The first tyrosyl radical intermediate formed in the S2–S3 transition of photosystem II†

The EPR ‘‘split signals’’ represent key intermediates of the S-state cycle where the redox active D1-Tyr161 (YZ) has been oxidized by the reaction center of the photosystem II enzyme to its tyrosyl radical form, but the successive oxidation of the Mn4CaO5 cluster has not yet occurred (SiYZ ). Here we focus on the S2YZ state, which is formed en route to the final metastable state of the catalyst, the S3 state, the state which immediately precedes O–O bond formation. Quantum chemical calculations demonstrate that both isomeric forms of the S2 state, the open and closed cubane isomers, can form states with an oxidized YZ residue without prior deprotonation of the Mn4CaO5 cluster. The two forms are expected to lie close in energy and retain the electronic structure and magnetic topology of the corresponding S2 state of the inorganic core. As expected, tyrosine oxidation results in a proton shift towards His190. Analysis of the electronic rearrangements that occur upon formation of the tyrosyl radical suggests that a likely next step in the catalytic cycle is the deprotonation of a terminal water ligand (W1) of the Mn4CaO5 cluster. Diamagnetic metal ion substitution is used in our calculations to obtain the molecular g-tensor of YZ . It is known that the gx value is a sensitive probe not only of the extent of the proton shift between the tyrosine–histidine pair, but also of the polarization environment of the tyrosine, especially about the phenolic oxygen. It is shown for PSII that this environment is determined by the Ca ion, which locates two water molecules about the phenoxyl oxygen, indirectly modulating the oxidation potential of YZ.


Introduction
All oxygenic life on Earth is sustained by biological water oxidation performed in higher plants, algae and cyanobacteria. [1][2][3][4][5] Understanding and mimicking this process in artificial systems is a central challenge for energy research. [6][7][8][9][10][11] Photosystem II, the enzyme responsible for water splitting, functions by coupling oneelectron photochemical charge separation with the four-electron process of water oxidation by storing oxidizing equivalents (''electron holes'') on the inorganic Mn 4 CaO 5 cluster known as the oxygen evolving complex (OEC), the active site of catalysis. The OEC cycles through five oxidation states S i , where i = 0-4 indicates the number of stored oxidizing equivalents. S 4 is not observable and spontaneously decays to S 0 with evolution of triplet dioxygen. As shown in Fig. 1, absorption of sunlight by PSII results in charge separation at the reaction center of the enzyme, with the electron being transported through several cofactors to the second exchangeable acceptor, the plastoquinone Q B . The highly oxidizing radical cation (P680 + ) is re-reduced by D1-Tyr161 (Y Z ), which subsequently extracts one electron from the Mn 4 Ca cluster, advancing it to the next S i state of the catalytic cycle. Y Z oxidation by P680 + occurs on a 1-10 ns timescale. The formed transient radical state (Y Z ) then decays on the ms to ms timescale, with the concomitant oxidation of the Mn 4 CaO 5 cluster under physiological conditions. Cryogenic temperatures (o20 K) alter this single electron progression. At these temperatures, electron donation to P680 + by Y Z can still occur in a minority of centers (o50%), but the subsequent electron transfer from the Mn 4 CaO 5 cluster to Y Z is blocked, arresting catalytic progression from S i Y Z to S i+1 Y Z . 12 [30][31][32] acetate treated PSII, 13,33 and in PSII poised at high pH. 34,35 It is also noted that in the higher S-states (S 2 , S 3 ) infrared light excitation of the Mn 4 CaO 5 cluster also allows the ''backwards'' cycling, i.e. inducing the one electron oxidation of Y Z by the Mn cluster. 15,20,[36][37][38] From the perspective of electron paramagnetic resonance (EPR) spectroscopy, the tyrosine radical is not isolated from the Mn 4 CaO 5 cluster. 