Structure, ligands and substrate coordination of the oxygen-evolving complex of photosystem II in the S 2 state: a combined EPR and DFT study †

The S 2 state of the oxygen-evolving complex of photosystem II, which consists of a Mn 4 O 5 Ca cofactor, is EPR-active, typically displaying a multiline signal, which arises from a ground spin state of total spin S T = 1/2. The precise appearance of the signal varies amongst diﬀerent photosynthetic species, preparation and solvent conditions/compositions. Over the past five years, using the model species Thermosynechococcus elongatus , we have examined modifications that induce changes in the multiline signal, i.e. Ca 2+ /Sr 2+ -substitution and the binding of ammonia, to ascertain how structural perturbations of the cluster are reflected in its magnetic/electronic properties. This refined analysis, which now includes high-field (W-band) data, demonstrates that the electronic structure of the S 2 state is essentially invariant to these modifications. This assessment is based on spectroscopies that examine the metal centres themselves (EPR, 55 Mn-ENDOR) and their first coordination sphere ligands ( 14 N/ 15 N- and 17 O-ESEEM, -HYSCORE and -EDNMR). In addition, extended quantum mechanical models from broken-symmetry DFT now reproduce all EPR, 55 Mn and 14 N experimental magnetic observables, with the inclusion of second coordination sphere ligands being crucial for accurately describing the interaction of NH 3 with the Mn tetramer. These results support a mechanism of multiline heterogeneity reported for species diﬀerences and the eﬀect of methanol [ Biochim. Biophys. Acta, Bioenerg. , 2011, 1807 , 829], involving small changes in the magnetic connectivity of the solvent accessible outer Mn A4 to the cuboidal unit Mn 3 O 3 Ca, resulting in predictable changes of the measured eﬀective 55 Mn hyperfine tensors. Sr 2+ and NH 3 replacement both aﬀect the observed 17 O-EDNMR signal envelope supporting the assignment of O5 as the exchangeable m -oxo bridge and it acting as the first site of substrate inclusion. N- N-three


Introduction
In oxygenic photosynthesis, light-driven water oxidation is catalysed by the oxygen-evolving complex (OEC) of the transmembrane protein complex photosystem II (PSII). The OEC consists of a m-oxo-bridged tetramanganese-calcium cofactor (Mn 4 O 5 Ca), embedded in its protein matrix. This matrix includes the redox-active tyrosine residue Y Z (D1-Tyr161), which couples electron transfer between the Mn 4 O 5 Ca cluster and P680 + , the photooxidant of the PSII reaction centre. The four-electron water oxidation reaction requires four consecutive light-induced charge separation events, driving the catalytic cycle of the OEC. 1,2 This cycle involves five redox intermediates, the S n states, where n = 0-4 indicates the number of stored oxidizing equivalents. All S-state transitions represent oxidations of the Mn 4 O 5 Ca cluster by Y Z with the exception of the regeneration of S 0 from S 4 , which proceeds spontaneously under the release of molecular triplet oxygen and the rebinding of at least one substrate water molecule. The ratelimiting step, oxidation of S 3 by Y Z , has prevented the transient, fast-decaying S 4 state from being characterized yet. For a general introduction into water oxidation by the OEC, see ref. 3-8. The structure of the Mn 4 O 5 Ca inorganic core 9 resembles a 'distorted chair' where the base is formed by a m-oxo-bridged cuboidal Mn 3 O 4 Ca unit (Fig. 1), while the fourth, 'outer' Mn, Mn A4 is attached to this core structure via two m-oxo bridges, O4 and O5. Spectroscopic evidence suggests that all five oxygen bridges are deprotonated in the S 2 state. [10][11][12][13][14][15] In addition to the m-oxo-bridged network, the Mn 4 O 5 Ca scaffold is held together by six carboxylate ligands, four of which (D1-Asp342, D1-Asp170, D1-Glu333 and CP47-Glu354) form m-carboxylato bridges between Mn sites. The cluster has only one immediate nitrogen ligand, the imidazole imino-N of D1-His332.
An understanding of the mechanism of water oxidation catalysis is intimately linked to an understanding of the electronic states of the paramagnetic tetranuclear Mn complex. Electron paramagnetic resonance (EPR) spectroscopy represents a powerful methodology in this endeavour. [16][17][18][19] The S 2 state, being readily generated from the dark-stable S 1 state, is the most extensively characterized S state by EPR. It is intrinsically heterogeneous, [20][21][22] existing in two ground state configurations. The lowest energy form exhibits a ground state of total spin S T = 1/2. This spin configuration gives rise to the well-known S 2 state EPR multiline signal centred at g E 2.0, with a hyperfine pattern of at least 24 peaks, with a line spacing around 87 G. 23 Depending on the conditions used, the S 2 state can also exhibit other broad EPR signals, centred at g Z 4.1, which have been assigned to (an) S T Z 5/2 spin state(s). 20,[24][25][26][27][28] The g Z 4.1 signal can also be induced by near-infrared illumination of the S 2 multiline state at temperatures r160 K. 24,25,29 Pulse electron nuclear double resonance (ENDOR) experiments, 30 especially at Q-band frequencies, [31][32][33][34] have enabled a more detailed examination of the electronic structure by uniquely determining four 55 Mn hyperfine interactions that give rise to the multiline pattern of the corresponding S T = 1/2 EPR signal. This ENDOR analysis has strongly suggested that the OEC contains one Mn III ion and three Mn IV ions when poised in the S 2 state. [30][31][32][33][34] OEC models developed from broken-symmetry (BS) density functional theory (DFT) have been crucial for the interpretation of EPR and related magnetic resonance results. 12,14,18,33,[35][36][37][38][39][40][41][42][43][44] These calculations now allow different S-state models to be assessed based on the spin multiplicities and relative energies of their lowest magnetic levels, and, by means of the spin projection formalism (see ref. 18 and 36), the metal and ligand hyperfine couplings. This strategy enables the assignment of individual Mn oxidation states and coordination geometries and represents a method to develop unified models of the OEC that combine constraints from X-ray diffraction, EXAFS and magnetic resonance.
