Structural and energetic insights into Mn-to-Fe substitution in the oxygen-evolving complex

Summary Manganese (Mn) serves as the catalytic center for water splitting in photosystem II (PSII), despite the abundance of iron (Fe) on earth. As a first step toward why Mn and not Fe is employed by Nature in the water oxidation catalyst, we investigated the Fe4CaO5 cluster in the PSII protein environment using a quantum mechanical/molecular mechanical (QM/MM) approach, assuming an equivalence between Mn(III/IV) and Fe(II/III). Substituting Mn with Fe resulted in the protonation of μ-oxo bridges at sites O2 and O3 by Arg357 and D1-His337, respectively. While the Mn4CaO5 cluster exhibits distinct open- and closed-cubane S2 conformations, the Fe4CaO5 cluster lacks this variability due to an equal spin distribution over sites Fe1 and Fe4. The absence of a low-barrier H-bond between a ligand water molecule (W1) and D1-Asp61 in the Fe4CaO5 cluster may underlie its incapability for ligand water deprotonation, highlighting the relevance of Mn in natural water splitting.

INTRODUCTION O 2 evolution occurs at the catalytic center, the Mn 4 CaO 5 cluster in photosystem II (PSII). 1 The Mn 4 CaO 5 cluster comprises the Mn 3 CaO 4 cubane region (Mn1, Mn2, Mn3, Ca 2+ , O1, O2, O3, and O5) and the dangling region (Mn4 and O4). 2 The most stable oxidation state, S 1 , is Mn(III) 2 Mn(IV) 2 in the high oxidation state model (corresponding to S 3 in the low oxidation model 3 ). Light-induced electron transfer to redoxactive D1-Tyr161 (TyrZ) increases the oxidation state of the Mn 4 CaO 5 cluster from S 1 via S 2 and S 3 to S 0 . The release of the proton occurs with a stoichiometry of 1:0:1:2 for the S 0 to S 1 , S 1 to S 2 , S 2 to S 3 , and S 3 to S 0 transitions. O 2 evolves during the S 3 to S 0 transition. 4 The release of the proton occurs along the O4-water chain in the S 0 to S 1 transition, [5][6][7] whereas it occurs via D1-Asp61 along the D1-Glu65/D2-Glu312 channel in the S 2 to S 3 transition. 8,9 D1-Asp61 forms an H-bond with a ligand water molecule at the dangling Mn4 site, W1. 2 S 2 is a characteristic intermediate state, as it proceeds to the S 2 to S 3 transition via proton-coupled electron transfer 7,8,[10][11][12][13][14][15][16][17] and substrate-water incorporation. [18][19][20][21][22] Electron spin echo envelope modulation (ESEEM) and electron nuclear double resonance (ENDOR) studies have suggested that all of the m-oxo bridges of the Mn 4 CaO 5 cluster are deprotonated in S 2 (i.e., Mn(III)Mn(IV) 3 ). 23,24 Theoretical studies suggest that S 2 has two distinct conformations, the open-and closed-cubane conformations with (Mn1, Mn2, Mn3, Mn4) = (III, IV, IV, IV) and (IV, IV, IV, III), respectively. The open-cubane S 2 conformation was identified in the X-ray free-electron laser (XFEL) structures, but not the closed-cubane S 2 conformation, [18][19][20][21][22] which may be due to the open-cubane S 2 conformation being energetically more stable than the closed-cubane S 2 conformation. [25][26][27][28] Recent theoretical studies performed in the presence of the PSII protein environment suggested that the g = 4.1 signal observed for plant PSII in electron paramagnetic resonance (EPR) spectroscopy 29,30 corresponds to the closed-cubane S 2 conformation with W1 = OH -. 31 Mn1(III) . O5 is long and Mn4(IV) . O5 is short in the open-cubane S 2 conformation, whereas Mn1(IV) . O5 is short and Mn4(III) . O5 is long in the closed-cubane S 2 conformation. 32,33 Theoretical studies proposed that the substrate water molecule could be incorporated into the O5 moiety. [10][11][12][14][15][16] According to the XFEL structures, a water molecule is incorporated into the O5 moiety during the S 2 to S 3 transition. [18][19][20][21][22] Notably, Cl À and Ca 2+ are required to proceed the S 2 to S 3 transition. The PSII crystal structure shows that a chloride ion, Cl-1, is located at D1-Asn181 and D2-Lys317. 2 When Cl À is depleted, the S-state transition is inhibited at the S 2 TyrZ , formation. 34 Indeed, electron transfer from S 2 /S 3 to TyrZ is energetically uphill due to an increase in the S 2 /S 3 redox potential in Cl À -depleted PSII. 35 A salt bridge forms between D1-Asp61 and D2-Lys317 upon the depletion of Cl -, 36,37 which may also inhibit proton transfer from W1 via D1-Asp61 toward the bulk region. The salt bridge and Cl À (Cl-1) are absent in D2-Lys317Ala PSII. Nevertheless, electron transfer can occur during the S 2 to S 3 transition even in the absence of Cl-1, 37 which suggests that Cl-1 is not necessarily required for the electron transfer process in the S 2 to S 3 transition.