39,40 This means that the tyrosyl radical senses the structure of the Mn 4 CaO 5 cluster, and hence each metastable S i Y Z state has a characteristic spectral lineshape. Importantly, these ''metallo-radical'' states have the potential to resolve the concerted structural changes that occur in-between S-state transitions, which are coupled to cluster deprotonation and substrate binding. [41][42][43][44][45] This is particularly important for the most complicated S-state transition, from S 2 to S 3 . While experimental and theoretical efforts have produced a quite robust understanding of the S 2 state itself, 2 the S 3 state is arguably the least well understood state of the cycle, with even the nature of the oxidation event, Mn-centered vs. ligand-centered oxidation, still being debated. [46][47][48] As shown in Fig. 1, S 2 -S 3 is a complex transition that must involve several intermediates because it likely combines water binding 43 with deprotonation of the OEC. 41,42 It is the most crucial step of the cycle, because it activates the catalyst allowing the O-O bond formation step to occur. The precise sequence of events within the S 2 -S 3 transition is currently unknown and several strands of observations must be rationalized simultaneously for it to be understood. 49 A quantum chemical description of the electronic structure of the split signal states has been lacking because of (a) the large size of molecular models which must include both the OEC with its first and second coordination spheres and the Y Z residue with its environment and (b) the sufficiently high levels of theory required to obtain an adequate description of the electronic structure of the coupled system and to predict its spectroscopic properties. In the present work we employ large quantum cluster models to describe the S 2 Y Z state. We report its electronic structure and spectroscopic properties, including the g-tensor of the tyrosine radical. The present results suggest deprotonation of a terminal water ligand as the most likely next step in the S 2 -S 3 transition. We also resolve the influence of hydrogen bonding on the phenolic oxygen of the Y Z , suggesting a role for the Ca 2+ ion in modulating the oxidation potential of the tyrosine by structuring its local hydration environment.

Methodology
Optimized models of the S 2 state were used as starting points for geometry optimizations of the one-electron oxidized species. 50 The structures contain all first-sphere residues of the Mn 4 CaO 5 core, second-sphere residues that hydrogen-bond to the inorganic core or its first-sphere residues, several vicinal water molecules and the D1-Tyr161-His190 pair, along with complete backbone loops where necessary. The BP86 functional 51,52 was employed for optimizations, while the TPSSh functional 53,54 was used for single-point calculations of different spin configurations in the context of broken-symmetry density functional theory [55][56][57][58][59][60][61] and for the calculation of all spectroscopic properties reported in this work. D3 dispersion corrections 62 were used throughout, and the influence of the environment was simulated using the conductor-like screening model (COSMO) 63 assuming a dielectric constant of e = 8, as in previous studies. The scalar relativistic effects were included with the zeroth order regular approximation (ZORA) 64,65 with one-center terms and using ZORA-recontracted polarized triple-z basis sets. 66,67 Fully decontracted auxiliary basis sets were used for the resolution of the identity (RI) and chain-of-spheres (COSX) approximations 68 for the Coulomb and exact exchange, as implemented in ORCA. 69 Fine general integration grids with additionally increased radial accuracy and tight convergence criteria were used for all calculations. The extraction of pairwise exchange coupling constants from the individual broken-symmetry solutions and calculation of the complete energy ladder of spin eigenstates through diagonalization of the Heisenberg Hamiltonian followed established procedures. 50,[70][71][72][73][74][75][76][77] The same level of theory was used for the calculation of the magnetic and spectroscopic parameters. The g-matrix describes the coupling between the molecular magnetic moment and the external field. In the framework of DFT and Hartree-Fock theory the g-matrix can be evaluated using a coupled-perturbed selfconsistent field approach. 78 For the evaluation of the spin-orbit coupling operator we used an efficient implementation of the spin-orbit mean-field approximation to the Breit-Pauli operator. 79 In determining the g-matrix components of the Y Z radical we followed the pragmatic approach of substituting the open-shell Mn ions of the cluster with diamagnetic ions of the same formal charge (Ga 3+ and Ge 4+ ), as the interaction between Y Z and the Mn cluster is small (less than 500 MHz). We note that more elaborate approaches have been proposed for the calculation of g-values in magnetically strongly coupled dimers. 80 In the present case, namely the very weak interaction between the hypothetical spin pair, the Y Z and the Mn cluster, only the electrostatic influence of the inorganic cluster on the tyrosine needs to be considered and can be fully captured using diamagnetic ion substitution.

