The Mn Cluster in the S₀ State of the Oxygen-Evolving Complex of Photosystem II Studied by EXAFS Spectroscopy: Are There Three Di-μ-oxo-bridged Mn₂ Moieties in the Tetranuclear Mn Complex?

Abstract

A key component required for an understanding of the mechanism of the evolution of molecular oxygen by the photosynthetic oxygen-evolving complex (OEC) in photosystem II (PS II) is the knowledge of the structures of the Mn cluster in the OEC in each of its intermediate redox states, or S-states. In this paper, we report the first detailed structural characterization using Mn extended X-ray absorption fine structure (EXAFS) spectroscopy of the Mn cluster of the OEC in the S₀ state, which exists immediately after the release of molecular oxygen. On the basis of the EXAFS spectroscopic results, the most likely interpretation is that one of the di-μ-oxo-bridged Mn–Mn moieties in the OEC has increased in distance from 2.7 Å in the dark-stable S₁ state to 2.85 Å in the S₀ state. Furthermore, curve fitting of the distance heterogeneity present in the EXAFS data from the S₀ state leads to the intriguing possibility that three di-μ-oxo-bridged Mn–Mn moieties may exist in the OEC instead of the two di-μ-oxo-bridged Mn–Mn moieties that are widely used in proposed structural models for the OEC. This possibility is developed using novel structural models for the Mn cluster in the OEC which are consistent with the structural information available from EXAFS and the recent X-ray crystallographic structure of PS II at 3.8 Å resolution.

Introduction

The biological generation of oxygen by the oxygen-evolving complex (OEC) in photosystem II (PS II) is arguably one of nature's most important reactions. To perform the four-electron oxidation of water to dioxygen, the Mn–containing OEC cycles through five intermediate S-states, S₀ through S₄, as shown in Figure 1. This proposal by Kok et al.¹ in 1970 has led to intense study to identify the nature of each of the quasi-stable S-states S₀, S₁, S₂, and S₃ to derive insight about the mechanism of water oxidation in the OEC. Because the S₁ state is the dark-stable state,¹ this S-state is relatively easily studied in the form of concentrated, dark-adapted samples. The discovery that the S₂ state could be prepared essentially quantitatively by low-temperature (190 K) illumination² allowed similar studies to be performed on the S₂ state as had been done on the S₁ state. However, the remaining S-states, the S₀ state and the S₃ state, required single-flash saturation techniques. Because this was possible only with dilute samples (e5 mg Chl/mL, prior to the experiments detailed in Messinger et al.³), the experimental horizons for these S-states were not very promising, especially for X-ray spectroscopic experiments. This has severely limited experimental studies; hence, much less is known about the S₀ and S₃ states than is known about the S₁ and S₂ states.

Figure 1.

S-state scheme for oxygen evolution as proposed by Kok et al.¹

Most of the information about the structure of the OEC has come from electron paramagnetic resonance (EPR) and X-ray spectroscopic studies of the S₁ and S₂ states. One form of X-ray spectroscopy, extended X-ray absorption fine structure (EXAFS) spectroscopy, has provided a wealth of information about the structure of the OEC in the S₁ and S₂ states. This is largely due to the inherent element specificity of X-ray spectroscopy. EXAFS studies^4–13 have firmly established that the OEC in PS II is comprised of di-μ-oxo-bridged Mn₂ clusters which show Mn–Mn scattering at a distance of 2.7 Å. Furthermore, an additional scatterer has been shown to be present at 3.3 Å from Mn,^7,8,10–12 which has been interpreted as containing contributions from both Mn–Mn and Mn–Ca interactions at 3.3–3.4 Å.^14,15 These EXAFS-derived structural building blocks have been prerequisite structural elements in virtually any proposed model.^16–22 However, the models, some of which are shown in Figure 2, are adversely affected by relative paucity of structural information about the S₀ and S₃ states (vide supra). In addition, uncertainty about whether two or three di-μ-oxo-bridged Mn–Mn moieties exist in the Mn cluster has been reflected in the proposed models. To better understand the structure of the Mn cluster, it is therefore important to determine whether it is comprised of two or three di-μ-oxo-bridged Mn–Mn moieties. This is difficult for the S₁ and S₂ states because an accurate determination of the total number of such interactions (the N value from EXAFS curve fitting) is complicated by the inherent error of ±30% in determining the total N value.²³ However, if the samples contain distance heterogeneity in a certain S-state, then the ratio of N values for these slightly different distances, which can be expected to be significantly more accurate than the total N value, can provide additional information on the question of whether two or three di-μ-oxo-bridged Mn–Mn moieties exist in the OEC.

Figure 2.

Possible structural models for the active site of the OEC in PS II. Adapted from DeRose et al.,¹¹ Cinco et al.,⁵¹ and Robblee et al.²²

Early XAS experiments by Guiles et al.²⁴ with the S₀ state used chemical treatments to get around the problem of low concentrations. However, the S₀ state generated in this manner was designated as S₀* to emphasize that it is generated through chemical treatment and is thus not a native S-state. Although hampered by a low signal-to-noise ratio and the uncertainty about the relationship between the chemically generated S₀* state and the native S₀ state, those experiments provided the first evidence from XAS that heterogeneity may exist in the 2.7 Å Mn–Mn distances in the S₀ state in the form of a reduced amplitude of Fourier peak II in the S₀* state relative to that in the S₁ state; this heterogeneity is not seen in the S₁- or S₂-state EXAFS spectra (but is seen in the S₃-state EXAFS spectra²⁵).

A separate study by Riggs-Gelasco et al.¹³ examined reduced S-states of the OEC and observed a decrease in the amplitude of the 2.7 Å Mn–Mn Fourier peak. This was interpreted by Riggs-Gelasco et al. as a reduction in the number of Mn–Mn vectors instead of the appearance of distance heterogeneity.

The native S₀ state prepared using single-flash turnover has not been extensively examined using EXAFS spectroscopy due to the difficulties in collecting EXAFS spectra from single-flash saturable samples, although preliminary reports have appeared.^26,27 In this paper, the method of preparing samples in the native S₀ state described by Messinger et al.²⁸ has been extended to perform EXAFS spectroscopic experiments on the S₀ state of PS II generated through single-flash turnover. These experiments show that, in the S₀ state, heterogeneity most likely exists in the 2.7 Å Mn–Mn distances, which can be explained through the protonation of a di-μ-oxo-bridged Mn–Mn moiety and/or the presence of Mn(II). The presence of distance heterogeneity in the S₀ state has been exploited in the curve-fitting procedure, whose results are suggestive of the possibility that three di-μ-oxo-bridged Mn–Mn moieties exist in the OEC instead of the two di-μ-oxo-bridged Mn–Mn moieties that are widely used in proposed structural models for the OEC.^{10,15–17,19,22,29–32} Possible topological models for the structure of the OEC in the S₀ and S₁ states that are consistent with the EXAFS data from the current study and also with the recent X-ray crystallographic structure of PS II at 3.8 Å resolution³³ are discussed.

Materials and Methods

PS II membranes were prepared from fresh spinach leaves by a 2 min incubation of the isolated thylakoids with the detergent Triton X-100.^34,35 The samples were then resuspended to a chlorophyll (Chl) concentration of 6.5 mg Chl/mL in sucrose buffer (pH 6.5, 400 mM sucrose, 50 mM MES, 15 mM NaCl, 5 mM MgCl₂, 5 mM CaCl₂) and stored as aliquots at –80 °C until used. Chl concentrations were calculated as described in Porra et al.³⁶

To prepare samples enriched in the S₀ state, a frequency-doubled Nd:YAG laser system (Spectra-Physics PRO 230–10, 800 mJ/pulse at 532 nm, 9 ns pulse width) was used to illuminate the PS II samples. The laser was operated continuously at 10 Hz, and flashes were selected using an external shutter (model LSTX-Y3, nm Laser Products, Inc.).

Before flash illumination, the PS II membranes were diluted to a concentration of 1 mg Chl/mL in sucrose buffer, and 3 mL of this solution was transferred in darkness into each of 20 tissue culture flasks (Falcon 3014, 50 mL, 25 cm² growth area) that were kept on ice. The Nd:YAG beam was redirected and diffused such that the laser beam could illuminate the entire growth area of the flask from below. Sample illumination under these conditions was proven to be saturating by separate experiments in which the Chl concentration was reduced to 0.5 and 0.25 mg Chl/mL, and no increase in the yield of the S₀ state formed in three-flash samples (referred to as 3F sample hereafter) relative to the experiments described herein was seen (data not shown). Each sample was given one pre-flash and was dark-adapted for 90 min on ice.