The inhibition of the S 2 to S 3 transition in Cl À -depleted PSII resembles that in Ca 2+ -depleted PSII. [38][39][40][41] Ca 2+ depletion not only causes the alteration of the H-bond network at the Mn 4 O 5 and TyrZ moieties 26 but also decreases the redox potential of TyrZ significantly due to reorientation of the water molecules in the H-bond network, making electron transfer from the Mn 4 CaO 5 cluster to TyrZ uphill. 42 Replacement of Ca 2+ with any metals except Sr 2+ inhibits O 2 evolution. 38,[43][44][45][46] The geometry of the catalytic site in Sr 2+substituted PSII (Sr 2+ -PSII) resembles that of native PSII (Ca 2+ -PSII). 47,48 The E m values for the artificial clusters with Sr 2+ are also similar to those with Ca 2+ . [49][50][51] Thus, Ca 2+ is a prerequisite for the O 2 evolving activity in native PSII but can functionally be substituted with Sr 2+ . 52 In contrast, Mn is indispensable for the light-driven catalytic activity of the Mn 4 CaO 5 cluster in the PSII protein environment. In the artificial cluster with the Fe 5 O core, electrocatalytic O 2 evolution was reported. 53 However, no O 2 evolution was observed in the PSII membrane when two of the four Mn sites were substituted with Fe. 54 Understanding why Mn is chosen over Fe as the redox-active metal in the PSII catalytic center raises several fundamental questions: Why did Nature pick Mn and not Fe for the water-oxidation catalyst? Would there be any advantageous properties in the electronic state of Mn compared to Fe in the context of the PSII catalytic center? What would be the protonation states of the bridges and ligands, considering the evolutionary optimization for higher valence of Mn(III/IV)? What would be the activation energies for S state transitions? Would Fe also exhibit two characteristic conformations, corresponding to the open-and closed-S 2 conformations observed in the Mn 4 CaO 5 cluster? What would the active site look like if it were optimized by evolution to operate with Fe instead of Mn? Is there a chance that it would function effectively in that scenario? Some of these questions were put forwarded by Armstrong previously. 52 To the best of our knowledge, the detailed characterization of the Fe 4 CaO 5 cluster in the PSII protein environment has not been reported. Investigating the properties of the Fe 4 CaO 5 cluster in the PSII protein environment could provide valuable insights into the reasons behind the selection of Mn instead of the more abundant Fe 55 for the catalytic center. Here, we investigate the structural and energetic details of the Fe 4 CaO 5 cluster in the characteristic intermediate state, S 2 , using a quantum mechanical/molecular mechanical (QM/MM) approach, considering the entire PSII protein environment (i.e., all protein subunits and cofactors), and assuming that Mn(III/IV) corresponds to Fe(II/III). During QM/MM calculations, the decrease in the cluster net charge caused by substitution of Mn with Fe can be compensated for by the release of protons from titratable groups, 56 including adjacent water molecules, 57 if energetically favorable. The protonation states and oxidation states of the Fe 4 CaO 5 cluster are currently unknown, unlike the extensively studied Mn 4 CaO 5 cluster. By investigating these protonation states and oxidation states, our aim is to gain a deeper understanding of the unique characteristics and potential differences between Mn and Fe in the context of the water-oxidizing catalyst. Additionally, the analysis of the resulting protonation state of the Fe 4 CaO 5 cluster may provide insights into why only two Mn sites were replaced with Fe in the Ca 2+ -depleted PSII. 54

Protonation state of the Fe 4 CaO 5 cluster
To appropriately model and define the Fe 4 CaO 5 cluster, the presumed oxidation and protonation states of the Fe 4 CaO 5 cluster must be considered. While the Mn 4 CaO 5 cluster utilizes two typical valence states, Mn(III) and Mn(IV), it is assumed that the Fe 4 CaO 5 cluster employs Fe(II) and Fe(III).