Results and discussion
3.1 The S 2 state prior to the oxidation of Y Z The models used in the present work are based on previously optimized models for the S 2 state of the OEC. 50 Recent efforts towards refining the 1.9 Å resolution crystallographic model 81 of PSII for the S 2 state led to the realization that the S 2 state of the cluster can exist in two interconvertible and approximately isoenergetic forms. These two forms of S 2 , which can be described as ''open cubane'' (S 2 A ) and ''closed cubane'' (S 2 B ) owing to their core connectivity (see Fig. 2), are valence isomers that differ in the position of the unique Mn(III) ion. This difference in the oxidation state distribution results in distinct magnetic topologies and hence distinct ground spin states for the two structures (S = 1/2 for the open cubane and S = 5/2 for the closed cubane). These two states correspond to the wellknown S 2 state EPR signatures, the g E 2.0 multiline signal for the open cubane and the g E 4.1 signal for the closed cubane, with the computed hyperfine coupling constants of the metal ions and the first coordination sphere ligands matching experiment. These two structures also rationalize a series of recent experiments aimed at determining the protonation states of Mn-bound water-derived ligands and identifying one of the substrate oxygen atoms. 2,73,74,76,[82][83][84] The protonation state assignment for the cluster in the S 2 state is as follows: none of the five oxo bridges is protonated (O1-O5) and all four terminal waterderived ligands (W1 and W2 on Mn4, W3 and W4 on Ca 2+ ) represent H 2 O ligands, with the exception of W2, which is bound as OH À . 72 These experimentally validated models represent a welldefined starting point for the search of the corresponding split signal state. This was initiated by oxidation of the open and closed cubane forms, thus mimicking the first step upon illumination of PSII poised in the S 2 state, i.e. the oxidation of Y Z by P680 + . This forms S 2 Y Z , where the tyrosine radical is weakly coupled to the S 2 -state Mn cluster (i.e. with the same Mn oxidation states as in S 2 ). As mentioned in the Introduction, this triggers a series of events that results in the formation of the S 3 state, which involves in an unspecified sequence: (i) the reduction of the Y Z radical by either the last remaining Mn(III) ion of the cluster to yield an all-Mn(IV) configuration or the oxidation of a first coordination sphere ligand; (ii) the binding of the second H 2 O substrate molecule; and (iii) the loss of a proton either from the incoming H 2 O molecule or from a preexisting titratable group. As described in the Introduction, under physiological advancement Y Z oxidation must occur prior to OEC deprotonation as Y Z oxidation by P680 + is 2-3 orders of magnitude faster than OEC deprotonation. However, it has been suggested 25   Y Z formation at this temperature is thought to trigger cluster deprotonation, as the protein is flexible enough for proton transfer but not enough to allow S 3 formation, which requires a protein conformational change. 24,30,31,85 While we cannot exclude this possibility, we consider it unlikely. The S 2 state multiline EPR signal is the same with or without this pre-illumination. As demonstrated by Ames et al., 72 deprotonation of the S 2 -state cluster should clearly manifest itself in terms of an altered 55 Mn-hyperfine structure, but no such drastic change is observed experimentally. Therefore, we assume that no modification of the OEC needs to occur to trap the first S 2 Y Z intermediate state, but instead occurs at a later stage (see Section 3.4).