Site perturbation of the OEC provides a means to further characterize the global geometric and electronic structure of the Mn tetramer and obtain information about localized changes associated with the introduced modification. The two best described OEC perturbations in the literature are (i) the replacement of the Ca 2+ with a Sr 2+ ion 33,[45][46][47] and (ii) the binding of NH 3 to the cluster. [48][49][50][51][52][53][54][55] Both of these perturbations do not inhibit functional turnover of the catalyst, but do modify the kinetics of O-O bond formation, substrate water uptake and its subsequent deprotonation. A brief description of these two modified OEC forms is given below: (i) Sr 2+ can be introduced into the OEC through chemical removal of the Ca 2+ ion 45,46,56 or alternatively by biosynthetic incorporation in cyanobacterial cultures grown in the presence of SrCl 2 instead of CaCl 2 . 47 Without Ca 2+ (or Sr 2+ ) the OEC is inhibited, [56][57][58][59][60][61][62][63] not progressing further than a modified form of the S 2 state, i.e. the S 2 0 Y Z state. [60][61][62] Sr 2+ is unique as it is the only ion which can replace the Ca 2+ ion while retaining catalytic activity, albeit at a lower enzymatic rate. 45,47,56 Presumably this is because Sr 2+ has a similar size and Lewis acidity as Ca 2+ . 64 While slowing the turnover rate of the catalyst, Sr 2+ substitution at the same time enhances the exchange rate with bulk water of at least one of the bound substrates, 65,66 as observed by timeresolved membrane inlet mass spectrometry (MIMS). 67,68 This behaviour suggests the Ca 2+ ion may play an important role in substrate water binding and possibly proton release (for reviews, see ref. 3 and 69).
(ii) Ammonia binding to the Mn cluster (in the presence of high Cl À concentrations) 48,50,51,54 only occurs upon formation of the S 2 state. It is subsequently released at some point during the S-state cycle (S 3 -S 0 -S 1 ), such that it is not bound upon return to S 1 . 51 As with Sr 2+ replacement, NH 3 binding does not inhibit catalytic function. In the higher plant electron spin echo envelope modulation (ESEEM) study of Britt et al., 70 NH 3 was shown to bind as a direct ligand of the Mn tetramer. The precise binding site and coordination mode of the NH 3 molecule was the subject of a recent study on cyanobacterial PSII from our laboratory. 44 By employing electron electron double resonance (ELDOR)-detected NMR (EDNMR), it was concluded that NH 3 replaces the water ligand of Mn A4 trans to the O5 bridge (W1, Fig. 1b). As the binding of NH 3 was also shown not to affect substrate exchange rates, these results suggest W1 does not represent a substrate water. One or more additional NH 3 binding sites, which are inhibitory, are known but are less well characterized. 49,50,71 Here, we present an extension of our earlier multifrequency EPR studies 33 55 Mn and 14 N magnetic spectroscopic observables for the native and the modified systems, a feature not achieved previously. The experimental results and calculations for 14 N/ 15 N ligands of the various S 2 state forms serve to prove that the basic electronic structure is not perturbed by these modifications, a result crucial for the interpretation of concomitant perturbations of the 17 O EDNMR signal envelope. This combined experimental and theoretical approach supports our qualitative model for multiline heterogeneity, demonstrating that the magnetic connectivity between the two subunits and also within the trimeric moiety governs the structure of the multiline signal. 34 This basic structural template also explains the apparent orientations of the 55 Mn hyperfine tensors, as inferred from spectral simulations and single crystal measurements, 32 and potentially provides a framework to further examine substrate binding. The different OEC forms represent a starting point to examine the energetics of higher S-state transitions, as they differ with regard to substrate binding and the kinetics of O-O bond formation and O 2 release.

PSII sample preparation
Ca 2+ -and Sr 2+ -containing PSII core complex preparations from T. elongatus 72 were isolated as reported before 47,73,74 with the same modifications for the X-band samples as described in ref. 44. Universal 15 N-labelling of the PSII proteins was achieved by growing the cyanobacteria in a modified BG11 or DTN medium that contained 15  . In order to ensure broadband microwave excitation and minimize distortions, the loaded quality factor Q L was lowered to 700 to obtain a microwave frequency bandwidth of 130 MHz. Electron spin echo-detected (ESE) field-swept EPR spectra were measured using the pulse sequence t p -t-2t p -t-echo, 77 three-pulse ESEEM spectra by use of t p -t-t p -T-t p -t-echo 78 and hyperfine sublevel correlation (HYSCORE) spectra by employing t p -t-t p -T 1 -2t p -T 2 -t p -t-echo. 79 The lengths of the p/2 microwave pulses were generally set to t p = 16 ns (X-band), 12 ns (Q-band) and 24 ns (W-band), respectively. For ESE-detected EPR experiments, inter-pulse distances were t = 260 ns (Q-band) and 300 ns (W-band). For the threepulse ESEEM measurements, multiple t values in the ranges t = 136-248 ns (X-band) and 200-356 ns (Q-band) and an optimum t = 260 ns for the HYSCORE experiments were chosen to account for blind-spotting artefacts. Q-band 55 Mn-ENDOR spectra were acquired employing the Davies-type pulse sequence t inv -t RF -T-t p -t-2t p -t-echo 80 using a length t inv = 24 ns for the p inversion microwave pulse and a radio frequency p pulse length t RF = 3.5 ms. The length of the p/2 microwave pulse in the detection sequence was generally set to t p = 12 ns and the inter-pulse delays to T = 2 ms and t = 268 ns. A shot repetition time of 1 ms was used for all measurements. EDNMR measurements were collected using the pulse sequence t HTA -T-t p -t-2t pt-echo. 81 The high turning angle (HTA) microwave pulse was applied at microwave frequencies n mw . The Hahn echo detection pulse sequence t p -t-2t p -t-echo, at a microwave frequency n (0) mw matched to the cavity resonance, was set at a sufficient time T after the HTA pulse to ensure near-complete decay of the electron spin coherencies. The p/2 pulse length used for detection was t p = 200 ns ( 14 N, 17 O) or 80 ns ( 15 N) and an inter-pulse separation of t = 500 ns was used. The echo was integrated E600 ns around its maximum. The spectra were acquired via continuously sweeping the HTA frequency n mw at a fixed magnetic field in steps of 78.1 kHz ( 14 N), 128.9 kHz ( 15 N) or 162.1 kHz ( 17 O). A HTA microwave pulse of length t HTA = 14 ms ( 14 N, 17 O) and 8 ms ( 15 N) and an amplitude o 1 = 12-16 Â 10 6 rad s À1 was used.

Spectral simulations
Spectra were fit assuming an effective spin S T = 1/2 ground state (Section S4.2, ESI †). The basis set that describes the 55 Mn tetramer-single electron spin manifold (eqn (1)) and the 14 N, 15 N and 17 O single nucleus-single electron spin manifolds (eqn (2)) can be built from the product of the eigenstates of the interacting spins: Here, M refers to the electronic magnetic sublevel, AE 1 2 ; I takes the values 5/2 for 55 Mn, 1 for 14 N, 1 2 for 15 N and 5/2 for 17 O; m i takes the values ÀI i , 1 À I i ,. . ., I i À 1, I i . The spin manifolds can be described by the following spin Hamiltonian: It contains (i) the Zeeman term for the total electronic spin, (ii) the hyperfine and (iii), except for the EPR spectra, nuclear Zeeman terms for either the metal 55 82 For further information on data processing, details of the simulations and theory, see Sections S2, S3 and S4 (ESI †), respectively.