After dark-adaptation, PPBQ (phenyl-1,4-benzoquinone; 50 mM in MeOH) was added to each flask to a final concentration of 25 μM, and then each sample was illuminated with three flashes at 1 Hz frequency. Immediately after flashing all samples, FCCP (carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone; 5 mM in MeOH) was added to each flask to a final concentration of 1 μM. This accelerated the deactivation of the S₂ and S₃ states of PS II to the S₁ state³⁷ and reduced Y_D^ox, the stable tyrosine radical of PS II.³⁸ The latter reaction essentially eliminates the main path for the decay of the S₀ state, which is the oxidation of the S₀ state to the S state by Y_D^ox.³⁹ This reaction has a half-life of 30 min at 5 °C.⁴⁰ In addition to FCCP, MeOH was added to a final concentration of 3% v/v, which enabled the detection of the S₀ EPR multiline signal (MLS) in these samples (data not shown).^28,41,42 The samples were collected and centrifuged at 4 °C for 30 min at 48 000g. The concentrated PS II membranes were then put into Lexan sample holders (22 × 3.2 × 0.8 mm inner dimensions) and were frozen at 77 K for EPR and X-ray experiments. It took a total of 50 min to complete the protocol from the point of flash-induced S₀-state formation to freezing the samples at 77 K. Control samples in the S₁ state were prepared in an identical fashion, except that the application of three laser flashes to each tissue culture flask was omitted.

EPR spectra were collected on a Varian E-109 spectrometer with an E-102 microwave bridge and stored using Labview running on a Macintosh G3 computer. Samples were maintained at cryogenic temperature using an Air Products Heli-tran liquid helium cryostat. Spectrometer conditions were as follows: S₂ EPR multiline signal, 3300 ± 1000 G scan range, 6300 gain, 30 mW microwave power, 8 K temperature, 32 G modulation amplitude, 100 kHz modulation frequency, 4 min/scan, 1 scan per sample, 0.25 s time constant, 9.26 GHz microwave frequency. MLS amplitudes were determined from the low-field and high-field peak-to-trough measurements for each designated peak in Figure 3. PS II centers in the S₁ state were advanced to the S₂ state by continuous illumination performed at low temperature using a 600 W lamp and a 5% w/v CuSO₄ solution as an IR filter. Samples were placed in a tall test tube which was suspended in a 200 K solid CO₂/ethanol bath. An unsilvered dewar contained the 200 K bath and enabled the low-temperature illuminations.

Figure 3.

Use of the S₂-state multiline EPR signal to quantitate the S₀:S₁ ratio in three-flash (3F) samples. A 3F sample and a control S₁-state sample were continuously illuminated at 200 K for 60 min; the spectra shown above are light-minus-dark difference spectra. The illumination temperature was low enough that only the S₁ → S₂ transition could proceed; the S₀ → S₁ and S₂ → S₃ transitions were cryogenically blocked. The amplitude of the S₂ multiline EPR signal was then measured using the marked peaks. The central region corresponding to Y_D^ox has been deleted for clarity. The ratio of the induced S₂-state multiline EPR signal in the 3F sample relative to that induced in the control S₁-state sample (50%) corresponds to the percentage of centers in the S₁ state in the three-flash sample before continuous illumination. The addition of FCCP ensures that the only S-states in the 3F sample are the S₀ state and the S₁ state; therefore, the remaining 50% of the centers are poised in the S₀ state. This S-state distribution is used when deconvoluting the EXAFS spectra of the 3F samples.

EXAFS spectra were recorded on beamline 7–3 at SSRL (Stanford Synchrotron Radiation Laboratory) essentially as described in DeRose et al.¹¹ and Latimer et al.¹⁴ Additional details are provided in the Supporting Information. Two sample regions of 1.4 mm height were used, and eight scans were collected from each separate region. The samples were protected with a shutter from the X-ray beam at all times unless a measurement was in progress. Sixteen scans (eight scans per region, two regions per sample) were averaged per sample for each of six 3F samples and six control samples in the S₁ state.

Data reduction of the EXAFS spectra was performed essentially as described in DeRose et al.¹¹ and Latimer et al.¹⁴ Curve fitting was performed using ab initio-calculated phases and amplitudes from the program FEFF 7.02 from the University of Washington.^43–46 These ab initio phases and amplitudes were used in the EXAFS equation,^47–50 shown as eq 1:

$$\chi \left(k\right)={S_0}^2\sum _j\frac{N_j}{k{R_j}^2}{f}_{{\mathrm{eff}}_j}(\pi ,k,R_j)e^{-2{{\sigma }_j}^2{k}^2}{\mathrm{e}}^{-2R_j∕{\lambda }_j\left(k\right)}\sin (2k{R}_j+{\varphi }_{ij}\left(k\right))$$ (1)

S₀² is an amplitude reduction factor due to shake-up/shake-off processes at the central atom(s). This factor was set to 0.85 for all fits, on the basis of fits to model compounds.⁵¹ The neighboring atoms to the central atom(s) are then divided into j shells, with all atoms with the same atomic number and distance from the central atom grouped into a single shell. Within each shell, the coordination number N_j denotes the number of neighboring atoms in shell j at a distance of R_j from the central atom. f_effj(π,k,R_j) is the ab initio amplitude function for shell j, and the Debye–Waller term e^–2σ_j2k2 accounts for damping due to static and thermal disorder in absorber–backscatterer distances. A larger Debye–Waller factor σ_j reflects increased disorder, and leads to an exponential damping of the EXAFS oscillations. The mean free path term e^{–2R_j/λ_j(k)}, which depends on k (the momentum of the photoelectron), reflects losses due to inelastic scattering, where λ_j(k) is the electron mean free path.⁵⁰λ_j(k) was calculated by ab initio methods using FEFF 7.02. The oscillations in the EXAFS spectrum are reflected in the sinusoidal term sin (2kR_j + ϕ_ij(k)), where ϕ_ij(k) is the ab initio phase function for shell j. This sinusoidal term shows the direct relation between the frequency of the EXAFS oscillations in k-space and the absorber-backscatterer distance.

Equation 1 was used to fit the experimental Fourier isolates using N, R, and σ² as variable parameters. No firm theoretical basis exists to guide the choice of E₀ for Mn K-edges; thus, uncertainty in E₀ translates into uncertainty in k-space values (see eq S1 in the Supporting Information). Therefore, E₀ was also treated as a variable parameter. To reduce the number of free parameters in the fits, the value of E₀ was constrained to be equal for all shells in the fit. This was shown by O’Day et al.⁵² to be a valid constraint when using FEFF phases and amplitudes.

N values are defined as shown in eq 2:

$$N=\frac{\text{total number of Mn-backscatterer vectors}}{\text{number of Mn atoms per OEC}}$$ (2)

Hence, coordination numbers are evaluated on a per Mn basis and are dependent on the stoichiometry of Mn atoms in the OEC. It is well established that an active OEC contains four Mn atoms;^6,53–64 a report of six Mn per PS II⁶⁵ was most likely due to an uncorrected high residual Mn content in inactive centers present in the samples.⁶⁶ On a four Mn/PS II basis, N values for Mn–Mn interactions in the OEC come in multiples of 0.5, because each Mn–Mn interaction contains two Mn–backscatterer interactions. Other Mn–backscatterer interactions come in multiples of 0.25; that is, an N value of 2 in a Mn–O shell is interpreted as two O neighbors to each Mn atom.

The ability of the EXAFS technique to resolve the presence of similar backscatterers at closely separated distances is well known to be dependent on ΔR, the difference in absorber-backscatterer distances (Å), and Δk, the width of the k-space EXAFS data set (Å^–1). When EXAFS oscillations from two backscatterers at closely separated distances are superimposed, the addition of the sinusoidal terms from eq 1 (sin[2kR₁] and sin[2kR₂], if the phase shifts are identical) generates a local amplitude minimum in the k-space spectrum from the addition of two sine waves with different frequencies; this is commonly known as a beat. The magnitude of R₁ - R₂ determines at what value of k the beat will appear. This is shown by the trigonometric identity in eq 3:

$$\sin a+\sin b=2\cos \left(\frac{a-b}{2}\right)\sin \left(\frac{a+b}{2}\right)$$ (3)

Equation 4 shows the application of eq 3 to the current problem:

$$\sin 2k{R}_1+\sin 2k{R}_2=2\cos \left(k[R_1-R_2]\right)\sin \left(k[R_1+R_2]\right)$$ (4)

The beat in the k-space spectrum arises from the cos(k[R₁ – R₂]) term in eq 4. When the cosine function equals zero, the first beat will appear. This occurs when eq 5 is true:

$$k[R_1-R_2]=\frac{\pi }{2}$$ (5)

Thus, the minimum ΔR necessary to see a beat in the k-space spectrum is most accurately depicted by eq 6, assuming that the k-space window is wide enough to see the beat:^67–71

$$\Delta R=\frac{\pi }{2k_{\max }}$$ (6)

It should be noted that other formulas, such as ΔRΔk ≈ 1 and ΔRΔk = (π/2), are often seen in the EXAFS literature^50,72–76 and do not predict the correct resolution limit.