Geometric structure of the S 2 Y Z states
The optimized open and closed cubane forms of the S 2 -state (S 2 A and S 2 B ) described above 50  3.3 Electronic structure of the S 2 Y Z states Fig. 3(a) shows the frontier spin-a and spin-b orbitals of the S 2 state of the OEC in the multiline g E 2 (S 2 A ) conformation. For both spin manifolds, the HOMO À 1 is a p-type orbital localized In the context of this one-particle picture, the tyrosine-based HOMO should be the orbital involved in the one-electron oxidation of the system. This is precisely what the calculations on the S 2 Y Z states demonstrate: Fig. 3 It is stressed that no valence isomers could be found where the Mn(III) ion is oxidized instead of Y Z . This suggests that the S 2 state cannot proceed to the S 3 state but instead is arrested in the S 2 Y Z state. It is also important to note that the S 2 state structures were developed using backbone constraints from X-ray crystallographic data and are fully consistent with the EPR results which constrain the protonation state. Thus our results require a conformational change during the S 2 to S 3 transition to allow the oxidation of the cluster, in line with earlier experimental results. It is noted that if the Tyr161-His190 pair is absent from the S 2 state model, the HOMO is fundamentally different by construction from the one described above. Oxidation of such an S 2 -state model without further modifications (water binding and/or deprotonation) leads, by definition, to an ''unphysical'' oxidation event that cannot have any correspondence to the natural system.

Y Z oxidation reorients the dipole moment of the OEC
An important question regarding the mechanistic details of the S 2 -S 3 transition is what happens after the oxidation of Y Z by P680 + and the formation of S 2 Y Z . Although the next steps are not explicitly studied in the present work, an important observation is that the dipole moment of the model is reoriented after Y Z oxidation. Fig. 5 compares the dipole moment vector before and after oxidation of Y Z . It can be seen that in the ''split signal'' state the dipole moment is directed in such a way that it points approximately from the cationic imidazolium of His190 to Asp61. The same result is seen for both S 2 A Y Z and S 2 B Y Z , suggesting that this region of the OEC is now the locus of the negative charge regardless of the formal oxidation state of Mn4.
In the context of the putative requirements for progression to the S 3 state (see Fig. 1), this dipole reorientation can be interpreted as indicating the likely direction of proton removal from the system. Given that the dipole is almost coincident with the Mn4-W1 bond and Asp61, it is likely that Y Z oxidation triggers the loss of a proton from W1 to the acceptor Asp61. This was suggested also in previous computational studies [91][92][93] and would be consistent with the proposal that: (i) proton release occurs prior to formation of the S 3 state 42 and (ii) Asp61 participates in a proton-transfer pathway involving the proximal chloride ion. [94][95][96][97] Note that a continuous network of hydrogen bonds formed by a chain of water molecules between Y Z and Mn4 establishes a communication pathway across the cluster. 89,98-101

Magnetism of the S 2 Y Z states
The magnetism of each S 2 Y Z model is defined by Mn-Mn ion exchange interactions within the inorganic cluster, and the tyrosine-OEC interaction which is orders of magnitude smaller. For the two configurations discussed above (S 2 A Y Z and S 2 B Y Z ) an overdetermined system of equations, which is derived from the set of broken-symmetry solutions, can be solved by singular value decomposition to yield pairwise exchange coupling constants, J ij , shown in Fig. 6. Based on these values, diagonalization of the Heisenberg Hamiltonian then yields the complete spectrum of energy levels produced by the coupling of the four  It is noted though that the closed cubane conformation may be associated with the NIR-induced split signals (see Introduction). Turning now to the coupling between the inorganic cluster and the oxidized tyrosine, according to simulations of EPR spectra, this interaction is of the order of À400 MHz. 13 Since this value is very small, it becomes important to eliminate all numerical noise from the calculations and for this reason a higher threshold for convergence was used for the computed broken-symmetry energies. Nevertheless, when confronted with such energy differences, agreement of the calculation with experiment even in terms of the order of magnitude should be considered satisfactory. In the present case the calculated values for the coupling of Y Z with the manganese cluster are À819 MHz for S 2 A Y Z and À626 MHz for S 2 B Y Z , when the lowest-energy spin configuration is used in the broken-symmetry calculations for each of the manganese clusters (7abba4, M S = 1/2 for the open cubane and 7aaab4, M S = 5/2 for the closed cubane form). If instead the high-spin (M S = 13/2) configuration of the clusters is used, then the corresponding values become À149 MHz for S 2 A Y Z and À324 MHz for S 2 B Y Z . In either case the models used in this work correctly capture the essential physics of the interaction.

