78	2.72	2.76
538–393	2.71	2.60	2.64	2.66	2.70	2.68	2.65
393–397	2.73	2.62	2.64	2.67	2.75	2.72	2.71
397–477	2.69	2.83	2.83	2.87	2.68	2.66	2.64
477–545b	2.87	2.72	2.71	2.71	2.71	2.70	2.69
545–1047	2.50	2.61	2.61	2.60	2.66	2.66	2.66
1047–399	2.52	2.71	2.70	2.68	2.53	2.54	2.56
RMSD : Mn4Ca		0.11	0.20	0.18	0.16	0.24	0.16
RMSD : Mn4CaO5		0.27	0.31	0.30	0.30	0.30	0.30

^aHigh oxidation state model in normal script (low oxidation state model in italics and parenthesis ¹⁵).

^bThe large deviation of the W477–W545 length from the 1.9-Å structure ¹⁴ originates from the poorer density specifically for W477 in the crystal (Supplementary Figure 8). In the O4-water chain, W477 is the only water molecule that does not donate an H-bond to a backbone C=O but accepts an H-bond from the backbone NH of CP43-G353 (Figure 1) and this seems to be associated with the greater disorder.

In S₀ [(Mn1, Mn2, Mn3, Mn4) = (III, IV, III, III)], QM/MM calculations reproduced a short H-bond distance of ~2.5 Å (2.55 Å, Table 1) for O4–O_W539 only when O4 was OH⁻ and donated an H-bond to W539 (“pre-PT” pattern in Figure 2; Table 1). In S₁, the energy profile of the O4–O_W539 bond (Figure 3a) resembles an energetically unstable single-well H-bond (Supplementary Figure 2). These results indicate that matching of the pK_a values of the donor (i.e., O4, Supplementary Figure 1) and acceptor (i.e., W539) moieties occurs in S₁. Due to the presence of the single-well H-bond (e.g., ⁴⁷), the “pre-PT” H-bond pattern (Figure 2) was energetically unstable in S₁ (discussed later). Note that even a single-well H-bond has the only energy minimum at the O–O distance of ~2.5 Å (e.g., not 2.6 or 2.7 Å ⁴³).

Figure 2. H-bond pattern change of the entire water chain induced by a release of a proton from O4 in response to the formation of the S₁ state.

(a, b) The two H-bond patterns (pre- and post-PT) possess the same net charge and the same number of H atoms. Altered and unaltered H-bonds are indicated by red and blue lines and balls, respectively. (c) QM/MM optimized geometries in the pre-PT, (d) post-PT and (e) both conformations. Initial donor-to-acceptor orientations are indicated by black arrows, whereas altered donor-to-acceptor orientations in the post-PT conformation are indicated by magenta arrows.

Figure 3. The energy profiles along the proton transfer coordinate for the O4–O_W539 bond.

(a) pre-Tand (b) post-PT in S₁ (blue solid curve), S₀ (black dotted curve), and S₋₁ (blue thin solid curve). For comparison, the energy minimum in the O4 moiety was set to zero for all S states.

In the over-reduced state, S₋₁, QM/MM calculations also reproduced the short H-bond distance of ~2.5 Å for O4–O_W539 (Table 1) but only when O4 was O²⁻ and when it accepted an H-bond from W539 (“post-PT” pattern in Figure 2).

Proton transfer from OH⁻ at O4 along the O4-water chain

In the QM/MM calculations, OH⁻ at O4 was stable as an H-bond donor to W539 in S₀ and lower S states (Figure 3). However, in the S₁ state, OH⁻ at O4 was energetically unstable due to formation of the single-well H-bond, leading to the release of a proton from OH⁻ at O4 and the formation of O²⁻ at O4. Remarkably, the released proton was stabilized at W1047 in the form of H₃O⁺, which is 13.5 Å away from O4 (Figure 1). Overall, the pre-PT pattern (Figure 2) in the initial state completely transformed into the post-PT pattern (Figure 2) as a result of proton transfer.