DFT calculations
All calculations were performed with ORCA. 83 The DFT models of the OEC systems consist of 238 or 239 atoms (with H 2 O or NH 3 at the W1 position, respectively) and were constructed as described in Pantazis et al. 40 Alternative ammonia binding modes, including terminal or bridging amido and imido substitution, can be rejected on energetic grounds alone (see Fig. S2, ESI †). Geometry optimizations of the cluster models used the BP86 density functional 84,85 with the zeroth-order regular approximation (ZORA) [86][87][88] and specially adapted segmented all-electron relativistically recontracted basis sets 89 and all EPR properties were computed with the TPSSh hybrid meta-GGA functional 92,93 from BS-DFT calculations. 35,[94][95][96][97][98] The resolution of identity (RI) 99 approximation was used in the calculation of Coulomb integrals and the chain-of-spheres approximation (COSX) 100 was used for Hartree-Fock exchange, employing completely decontracted def2-TZVP/J auxiliary basis sets. 101 Tight SCF convergence criteria and increased integration grids (Grid6 and GridX6) were applied throughout. For the calculation of the hyperfine tensors, triple-zeta ZORA-recontracted basis sets were used on all atoms, while locally dense radial grids were used for Mn, N and O atoms (integration accuracy of 11 for Mn and 9 for N and O as per ORCA nomenclature). Picture change effects were applied for the calculation of EPR parameters and the complete mean-field approach was used for the spin-orbit coupling operator. The results were transformed into on-site or spin-projected values as detailed previously. 36 To compare computed 55 Mn hyperfine coupling constants using the methods described above with experimental results, a scaling factor of 1.78 was calculated from a set of twelve Mn III Mn IV mixed-valence dimers. 102 3 Results and discussion

DFT models of different OEC forms in the S 2 state
Geometric parameters of optimized DFT cluster models of the S 2 state of the OEC in the S T = 1/2 configuration 40 are shown in Fig. 2 (for coordinates, see Section S5, ESI †). Four variants were considered in this study: (i) the native cofactor system (Mn 4 O 5 Ca, also see Fig. 1), (ii) the Sr 2+ -substituted system obtained by replacing Ca 2+ with Sr 2+ (Mn 4 O 5 Sr), (iii) the NH 3 -modified system obtained by replacing the H 2 O in the W1 position with NH 3 (Mn 4 O 5 Ca-NH 3 ), and (iv) the combined Sr 2+ -substituted and NH 3 -modified system (Mn 4 O 5 Sr-NH 3 ). In all models, W2 was considered to be an OH À ligand, as determined previously. 12 Mulliken spin population analysis of all four variants confirms that the only Mn III ion of the tetramanganese complex is Mn D1 . The three Mn IV ions (Mn A4 , Mn B3 and Mn C2 ) represent coordinatively saturated, 6-coordinate octahedral sites, whereas the Mn D1 III is 5-coordinate square-pyramidal, with a Jahn-Teller elongation along the axis of the Mn D1 -Asp342 carboxylate ligand, opposite to its open coordination site.
In accordance with previous DFT and QM/MM structures, 12,37,38,40-42,103-106 the optimized Mn-Mn and Mn-Ca distances of the Mn 4 O 5 Ca model are consistent with those determined from EXAFS spectroscopy. [107][108][109][110] Only minor changes are observed between the Mn 4 O 5 Ca and the Mn 4 O 5 Sr models (Fig. 2). As a result of the larger radius of Sr 2+ , the O-Sr bond lengths increase by 0.04 Å, while the Mn-Sr distances also increase by 0.04 Å except for Mn D1 -Sr, which is 0.03 Å shorter than the Mn D1 -Ca distance. On average, this is in line with observations from EXAFS spectroscopy 108 and with the recent 2.1 Å resolution crystallographic model of Sr 2+ -substituted PSII. 111 The Mn-Mn distances are almost entirely unaffected, with the exception of Mn D1 -Mn C2 , which is shortened by 0.02 Å in the Mn 4 O 5 Sr model.
Upon NH 3 substitution of W1 (Mn 4 O 5 Ca-NH 3 ), only the Mn D1 -Mn B3 distance and the Ca 2+ distance from the terminal Mn ions change notably, albeit by less than 0.05 Å (Fig. 2). Only one structural element is more significantly perturbed, i.e. the position of O5, the m-oxo bridge trans to the binding position of NH 3 . The Mn A4 -O5 distance increases by 0.05 Å with concomitant decrease of the Mn D1 -O5 distance by 0.14 Å. Other ligands of Mn A4 , such as the second water-derived ligand, W2, remain unaffected. A similar modification to the connectivity of the Mn 4 O 5 core was seen for the smaller OEC models reported in our previous study. 44 We note that EXAFS data for the NH 3 -modified OEC have only been reported for samples purified from spinach, not from the cyanobacterial model systems T. elongatus/vulcanus, and suggested an elongation of one of the short Mn-Mn distances of 0.02 Å. 113 This type of perturbation is not observed in our optimized Mn 4 O 5 Ca-NH 3 model. The structural model including both Sr 2+ and NH 3 (Mn 4 O 5 Sr-NH 3 ) is found to replicate both effects seen in the singly modified structures.