Fit quality was evaluated using two different fit parameters, Φ and ε ². Φ is described in eq 7:

$$\Phi =\sum _1^{N_{\mathrm{T}}}{\left(\frac{1}{s_i}\right)}^2{[{\chi }^{\mathrm{expt}}\left(k_i\right)-{\chi }^{\mathrm{calc}}\left(k_i\right)]}^2$$ (7)

where N_T is the total number of data points collected, χ^expt(k_i) is the experimental EXAFS amplitude at point i, and χ^calc(k_i) is the theoretical EXAFS amplitude at point i. The normalization factor s_i is given by eq 8:

$$\frac{1}{s_i}=\frac{{k_i}^3}{\sum _j^N{k_j}^3\mid {\chi }^{\mathrm{expt}}\left(k_j\right)\mid }$$ (8)

The ε² error takes into account the number of variable parameters p in the fit and the number of independent data points N_ind, as shown in eq 9:^77,78

$${∊}^2=\left[\frac{N_{\mathrm{ind}}}{N_{\mathrm{ind}}-p}\right]{N_{\mathrm{T}}}^{-1}\Phi$$ (9)

N_T is the total number of data points collected, and the number of independent data points N_ind is estimated from the Nyquist sampling theorem, as shown in eq 10:

$$N_{\mathrm{ind}}=\frac{2\Delta k\Delta R}{\pi }$$ (10)

Δk is the k-range of the data (3.5–11.5 Å^–1), and ΔR is the width of the Fourier-filtered peak in Å. ε² provides a gauge of whether the addition of another shell to the fit is justified. If, upon addition of a second shell, ε² becomes negative, then there are not enough free parameters available to statistically justify the inclusion of the additional shell. An acceptable model, of which there may be more than one, is any one for which ε² differs by a factor of less than two.

Results

EPR

It is critical that an independent determination of the S-state distribution of the three-flash (3F) samples is performed using EPR spectroscopy; this information is required for the deconvolution of the 3F EXAFS spectra to obtain the spectra of the “pure” S₀ state. Because FCCP was added to the samples immediately after the flashes, any centers that remain in the S₂ or S₃ states after the flash treatment are rapidly deactivated to the S₁ state,³⁷ as explained in the Materials and Methods section. Therefore, the calculated S-state distribution will contain PS II centers in only the S₀ and S₁ states. In separate experiments, it was discovered that increasing the FCCP concentration 10-fold (to 10 μM) caused an ~25% conversion of the S₁ state in the dark to the S₀ state and possibly the S_–1 state (data not shown). To confirm that the FCCP concentration used in this study (1 μM) was not of sufficient concentration to cause reduction of the S₁ state, the S₂-state EPR multiline signal was measured from parallel S₁-state samples with and without FCCP that had been continuously illuminated at 200 K. The fact that identical normalized S₂-state EPR multiline signal amplitudes were obtained from both types of samples argues against reduction of the S₁ state by FCCP under the conditions used in this study.

During the preparation of the XAS samples, eight identical 3F samples and eight identical S₁-state samples were prepared. From these 16 samples, two S₁-state samples and two 3F samples were set aside for EPR characterization. Because the samples that were set aside for EPR characterization and the respective samples used in the XAS experiments are all aliquots from the same PS II solution, the S-state distribution that was determined for the samples characterized by EPR can also be used to deconvolute the EXAFS spectra for the XAS samples.

The S₂-state multiline EPR signal generated by continuous illumination (CI) at 190 K of the 3F and S₁-state samples is shown in Figure 3. At 190 K, all S-state transitions except for the S₁ → S₂ transition are blocked.⁷⁹ This means that, in the 3F sample, the S₀ state cannot advance to the S₁ or S₂ states during this illumination procedure; thus the 3F CI samples will have a smaller normalized S₂-state multiline EPR signal than will the S₁-state CI samples. In addition, because the spectra shown in Figure 3 are difference spectra (after CI – before CI), the presence of the S₀-state multiline EPR signal,^28,41,42 which was visible in the 3F samples before CI (data not shown), will not affect the quantitations of the S₂-state multiline EPR signal. To correct for differences in sample volume, the measured S₂-state multiline EPR signals for each sample were normalized by the magnitude of their respective rhombic Fe²⁺ signal.

The results of the EPR characterization study are shown in Table 1. The ratio of the induced S₂-state multiline EPR signal in the 3F sample relative to that induced in the control S₁-state sample (50%) corresponds to the percentage of centers originally in the S₁ state in the 3F sample. Therefore, the S-state distribution assigned to the samples in the current study is 50% S₀, 50% S₁. This S-state distribution will be used in subsequent deconvolutions of the EXAFS spectra of 3F samples. To examine the effects of possible deconvolution error, two additional S-state distributions were also considered during the fitting procedure: 40% S₀, 60% S₁ and 60% S₀, 40% S₁.

Table 1.

S-State Distribution Results from Measurements of the Light-Induced S₂-State Multiline EPR Signal in Three-Flash Samples and Control S₁-State Samples^a

sample	(3-flash + hv(200 K))/(S1 + hv(200 K))	S0-state population
S0 #1	0.487	51.3%
S0 #2	0.498	50.2%

^aThe spectra used in the calculation of the values for sample S₀ #2 are shown in Figure 3.

By using samples with a final Chl concentration of 9.5 mg Chl/mL and freezing the samples immediately after the flash protocol, a greater S₀-state percentage (65% according to Table 2 from Messinger et al.³) was obtained in the 3F samples in the experiments described in Messinger et al.³ than the S₀-state percentage reported in the current study. The smaller S₀-state percentage is most likely due to the oxidation of some of the centers in the S₀ state during the 30 min centrifugation and subsequent sample handling by Y_D^ox residues which have not yet been reduced by FCCP. However, having a lower percentage of centers in the S₀ state for the S₀-state EXAFS experiments is necessary to ensure that the only other S-state is the S₁ state (which is well characterized) and to obtain highly concentrated PS II samples for EXAFS studies.

Table 2.

One- and Two-Shell Simulations of Fourier Peak I of the S₀-State Samples (Fit #7 Corresponds to the Average of the Fit Parameters from Fits #1–6, while Fits #8 and #10 Are Fits to the Average S₀-State Spectrum)

fit #	sample	shell	R (Å)	N	σ2 (Å2)	Δ E 0 b	Φ (×103)	ε2 (×105)
One Shell
1	S0 A	Mn–O	1.86	2.5c	0.005	–20	0.83	0.75
2	S0 B	Mn–O	1.85	2.5c	0.005	–20	0.47	0.40
3	S0 C	Mn–O	1.87	2.5c	0.003	–19	0.58	0.49
4	S0 D	Mn–O	1.87	2.5c	0.006	–19	0.39	0.29
5	S0 E	Mn–O	1.88	2.5c	0.004	–15	0.66	0.62
6	S0 F	Mn–O	1.85	2.5c	0.004	–20	0.59	0.55
7	average	Mn–O	1.86	2.5c	0.004	–19	0.59	0.52
8	S0 grand add	Mn–O	1.86	2.5c	0.005	–19	0.56	0.48
9	S1	Mn–O	1.84	2.5c	0.005	–20	1.47	1.61
Two Shells
10e	S0 grand add	Mn–O	1.86	2.5c	0.005	–20	0.37	0.61
		Mn–O	2.24	3.5c	0.025d
11e	S1	Mn–O	1.84	2.5c	0.004	–20	1.13	5.93
		Mn–O	2.05	3.5c	0.025d

^a Fit parameters and quality-of-fit parameters are described in Materials and Methods.

^bΔE₀ was constrained to be equal for all shells within a fit.

^cParameter fixed in fit.

^dUpper limit for parameter.

^eSee text for details.

