Structural Change of the Mn Cluster during the S₂→S₃ State Transition of the Oxygen-Evolving Complex of Photosystem II. Does It Reflect the Onset of Water/Substrate Oxidation? Determination by Mn X-ray Absorption Spectroscopy

Abstract

The oxygen-evolving complex of Photosystem II in plants and cyanobacteria catalyzes the oxidation of two water molecules to one molecule of dioxygen. A tetranuclear Mn complex is believed to cycle through five intermediate states (S₀–S₄) to couple the four-electron oxidation of water with the one-electron photochemistry occurring at the Photosystem II reaction center. We have used X-ray absorption spectroscopy to study the local structure of the Mn complex and have proposed a model for it, based on studies of the Mn K-edges and the extended X-ray absorption fine structure of the S₁ and S₂ states. The proposed model consists of two di-μ-oxo-bridged binuclear Mn units with Mn–Mn distances of ~2.7 Å that are linked to each other by a mono-μ-oxo bridge with a Mn–Mn separation of ~3.3 Å. The Mn–Mn distances are invariant in the native S₁ and S₂ states. This report describes the application of X-ray absorption spectroscopy to S₃ samples created under physiological conditions with saturating flash illumination. Significant changes are observed in the Mn–Mn distances in the S₃ state compared to the S₁ and the S₂ states. The two 2.7 Å Mn–Mn distances that characterize the di-μ-oxo centers in the S₁ and S₂ states are lengthened to ~2.8 and 3.0 Å in the S₃ state, respectively. The 3.3 Å Mn–Mn and Mn–Ca distances also increase by 0.04–0.2 Å. These changes in Mn–Mn distances are interpreted as consequences of the onset of substrate/water oxidation in the S₃ state. Mn-centered oxidation is evident during the S₀→S₁ and S₁→S₂ transitions. We propose that the changes in Mn–Mn distances during the S₂→S₃ transition are the result of ligand or water oxidation, leading to the formation of an oxyl radical intermediate formed at a bridging or terminal position. The reaction of the oxyl radical with OH⁻, H₂O, or an oxo group during the subsequent S state conversion is proposed to lead to the formation of the O–O bond. Models that can account for changes in the Mn–Mn distances in the S₃ state and the implications for the mechanism of water oxidation are discussed.

Introduction

The oxidation of water to dioxygen in plants and cyanobacteria involves the stepwise transfer of oxidizing equivalents by the Photosystem II (PS II)¹ reaction center to the quinone electron acceptors. Photon absorption at PS II results in the transfer of electrons sequentially until four electrons are removed from two water molecules to produce one molecule of dioxygen. The oxidizing potential produced by photosynthetic charge separation at PS II is thought to accumulate at a site in the membrane-bound proteins denoted the oxygen-evolving complex (OEC).^2–5 A tetranuclear Mn complex in the OEC is believed to cycle through five intermediate states S_i (i = 0–4), with i representing the number of oxidative equivalents accumulated in the OEC.⁶ After long-term dark adaptation (tens of minutes), nearly all of the PS II centers are present at the S₁ state. The postulated S₄ state is a transient intermediate that spontaneously returns to the S₀ state with the release of dioxygen.

Investigation of the structure of the Mn complex at each S state is essential for understanding the catalytic mechanism of photosynthetic oxygen evolution. We have used X-ray absorption spectroscopy (XAS) in conjunction with electron paramagnetic resonance spectroscopy (EPR) to study the oxidation state(s) and structural changes in the Mn complex as the OEC advances through the S state cycle.⁵ XAS is a convenient technique for studying the structure of the Mn complex from PS II preparations that are prepared in the various S states and frozen as solutions. The energy of the incoming X-ray beam and of the outgoing fluorescence is specific to the Mn atom; hence, other metals or the protein matrix normally copurified with the OEC in a PS II preparation do not interfere. X-ray absorption near-edge structure (XANES) is sensitive to both the oxidation state and the symmetry of the Mn cluster. Extended X-ray absorption fine structure (EXAFS) provides information about the types, numbers, and distances of the neighboring backscattering atoms from the absorbing Mn atoms.^7–9

We have proposed a “dimer-of-dimers” model for the structure of the Mn cluster based on studies of the Mn K-edges and the extended X-ray absorption fine structure (EXAFS) of the S₁ and S₂ states.¹⁰ This model represents one of the simplest explanations of the EXAFS data but is not unique.^5,11,15 The Mn cluster is proposed to consist of two di-μ-oxo-bridged Mn–Mn binuclear units which are linked to each other by a mono-μ-oxo-bridging ligand. The Mn–Mn distance is ~2.7 Å within each of the di-μ-oxo units, and is ~3.3 Å when the Mn atoms are linked by the mono-μ-oxo-bridging ligand.