The short O4–O_W539 in the pre-PT conformation lengthened to 2.58 Å in the post-PT pattern in S₁ (Table 1) from 2.50 Å in the 1.9-Å structure. This longer H-bond in S₁ is consistent with O4–O_W539 distances of 2.59 Å in both PSII monomers in the recent radiation-damage-free PSII crystal structure obtained using the X-ray free electron laser (Supplementary Table 1) ³⁶. These results seem to correspond to a situation in which the free electron laser structure ³⁶ has the S₁ state in which the O4 is deprotonated, while in the earlier conventional X-ray diffraction structure (1.9-Å structure) ¹⁴ the Mn₄CaO₄ is in a more reduced state and exhibiting a single-well H-bond between O4–O_W539.

In the 1.9-Å structure, W1047 has unusually short H bonds, O_W1047–O_W545 (2.50 Å) and O_W1047–O_W399 (2.52 Å) ¹⁴. Intriguingly, the short O–O distances were reproduced only when we assumed the post-PT H-bond pattern of the O4-water chain (2.46~2.54 Å, Table 1 and Supplementary Table 2). In general, typical H₂O…H₃O⁺ (i.e., the Zundel cation) has an unusually short O–O distance of ~2.4 Å ^39,40. Therefore, the short H-bond distances of W1047 in the 1.9-Å structure imply that W1047 is capable of forming H₃O⁺. Remarkably, the X-ray free electron laser structure ³⁶ also has unusually short H-bonds at the corresponding positions (Supplementary Table 1), a further indication that the X-ray free electron laser structure represents the post-PT S₁ state. The presence of these single-well H-bonds indicates that local movement of a proton within the O4–O_W539 bond of ~2.5 Å results in a sequential downhill PT from O4 to W1047 over a distance of 13.5 Å (Figure 2 and Supplementary Figure 3).

A working model

The results indicate that proton transfer occurs through the O4-water chain and thence toward the lumenal bulk surface via PsbU (see Supplementary Discussion). This proton transfer reaction is specifically associated with the S₀ to S₁ transition. This is supported by the following arguments.

The O4-water chain is the only H-bond network in the 1.9-Å structure ¹⁴ that is directly linked with the O atoms of the Mn₄CaO₅ cluster (via the O4–O_W539 bond). The energy barrier for proton transfer along the O4–O_W539 bond is clearly the lowest among all the possible Mn₄CaO₅ deprotonation sites in the 1.9-Å structure ¹⁴. This should contribute to a smaller activation energy for deprotonation, a property consistent with the known characteristics of the S₀ to S₁ transition, which include its rate being insensitive to H₂O/D₂O exchange and its activation energy being low compared that of the S₂ to S₃ transition ^21,48,49. In the energy profiles of the H-bonds studied here, a proton became energetically more stable as it moved away from Mn₄CaO₅, indicating that a downhill proton transfer reaction occurs specifically on the S₀ to S₁ transition (Figure 4). The driving force for the PT toward the protein surface in S₁ is attributed to the increased positive charge on Mn₄CaO₅ on the S₀ to S₁ transition and this conclusion was supported by the observation that lower S-states resulted in the uphill proton transfer reaction (Figure 4 and Supplementary Figure 4).

Figure 4. The energy profiles along the proton transfer coordinate for all of the H-bonds along the O4-water chain in the pre-PT conformation.

S₁ (colored solid curves) and S₀ (black dotted curves). For comparison, the energy minimum in the O4 moiety was set to zero for all S states. For efficient analysis, the small QM region was adopted.