Multifrequency EPR and 55 Mn-ENDOR of the S 2 states of the OEC variants
Multifrequency EPR/ 55 Mn-ENDOR experiments spanning the microwave range from E9 to E90 GHz were employed to experimentally characterize the electronic structures of the different S 2 state forms described above. Fig. 3A (black solid traces) depicts X-band CW EPR spectra of the S 2 state of native PSII (Mn 4 O 5 Ca), 33 Sr 2+ -substituted PSII (Mn 4 O 5 Sr), 33 NH 3 -modified (annealed) native PSII (Mn 4 O 5 Ca-NH 3 ) 44 and NH 3 -modified Sr 2+ -substituted PSII (Mn 4 O 5 Sr-NH 3 ). Shown are light-minusdark spectra, generated by taking the difference between the illuminated spectrum (S 2 ) and the dark-state spectrum (S 1 ) in order to remove background signals, such as from the cytochromes b 559 and c 550 . The modified multiline (Mn 4 O 5 Sr, Mn 4 O 5 Ca-NH 3 ) displays 26 lines of altered intensity as compared to the native multiline signal with 24 lines. The Mn 4 O 5 Sr-NH 3 S 2 state yields essentially the same modified multiline signal; the simulation superimposing this data trace uses the parameters that fit the Mn 4 O 5 Sr dataset. Fig. 3B shows the corresponding Q-band ESE-detected S 2 state multiline EPR signals of the Mn 4 O 5 Ca, Mn 4 O 5 Sr and Mn 4 O 5 Ca-NH 3 OEC forms. Pseudo-modulated (CW-like) spectra are shown in order to more clearly visualize differences in the hyperfine structures between the three forms. Compared to earlier published data by Cox et al., 33 the spectra are essentially free of contaminating hexaquo-Mn 2+ signals. Furthermore, there is a small difference of the centre positions of the multiline spectral envelopes, presumably due to inaccuracy in the microwave frequency calibration of this earlier study. contaminations. This contamination manifests itself as six inverted hyperfine lines centred at g E 2, as the high-spin Mn 2+ signal (S = 5/2) is over-rotated when using optimal instrumental settings to visualize the S = 1/2 multiline signal. No 55 Mn hyperfine structure is observable in the W-band multiline EPR spectra. Thus, these signals provide no additional information on the hyperfine matrices. The utility of these high-frequency data instead is to constrain the G tensor. While the spectra of the Mn 4 O 5 Ca and the Mn 4 O 5 Ca-NH 3 cluster show similar signals, centred at g = 1.975, the Mn 4 O 5 Sr spectrum is shifted to higher field and centred at g = 1.980, similar to the shift observed in W-band spectra from the Bittl laboratory. 114 The NH 3 -modified Sr 2+ -substituted S 2 state signal is centred at g = 1.979. The almost identical high-field shift, illustrated by the superimposed Mn 4 O 5 Sr simulation, indicates that the G-tensor shift is dependent only on the presence of Sr 2+ but not NH 3 . This is in contrast to the result at X-band, which showed that the hyperfine structure is approximately the same for both modifications. Fig. 3D shows the Q-band Davies 55 Mn-ENDOR light-minusdark spectra, measured at the central magnetic field position of the corresponding EPR spectra (Fig. 3B). The line shape of the 55 Mn-ENDOR signal shows only a small field dependence over the 1190 to 1260 mT range (not shown), consistent with its assignment to the tetranuclear Mn cluster. 31 In contrast to the X-band CW EPR spectra described above (Fig. 3A), the S 2 states of the native, Sr 2+ -substituted, NH 3 -annealed and doubly modified (not shown) OEC give rise to highly similar 55 Mn-ENDOR spectra. Five peaks are observed for all three sample types appearing at approximately the same frequency positions and of similar intensities. Small differences in the region of the largest peak (E115 MHz) may represent residual Mn 2+ contaminations. Importantly, no large difference is seen with regard to the total spectral breadth of the signal (E55 to E195 MHz). For the NH 3 -modified S 2 state, the results are nominally consistent with the earlier X-band ENDOR data of Peloquin et al. 30 The Q-band ENDOR spectra presented here do slightly differ from those presented in ref. 33,34,44 and 115 with regard to line intensities, discussed in detail in the Section S6 (ESI †).
Spectral simulations of the complete EPR and 55 Mn-ENDOR datasets using the spin Hamiltonian formalism are also shown in Fig. 3 (red dashed lines); the fitted effective G and hyperfine tensors A of the Mn clusters in the low-spin S 2 state are listed in Table 1 117 As found previously, 30,31,33,34,44 the inclusion of four hyperfine tensors of approximately the same magnitude and near-axial symmetry is required to simultaneously fit the X-, Q-and W-band EPR and Q-band 55 Mn-ENDOR line widths and shapes. The z component represents the principal component for the fitted G and all four hyperfine tensors. Comparison of the fitted parameters demonstrates that the three samples basically exhibit the same electronic structure. The sets of the four isotropic values A i,iso deviate only by r4% between the three different systems and the signs and magnitudes of the anisotropies A i,aniso are broadly similar, suggesting that there are no significant differences in the electronic exchange coupling schemes of the Mn 4 O 5 Ca/ Sr(-NH 3 ) clusters.  Table 1.
(A) X-band CW EPR. In the Ca and Sr samples, Y D had been replaced by a phenylalanine, removing the Y D signal from the spectra, 112 which were taken from Cox et al. 33 The CaNH 3 spectrum was originally published in ref. 44

Calculated magnetic properties for the native and modified S 2 states of the OEC
The electronic structure of the coupled OEC spin system is defined by the set of six pairwise Mn-Mn exchange interaction terms J ij , which can be calculated using BS-DFT. For all four computational models describing the set of native and chemically perturbed S 2 state clusters, the calculations reveal that the abba spin configuration (Fig. 4A and B) is the lowest in energy. Sets of J ij coupling constants are given in Table S1 of the ESI. † Diagonalization of the Heisenberg Hamiltonian to obtain the complete spin ladder confirms that all four models exhibit an effective total spin S T = 1/2 ground state, as observed experimentally, and an S T = 3/2 first excited state. The estimated energy differences between the two lowest states of the spin ladder are on the order of 24-26 cm À1 for the Mn 4 O 5 Ca/Sr S 2 state structures, lowering by 7 cm À1 upon exchange of W1 for NH 3 (Table S1, ESI †). These values are in the range inferred from experiments. ‡ 33,34,118,119 For all four S 2 state OEC forms, the J-coupling topology consists of three main coupling pathways (Table S1, ESI † and Fig. 4B): (i) an antiferromagnetic coupling pathway between Mn D1 and Mn C2 ( J CD ); (ii) a ferromagnetic coupling pathway between Mn C2 and Mn B3 ( J BC ); (iii) and an antiferromagnetic coupling pathway between Mn B3 and Mn A4 (J AB ). The ferromagnetic exchange pathway J BC = 19-28 cm À1 is the largest in absolute magnitude, while the antiferromagnetic pathways J CD = À16 to À18 cm À1 and J AB = À12 to À16 cm À1 are slightly weaker. The remaining exchange coupling constants J AC , J AD and J BD are small, as can be expected from geometric considerations (see Table S1, ESI †). J CD and J BC represent the two largest exchange interactions within the cuboidal trimer unit (Mn B3 Mn C2 Mn D1 ) of the cluster, whereas J AB can be considered to a good approximation as being representative of an effective exchange interaction between this cuboidal unit and the outer Mn A4 , as shown in Fig. 4C.
Systematic differences are observed for the exchange pathways upon the two chemical perturbations, replacement of Ca 2+ by Sr 2+ and NH 3 exchange at W1 (Table S1, 3 , it is seen that only the major coupling pathways J CD and J BC are modified, decreasing by 2 cm À1 and 5 cm À1 , respectively. J AB remains unchanged. By contrast, for the corresponding structure pairs where NH 3 is exchanged for W1, the J CD pathway is unchanged, while J BC and J AB increase by 4 cm À1 . It is noted that the perturbation of the O5 position upon NH 3 substitution, as shown in Fig. 2, results also in an enhancement of J BD by 3 cm À1 . In both cases, the changes in the magnetic interactions can be understood within the geometric changes discussed in Section 3.1 (see Fig. 2): Sr 2+ substitution mostly affects the structure of the cuboidal unit, thus perturbing principally the exchange pathways within the Mn-trimer unit, whereas NH 3 binding perturbs mostly the connectivity between the trimeric moiety and the outer Mn A4 (Fig. 4C).