EXAFS. EXAFS Spectra (k³-weighted) and Fourier Transforms

The average Mn K-edge EXAFS spectra from six 3F samples and six samples in the S₁ state were used to deconvolute the 3F spectrum into the EXAFS spectrum of the pure S₀ state, using the S-state distribution determined from Figure 3 (shown in Table 1 as 50% S₀, 50% S₁). These average k³-weighted spectra for the S₁ state and the deconvoluted S₀ state are shown in Figure 4. The deconvolution was performed using normalized E-space spectra before conversion into k-space. When the deconvolution was performed after both the 3F spectrum and the S₁ state were converted into k-space, the results were virtually identical (see Figures S1 and S2 in the Supporting Information). Furthermore, there is almost no difference in the k³-weighted spectra if the first two EXAFS scans from each region of the sample are compared to the last two EXAFS scans from each region of the sample (see Figure S3 in the Supporting Information). In addition, there was no discernible difference in the fits to the EXAFS data for the two k³-weighted spectra (data not shown). Therefore, the effects of photoreduction are assumed to be negligible.

Figure 4.

Average Mn K-edge EXAFS spectra from samples in the S₀ state (red) and samples in the S₁ state (black). The deconvoluted S₀-state spectrum was calculated from the 3F spectrum and the S₁-state spectrum using the quantitations from Figure 3 and Table 1 (50% S₀, 50% S₁). Six 3F samples and six S₁-state samples (16 scans per sample) were averaged for each spectrum.

The most noticeable change between the S₀-state and S₁-state spectra is that the resolution of EXAFS oscillations in the S₁ state between k = 8.5 and 11 Å^–1 is decreased in the S₀ state and the frequency of the oscillations is different between the two S-states. This can be explained by an increase in distance heterogeneity in the S₀ state relative to the S₁ state, which leads to destructive interference at higher k values of EXAFS oscillations of slightly different frequencies. This phenomenon has also been seen in other S-states that have heterogeneity in the Mn–Mn distances, such as the S₀* state,²⁴ the S₂ (g = 4.1) state,⁸⁰ the NH₃-treated S₂ state,⁸¹ the F^–-treated S₂ state,⁸² and the S₃ state.^25,83 Each of the six 3F samples that were used to generate the average spectrum shown in Figure 4 was deconvoluted separately using the average S₁-state EXAFS spectrum to generate six independent S₀-state EXAFS spectra; the results are shown in Figure S4 in the Supporting Information. This shows that the loss of resolution of the EXAFS oscillations that is seen in the average spectrum is also seen in the spectra from the six individual samples.

By performing a Fourier transform on the k³-weighted spectra, a graphical representation of the environment surrounding the Mn atoms in PS II is obtained as a radial distribution function.⁸⁴ Figure 5 shows the Fourier transforms from the S₀-state and S₁-state k³-weighted spectra shown in Figure 4. Three prominent peaks, labeled as peaks I, II, and III in Figure 5, exist in the Fourier transforms. Peak I corresponds to first-shell Mn–O interactions arising from μ-oxo-bridging and terminal ligands. Peak II arises from Mn–Mn backscattering in di-μ-oxo-bridged Mn₂ moieties. Peak III has been proposed to contain contributions from both mono-μ-oxo-bridged Mn–Mn and mono-μ-oxo-bridged Mn–Ca moieties, although other bridging motifs, such as μ-1,1-carboxylato and μ-hydroxo, are possible.

Figure 5.

Fourier transforms of the average Mn K-edge EXAFS spectra shown in Figure 4. The Fourier transform corresponding to the S₁ state is shown in black, and the Fourier transform corresponding to the pure S₀ state is shown in red.

Although peak III is relatively invariant between the S₀ and S₁ states, peaks I and II show significant differences between the S₀ and S₁ states. Peak I is at a longer apparent distance and has a slightly lower peak amplitude in the S₀ state relative to the S₁ state. This implies an increase in distance and/or distance heterogeneity in the Mn–O distances in the S₀ state relative to the S₁ state. In the S₀ state, peak II is approximately 30% lower in amplitude than in the S₁ state. This consequence of the aforementioned loss in resolution of EXAFS oscillations in the k-space spectrum of the S₀ state was also seen in the previously mentioned studies of other S-states that have heterogeneity in the Mn–Mn distances.^{24,25,80–83} This provides compelling evidence that structural changes occur during the S₀ → S₁ transition that reduce the heterogeneity of the 2.7 Å Mn–Mn distances.

Curve Fitting of EXAFS Spectra

The Fourier transforms shown in Figure 5 provide the basis for drawing qualitative conclusions about structural changes during the S₀ → S₁ transition. Reliable quantitative results can be obtained by fitting the experimental data using the EXAFS equation (eq 1), as described in the Materials and Methods section. The Fourier isolates of peaks I, II, and III are shown in Figure 6, Figure 7, and Figure 8, respectively; these isolates show the k³-weighted contributions to each Fourier peak. The increase in the apparent distance of peak I in the Fourier transforms of the S₀ state relative to the S₁ state is evident in the increase in frequency of the EXAFS oscillations in the S₀-state Fourier isolate relative to that of the S₁-state, as shown in Figure 6. For peak II, the amplitude of the S₀-state Fourier isolate in Figure 7 is significantly smaller than that of the corresponding S₁-state Fourier isolate. Because the amplitude envelope for the S₀-state Fourier isolate is significantly different from that of the S₁ state (maximum for the S₀ state is at ~8 Å^–1 and S₁ state is at ~9 Å^–1; see Figure 7), it is unlikely that the decrease in peak II amplitude can be explained by a decrease in the number of Mn–Mn interactions in the S₀ state relative to the S₁ state, which should not change the amplitude envelope. A more likely origin, which was stated earlier, is the presence of distance heterogeneity in the Mn–Mn distances in the S₀ state. If the ΔR value is smaller than the theoretical resolution limit of 0.14 Å (see Materials and Methods for details), the observed damping of the amplitude function would occur. The fit results shown in the Curve Fitting of EXAFS Spectra section show that the calculated distance separation from the fits is, in fact, close to the theoretical resolution limit (vide infra). To see a beat in the Fourier isolates at the proposed distance separation in the S₀ state, the EXAFS spectrum would have to be collected to higher k values. Unfortunately, the Fe K-edge occurs at a k value of approximately 12 Å^–1. Because there are two to three Fe atoms per OEC (one or two copies of cytochrome b₅₅₉ and one non-heme Fe²⁺)⁸⁵ and extra Fe can be present in the preparation depending on isolation conditions,⁶⁰ this makes it very difficult to collect data at higher k values.

Figure 6.

Fourier isolates from peak I (k³-weighted from 3.5 to 11.5 Å^–1) of the Fourier transforms shown in Figure 5. The S₀ state is shown in red, and the S₁ state is shown in black. The difference in the frequency of the EXAFS oscillations between the two S-states is evident. The individual Fourier peaks I, II (Figure 7), and III (Figure 8) were isolated by applying a Hamming window to the first and last 15% of the chosen range, leaving the middle 70% untouched.

Figure 7.

Fourier isolates from peak II (k³-weighted from 3.5 to 11.5 Å^–1) of the Fourier transforms shown in Figure 5. The S₀ state is shown in red, and the S₁ state is shown in black. The difference in the amplitude envelope of the EXAFS oscillations between the two S-states is evident, and can be explained by the presence of two different Mn–Mn distances with a small (<0.2 Å) separation in distance.

Figure 8.

Fourier isolates from peak III (k³-weighted from 3.5 to 11.5 Å^–1) of the Fourier transforms shown in Figure 5. The S₀ state is shown in red, and the S₁ state is shown in black.

The Fourier isolates from peak II from each individual S₀-state data set are shown in Figure S5 in the Supporting Information. These Fourier isolates show that the trends seen in the Fourier isolate generated from the Fourier transform of the average S₀-state spectrum are also seen in the Fourier isolates generated from each individual sample.

Fourier Peak I

Fits to peak I are known to be dominated by the 1.8 Å bridging Mn–O distances,¹¹ although more terminal Mn–O distances exist than bridging Mn–O distances. This is due to the two types of distances having different disorder parameters.¹¹ Thus, using one Mn–O shell, fits #1–8 in Table 2 show that the S₀ state is best fit by a 1.86 Å Mn–O distance. This is a longer distance than the corresponding fit to the S₁-state spectrum (fit #9 in Table 2), which is 1.84 Å.