The remarkable linear, 8-water molecular chain acting as Grotthuss-like proton conduit, which is reported here, might be expected to work with very fast kinetics. However, proton release on this step occurs relatively slowly (tens of microseconds ²¹). This is easily understandable since the electron transfer step occurs prior to H⁺ release and is thus rate-limiting.²¹

Energetics of proton release from O5

In the present work we also analyzed the energy profiles of the proton transfer along the O5 path that proceeds from O5 to D1-Asp61 via the water molecules, W2, W446, and W442 (Figure 5). W2 is the only possible proton acceptor from OH⁻ at O5 when the original geometry of the 1.9-Å ¹⁴ or X-ray free electron laser ³⁶ structure is maintained (see however below and Supplementary Discussion. When OH⁻ at W2 becomes H₂O by accepting the proton from OH⁻ at O5 (label b in Figure 6 and Supplementary Figure 5b), W2 cannot release a proton easily to its H-bond acceptor W446 (labels b and c in Figure 6, and Supplementary Figures 5b and 5c, see Supplementary Discussion for details), thus causing a significant energy barrier for further proton transfer (Figure 6). Note that the proton transfer from O5 to W2 is more unlikely when W2 = H₂O (Supplementary Figure 6) due to the even less appropriate H-bond angle for O5…W2 (Figure 7).

Figure 5. QM/MM optimized geometries.

(a) pre-PT-like, (b) post-PT-like and (c) both conformations. Initial donor-to-acceptor orientations are indicated by black arrows, whereas altered donor-to-acceptor orientations in the post-PT conformation are indicated by magenta arrows.

Figure 6. The energy profiles along the proton transfer coordinate for all of the H-bonds along the O5 path in the pre-PT conformation.

S₁ (solid curves) and S₀ (dotted curves) for W2 = OH⁻ (see Supplementary Figure 6 for W2 = H₂O). For comparison, the energy minimum in the O5 moiety was set to zero for all S states. Labels a, b, and c correspond to the states in Supplementary Figures 5a, b, and c, respectively.

Figure 7. H-bond patterns of the O5 and W2 moiety in the QM/MM geometry.

(a) W2 = H₂O and (b) W2 = OH⁻. For clarity, only the Mn₄CaO₅ cluster, W1, and W2 are shown. In QM/MM calculations the electrostatic interactions, which are calculated from parametrized partial atomic charge, are sometimes intentionally overestimated. The correct treatment of long-range electrostatics in QM/MM calculations is hard to achieve (see e.g., ⁶⁹) and some improvements are in progress (e.g., ⁷⁰). The present results are to be considered qualitatively.

For a proton transfer, formation of a proton conducting wire (i.e., proton donor and acceptor pairs), which is itself an activated process, must occur first ⁵⁰. The activation-less “pre-organized” proton conducting wire possesses well-arranged water groups along the O4-water chain (Figure 2e) and appears to be achieved by the “pre-organized protein dipole ⁵¹” of PSII. A corresponding feature is absent (e.g., interrupted H-bond network) in the O5 path (Figure 5), suggesting that a sequential downhill PT from O5 is unlikely. It seems clear that deprotonation of a putative OH⁻ at O5 is less likely than OH⁻ at O4. The activation barrier of the O5 path appears to remain high in S₁ (Figure 6); it might be possible however that this path becomes active in proton conduction in higher S states.

Alternative O5 deprotonation models

Although some studies have suggested or assumed that O5 (=OH⁻) deprotonation occurs on the S₀ to S₁ transition ^33-35, it should also be noted that O5 has no direct H-bond partner in the original geometry of the 1.9-Å structure ¹⁴ (Figure 7) and in the free electron laser structure ³⁶. Recent QM/MM studies by Pal et al. also demonstrated that OH⁻ at O5 has no direct-H-bond partner in their S₀ model (see Ref. ³⁴); this implies that the “energy barrier” for proton transfer with O5 must be high. Indeed, Pal et al. demonstrated that O5 deprotonation only occurred when using “a small model of the oxidized S₀’ state” (i.e. not “the QM/MM S₀ model”, see Ref. ³⁴), in line with our conclusions.