XAS experiments on the S₃ and S₀ states have been difficult for the PS II samples generated by single flashes. The requirement of optically dilute samples to ensure saturation by single actinic flashes generally results in a low signal-to-noise ratio of the Mn X-ray fluorescence. To explore the structures beyond the S₂ state, X-ray absorption spectra have been collected on chemically treated S₃ and S₀ states accumulated by continuous illumination at cryogenic temperatures.^12,13 Reduced derivatives of the Mn cluster prepared by treatment with NH₂OH or hydroquinone have also been investigated using XAS methods.^14,15 Recently, we have been able to collect a series of Mn K-edge and EXAFS data from flash-induced samples prepared under physiological conditions. Furthermore, pure S state XANES spectra of high quality were extracted from linear combinations of the flash-induced XAS spectra.¹⁶ Each of the S₀→S₁ and S₁→S₂ transitions is accompanied by a shift of the edge to higher energies, indicating that during those transitions the Mn cluster undergoes oxidation. The S₂→S₃ transition, however, results in a very small edge shift, consistent with the absence of a direct Mn oxidation step. This suggests that the observed edge shift during the S₂→S₃ transition may be attributed to ligand/substrate oxidation of the Mn cluster rather than to direct Mn oxidation. However, on the basis of Mn K-edge studies, other groups have proposed that Mn is oxidized during the S₀→S₁, S₁→S₂, and also during the S₂→S₃ transitions.^17,18

The requirement of a higher signal-to-noise ratio for the EXAFS relative to the XANES experiments has made EXAFS measurements more difficult for the flash-induced S state samples. Earlier EXAFS data from the chemically treated S₃ state samples produced by a double-turnover method indicated that the two di-μ-oxo-bridged Mn–Mn dimer units may become nonequivalent.¹² The EXAFS spectra from calcium-depleted S₃′ samples (S₂Y_Z^•) prepared by low-pH treatment in a citrate buffer do not exhibit similar heterogeneity in the 2.7 Å Mn–Mn distances.^19,20 However, when NaCl treatment is used for Ca depletion, Evans and co-workers have shown that the 2.7 Å Fourier feature attributed to Mn–Mn interaction is split into 2.7 and 3.0 Å features that are most apparent in the S₃′ state.²¹ We have previously reported preliminary Fourier transforms of the S₃ state samples prepared under physiological conditions with saturating flashes.¹⁹ Since then, data have been collected and analyzed on more samples. In this report we present a detailed analysis of the structural changes observed for the S₃ state and the implications for the mechanism of water oxidation.

Material and Methods

Preparation of PS II-Enriched Membranes

PS II-enriched membranes were prepared from spinach by a modified BBY protocol.²² The incubation time for detergent treatment was 1 min. The pelleted PS II-enriched membranes were washed in medium A (15 mM NaCl, 5 mM MgCl₂, 5 mM CaCl₂, and 50 mM MES at pH 6.5) with 0.4 M sucrose and centrifuged for 30 min at 37000g. After one more wash and centrifugation at 37000g for 30 min, pelleted PS II was resuspended into medium A with 50% (v/v) glycerol to a Chl concentration of 4–5 mg/mL. Typical oxygen evolution activities were 400–550 μmol of O₂ mg⁻¹ of Chl h⁻¹.

A freshly prepared stock solution (50 mM in dimethyl sulfoxide) of phenyl-p-benzoquinone (PPBQ) was added to the PS II samples to a final concentration of 500 μM just before flash illumination. PPBQ (Eastman Kodak) was recrystallized from ethanol. PS II membrane suspensions were loaded directly into Lucite sample holders with inner dimensions of 18 × 2.5 × 0.8 mm. All of the flash illuminations, EPR, and X-ray absorption measurements were performed directly on samples mounted in these holders. The samples were dark-adapted for at least 1 h at 4 °C before being frozen to 77 K in liquid nitrogen. The amplitude of the Y_D^ox EPR signal was measured and used to correct for sample volume fluctuations. The samples were then equilibrated to room temperature (22–24 °C) in the dark for at least 20 min before the flash illumination procedure was applied.

Oxygen Evolution Measurements

Oxygen evolution activity was measured by using a Clark-type oxygen electrode (Yellow Springs Instruments, Yellow Springs, OH) as described by DeRose.²³ A 5% CuSO₄ solution in a spherical flask was used to focus the light onto the cuvette and to serve as a heat filter.