The anisotropy in the G and A i values is expressed as the difference  Table 2 for the four S 2 state variants. The calculated isotropic hyperfine values a i,iso for the three Mn IV ions fall within the range seen in Mn IV model compounds experimentally, i.e. |a iso | = 187-253 MHz (see ref. 30 and 33). The anisotropy of the calculated hyperfine tensors for these three sites is also small, of the order seen in octahedral Mn IV model complexes, i.e. |a aniso | o 30 MHz. 33 For the Mn D1 III ion, the calculated isotropic hyperfine value (E130 MHz) is smaller than that for Mn IV , as expected, and lies just outside the range seen in Mn III compounds, i.e. |a iso | = 165-225 MHz. 30,33 As typical for Mn III , it exhibits a significant hyperfine anisotropy, more pronounced than for the Mn IV ions. However, it is noted that the calculated values for the Mn D1 III site are unexpectedly large.
Nevertheless, the computed parameters correlate with the inferred site geometry of Mn D1 , namely that of a square-pyramidal 5-coordinate Mn III ion. Such a coordination environment generally yields a small isotropic 55 Mn on-site hyperfine coupling and a negative anisotropy (see Table 2), consistent with an effective local 5 B 1 electronic ground state for the Mn D1 III ion. 30,33,120 The effective hyperfine couplings measured by EPR spectroscopy for oligonuclear metal complexes reflect the on-site hyperfine couplings of the individual metal ion nuclei scaled by the contribution of the electronic spin of each metal ion to the effective spin state: A i = r i a i . The set of scaling factors r i , termed spin projection coefficients, are primarily determined by the set of pairwise exchange couplings as detailed in ref. 18, 30, 31, 33, 34, 36, 37 and 63. However, additional terms must be included to correctly estimate such spin projections for the OEC, specifically the relevant on-site fine structure parameters d i for the individual Mn i ions, 18,30,33,34,63 yielding what are more accurately described as spin projection tensors. As the coordination geometries of the three Mn IV ions of the S 2 state are all octahedral, their local electronic structure should be of approximate spherical symmetry, their orbitals of t 2g origin (d xy , d xz and d yz ) being half-filled (local high-spin d 3 configuration). As such, the Mn IV ions are expected to only display small fine structure parameters d i (o0.3 cm À1 ) 121 and hence do not need to be explicitly considered. Thus, the set of parameters which define the spin projection tensors in the S 2 state are the six pairwise exchange interaction terms and the fine structure parameter of the Mn III ion, d D1 .
Using these spin projection tensors, the fitted projected 55 Mn hyperfine tensors were scaled back to on-site hyperfine tensors to allow comparison to the BS-DFT values discussed above ( Table 2). The only plausible assignment for all three forms of the OEC is that A 1 , A 2 , A 3 , and A 4 correspond to a D1 , a A4 , a C2 and a B3 , respectively. In our previous work, 33 using BS-DFT structural models predating the latest crystal structure, 9 values of À1.2 to À1.3 cm À1 were estimated for a supposedly axially symmetric d D1 in the native and Sr 2+ -substituted S 2 states. Using the same approach, d D1 was re-estimated. It was possible to obtain on-site hyperfine anisotropies in the ranges characteristic for Mn III and Mn IV ions employing a single value of À1.43 cm À1 for the three OEC systems, well within the range typically seen for Mn III model complexes. As discussed above with regard to the hyperfine tensor anisotropy of Mn D1 III , a negative d value requires an effective local 5  , a D1,aniso o À40 MHz is less negative than calculated. Overall, the experimental results confirm that the computed spin coupling schemes serve as a valid description of the native and modified S 2 states.

The Mn D1 -His332-imino-N interaction
Three-pulse ESEEM measurements were performed to characterize the imino-N signal of His332 associated with the OEC variants in the S 2 state. Fig. S4 and S5 (ESI †), respectively, depict tand magnetic-field-dependent (g E 2.10-1.  a The isotropic r i,iso and a i,iso values are the averages of the individual tensor components r i,iso = (2r i,> + r i,J )/3 and a i,iso = (a i,x + a i,y + a i,z )/3. b The anisotropies of the r i and a i tensors are expressed as the differences r i,aniso = r i,> À r i,J and a i,aniso = a i,> À a i,J , i.e. between the perpendicular and parallel tensor components. c The intrinsic fine structure values of Mn IV ions were assumed to be d A4 = d B3 = d C2 = 0 cm À1 . For the Mn D1 III ion, a value of d D1 = -1.43 cm À1 was fitted, with e D1 /d D1 = 0. from T. elongatus is very similar to that measured in PSII from both higher plants (spinach) 122 and the mesophilic cyanobacteria Synechocystis sp. PCC 6803, 123 assigned to the imino-N of His332 via mutagenesis. 123,124 The signals are essentially the same in the native, Sr 2+ -substituted and NH 3 -modified OEC clusters with regard to both their t and magnetic-field dependence. The His332 imino-14 N signal at Q-band nearly fulfils the cancellation condition, where A iso is twice the 14 N nuclear Larmor frequency (n n = 3.75 MHz at 1.22 T). The spectra are characterized by three features: the lines centred at frequencies below 2.5 MHz (n a = n n À |A iso |/2), single-quantum transitions around 7.5 MHz (n b = n n + |A iso |/2) and smaller double-quantum resonances around 15 MHz (n 2b = 2n n + |A iso |). The line structuring is defined both by the 14 N hyperfine anisotropy and the NQI.
HYSCORE spectroscopy (a two-dimensional ESEEM technique) was performed on the three S 2 state OECs at different magnetic-field positions (g E 2.07-1.93) of the corresponding Q-band EPR envelopes to further constrain the 14 N hyperfine and quadrupolar interaction matrices. Panels A, C and E in Fig. 5 show the Fourier-transformed spectra and simulations at the centre field position; low-and high-field spectra and simulations are presented in Fig. S6 and S7 in the ESI. † As seen for the three-pulse ESEEM spectra, their appearance is highly similar for all three variants of the OEC in the S 2 state. In two dimensions, the three features that comprise the Q-band ESEEM spectra appear as cross peaks at corresponding frequencies both in the (À,+) and the (+,+) quadrants. As the 14 N hyperfine coupling matches the cancellation condition, the cross peaks are shifted away from the diagonal, instead appearing near the  Table 3 and, in detail, in Table S2 (ESI †) frequency axes. Overall, virtually no orientation dependence is seen comparing the spectra at the three different magnetic fields (Section S8.2, ESI †), consistent with the electron-nuclear interaction being dominated by the isotropic component of the hyperfine coupling as compared to the anisotropic part and the traceless NQI, as in ref. 122, 123 and 125. Thus, the orientation of the His332 imino-14 N hyperfine tensor relative to the G tensor cannot be determined from this dataset.