The stable oxidation and protonation states of the Fe 4 CaO 5 cluster remain unknown. In most studies on the Mn 4 CaO 5 cluster, the lower S-states, namely, S 0 , S 1 , and S 2 , are defined based on the number of oxidized Mn ions (Mn(IV)), i.e., Mn(III) 3  When all of the m-oxo bridges of the Fe 4 CaO 5 cluster are deprotonated in the corresponding Fe(II)Fe(III) 3 state, CP43-Arg357 and D1-His337 release the protons toward O2 and O4, respectively ( Figure 1A, top). Consequently, the Fe 4 CaO 5 cluster is doubly protonated with OH À at O2 and O3 in S 2 , which partially compensates for the loss of the positive charge of Fe(II/III) with respect to Mn(III/IV) ( Figure 1B). The corresponding protonation events were also reported for the overreduced Mn 4 CaO 5 cluster. 57 Thus, the Fe 4 CaO 5 cluster may resemble the overreduced Mn 4 CaO 5 cluster.
Deprotonated CP43-Arg357 is likely to eventually reprotonate in equilibrium, as the pK a value of CP43-Arg357 is sufficiently high even in Mn-PSII. 58 In the Mn 4 CaO 5 cluster, the release of the proton preferentially occurs from the m-oxo bridges of the cluster (e.g., O4 in the S 0 to S 1 transition 5-7 ) with respect to the ligand water molecule (e.g., W1 in the S 2 to S 3 transition 8,9,[59][60][61]. The presence of the protonated O2 site in the Fe 4 CaO 5 cluster suggests that the release of the proton from the ligand water molecule cannot proceed even in S 2 . Thus, the protonation of the Fe 4 CaO 5 cluster is disadvantageous for water oxidation even if the loss of the net charge of the cluster is partially compensated for by the proton.

Effect of metal substitution on molecular symmetry
Although the average angles formed by the Fe, Ca, and O sites in the Fe 4 CaO 5 cluster and the corresponding angles in the Mn 4 CaO 5 cluster are the same, there is a notable difference in the angle distribution. The Fe 4 CaO 5 cluster exhibits a narrower angle distribution (89.3 G 9.2 ) than the Mn 4 CaO 5 cluster (89.9 G 14 ) ( Figure S1). This narrower angle distribution contributes to the Fe 4 CaO 5 cluster appearing more cubic and symmetric in its molecular structure when compared to the Mn 4 CaO 5 cluster (Figure 1). The contrasting angle distributions between the Fe and Mn clusters highlight the distinct molecular symmetries resulting from metal substitution.

Absence of the open-and closed-cubane S 2 conformations in the Fe 4 CaO 5 cluster
In the Mn 4 CaO 5 cluster, either Mn1 (closed-cubane conformation) or Mn4 (open-cubane conformation) is the reduced Mn(III) site in S 2 , as indicated by the spin density (Table 1). In the Fe 4 CaO 5 cluster, however, the spin density is at the same level ($4) for all four Fe sites. In addition, the spin state is equally distributed over the entire cluster, including the five O sites ( Table 1). The equal distribution of the spin state over several Fe sites may be in line with the cubic and symmetric Fe 4 CaO 5 structure when compared to the Mn 4 CaO 5 cluster (Figures 1 and S1). The equal distribution of the spin state suggests that the Fe sites cannot be fully oxidized due to partial oxidation of the O sites, which is disadvantageous for decreasing the pK a values of the substrate water molecules at the Fe moieties.
The spin exchange coupling between the two transition metals increases as the distance between the two metals decreases. 64 The Fe . Fe distances in the Fe 4 CaS 5 cluster are longer than those in the Fe 4 CaO 5 cluster, as the Fe . S distances are longer than the Fe . O distances (Table S1). However, the spin state is more equally distributed over the five S sites in the Fe 4 CaS 5 cluster than over the five O sites in the Fe 4 CaO 5 cluster (Table S2). Thus, the equally distributed spin state in the Fe 4 CaO 5 cluster cannot be explained by the short Fe . Fe distances.