Siegbahn put forward a model in which O5 (=OH⁻) deprotonation occurred on the S₀ to S₁ transition by assuming an additional water molecule near O5 in DFT calculations ³⁵, a water molecule that is not visible not only in the original 1.9-Å structure ¹⁴ but also in the recent radiation-damage-free PSII crystal structure ³⁶. This extra water molecule was positioned so that it could accept an H-bond from O5 ³⁵. An appropriately oriented proton acceptor for O5 would significantly decrease the energy barrier for release of a proton from O5. In this case however, the actual 1.9-Å structure shows that the water molecule cannot be located at the corresponding position due principally to the presence of a conserved hydrophobic amino acid side chain, D1-Val185, (Cγ_Val185-O_water = ~2.1 Å in Supplementary Figure 7). This feature is also present in the free electron laser structure, which is assumed to be in the S₁ state ³⁶. In the O5 deprotonation model, D1-Val185 was not included in the calculation and hence was absent from the model ³⁵. Furthermore, the O5 deprotonation model also included a significant structural modification in the vicinity of the H-bond accepting water: the side chain of D1-His332 was twisted (by ~90^○) along the Mn1-Nε_His332 axis ³⁵ compared to its position in the 1.9-Å structure. These two structural differences compared to the reference crystal structure thus allowed the water to be located close to O5 without the steric repulsion that would have occurred in the unmodified 1.9-Å structure (Supplementary Figure 7). It could be argued ad hoc that a conformational change may occur resulting in the situation modeled. However the most recent unreduced structure of the enzyme shows no sign of such changes ³⁶, so arguments for such a conformational change not compelling. Without specific justification for the ~90^○ twist of the Mn-ligated D1-His332 side chain and the relocation of the conserved hydrophobic D1-Val185 side chain away from the O5 moiety, it seems unjustified to place a water molecule at the position required for it to accept an H-bond from O5 for modeling the S₀ and S₁ transition.

The O5 deprotonation model also showed significantly large deviations of the atomic coordinates from those of the original 1.9-Å structure³⁵ not only for D1-His332 (RMSD = 1.28 Å) but also for CP43-Arg357 (1.19 Å) and D1-Glu189 (0.51 Å, Supplementary Table 3). Furthermore, the H-bond partner of O4, i.e., W539, and the entire O4-water chain were also absent in the earlier model ³⁵. These changes and absences are expected to affect the resulting energy of the system in that model.

Overall then based on the current data it seems clear that the deprotonation of a OH⁻ at O5, as suggested earlier ^34,35, is less likely than deprotonation of OH⁻ at O4, as suggested in the present work.

Deprotonation of the substrate oxygen and protonation of O4

It has been proposed that the exchangeable μ-oxo bridge is likely to be O5 ^24,29,33,52 (see also relevant articles ^19,20 published prior to the detailed 1.9-Å structure¹⁴), and thus O5 is a plausible candidate for the slow exchanging substrate water molecule W_s ⁵³. This is not incompatible with the presence of protonated O4 in S₀: it is possible that the liberation of O₂ upon formation of S₀ is associated with the release of a proton from the substrate ending up on μ-oxo bridge O4. A similar protonation of a μ-oxo bridge upon liberation of O₂ has been proposed in manganese catalases ⁵⁴.