The g-matrix of the Y Z radical
The calculation of the g-matrix in tyrosine radicals has been attempted previously using various approaches. The majority of these initial studies used simplified models such as the phenoxylwater or phenoxyl-imidazolium pair to obtain structural and spectroscopic properties of the tyrosine radical. [102][103][104][105][106][107][108] Recently, more elaborate models of PSII that include the surrounding residues were used to study the histidine-tyrosine pair for both Y Z and Y D . 88,89,109,110 In addition to considering the effects of the protein on the calculated properties, these larger models have allowed the environment of the two redox active tyrosine residues of PSII (Y D and Y Z ) to be differentiated. Our g-matrix calculations match earlier model system studies regarding the g orientation with respect to the phenoxyl ring: the g x axis is oriented almost perfectly along the C-O bond (y = 3.81) and g z is perpendicular to the plane of the ring (Fig. 7). The values of the individual g-matrix components are identical for the two forms of the cluster, indicating the absence of any effect derived from the different magnetic topologies of the inorganic core.  The dependence of the g-matrix elements on the position of the proton between the Y Z -His190 pair is depicted in Fig. 8. The g x component shows the strongest dependence: the largest calculated value corresponds to the largest OÁ Á ÁH distance (g x = 2.0057), while the smallest value is obtained for the structure in which the hydrogen is closest to the oxygen atom of the Y Z radical (g x = 2.0041). For the g y and g z components no such dependence is observed, in agreement with the previous tyrosyl radical g-tensor studies. EPR studies of Y Z have been performed by replacing Y D by phenylalanine and removing the Mn cluster (see ref. 39 and references therein for a review of the relevant literature). In contrast to Y D , the g-matrix anisotropy of Y Z (in the absence of the OEC cluster) has only recently been measured. 108 The values of g x = 2.00714 in frozen solution and 2.00705 in single crystals have been reported, 108 while the experimental values of g y and g z match the computed ones of Fig. 7a 114 resulting from the combined influence of three hydrogen bonds. 115 Similarly, in phenoxyl radical compounds that contain a strong intramolecular hydrogen bond the g x values fall within 2.0063-2.0067. 116 In the present case the low g x value of 2.0054 for S 2 Y Z can be interpreted as being due to the existence of three hydrogen-bonding interactions, two from adjacent water molecules (see Fig. 7b) and one from the proton that shifts to the imidazole of His190, which remains at a distance of approximately 1.6 Å from the O atom of Y Z .
To test this hypothesis, three additional calculations were carried out on fragments of the full model. First, the inorganic part of the model and all its associated ligands were removed, leaving a model containing only the Y Z -His190 pair and the two water molecules that form direct hydrogen bonds to the phenoxyl oxygen, the calcium-bound W4 (HOH540 in the 3ARC PDB structure) and the HOH542 that forms a hydrogen-bonding bridge between the other calcium-bound water (W3 or HOH541) and the tyrosine. The structure was not allowed to relax, essentially isolating the effect of the inorganic cluster while retaining the immediate environment of the tyrosine. As shown in Table 1, the effect of the Mn cluster on the low g x value is insignificant. Removing also the HOH542 molecule results in an increase of g x value to 2.0063. However, the largest effect was observed when both H-bonding water molecules were removed; the model that contains only the two amino acid residues displays a marked increase in the g x value to 2.0072, showing how the local electropositive environment plays a crucial role in suppressing the expected effect on g x of the proton shift to His190.