Fitted spin Hamiltonian parameters derived from the simultaneous simulation of both the ESEEM and HYSCORE datasets are given in Table 3 together with BS-DFT estimates. To directly compare DFT values with experiment, the calculated site hyperfine tensor for the His332 was multiplied by the axial Mn D1 spin projection tensor described in Section 3.3. All DFT models yield virtually the same hyperfine and quadrupole values. The calculated A iso underestimates experimental results by o20%, but the dipolar component A dip and the rhombicity A Z nominally agree with experiment. It is noted that the on-site 14 N hyperfine tensor a is expected to be axial with its unique component a 1 aligned along the Mn D1 -N bond, as seen in our calculations ( While the orientation of the hyperfine tensor relative to the G tensor cannot be determined using ESEEM/HYSCORE at Q-band frequencies, it can be measured at W-band, e.g. using EDNMR. In our earlier study, 14 it was found that the hyperfine tensor is orientated such that its principal, i.e. the smallest component A 1 is aligned such that it is mid-way between G x , and G z . Importantly though, it is noted that the set of spin Hamiltonian parameters deduced from Q-band ESEEM/HYSCORE (Table 3) does not reproduce the W-band data sets (Section S8.3, ESI †). This is not due to the inclusion/exclusion of the NQI term, which, for the W-band EDNMR data, mainly contributes to the spectral line width. To reproduce the field dependence of the 14 Nand 15 N-EDNMR signals ( Fig. S8 and S9, ESI †), the values determined from Q-band ESEEM/HYSCORE needed to be scaled: A iso was decreased by 10%, whereas A dip was increased by a factor of two (Table S2, ESI †). The same results were observed for all three S 2 state forms, which basically exhibit the same 14 N-EDNMR spectra. A possible reason for this difference comes from the observation that the ground spin state, an effective spin S T = 1/2 state, is not very well separated energetically from higher spin states in the regime of the W-band excitation energy (E3 cm À1 ), consistent with DFT estimates for the ground-to-first excited state energy splitting DE (Section S7, ESI †). Excited-state mixing due to a small DE has the consequence of altering spin Hamiltonian observables such as effective 55 Mn and 14 N hyperfine tensors. Alternatively, the rhombicity of the effective G tensor as inferred from the EPR/ 55 Mn-ENDOR simulations may be artificial, a consequence of using collinear G and 55 Mn hyperfine tensors. This latter suggestion would also explain why the G tensors inferred from W-band measurements on PSII single crystals 116,117 differ from those inferred from our multifrequency measurements on frozen solution PSII samples.
The lack of agreement between the two 14 N datasets brings into question whether the W-band 14 N/ 15 N-EDNMR signals can be used to assign the exchangeable m-oxo bridge 17 O signal based on the relative orientations of the 14 N and 17 O hyperfine tensors, as suggested by Rapatskiy et al. 14 Thus, further experimental results, particularly from single crystals of PSII, are needed to test this proposal (see Section 3.6). Table 3 Fitted and calculated effective/projected 14 N hyperfine and NQI tensors in MHz for the electron-nuclear couplings of the His332 imino-N and of NH 3 with the various cluster forms studied in the S 2 state in PSII from T. elongatus a A iso is defined as the average of the principal components of the hyperfine tensor: A iso = (A 1 + A 2 + A 3 )/3. b A dip is defined in terms of T 1 , T 2 , and T 3 as A dip = (T 1 + T 2 )/2 = -T 3 /2. c The rhombicity is defined by A Z or Z = (T 1 À T 2 )/T 3 , respectively. T 1 , T 2 , and T 3 represent the three principal components of the hyperfine tensors minus A iso and of the NQI tensors and are labelled such that |T 1 | r |T 2 | r |T 3 |. 3.5 NH 3 binding to the Ca 2+ -and the Sr 2+ -containing OEC In the NH 3 -modified S 2 state, a second nitrogen nucleus is bound to the Mn cluster as a terminal ligand, as described in Pérez Navarro et al. 44 Its binding can be observed using X-band (three-pulse) ESEEM, as shown in Fig. S11 (ESI †) for 14 NH 3 / 15 NH 3bound, Ca 2+ -and Sr 2+ -containing PSII. The 14 NH 3 resonances comprise three characteristic single-quantum lines at 0.5, 0.95 and 1.45 MHz split by the NQI and smaller double-quantum transitions centred at 4.9 MHz, highly similar to the higher plant data. 70 Due to the lack of the NQI, the 15 NH 3 signal is clearly less complicated, consisting only of one single-quantum hyperfine peak centred at 0.3 MHz. As seen for the His332 imino-14 N signal at Q-band, the NH 3 interaction at X-band fulfils the cancellation condition, leading to a narrow n a line while the n b line is broadened beyond detection. 70 Most importantly, the spectra of the 14 NH 3 -modified Ca 2+ -and Sr 2+ -containing 14 N-PSII samples are essentially identical. Thus, NH 3 binding to the Sr 2+ -substituted S 2 state cluster is the same as in the native S 2 state. In our first report on NH 3 binding to the OEC, only the 14 NH 3 interaction was considered. 44 Here, we simultaneously fit the spectra of both the 14 NH 3 -modified 15 N-PSII and the 15 NH 3 -modified 14 N-PSII in the S 2 state (Table 3 (Table 3).
An axial projected 14 NH 3 hyperfine tensor is obtained from BS-DFT calculations, as seen in the experiment. This is because (i) the on-site 14 NH 3 hyperfine tensor is axial, and (ii) its axial and equatorial components are essentially coincident with those of the Mn A4 spin projection tensor (Table S3, ESI †), unlike the case for the His332 imino-14 N a tensor (see Section 3.4). The BS-DFT calculations also reproduce the comparably large and rhombic NQI parameters (Table 3), although the sign of the hyperfine anisotropy is inverted compared to experiment. For more details, see Section S9 in the ESI. †

Interactions with exchangeable 17 O species
As we have recently shown, 14,19 EDNMR spectroscopy at W-band, due to its comparatively high sensitivity, is the preferred method to measure the interactions of exchangeable 17 O nuclei with fastrelaxing electronic species such as the S 2 state of the OEC. Fig. 6 shows these spectra and simulations (see Section 2.3 and Sections S3 and S4, ESI † for details) of the single-quantum region for the native, the Sr 2+ -substituted, the NH 3 -annealed and the Sr 2+ -and NH 3 -modified S 2 state variants after H 2 17 O buffer exchange in the S 1 state (see Fig. S12 (ESI †) for the double-quantum region). The spectrum of the native system exhibits the single-and double-quantum resonances of the imino-14 N of His332 (blue) and of three different classes of 17 O species, 14 i.e. (i) a strongly coupled, bridging species (green), (ii) an intermediately coupled terminal O-ligand (orange), and (iii) a weakly coupled terminal class (cyan). These were assigned to the m-oxo bridge O5, the hydroxide ion W2 12 and the H 2 O matrix (comprising ligand W1 of Mn A4 and two H 2 O ligands at the Ca 2+ ion), respectively. NH 3 binding causes a narrowing of the 17 O single-and double-quantum envelopes, reproduced by a decrease of the hyperfine couplings of O5 and W2 and concomitant reduction of the matrix line intensity, which was interpreted by NH 3 binding to Mn A4 in exchange for W1. 44 Comparing these two spectral forms to those of the corresponding Sr 2+ -substituted W1-and NH 3 -containing clusters (Fig. S13, ESI †), we see a systematic narrowing of the single-quantum envelope by E0.5 MHz and a corresponding narrowing of the double-quantum envelope. This can be reproduced by spectral simulations in which the hyperfine couplings of the m-oxo bridge are reduced accordingly (W1: 9.2 MHz vs. 9.7 MHz, NH 3 : 6.5 vs. 7.0 MHz), while the other 17 O interactions remain unaltered (for a complete set of hyperfine parameters, see Table S4, ESI †). Although weaker than the NH 3 effect, the narrowing was found to be reproducible in all Sr 2+ -substituted PSII samples. It clearly shows that Ca 2+ /Sr 2+ exchange perturbs the m-oxo bridge, in addition to a simultaneous modification by NH 3 binding.