To account for the two different types of Mn–O distances (bridging and terminal), two-shell fits to peak I were attempted, although some previous studies have been unable to detect the presence of the Mn–O (terminal) shell.⁵¹ Fits #10 and #11 in Table 2 show one fit minimum for a two-shell fit to peak I for the S₀ state and one for the S₁ state. A significant improvement in fit error Φ is seen for the two-shell fit. However, these two-shell fit minima are quite shallow; essentially identical fit error values were obtained if the N and Debye–Waller values for the Mn–O (terminal) shell were both much smaller (0.5–1.0 and ~0.002 Ų, respectively), or were both at an intermediate value. Irrespective of the N and σ² values, however, the R values for the Mn–O (terminal) distances in the S₀ and S₁ states were much more precise. The Mn–O (terminal) shell for the S₀ state required a long 2.2–2.3 Å distance for an acceptable fit, as compared to the 2.0–2.1 Å distance that was required for the S₁ state. A logical extension of this fit, if Mn(II) is present in the S₀ state, is to attempt a three-shell fit that separates the Mn–O (bridging) distances, the Mn(II) Mn–O (terminal) distances, and the Mn–O (terminal) distances from the other three Mn atoms into separate shells. However, this resulted in an underdetermined fit as calculated by eq 9 because of the limited width of the Fourier peak (see eq 10).

A reasonable conclusion from the fits to peak I is that there is an increase in the Mn–O distances in the S₀ state relative to the S₁ state, which is also seen in the Fourier isolates shown in Figure 6 and the Fourier transforms shown in Figure 5. This can be accounted for by the 0.02 Å increase in the average Mn–O (bridging) distance in the S₀ state relative to the S₁ state. This conclusion is consistent with, but cannot prove, the presence of a Mn(II) atom in the S₀ state.

Fourier Peak II

Relative to peaks I and III, fits to peak II are well known to have deep fit minima and thus produce the most reliable information about neighbors to Mn in the OEC.¹¹ The qualitative analysis of the S₀- and S₁-state EXAFS data (see above) indicates that a more pronounced Mn–Mn distance heterogeneity exists in S₀ state than in the S₁ state. In principle, this heterogeneity can be caused by either increased disorder of one Mn–Mn distance or the presence of two (or more) distinct Mn–Mn distances of slightly different length. We tested these two possibilities by comparing one-shell fits with two-shell fits.

Fits #1–9 in Table 3 show the results from fitting one Mn–Mn shell to peak II in both the S₀ and the S₁ states. A high disorder of one Mn–Mn distance is expected to reveal itself by high Debye–Waller factor in the one-shell fits. Table 3 shows indeed that in fits of the S₀-state data, the Debye–Waller factor is about 2–2.5 times as large as for the S₁-state data. No differences in the Mn–Mn distance are observed under these fit conditions between the S₀ and S₁ states.

Table 3.

One- and Two-Shell Simulations of Fourier Peak II from the S₀-State Samples (Fits #7 and #16 Correspond to the Average of the Fit Parameters from Fits #1–6 and #10–15, Respectively, while Fits #8 and #17 Are Fits to the Average S₀-State Spectrum)

fit #	sample	shell	R (Å)	N	σ2 (Å2)	Δ E 0 b	Φ (×103)	ε2 (×105)
One Shell
1	S0 A	Mn–Mn	2.71	0.94	0.003	–20	0.66	0.92
2	S0 B	Mn–Mn	2.72	1.31	0.005	–20	0.39	0.51
3	S0 C	Mn–Mn	2.73	1.38	0.004	–20	0.19	0.28
4	S0 D	Mn–Mn	2.74	1.56	0.005	–17	0.24	0.36
5	S0 E	Mn–Mn	2.73	1.26	0.005	–18	0.24	0.39
6	S0 F	Mn–Mn	2.74	1.38	0.004	–17	0.23	0.41
7	average	Mn–Mn	2.73	1.30	0.004	–19	0.32	0.48
8	S0 grand add	Mn–Mn	2.72	1.30	0.005	–20	0.27	0.40
9	S1	Mn–Mn	2.72	1.26	0.002	–20	0.58	0.63
Two Shells
10	S0 A	Mn–Mn	2.70	0.98	0.002c	–20	0.36	0.91
		Mn–Mn	2.86	0.35	0.002c
11	S0 B	Mn–Mn	2.71	1.02	0.002c	–18	0.24	0.54
		Mn–Mn	2.85	0.48	0.002c
12	S0 C	Mn–Mn	2.72	1.13	0.002c	–16	0.12	0.35
		Mn–Mn	2.86	0.45	0.002c
13	S0 D	Mn–Mn	2.74	1.17	0.002c	–14	0.14	0.44
		Mn–Mn	2.87	0.56	0.002c
14	S0 E	Mn–Mn	2.73	0.99	0.002c	–14	0.13	0.48
		Mn–Mn	2.87	0.46	0.002c
15	S0 F	Mn–Mn	2.74	1.16	0.002c	–13	0.12	0.65
		Mn–Mn	2.88	0.48	0.002c
16	average	Mn–Mn	2.72	1.08	0.002c	–16	0.19	0.56
		Mn–Mn	2.86	0.46	0.002c
17	S0 grand add	Mn–Mn	2.72	1.05	0.002c	–16	0.17	0.50
		Mn–Mn	2.86	0.46	0.002c
18	S1	Mn–Mn	2.74	1.51	0.002c	–14	0.47	0.73
		Mn–Mn	2.90	0.55	0.002c

^a Fit parameters and quality-of-fit parameters are described in Materials and Methods.

^bΔE₀ was constrained to be equal for all shells within a fit.

^cParameter fixed in fit.

Fits #10–17 in Table 3 show improvements in the fits for the S₀ state if peak II is treated as two separate Mn–Mn distances. Attempts to fit the second shell as a Mn–C shell and the first shell as a Mn–Mn shell produced significantly (2–3-fold) higher fit error values than the one-shell Mn–Mn fits (data not shown); thus, this scenario is considered unlikely. With two separate Mn–Mn distances, the fit error Φ decreased by 40% for the two-shell fit in the S₀ state relative to the one-shell fit, indicating that splitting the Mn–Mn distances is a valid approach. However, the improvement in Φ for the S₁ state is only 19% for the same approach.

Furthermore, the fit results from the two-shell fit (fit #18 from Table 3), which show distance heterogeneity in the Mn–Mn distances in the S₁ state for this fit, are inconsistent with the low (0.002 Ų) Debye–Waller factor for the one-shell fit in the S₁ state. Curve-fitting results from Mn model compound EXAFS data⁵¹ show that Mn–Mn shells with no distance heterogeneity exhibit Debye–Waller factor values of 0.002 Ų. Thus, it may not be meaningful to apply a two-shell fit to peak II in the S₁ state, and Mn–Mn distance heterogeneity in the S₁ state is marginal at best.

Table 4 shows the results from two-shell fits to peak II. The results from Table 4 expand on the possible inequality between N₁ and N₂ by using the constraint that there must be an integral number of Mn–Mn interactions. To test the case of constant disorder of two separate Mn–Mn distances, we fixed the Debye–Waller factor to 0.002 Ų, the value observed for the S₁-state data and in model compound studies.⁵¹ We furthermore varied only N₁ and fixed the N₁:N₂ ratio to chemically reasonable values of 1:1 (fits #1–9) or 2:1 (fits #10–18) (Table 4). This corresponds to the scenarios of two Mn–Mn interactions (N₁:N₂ = 1:1) and three Mn–Mn interactions (N₁:N₂ = 2:1), respectively. E₀ was restrained to be identical for all subshells. This allows a direct comparison of Φ (and ε² values) between these different fits. For the S₁ state, the fit quality was essentially identical in all cases, so there is no compelling reason to invoke a distance heterogeneity. This is true especially also in light of the small Debye–Waller factor of the one-shell fit (fit 9, Table 3). It is important to note that, regardless of which N₁:N₂ ratio was applied, there was no improvement in Φ or ε² for the S₁-state two-shell fits relative to the single-shell fits, indicating that, if three Mn–Mn 2.7 Å distances exist in the S₁ state, there is no resolvable distance heterogeneity in these experiments. An almost identical fit quality is also observed for the one-shell and the two-shell fits of the S₀ state, when the N₁:N₂ ratio was fixed to 1:1. However, an improvement of the fit quality by about 30% is observed for the S₀ state, if the N₁:N₂ ratio was fixed to 2:1 (fits #10–17 from Table 4). Furthermore, the N₁ and N₂ values for all of the S₀-state fits are close to 1 and 0.5, respectively. The Mn–Mn distances for these S₀-state fits are quite similar to those from Table 3, with R₁ = 2.72 Å and R₂ = 2.85 Å (ΔR = 0.13 Å).

Table 4.