The equal distribution of the spin state over several Fe sites is also observed in Fe 4 S 4 clusters F A and F B located at the typical ferredoxin-like binding motif (CxxCxxCxxxCP) in photosystem I, in which a mixed  64 However, it should also be noted that equal distribution of the spin state over several Fe sites is not a characteristic of all Fe complexes. For example, a mixed valence Fe 2.5+ . Fe 2.5+ pair has not been reported for the Fe 4 S 4 cluster F X located on the preudo-C 2 axis of the reaction center in photosystem I. 65 As the size of the Fe 4 CaS 5 cluster differs significantly from those of the Fe 4 CaO 5 and Mn 4 CaO 5 clusters (Table S1), below we focus on the Fe 4 CaO 5 and Mn 4 CaO 5 clusters ( Figures 1C and 1D) if not otherwise specified.
Intriguingly, the potential-energy profile for the O5 position along the Fe1 .   iScience Article binding site in the closed-cubane S 2 conformation (e.g., [10][11][12][14][15][16]. The absence of the two S 2 conformations suggests that O6 incorporation is unlikely to proceed in the Fe 4 CaO 5 cluster during the S 2 to S 3 transition. It should also be noted that Mn(III) has a low redox-potential value, as Mn(III) aqua species tend to undergo disproportionation to form Mn(II) and Mn(IV), 52 while Fe(II) does not exhibit similar disproportionation behavior. Additionally, Fe(IV) is highly oxidizing, making it challenging to stabilize a Fe(IV) species in the S state. 52 This difference in redox properties between Mn and Fe may also pose a disadvantage for the Fe 4 CaO 5 cluster compared to the Mn 4 CaO 5 cluster in terms of its stability and suitability for water oxidation.
Absence of the release of the proton from W1 in the Fe 4 CaO 5 cluster W1 at the dangling Mn4(IV) site forms a low-barrier H-bond in the Mn 4 CaO 5 cluster, facilitating the release of the proton from W1 via D1-Asp61 toward the lumenal protein surface 8,9,60,61 ( Figure 3B). However, the corresponding low-barrier H-bond does not form in the Fe 4 CaO 5 cluster. The proton of H 2 O at W1 is localized at the W1 moiety, which suggests that proton transfer from W1 to D1-Asp61 is significantly inhibited in the Fe 4 CaO 5 cluster with respect to in the Mn 4 CaO 5 cluster ( Figure 3A). It seems likely that the oxidation of the Fe sites does not provide a sufficient driving force for deprotonation of the ligand water molecule owing to the low valence, Fe(III), with respect to Mn(IV), inhibiting proton transfer.  Figure 4B). However, W3 donates H-bonds to W2 and D1-Glu189 in the Fe 4 CaO 5 cluster, leading to the deformation of the diamond-shaped cluster of water molecules ( Figure 4A). The observed deformation of the H-bond network due to the reorientation of W3 resembles the deformation of the H-bond network due to the reorientation of W3 in Mg 2+substituted PSII. 68 The Fe 4 CaO 5 cluster is slightly smaller than the Mn 4 CaO 5 cluster, as the Ca 2+ . O distances are shorter in the Fe 4 CaO 5 cluster than in the Mn 4 CaO 5 cluster (Table S1). The Mn 4 MgO 5 cluster in Mg 2+ -PSII is slightly smaller than the Mn 4 CaO 5 cluster in native PSII, as the Mg 2+ radius is smaller than the Ca 2+ radius. 68 These characteristics suggest that the reorientation of W3 and the deformation of the H-bond network of TyrZ observed in the two clusters originate from the large cavity space with respect to the original Mn 4 CaO 5 cluster.