The inset in Fig. 6 depicts a section of the X-Band CW EPR spectra of the Sr 2+ -substituted S 2 state, which exhibits an intrinsically smaller average line width (E3.6 mT peak-topeak) than the native form (E4 mT), in the presence and absence of 17 O (see Fig. S14 (ESI †) for the entire spectra). No EPR line broadening is observed upon 17 O exchange. This demonstrates that the largest 17 O coupling represents only one exchangeable oxo bridge. In the case of two hyperfine interactions of E10 MHz, the effective line broadening due to the combined 17 O couplings would be larger than 120 MHz or 4.3 mT, exceeding the actual line width.

A common electronic structure of the S 2 state variants
Our DFT results show that the Sr 2+ -substituted, the NH 3 -annealed and the Sr 2+ -and NH 3 -modified low-spin S 2 states basically represent the same structure on both a geometric and electronic level. This result is not immediately obvious from their X-band EPR signals. Indeed historically, the Sr 2+ -substituted and NH 3 -modified forms were explained in terms of a change of the valence state distribution within the Mn tetramer and thus of the coordination environment of the Mn III ion. 126 The comprehensive approach pursued in this study conclusively rules out such a mechanism for electronic structure perturbation. Instead, as proposed by our group, 33,34,63 multiline heterogeneity reflects rather subtle changes of the Mn-tetramer structure. The similarity of the perturbed multiline forms suggest a common mechanism for electronic perturbation, which probably also explains S 1 state heterogeneity. This is discussed below, with reference to solvent access, substrate binding and exchange.
4.1.1 The mechanism of structural perturbation. In Su et al., 34 a qualitative model for multiline heterogeneity was proposed. In this model, the electronic structure of the S 2 state was considered in terms of a simplified model of two spin fragments: (i) a cuboidal trimer unit, made up of Mn D1 , Mn C2 and Mn B3 , and (ii) a 'monomeric' Mn unit, consisting of the outer Mn A4 . Therein, variation of the electronic structure in S 2 was attributed to changes in the connectivity of the outer Mn A4 to the cuboid, changing the properties of the electronic ground state by altering the mixing-in of excited spin state character. The physical rationale for this observation was that the outer Mn A4 represents the solvent accessible end of the cluster. It has two water-derived ligands and solvent channels that begin or terminate at this site. 9,127 As such, it is this site and its connection to the rest of the cluster that is most likely to vary amongst different sample conditions and possibly different PSII species. In terms of this 'monomer-trimer' model with regard to the two modifications discussed here, Ca 2+ /Sr 2+ forms part of the linkage between the cuboidal and outer fragments, as mediated by the m-oxo bridge O5, whereas NH 3 binds to the outer Mn A4 , also perturbing O5, the bridge trans to its binding position W1. 44 The magnetic observable that is altered upon Ca 2+ /Sr 2+ replacement and/or W1/NH 3 exchange, leading to the perturbed multiline forms is the 55 Mn hyperfine anisotropy. Small perturbations of the four hyperfine tensors result in a change in the hyperfine peak superposition, altering the apparent structure of the X-band EPR signal (Fig. 3A). Importantly these changes are subtle, as demonstrated by the invariance of the 55 Mn-ENDOR spectra (Fig. 3D). The 55 Mn hyperfine anisotropy is not a site property, but instead an indirect measure of the fine structure splitting of the Mn III ion 30,33,34,63 or, in the 'monomer-trimer' model, the zero-field splitting of the whole trimer unit. Within this model, its contribution is modulated by the electronic connectivity between the two fragments, predominantly the exchange pathway J AB , the coupling that mostly defines the energy splitting DE between the ground state and the first excited state (Fig. 4C). Our BS-DFT results support this basic mechanism for electronic structure perturbation and, for the first time, describe the changes on the molecular level that impart this variation, and which differ for the two modifications. Upon replacement of Ca 2+ by Sr 2+ , the slight distortion of the cuboidal moiety leads to a perturbation of the intra-cuboidal exchange network and possibly the Mn D1 III site fine structure splitting. It is noted that this, besides changing the 55 Mn hyperfine anisotropy, also manifests itself in terms of the G tensor, also contributing to the altered multiline appearance and the g shift of the W-band EPR signal. Exchange of W1 by NH 3 affects the connectivity of the outer Mn A4 to the cuboidal unit, as modulated by the m-oxo bridge O5, perturbing the J AB exchange pathway, thus changing the 55 Mn hyperfine anisotropy. In the case of the S 2 state variant that contains both these modifications, their effects on the electronic structure are additive. It is noted that it is the properties of the cuboidal unit that define the G tensor as opposed to the outer Mn A4 , which presumably has an isotropic on-site g value. This is expected, as it is the Mn D1 III ion, which is part of the trimer fragment in the S = 1/2 configuration, that should form the dominant contribution to the anisotropy of the G tensor in all four systems. 4.1.2 The Mn D1 -His332-imino-N and Mn A4 -NH 3 interactions as local probes for the electronic structure 4.1.2a The Mn D1 -His332-imino-N interaction. As described by Stich et al., 123 the magnitude of the Mn D1 -His332 imino-N hyperfine interaction, as compared against mixed-valence Mn III Mn IV model compounds and protein cofactors with imidazole ligands to Mn III (A iso r 13 MHz) and Mn IV ions (A iso = 1.5-3.3 MHz), [128][129][130][131][132][133][134] favours assigning Mn D1 as the only Mn III ion of the S 2 state, consistent with the EPR/ 55 Mn-ENDOR/DFT results already reported in the literature and detailed above (Sections 3.1-3.3). The large hyperfine couplings seen for ligands coordinating to Mn III in S 2 (and model systems) comes from the fact that the Mn III ion carries the largest spin projection coefficient, i.e. in Mn dimers r iso (Mn III ) = 2 and r iso (Mn IV ) = -1. Interestingly, the hyperfine and quadrupole couplings of imidazole ligands of Mn III ions differ depending on whether they represent axial (A iso = 9-13 MHz, e 2 Qq/h = 2.1-3.0 MHz) [128][129][130][131][132] or equatorial ligands (A iso = 1.5-6.6 MHz, e 2 Qq/h = 1.5-2.5 MHz). [131][132][133][134] The values seen for the His332 imino-N (A iso = 7.1 MHz, e 2 Qq/h = 1.97 MHz) fall closer to the equatorial range supporting its assignment as an equatorial ligand consistent with DFT structural models. In such models, 12,37,40,103,106 the local Jahn-Teller axis of the Mn D1 III ion is aligned along the open coordination site, thus considered a pseudo-Jahn-Teller axis, perpendicular to the Mn D1 -N bond. It is supposed that the reason why the 14 N couplings measured for the His332 do not exactly fall within the range seen in model complexes is that all model complexes measured thus far represent 6-coordinate Mn III ions whereas the Mn D1 III ion in the S 2 state is 5-coordinate.