Two-Shell Simulations of Fourier Peak II from the S₀-State Samples in which N₁ and N₂ for the Two Shells Were Fixed in Either a 1:1 or a 2:1 Ratio (Fits #7 and #16 Correspond to the Average of the Fit Parameters from Fits #1–6 and #10–15, Respectively, while Fits #8 and #17 Are Fits to the Average S₀-State Spectrum)

fit #	sample	shell	R (Å)	N	σ2 (Å2)	Δ E 0 b	Φ (×103)	ε2 (×105)
N1:N2 = 1:1
1	S0 A	Mn–Mn	2.67	0.48d	0.002c	–20	0.64	0.89
		Mn–Mn	2.75	0.48d	0.002c
2	S0 B	Mn–Mn	2.68	0.66d	0.002c	–20	0.35	0.46
		Mn–Mn	2.78	0.66d	0.002c
3	S0 C	Mn–Mn	2.68	0.69d	0.002c	–19	0.17	0.26
		Mn–Mn	2.78	0.69d	0.002c
4	S0 D	Mn–Mn	2.70	0.77d	0.002c	–17	0.21	0.32
		Mn–Mn	2.80	0.77d	0.002c
5	S0 E	Mn–Mn	2.68	0.62d	0.002c	–18	0.22	0.36
		Mn–Mn	2.79	0.62d	0.002c
6	S0 F	Mn–Mn	2.69	0.68d	0.002c	–17	0.21	0.38
		Mn–Mn	2.79	0.68d	0.002c
7	average	Mn–Mn	2.68	0.65d	0.002c	–18	0.30	0.44
		Mn–Mn	2.78	0.65d	0.002c
8	S0 grand add	Mn–Mn	2.68	0.65d	0.002c	–20	0.25	0.37
		Mn–Mn	2.77	0.65d	0.002c
9	S1	Mn–Mn	2.70	0.63d	0.002c	–20	0.58	0.63
		Mn–Mn	2.74	0.63d	0.002c
N1:N2 = 2:1
10	S0 A	Mn–Mn	2.69	0.97e	0.002c	–20	0.45	0.63
		Mn–Mn	2.83	0.49e	0.002c
11	S0 B	Mn–Mn	2.71	1.02e	0.002c	–18	0.24	0.32
		Mn–Mn	2.85	0.51e	0.002c
12	S0 C	Mn–Mn	2.71	1.03e	0.002c	–17	0.13	0.19
		Mn–Mn	2.83	0.51e	0.002c
13	S0 D	Mn–Mn	2.74	1.16e	0.002c	–14	0.14	0.22
		Mn–Mn	2.87	0.58e	0.002c
14	S0 E	Mn–Mn	2.73	0.98e	0.002c	–14	0.13	0.21
		Mn–Mn	2.86	0.49e	0.002c
15	S0 F	Mn–Mn	2.73	1.10e	0.002c	–14	0.14	0.26
		Mn–Mn	2.86	0.55e	0.002c
16	average	Mn–Mn	2.72	1.04e	0.002c	–16	0.20	0.30
		Mn–Mn	2.85	0.52e	0.002c
17	S0 grand add	Mn–Mn	2.72	1.02e	0.002c	–16	0.17	0.26
		Mn–Mn	2.85	0.51e	0.002c
18	S1	Mn–Mn	2.70	0.93e	0.002c	–19	0.57	0.62
		Mn–Mn	2.77	0.46e	0.002c

^a Fit parameters and quality-of-fit parameters are described in Materials and Methods.

^bΔE₀ was constrained to be equal for all shells within a fit.

^cParameter fixed in fit.

^dThe N₁:N₂ ratio was fixed to 1:1 for this fit.

^eThe N₁:N₂ ratio was fixed to 2:1 for this fit.

The above analysis favors the case where two distinct Mn–Mn distances of 2.72 and 2.85 Å contribute to peak II of the S₀-state samples. This observation is in agreement with earlier data of other samples with a decreased peak II amplitude: the NH₃-treated S₂-state⁸¹ and the S₃-state⁸³ dichroism studies on oriented PS II membranes have confirmed the presence of distance heterogeneity and two separate Mn–Mn distances in peak II, because significant differences are seen in the dichroism properties of the different Mn–Mn vectors.

The findings that a N₁:N₂ ratio of 2:1 gives clearly a better fit than a ratio of 1:1 and that the total N = N₁ + N₂ is about 1.5 have the intriguing implication that three and not two di-μ-oxo-bridged type Mn–Mn distances may exist in the OEC, and that one of them is longer in the S₀ state relative to those in the S₁ state.

The two-shell fit to peak II can also be displayed graphically as a contour plot, as shown for the S₀ state in Figure 9. In this plot, the N₁ and N₂ values corresponding to the Mn–Mn distances from each shell are fixed at the designated values for each fit. The S₀-state contour plot in Figure 9 shows that a well-defined fit minimum exists at N₁ = 1.05 and N₂ = 0.46, which corresponds to 2.1 Mn–Mn interactions at ~2.7 Å and 0.92 Mn–Mn interactions at ~2.85 Å. As expected, this is the same fit minimum that exists in fit #17 from Table 3. Thus, Figure 9 provides strong visual confirmation that the two-shell fits to peak II from the S₀ state asymmetrically distribute N₁ and N₂ values, and therefore the 2.7 and 2.85 Å Mn–Mn distances, in a 2:1 ratio. Although a clear minimum for Φ is observed for the case of two 2.7 Å and one 2.85 Å Mn–Mn vectors, the case of one 2.7 Å and one 2.85 Å Mn–Mn vector, or a site inhomogeneity are less likely, but may not be ruled out by our current data.

Figure 9.

(A) Contour plot of the Φ error parameter, showing the minimum for the two-shell peak II fit for the S₀ state. For each fit, σ² was fixed at 0.002 Ų for each shell, and N₁ and N₂ were fixed at the values corresponding to the designated number of Mn–Mn interactions; according to eq 2, the number of Mn–Mn interactions is twice the N value. The maximum value of Φ shown is 1.0 × 10^–3, and the contour step size is 2.5 × 10^–5 error units per contour. The dashed lines designate the fit minimum from fit #17 in Table 3. (B) A vertical section of the contour plot in part A taken at the fixed value of 0.92 Mn–Mn interactions at 2.85 Å (designated by a dashed line in part A). (C) A vertical section of the contour plot in part A taken at the fixed value of 2.1 Mn–Mn interactions at 2.7 Å (designated by a dashed line in part A).

Because the proposed distance heterogeneity in the S₀ state (0.13–0.14 Å) is so close to the theoretical resolution limit of 0.14 Å, it is not possible to unequivocally prove that there are two different Mn–Mn distances present in the S₀ state. This resolution limit is defined as the ability to see a k³-weighted beat in the peak II Fourier isolate, and can be improved if data can be collected to higher k values (see Materials and Methods for a discussion of the calculation of theoretical resolution limits). Thus, it is possible to explain the EXAFS data from the S₀ state in terms of an overall increase in site inhomogeneity for the 2.7 Å distances. This is a different disorder scenario from the presence of two different Mn–Mn distances, which would be expected to have very little site inhomogeneity.

Fourier Peak III

Curve-fitting results for peak III are shown in Table 5. The fit results for the individual S₀-state samples were essentially identical to those shown in Table 5 for the S₀ grand add fit (data not shown). Although the fit minima are much more shallow than those for fits to peak II, it is still possible to address some relevant questions about peak III, specifically the chemical nature of the backscatterers that contribute to this peak. A comparison of fits #1 and #2, as well as fits #4 and #5, shows that the fit results are slightly better if peak III is fit with a Mn–Mn rather than a Mn–Ca distance. Fits #3 and #6 show that the fit error parameters Φ and ε² drop by a sizable amount, approximately 50%, if an additional Mn–Ca interaction is added to the mono-μ-oxo Mn–Mn interaction. Attempts to fit peak III with a combination of a Mn–Mn shell and a Mn–C shell or a combination of a Mn–Ca shell and a Mn–C shell resulted in fit errors that were 2–3-fold higher than those for a combination of a Mn–Mn shell and a Mn–Ca shell (data not shown). A similar increase in error was observed if the N value for fit #1 or fit #4 in Table 5 was fixed to 1.0 to assess the possibility of two or more 3.3 Å Mn–Mn interactions or if peak III was fit with only a Mn–C shell (data not shown). In addition, attempts to fit peak III with one Mn–Mn shell (N = 0.5) and two Mn–Ca shells (N = 0.5) resulted in fit errors that were 20–50% higher than those shown for fits #3 and #6 in Table 5. As shown in the Fourier transforms from Figure 5, peak III is essentially invariant between the S₀ state and the S₁ state. Thus, it is not surprising that the fits are quite similar between the S₀ and the S₁ states. The Fourier isolate from Figure 8 shows a very small frequency shift between the S₀ and the S₁ states which is manifested in slightly smaller (0.02–0.03 Å) distances in the S₀ state relative to the S₁ state. However, these changes may be too small to be significant.