The TyrZ . D1-His190 distance is slightly longer in the Fe 4 CaO 5 cluster (2.55 Å , Figure 4A) than in the Mn 4 CaO 5 cluster (2.52 Å , Figure 4B). Nevertheless, the potential-energy profile for the H-bond suggests that the formation of a low-barrier H-bond between TyrZ and D1-His190 is not impaired even in the Fe 4 CaO 5 cluster ( Figure 5). Thus, the presence of low-barrier H-bonds cannot be judged by the H-bond distances but can be judged only by the shape of the potential-energy curve, as suggested by Schutz and Warshel. 69 Although the redox potential of TyrZ is slightly affected by the S-state change, 70 Figures 4C and 4D). The slight increase in the Cl À . O W446 distance (3.10 Å for the Mn 4 CaO 5 cluster and 3.34 Å for the Fe 4 CaO 5 cluster) is due to the increase in the cavity space, which in turn is compensated for by the decreases in the Cl À . D1-Asn181 and Cl À . D2-Lys317 distances. It seems likely that the binding affinity of Cl À is at the same level in Mn-PSII and Fe-PSII.
It was reported that photoelectrochemical water oxidation in NaCl aqueous solution using most metal oxides, including Fe, suffers from the production of toxic HClO À due to the oxidation of Cl -. 73 HClO À production was suppressed only when using Mn oxides. 73 This is not the case for PSII. The present result indicates that the PSII protein environment does not allow Cl À to approach even the Fe 4 CaO 5 cluster ( Figure 4C). D2-Lys317 not only serves as the binding site but also increases the redox potential for Cl À/, , inhibiting oxidation of Cl À to HClO À . Instead, Cl À is required to proceed with the S 2 to S 3 transition, 34 i.e., ultimately O 2 evolution. In the S 2 to S 3 transition, electron transfer from the Mn 4 CaO 5 cluster to TyrZ is energetically uphill in the absence of Cl -. 35 Furthermore, removal of Cl À leads to the formation of a salt bridge between D1-Asp61 and D2-Lys317 in the proton transfer pathway, leading to inhibition of the release of the proton. 36,37 Based on these observations and the findings presented in this study, it can be inferred that PSII likely opted for Mn(III/IV) rather than Fe(II/III) primarily to lower the pK a values of the substrate water molecules (Figure 3). However, iScience Article the redox potential value for Mn(III/IV) in the Mn 4 CaO 5 cluster alone may not be sufficiently low to enable efficient electron transfer to TyrZ. 35 To overcome this energetic challenge, PSII employs Cl À , which effectively decreases the redox potential. This strategic utilization of Clions allows for downhill electron transfer. 35

Conclusions
As the spin state is equally distributed over the entire Fe 4 CaO 5 cluster in contrast to the Mn 4 CaO 5 cluster (Table 1), the difference in the valence state is unclear between the Fe1 and Fe4 sites. Thus, the open-cubane and closed-cubane S 2 conformations do not exist in the Fe 4 CaO 5 cluster (Figure 2A). Instead, both Fe1 . O5 (1.89 Å ) and O5 . Fe4 (2.24 Å ) distances are too short (Table S1) to allow an external water molecule (e.g., O6 [18][19][20][21][22] to enter the O5 moiety in the S 2 to S 3 transition. The low-barrier H-bond between W1 and D1-Asp61, which is observed in the Mn 4 CaO 5 cluster 8,60,61 ( Figure 3B), is absent in the Fe 4 CaO 5 cluster ( Figure 3A), inhibiting the release of the proton from W1 via D1-Asp61 toward the protein lumenal surface in the S 2 to S 3 transition. 9 The H-bond network of TyrZ shows the difference specifically in the orientation of W3 at Ca 2+ between the Fe 4 CaO 5 and Mn 4 CaO 5 clusters (Figures 4A and 4B). The deformation of the H-bond network does not affect the formation of the low-barrier H-bond between TyrZ and D1-His190 in the Fe 4 CaO 5 cluster ( Figure 5). The H-bond network of the Cl-1 binding site remains essentially unchanged upon the substitution of Mn with Fe, which suggests that harmful HClO -73 is unlikely to be generated in the Fe 4 CaO 5 cluster as long as the PSII protein environment exists. Based on these observations, the Fe 4 CaO 5 cluster may not function as a water-oxidation catalyst in the binding site of PSII primarily because it cannot support in this environment the efficient deprotonation of the substrate water molecules (Figure 3) or the incorporation of an external water molecule into the cluster (Figure 2), the events occurring during the S 2 to S 3 transition of the natural Mn 4 CaO 5 catalyst.

Limitations of the study
The results depend on the original atomic coordinates of the crystal structures. The original side-chain orientations may affect the results, although the geometries of the catalytic sites are quantum-chemically optimized.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

DECLARATION OF INTERESTS
The authors declare no competing interest.