The O4-water chain and the low S₀/S₁ redox potential

The conversion of the S₀ state into the S₁ state ⁵⁵ due to its oxidation by TyrD^• showed that the S₀/S₁ redox potential is lower than that of TyrD redox potential ⁵⁶, the latter was estimated to be of ~720–760 mV ^57,58. Given that all the S₀ present is converted to S₁ by TyrD^•, the S₀/S₁ redox potential can be taken to be ≤ 700 mV assuming the estimates for TyrD redox couple are reasonable. In contrast, the S₁/S₂ and S₂/S₃ redox potentials are clearly higher since they oxidize TyrD (see ⁵⁶) and they have been suggested to be ~900–950 mV ⁵⁸. Thus, the barrierless deprotonation and the subsequent downhill proton transfer reaction occurring on the S₀ to S₁ transition may contribute to the uniquely low redox potential of this redox couple, compared to those of the other S-state transitions. As demonstrated by Warshel and coworkers ^47,51, low-barrier H-bonds (including single-well H-bonds), where the pK_a difference for the donor and acceptor moieties is nearly zero (i.e. less polarized), are particularly unstable in polar protein environments. A O4–O_W539 H-bond will become a very unstable low-barrier H-bond upon oxidation to S₁, resulting in the release of the proton via the O4-water chain (Table 1). This is consistent with the observation that S₁ does not spontaneously back-react to form S₀ ⁵⁹. Proton transfer may proceed through different pathways depending on the S-state transitions ^5-7. Intriguingly, it has been reported that the rate constant for the S₀ to S₁ transition is essentially pH-independent, whereas that for S₂ to S₃ transition is pH-dependent ^21,48,49. Ionizable groups are more likely to be involved in a proton transfer pathway for a pH-dependent process than a pH-independent process; this fits with the suggestion that the uncharged O4-water chain may be active in the S₀ to S₁ transition.

Changes in Mn–Mn distances by Mn oxidation / deprotonation

Upon proton release from O4, the Mn ion undergoing oxidation was Mn3 (Figure 8). We note that the Mn undergoing oxidation is unlikely to be Mn4 because it appears to be the most reduced site, i.e. it has the highest potential (Mn II in S₋₁ ⁶⁰). Our results suggest that upon formation of S₁, oxidation of Mn3 from the III valence to IV favors the release of the proton from OH⁻ at O4 (Figure 8). From ENDOR, EPR and simulation of the EXAFS, it has been proposed that a Mn–Mn distance of ~2.85 Å in the S₀ state is decreased to ~2.7 Å in the S₁ state (2.83 to 2.73 Å ¹⁶, 2.85 to 2.72 Å ¹⁹, or 2.85 to 2.75 Å ²⁰). In the present calculations, Mn3–Mn4 was the only distance that was decreased by ~0.1 Å from 2.86 Å in S₀ to 2.70 Å in S₁ (Table 2 and Supplementary Table 4). Note that the corresponding change was 2.90, a slightly longer distance, to 2.71 Å when O5 was assumed to be the deprotonation site ³⁴. These results suggest that the decreased Mn–Mn distance reported in EXAFS studies ^16-21 corresponds to the Mn3–Mn4 distance, which is decreased by deprotonation of O4 in the S₀ to S₁ transition.

Figure 8. Oxidation states of the four Mn ions and protonation states of the O atoms.

The release of a proton from O4 occurs due to oxidation at O4 via the single-well H-bond with W539 upon formation of S₁.

Table 2.

Mn–Mn distances (Å).

	S0: pre-PT (O4 = OH−)	S1: post-PT (O4 = O2−)
Mn1-4	III,IV,IV,III	III,IV,IV,III
Mn1–Mn2	2.75	2.75
Mn2–Mn3	2.76	2.76
Mn1–Mn3	3.36	3.28
Mn3–Mn4	2.86	2.70a
Mn1–Mn4	4.74	4.81

^aEssentially consistent with the value of 2.71 Å in S₁ (O4 and O5 = O²⁻) in other studies (e.g., ³⁴).

Mn3–Mn4 is a key distance that can allow us to evaluate the reduction state of PSII crystal structures with respect to geometries obtained from ENDOR, EPR, or simulation of the EXAFS ^16-21 in both S₀ and S₁. The Mn3–Mn4 distance of 2.87 Å in the free electron laser structure ³⁶ , which was reproduced (2.91 Å with OH⁻ at O5 in S₁) in QM/MM studies by Shoji et al. ⁶¹, is close to the distance of S₀ (~2.85 Å), not the distance of S₁ (~2.73 Å), obtained from ENDOR, EPR, and EXAFS. This is in line with the presence of significant amount of the reduced S₀ state suggested from recent QM/MM studies by Askerka et al. ³⁸. It seems likely that the geometry of the free electron laser structure does not represent a pure S₁ state detected ENDOR, EPR, and EXAFS ^16-21 and is probably in a more reduced state.