Table 5.

One- and Two-Shell Simulations of Fourier Peak III from the S₀-State Samples

fit #	sample	shell	R (Å)	N	σ2 (Å2)	Δ E 0 b	Φ (×103)	ε2 (×105)
1	S0 grand add	Mn–Mn	3.33	0.5c	0.004	–13	0.28	0.76
2	S0 grand add	Mn–Ca	3.39	0.5c	0.002	–13	0.32	0.87
3	S0 grand add	Mn–Mn	3.35	0.5c	0.002c	–8	0.17	0.46
		Mn–Ca	3.57	0.25c	0.002c
4	S1	Mn–Mn	3.36	0.5c	0.004	–7	0.25	0.36
5	S1	Mn–Ca	3.42	0.5c	0.002	–7	0.28	0.40
6	S1	Mn–Mn	3.37	0.5c	0.002c	–3	0.08	0.11
		Mn–Ca	3.58	0.25c	0.002c

^a Fit parameters and quality-of-fit parameters are described in Materials and Methods.

^bΔE₀ was constrained to be equal for all shells within a fit.

^cParameter fixed in fit.

Controls

All of the previous fits to peaks I, II, and III were also applied to the average EXAFS spectrum from the 3F samples (50% S₀, 50% S₁) to examine the consequences of the subtraction procedure. Because of the high signal-to-noise ratio achieved in this experiment, it is unlikely that the fit results presented in Tables 2–5 are the result of noise introduced into the data during the subtraction procedure. The fits to the EXAFS spectra from the 3F samples confirmed this, because the fit results were halfway between those reported above for the S₀ state and those reported above for the S₁ state (data not shown). Furthermore, the same fitting approach was applied to S₀-state spectra obtained from 3F-sample spectra using the alternative deconvolutions mentioned earlier (40% S₀, 60% S₁ and 60% S₀, 40% S₁), and almost identical results were found between these fits and those shown in Tables 2–5 (data not shown). This makes it highly unlikely that the observed differences between the S₀-state and S₁-state EXAFS spectra, as well as the 2:1 ratio for N₁:N₂, arise from errors in deconvolution.

As mentioned in the Materials and Methods section, the possibility was examined that the Fourier isolation technique is generating artifacts in the data, although most of the Fourier peaks in Figure 5 are well separated, which minimizes any distortion artifacts from the isolation procedure. This possibility was addressed by isolating the Fourier peaks as pairs (peaks I + II and peaks II + III) and comparing the obtained curve- fitting results to those obtained when the peaks were isolated individually. As shown in Tables S1 and S2 in the Supporting Information, no major changes in the fit results are obtained by fitting the Fourier peaks as pairs relative to the fits to the individual Fourier peaks; thus, it is rather unlikely that the fit results presented in Tables 2–5 are affected by Fourier isolation artifacts.

Discussion

Mn–Mn Distance Heterogeneity in the OEC

As mentioned in the Introduction, EXAFS experiments have shown that a major structural motif in the OEC is the di-μ-oxo-bridged binuclear Mn₂ unit.^4–12 The additional presence of mono-μ-oxo Mn–Mn and Mn–Ca motifs at a longer distance has also been shown, although the bridging motif could also be mono-μ-carboxylato or mono-μ-hydroxo.^{7,8,10–12,14} EXAFS studies of the S₁ and S₂ states revealed that all of the di-μ-oxo-bridged Mn–Mn moieties have essentially the same Mn–Mn distance of ~2.7 Å. This distance is consistent with those found in numerous studies of di-μ-oxo-bridged Mn₂(III,IV) and Mn₂(IV,-IV) complexes.^4,86,87 This provides strong evidence that the 2.7 Å distance detectable in the EXAFS spectra of the S₁ and S₂ states originates from di-μ-oxo-bridged Mn₂(III,IV) and/or Mn₂-(IV,IV) moieties.

However, other states of the OEC, including those generated by treatments with oxygen-evolution inhibitors, show that many of these states contain Mn–Mn distance heterogeneity which affects the amplitude and, in some cases, the position of peak II in the Fourier transforms. EXAFS studies of the S₂ (g = 4.1) state,⁸⁰ the F^–-treated S₂ state,⁸² and the NH₃-treated S₂ state⁸¹ all show that one of the Mn–Mn distances has increased to 2.8–2.85 Å. In the native S₃ state, EXAFS detects the presence of at least two Mn–Mn distances which have increased in length to 2.85 and 3.0 Å relative to the S₁-state distances of 2.7 Å.^25,83

The results from the current study represent the first EXAFS characterization of the native S₀ state. A previous study by Guiles et al.²⁴ characterized a chemically reduced S-state called the S₀* state, and provided evidence for Mn–Mn distance heterogeneity in that S-state in the form of a lower peak II amplitude in the S₀* state. The results from the current study show that the S₀ state has a noticeable reduction in the amplitude of peak II in the Fourier transform relative to the S₁ state, as seen in Figure 5. This can be explained by the existence of Mn–Mn distance heterogeneity in the S₀ state in the form of Mn–Mn distances at ~2.7 and ~2.85 Å. These results provide the first evidence that structural changes occur during the S₀ → S₁ transition in the OEC, which can be rationalized by a combination of two effects: μ-oxo bridge protonation and/or the presence of Mn(II) in the S₀ state.

The plausibility of μ-oxo bridge protonation in the S₀ state comes from an EXAFS spectroscopic study by Baldwin et al.⁸⁸ of a series of Mn₂(IV,IV) di-μ-oxo-bridged complexes in which the μ-oxo bridges were successively protonated. The results from this study showed that the Mn–Mn distance increases from 2.7 to 2.8 to 2.9 Å with 0, 1, or 2 μ-oxo bridge protonations, respectively. This can be explained by the fact that protonation of a μ-oxo bridge lowers the Mn–O bond order, which causes an increase in the Mn–Mn distance. Protonation of a μ-oxo bridge could occur as a consequence of the lower Mn average oxidation state in S₀ and/or due to rebinding of substrate water to the Mn cluster in the S₄ → S₀ transition.

Another effect that could increase the Mn–Mn distance is the presence of Mn(II). It is well known that the Mn–ligand distances are longer for Mn(II) complexes than they are for Mn-(III) and Mn(IV) complexes,^89–91 and the XANES and Kβ spectroscopic data from Messinger et al.³ are consistent with the presence of Mn(II) in the S₀ state. However, no Mn model complexes have been reported which contain a di-μ-oxo-bridged Mn(II) atom. The closest analogues which have been structurally characterized are Mn₂(II,II) and Mn₂(II,III) di-μ-phenoxyl-bridged complexes which have Mn–Mn distances of 3.2–3.4 Å.^92–96 In addition, a di-μ-hydroxo-bridged Mn₂(II,II) complex with a Mn–Mn distance of 3.31 Å has been characterized.⁹⁷ The long Mn–Mn distance in these complexes is most likely due to a combination of the nature of the bridging ligands and the oxidation states of the Mn ions; a di-μ-oxo-bridged moiety that incorporates Mn(II) along with Mn(III) or Mn(IV) would be expected to have a shorter Mn–Mn distance than is seen in the above-mentioned complexes due to the presence of the μ-oxo bridges.

It is also possible that both of these features, the presence of Mn(II) and μ-oxo bridge protonation, occur in the S₀ state.

Are There Three Di-μ-oxo Bridges Present in the OEC?

Analysis of the results presented in this paper shows that it is possible that three, not two, di-μ-oxo Mn–Mn moieties are present in the OEC. This conclusion is a result of the finding that the fits to Fourier peak II in the S₀ state shown in Table 4 do not favor an equal distribution of the coordination numbers N₁ and N₂ between the 2.7 and the 2.85 Å shells; an equal distribution would be consistent with the presence of two di-μ-oxo Mn–Mn moieties. Instead, clearly better results are seen when the distribution of N values is 2:1 with N_tot ~ 1.5. Although EXAFS spectroscopic results have been consistent with two to three di-μ-oxo-bridged Mn–Mn moieties, proposed models^{17,19,30–32} (including our own^{10,15,16,22,29}) have emphasized two di-μ-oxo-bridged Mn–Mn moieties.