4.1.2b
The Mn A4 -NH 3 interaction. As recently shown in Pérez Navarro et al., 44 the binding site of NH 3 is likely the W1 site. The small effective isotropic 14 N hyperfine coupling (A iso = 2.36 MHz) and the axiality of the hyperfine tensor are both consistent with a terminal ligand to a Mn IV (d 3 ) ion. 44,70 The similar A iso in the Mn 4 O 5 Sr-NH 3 cluster confirms that the oxidation state of the Mn A4 ion is not altered by Sr 2+ substitution. Moreover, the binding mode and perturbation mechanism of NH 3 is the same in the Ca 2+and Sr 2+ -containing Mn clusters. The non-axiality of the electric field gradient (Z = 0.47) is characteristic for this ligand. A large asymmetry parameter is uncommon for a terminal ligand of Mn IV (although our value is already E20% smaller than that reported earlier 70 ). The latest crystal structure 9 suggests that such an asymmetric distortion could be present for the W1 site due to the charged residue D1-Asp61, in H-bonding distance to W1/NH 3 , as seen for other protein systems. 135 Indeed, upon inclusion of the Asp61 residue, which was not included in our previous, smaller BS-DFT model, 44 the asymmetric quadrupole tensor is now reproduced, and the hyperfine coupling constant shows better agreement with experiment ( Fig. 7 and Table 3). In contrast, such an asymmetric distortion is not seen for the W2 ligand as a similar charged amino-acid residue partner is not present to provide a H-bond.

The exchangeable l-oxo bridge
Both modifications investigated, Ca 2+ /Sr 2+ substitution and NH 3 /W1 replacement, perturb the 17 O-EDNMR signals of exchangeable oxygen species of the OEC, specifically the exchangeable m-oxo bridge. It is this bridge that likely represents one of the substrate water sites of the Mn tetramer. As the electronic structure of the OEC is essentially invariant for all four OEC forms, the change in hyperfine coupling for this m-oxo bridge must represent a site modification, near or at the oxygen nucleus. NH 3 binding primarily affects the connectivity of the outer Mn A4 to the cuboidal trimer, whereas Sr 2+ substitution instead perturbs the exchange network within the cluster. Thus, it can be surmised that the exchangeable m-oxo bridge must both coordinate to the outer Mn A4 and be associated with the Ca 2+ /Sr 2+ ion itself as a structural element of the cuboidal trimer. Only the bridge O5 fulfils both these criteria. As a ligand to the Ca 2+ /Sr 2+ ion, O5 is affected by the exchange of these ions of the same charge but different sizes. Similarly, as argued in Pérez Navarro et al., 44 NH 3 /W1 exchange perturbs O5 by binding trans to this bridge position, distorting the Mn A4 -O5 bond length.
It is noted that these results exclude the possibility that NH 3 displaces the exchangeable m-oxo bridge as a bridging -NH 2species, an alternative rationale for the narrowing of the 17 O signal envelope in line with earlier suggestions. 70 Ca 2+ /Sr 2+ and W1/NH 3 exchange are additive in terms of their effect on the width of the 17 O-EDNMR envelope, modelled here as defined by the m-oxo bridge hyperfine coupling A iso . This result mirrors the structural modifications observed for the doubly modified Mn 4 O 5 Sr-NH 3 OEC model; i.e., the model contains additive structural modifications reflecting both singularly modified Mn 4 O 5 Sr and Mn 4 O 5 Ca-NH 3 structures. If instead the NH 3 did indeed replace the bridge, the width of the 17 O-EDNMR envelope would be now defined by the W2 hyperfine coupling, and as such should be invariant to Ca 2+ /Sr 2+ substitution. It is also noted that NH 3 replacement of the exchangeable bridge O5 cannot quantitatively explain the virtually unaltered 14 NH 3 signal upon exchange of the O5-binding Ca 2+ , and the 17 O hyperfine changes. Assuming an unaltered spin density on the bridge position, as follows from the similar spin projection factors for the four Mn ions, the measured 14 N hyperfine coupling for the bound ammonia at this position is far too small. On the other hand, the 17 O coupling of 6.5-7 MHz seen for the NH 3 /W1-exchanged system is in the range of those observed in Mn model complexes with a N-ligand trans to the oxo bridge. 14

Conclusions
Time-dependent mass spectrometry experiments indicate that the early binding substrate (W S ) is associated with all intermediate states of the OEC. 66,136 Furthermore, the relatively slow exchange and the S-state dependence of this bound substrate with bulk water suggests that it represents a ligand of (a) Mn ion(s). As Ca 2+ /Sr 2+ substitution also perturbs its exchange rate, W S is also supposed to coordinate to the Ca 2+ ion. 65,66 Of the exchangeable oxygen species identified here by 17 O-EDNMR, only O5 is a ligand to both Mn and Ca 2+ . Similarly, only the O5 spectral signature is perturbed by Ca 2+ /Sr 2+ exchange. Thus, O5 is the most likely candidate for W S . This assignment limits the possible reaction pathways for photosynthetic water splitting, and lays a foundation for studies of higher oxidized S states, which will serve to identify the second, fast exchanging substrate and eventually elucidate the mechanism of O-O bond formation. Currently two pathways are envisaged: O-O bond formation could proceed as a coupling between O5 and either (i) Mn A4 -bound W2 or Ca 2+ -bound W3, or (ii) a further oxygen not present yet in the S 2 state.