An argument for the presence of a hydroxide ion at O5 for the free electron laser structure proposed by Suga et al. ³⁶ was the observation of significantly lengthened distances between O5 and the adjacent Mn ions. The presence of a hydroxide ion at O5 is likely, because recent QM/MM studies by Shoji et al. have reproduced two of the three significantly lengthened distances between O5 and the adjacent Mn ions, Mn1–O5 (2.70 ³⁶ and 2.73 ⁶¹ Å) and Mn4–O5 (2.33 ³⁶ and 2.34 ⁶¹ Å), by assuming OH⁻ at O5 in S₁ ⁶¹. However, it should also be noted that one of the three significantly lengthened distances in the free electron laser structure, Mn3–O5, could not be reproduced in this study ⁶¹; in fact Mn3–O5 was significantly shortened in the QM/MM geometry (2.20 ³⁶ to 1.96 ⁶¹ Å). Indeed, the 1.9-Å structure has an even longer Mn3–O5 distance of 2.38 Å ¹⁴. Notably, a similar Mn3–O5 distance of 2.37 Å was obtained, assuming reduced Mn(II) for Mn4 in DFT studies by Petrie et al. ⁶².

These features of the free electron laser structure, (i) the lengthened Mn3–Mn4 distance with respect to S₀ of ENDOR, EPR, and EXAFS ^16-21 and (ii) the lengthened Mn3–O5 distance with respect to the QM/MM geometry (S₁ with OH⁻ at O5 ⁶¹), suggest that Mn₄CaO₅ of the free electron laser structure is probably reduced and cannot be simply explained as representing a single oxidation (and protonation) state.

DISCUSSION

Based on the findings reported here, we are able to propose a mechanism for PT along the O4-water chain. (Note that the role of the O4 water chain as a proton channel as proposed here does not exclude a role as a substrate channel on other steps of the cycle.) On the S₀ to S₁ transition deprotonation of a μ-hydroxo group at the O4 position occurs due to oxidation of Mn3(III) to Mn3(IV) (Figure 8). This deprotonation results in a decrease in the Mn3–Mn4 distance from 2.86 to 2.70 Å (Table 2 and Supplementary Table 4); a similar decrease in the Mn–Mn separation has been reported in ENDOR, EPR, or EXAFS studies ^16,19,20. The nature of the O4-water chain, being composed exclusively of water molecules, is consistent with and may explain the pH-independence of proton transfer in the S₀ to S₁ transition ^21,48,49. At the start of the O4-water chain, the proton in the barrier-less O4–O_W539 H-bond is likely to move away from the positively charged S₁ when it forms. Thus, the O4–O_W539 H-bond is energetically less stable in S₁, resulting in the sequential (Figure 2) and downhill (Figure 4) proton transfer reaction. The presence of the O4-water chain may explain why formation of the S₁ state is less inhibited by Cl⁻ depletion ⁶³. It may also explain the apparently irreversibility of S₀ to S₁ step ⁵⁹ and the redox potential of S₀/S₁ being uniquely low, lower than the TyrD redox potential ^56-58, resulting in the conversion of S₀ to S₁ in the dark ^55,56.

It has been suggested that in bacteriorhodopsin, the proton transfer occurs to the Schiff base along a chain of three water molecules, by transforming from the pre-PT to post PT patterns ⁶⁴. As far as we are aware the O4-water chain is the longest PT channel of water molecules (~8 water molecules) identified in protein crystal structures. This unusually long and straight water chain, which seems to function as a proton transfer pathway in the water-oxidizing enzyme, is worthy of study in its own right.