Because the possibility that three di-μ-oxo Mn–Mn moieties exist in the OEC has not been seriously considered until now,⁷⁴ it is productive to reexamine the data already in the literature under the premise of three di-μ-oxo Mn–Mn moieties. This is most convincingly done for S-states in which Mn–Mn distance heterogeneity exists, because the N₁:N₂ ratio for these S-states can be examined. As mentioned earlier, we expect this to be more reliable than examining the total N value from EXAFS curve-fitting of the S₁ and S₂ states, which can be different from the actual N value by up to 30%.²³ The S-states of interest include the S₀* state,²⁴ the S₂ (g = 4.1) state,⁸⁰ the NH₃-treated S₂ state,⁸¹ the F^–-treated S₂ state,⁸² and the S₃ state.^25,83

The results of this reexamination are shown in Table 6, which show that none of the S-states which exhibit distance heterogeneity is best fit by an equal N value for both Mn–Mn shells; the fit results from these S-states are, in fact, more consistent with a 2:1 N₁:N₂ ratio (where N₁ corresponds to the shorter distance), as seen in the data from the current study.

Table 6.

Curve-Fitting Results for Fourier Isolates of Peak II from S-States that Exhibit Distance Heterogeneity in the Di-μ-oxo-bridged Mn–Mn Moieties

S-state	ref	R1 (Å)	R2 (Å)	N 1	N 2	N1:N2
S0* state	table III from Guiles et al.24	2.69	2.87	1.0	0.5	2:1
S2 (g = 4.1) state	table 4B from Liang et al.80	2.72	2.85	0.76	0.44	1.73:1
NH3-inhibited S2 state	table 1 from Dau et al.81	2.71	2.86	0.75	0.5	1.5:1
F–-inhibited S2 state	table 2 from DeRose et al.82	2.71	2.85	0.8	0.4	2:1
S3 state	table 1B from Liang et al.25	2.82	2.95	0.7	0.4	1.75:1
S0 state	table 3 from current study	2.72	2.86	1.05	0.46	2.28:1
S0 state	table 4 from current study	2.72	2.85	1.02	0.51	2:1

Mechanistic and Structural Consequences

The data from the current study and a reinterpretation of data already in the literature are consistent with two or three Mn–Mn di-μ-oxo bridges in the OEC, favoring three based on the N₁:N₂ ratios in samples with heterogeneity in the 2.7 Å distances. Figure 2 shows several structures that are consistent with this result. The number of possible structures can be largely reduced, when, in addition, the constraint of one 3.3 Å Mn–Mn distance is taken into account, which stems from fits to Fourier peak III in untreated and Ca-depleted PS II samples.^11,98 In this case, only structures A, E, F, G, and K are likely, because the others have either none (C, J), two (B, D, I), or three (H) ~3.3 Å Mn–Mn distances.

From the remaining structures with two di-μ-oxo-bridged Mn–Mn moieties (A, E, F, and K), structure A was favored for a long time because of its simplicity.¹⁰ However, on the basis of simulations of EPR spectra from the S₂ state, Peloquin et al.³² prefer structures E or F, because these structures provide better rationale for the strong exchange coupling between two di-μ-oxo Mn–Mn moieties.

Model G (Figure 2) is one of the few proposed structures for the OEC that contain three di-μ-oxo type Mn–Mn moieties and one 3.3 Å type Mn–Mn distance.¹¹ This is not surprising because only a few models can be constructed with four Mn ions under these geometrical constraints. However, by moving one of the μ-oxo bridges around, the two new structural models L and M can be derived (Figure 10). A structure similar to G was recently suggested on the basis of simulations of EPR spectra from the S₂ state by Kusunoki and co-workers.^99,100 In addition, all three structures shown in Figure 10 appear to be consistent with the dimensions of the electron density envelope seen in the 3.8 Å X-ray structure data of the Mn cluster in PS II.^33,101 However, as confirmed by Mn and Sr EXAFS studies,^14,15 the OEC is most accurately described as a Mn/Ca heteronuclear cluster; therefore, Ca should be incorporated into each of the proposed structures in Figure 10 so that 1–2 Mn–Ca vectors exist which are oriented close to the membrane normal.¹⁰² It should be noted that Ca has not yet been detected in PS II by X-ray crystallographic studies.

Figure 10.

Modified structural models for the active site of the OEC in PS II poised in the S₀ state. Structures L and M are modifications of I or G from Figure 2. Each model contains a μ-oxo or a μ-hydroxo bridge colored blue as a suggestion for the Mn–Mn moiety giving rise to the 2.85 Å Mn–Mn distance; the two remaining di-μ-oxo Mn–Mn distances are 2.7 Å. A possible site of the proposed oxyl radical in the S₃ state is denoted in red for each model (see text for details). Mn oxidation states in the S₀ state can be either Mn₄(III₃,IV) or Mn₄(II,III,IV₂).

S₀ State

As drawn, the models in Figure 10 are depicted in the S₀ state. On the basis of the detailed discussion in Messinger et al.,³ the Mn oxidation states in the S₀ state are most likely either Mn₄(II,III,IV,IV) or Mn₄(III,III,III,IV). The data from the current study are consistent with both of these options. In Figure 10, only models with three 2.7 Å type Mn–Mn distances and one 3.3 Å vector are shown. As discussed above, models A, E, F, and K are also still possibilities, and similar arguments apply to those given below.

In the S₀ state, the presence of Mn(II) and/or protonation of a μ-oxo bridge can account for the increase of one 2.7 Å Mn–Mn distance to 2.85 Å, as explained above. However, in G, the presence of Mn(II) in the S₀ state would cause two out of three 2.7 Å Mn–Mn distances to increase, not one out of three. Because this is inconsistent with the EXAFS fits presented in this paper, the use of G necessitates assigning the oxidation states of Mn in the S₀ state as Mn₄(III₃,IV) and protonating one of the μ-oxo bridges in the S₀ state (shown in blue in Figure 10).

Each of the remaining two structures in Figure 10, L and M, can incorporate Mn(II) in the S₀ state and increase only one of the Mn–Mn distances; one possibility is shown in Figure 10. Because it is unclear from model-compound chemistry whether protonation of a μ-oxo bridge would be required, the proton is denoted in blue in Figure 10 as optional for both models; however, protonation of a μ-oxo bridge would most likely be necessary if the Mn₄(III₃,IV) oxidation state option is invoked.

S₃ State

EXAFS experiments on the S₃ state suggest that all of the 2.7 Å Mn–Mn distances in the S₂ state increase in distance in the S₃ state.²⁵ This was explained by Yachandra et al.¹⁶ and Liang et al.²⁵ using A in Figure 2 as a structural framework and incorporating an oxyl radical in a μ-oxo bridging position in the S₃ state. This provided an explanation for an increase in distance of one di-μ-oxo-bridged Mn–Mn moiety from 2.7 Å in the S₂ state to 3.0 Å, but it was difficult to understand why, in A, the other di-μ-oxo Mn–Mn moiety also increased in distance in the S₃ state even though it was somewhat isolated from the proposed oxyl radical. Thus, it would be more logical if the structure of the OEC was, in fact, more “tied together” than is shown in A, which would more easily explain the lengthening of all di-μ-oxo Mn–Mn motifs in the S₃ state.

From the topological models in Figure 10, model G seems best suited to understand the structural changes during the S₂ to S₃ transition. If G is used, formation of an oxyl radical at the oxygen denoted in red in Figure 10, for example, would give rise to the longer ~3.0 Å Mn–Mn distance in the S₃ state. The lengthening of the other two di-μ-oxo Mn–Mn moieties can be explained if some of the spin density of the oxyl radical in G is present on the μ₃-oxo bridge or the other μ₂-oxo bridges.

Similarly, one possibility for the di-μ-oxo-bridged oxygen which becomes the oxyl radical and gives rise to the 3.0 Å Mn–Mn distance in the S₃ state is shown in red for structure L and M. In a fashion similar to what was proposed above for G, the increase in the other Mn–Mn distances can be rationalized by some of this spin density being present on the μ₃-oxo-bridged oxygen and/or the other μ₂-oxo-bridged oxygens.

Conclusions

The data in this study suggest that three di-μ-oxo Mn–Mn motifs may be present in the OEC. The new topological models shown in Figure 10 evolved from the results presented in this paper and represent new structural possibilities for the OEC that have not been widely considered in previous studies. It is expected that significant insights will come from evaluation of experimental data in terms of these new models. This will be particularly revealing for the interpretations of EXAFS data from oriented PS II membranes in various S-states^7,12,81,83 and the interpretation of the ENDOR, ESEEM, and continuous-wave EPR spectra from the S₂ and S₀ states.^32,103