As demonstrated previously for a series of synthetic phenoxyl radical models, the value of g x tracks closely the change in the unpaired spin population on the oxygen atom (Table 1 and Fig. S3, ESI †), the atom with the largest spin-orbit coupling constant. 116 An almost linear correlation is observed between the O spin population and g x , with R 2 = 0.993. The change in spin population can be interpreted as the result of increased hydrogen bonding ''pulling'' the electron density towards the oxygen and hence ''pushing'' the unpaired density over the phenyl ring, an effect also seen for the in-plane proton alone (Fig. 8). A detailed discussion of this effect is provided by Sinnecker et al. for hydrogen bonding in semiquinones 117 and for the case of the primary quinone (Q A /Q A À ) in bacterial reaction centers. 118 It is noted that the larger value of g x for the model where the hydrogen-bonded water molecules are removed agrees with the experimental value reported for Y Z in Mn-depleted PSII and with that reported for the Y D radical. The above observation regarding g x implies that the removal of the OEC cluster in the experiments also perturbs the native hydration environment of Y Z . As a corollary, and given the highly ordered hydrogen bonding network shown in Fig. 7, it can be suggested that in addition to other roles, 3,119-121 the Ca 2+ ion serves to structure the water molecules in that region and adjust their acidities in order to (a) modulate the electronic structure and hence the redox potential of the tyrosine residue and (b) optimize the hydrogen-bond-mediated communication between Y Z and the inorganic cluster. Given that the hydrogen bonding network into which Y Z is embedded must be highly optimized for a proper function, even small perturbations on the structure and pK A value of the constituent water molecules may have a significant impact on the function of the catalyst. The role of the structured water environment in modulating the redox potential of the tyrosyl   Fig. S4, ESI †). The above suggestion for the role of the Ca 2+ ion is consistent with the measurements of frequency shifts of the proximal peptide carbonyl bands as markers for hydrogen-bonding changes between the S 1 -S 2 transition in the presence of different cations. 122 It is also in line with the recent proposal of Boussac and coworkers regarding the Ca 2+ ion acting as an entropic regulator for the S 3 -S 0 transition based on its involvement in structuring the environment of Y Z in the S 3 state. 123

Conclusions
In this work we developed structural models for the S 2 Y Z ''split . In addition, no valence isomers could be found where the Mn(III) ion is oxidized instead of Y Z , implying that a chemical modification of the OEC is necessary for the subsequent S 2 to S 3 transition to occur. The reorientation of the dipole moment suggests that this most probably involves the deprotonation of W1. The calculations of the tyrosyl g-matrix provide estimates (g x = 2.0054, g y = 2.0042, and g z = 2.0022) that form a good predictive basis regarding the environment of the tyrosine. The response of the g-matrix components was examined with respect to the extent of proton transfer to the t-N of His190 and the contribution of different subsets of the model. The predicted g x value is due to the presence of three hydrogen bonding interactions in which calcium-bound water molecules are involved. Given that the existing experimental g-anisotropy values were obtained from measurements on Ca-depleted PSII samples, it is suggested that the measured values do not necessarily reflect the natural environment of Y Z because they presumably miss the structuring effect of the Ca 2+ ion which operates in the native system. Thus, the Ca 2+ ion, besides its possible role in adjusting the redox potential of the inorganic cluster itself, as also indicated in studies of synthetic complexes, 124 has an additional role; it organizes the water environment and optimizes the hydrogen-bonding network around Y Z . Since, as demonstrated by the present results, hydrogen bonding to W4 and HOH542 strongly affects the electronic structure of the tyrosyl radical through modulation of its spin and charge density distribution, this ordering effect of the Ca 2+ ion indirectly fine-tunes the function of the tyrosine residue by regulating its redox potential and electron transfer properties.