METHODS

Coordinates and atomic partial charges

The atomic coordinates of PSII were taken from the X-ray structure of PSII monomer unit “A” of the PSII complexes from Thermosynechococcus vulcanus at a resolution of 1.9 Å (PDB code, 3ARC) ¹⁴. Hydrogen atoms were generated and energetically optimized with CHARMM ⁶⁵, whereas the positions of all non-hydrogen atoms were fixed, and all titratable groups were kept in their standard protonation states (i.e., acidic groups were ionized and basic groups were protonated). For the QM/MM calculations, we added additional counter ions to neutralize the entire system. Atomic partial charges of the amino acids were adopted from the all-atom CHARMM22 ⁶⁶ parameter set. The atomic charges of cofactors were taken from our previous studies of PSII ⁴³.

QM/MM calculations

We employed the electrostatic embedding QM/MM scheme, in which electrostatic and steric effects created by a protein environment were explicitly considered, and we used the Qsite ⁶⁷ program code, as used in previous studies ⁴³. We employed the unrestricted DFT method with the B3LYP functional and LACVP** basis sets. To analyze H-bond potential energy profiles of the O4–O_W539 bond, the QM region was defined as [the Mn₄CaO₅ cluster (including the ligands), water molecules shown in Figure 1, side chains of D1-Asp61, D1-Asn338, D2-Asn350, and CP43-Thr335, and backbones of D1-Asp61, D1-Asn335, D1-Ala336, D2-Asn350, and CP43-Gly338], whereas other protein units and all cofactors were approximated by the MM force field. To analyze all of the H-bonds along the O4-water chain (Figure 1) in the pre-PT conformation, a slightly smaller QM region was defined as [the Mn₄CaO₅ cluster (including the ligands), water molecules shown in Figure 1, and the side chain of CP43-Thr335] for efficiency (small QM; Supplementary Tables 2 and 3). The two QM regions did not essentially alter the results and conclusions. The resulting geometries were essentially identical regardless of the size of the QM region (see Supplementary Data 1). The results obtained on the basis of the large QM region were described in the main text. To analyze all of the H-bonds along the O5 path (Figure 1) in the pre-PT-like conformation, the QM region was defined as [the Mn₄CaO₅ cluster (including the ligands), water molecules of W1, W2, W3, W4, W539, W538, W446, and W442, side chains of D1-Asp61, D1-Asn181, and D2-Lys317, and a chloride ion (Cl-1)]. The geometries were refined by constrained QM/MM optimization. Specifically, the coordinates of the heavy atoms in the surrounding MM region were fixed at their original X-ray coordinates, while those of the H atoms in the MM region were optimized using the OPLS2005 force field. All of the atomic coordinates in the QM region were fully relaxed (i.e., not fixed) in the QM/MM calculation. Note that the resulting atomic coordinates of the QM region essentially did not alter upon full relaxation of the entire atoms (including heavy atoms) in the surrounding MM region ⁶⁸. The cluster was considered to be in the S₁ state with ferromagnetically coupled Mn atoms; the total spin S = 7 and the resulting Mn oxidation state (Mn1, Mn2, Mn3, Mn4) = (III, IV, IV, III) (see Supplementary Discussion for further details). The potential-energy profiles of H-bonds were obtained as follows: first, we prepared the QM/MM optimized geometry without constraints and used the resulting geometry as the initial geometry. The H atom under investigation was then moved from the H-bond donor atom (O_donor) towards the acceptor atom (O_acceptor) by 0.05 Å, after which the geometry was optimized by constraining the O_donor–H and H–O_acceptor distances. The energy of the resulting geometry was calculated. This procedure was repeated until the H atom reached the O_acceptor atom. See SI for the atomic coordinates of the QM region (i.e., Mn₄CaO₅). As discussed later, OH⁻ at O5 has no direct-H-bond in the protein environment of PSII. Thus, to analyze the proton transfer from O5, we assumed W2 (O5…O_W2 = 3.1 Å ¹⁴) as the plausible proton acceptor, and analyzed the energy profile by constraining the H–O_W2 distance.