Flash-Induced Illuminations of PS II

The source was either a Xe flash lamp (CHH 174, ILC Inc.) which generated flashes of white light of 14 μs full-width at half-maximum (fwhm), or a frequency-doubled (532 nm) Nd:YAG laser of 8 ns fwhm. To maintain maximal synchronization of the PS II centers upon flash illumination, fast recombination reactions between both the S₂ and S₃ states and the reduced form of the redox-active tyrosine residue Y_D must be suppressed. This was achieved by the application of two preflashes, followed by a 60-min dark adaptation period at room temperature. This procedure synchronizes almost all PS II centers into the S₁Y_D^ox state. Each sample was next given 0, 1, 2, 3, 4, or 5 flashes, at room temperature, with 1.5 s intervals between individual flashes. The light was directed to the samples by a light guide when the Xe flash lamp was used for illumination, or focused using cylindrical lenses when the YAG laser was employed. After the last flash, the samples were frozen immediately (within 1 s) in liquid nitrogen. The EPR spectra (both Y_D^ox and the MLS) were collected, and the samples were stored at 77 K for further use in the XAS experiments (within 7–10 days).

Light saturation of the PS II samples was ensured in two ways. First, the S₂ state MLS amplitude induced by one flash was compared with that of an identical sample subjected to a low-temperature continuous illumination (10 min at 190 K) commonly used to generate a maximal S₂ state MLS. These amplitudes were indistinguishable within the accuracy of the MLS amplitude determination. We additionally determined the flash-intensity dependence of the MLS amplitude for a single flash. When only 60% of the maximal flash intensity was used, the MLS was 95% of the maximal amplitude. Reducing the energy to 25% resulted in 80% of the maximum MLS. Both observations indicate that the full-intensity flashes were nearly saturating, so that a reasonably small value for the miss parameter can be expected.

EPR Measurements

Low-temperature X-band EPR spectra were recorded using a Varian E109 EPR spectrometer equipped with a model 102 microwave bridge. For the MLS measurement, the sample temperature was maintained at 8 K using an Air Products LTR liquid helium cryostat. Spectrometer conditions were as follows: microwave frequency, 9.21 GHz; field modulation amplitude, 3.2 mT at 100 kHz; microwave power, 11 mW. EPR MLS amplitudes were quantitated by adding peak-to-trough amplitudes of three of the downfield and three of the upfield hyperfine lines, relative to g = 2. Sample temperature was kept at 18 K for the measurements of the Y_D^ox signal, with the following spectrometer conditions: microwave frequency, 9.21 GHz; field modulation amplitude, 3.2 mT at 100 kHz; microwave power, 0.5 μW. The field modulation amplitude of 3.2 mT gives rise to a modulation-broadened Y_D^ox signal that is easily quantitated. After the XAS experiments the EPR spectra of the samples were remeasured to ascertain the possible damage induced by the intense X-ray beam. The amplitude of the MLS was typically ≥80% relative to that measured before the X-ray exposure. In addition, no free Mn²⁺ EPR signal was detected in any of these samples. This indicated that no significant radiation damage occurred during the data collection.

X-ray Absorption Measurements

The manganese EXAFS spectra were collected at the beamlines 7-3, 4-2, 6-2, and 10-2 at the Stanford Synchrotron Radiation Laboratory (SSRL) (SPEAR operating at 40–100 mA at 3.0 GeV); some data were collected at the bending magnet beamline X-9B at the National Synchrotron Light Source (NSLS) (100–250 mA at 2.5 GeV). An unfocused X-ray beam was used on 7-3 at SSRL and at X9B at NSLS, while a focused beam was used for measurements at 4-2, 6-2, and 10-2 at SSRL. A Si 〈220〉 double-crystal monochromator was used, with the second crystal detuned to 50% of the maximum flux to reduce the transmission of the harmonics. The sample was maintained at 10 ± 1 K in a liquid helium cryostat (Oxford Instruments, CF1204). A 13-element Ge solid-state detector (Canberra Instruments) was used²⁴ to detect the Mn fluorescence signal.²⁵ We set different SCA windows for the each of the three concentric circles of detector elements, with shaping times of 1, 0.75, and 0.5 μs, based on the total number of counts each channel detects. The shaping time for the center channel, which “sees” the fewest total counts, is set at 1.0 μs, and for the outer channels that detect more total counts, because of increased scatter, the shaping time was set at 0.5 μs. This kind of attention to detail is necessary for collecting data from these dilute samples with good S/N ratios. The total number of fluorescence counts in the SCAs above the Mn K-edge (at 6600 eV) is ~2000/s from the dilute samples, compared to ~10000/s for the more concentrated samples. The dilute samples were 5–6 mg of Chl/ml, whereas the concentrated samples are estimated to be 25–30 mg of Chl/ml.

EXAFS spectra were recorded from 6500 to 7100 eV, with higher point density around the pre-edge region for better KMnO₄ energy referencing. Typically, 25–40 scans of about 20 min each were collected and averaged for each S₃ spectrum. The Mn K-edge was monitored closely to detect any photoreduction. Energy calibration and resolution were monitored by simultaneously measuring the absorption spectrum of KMnO₄, using its narrow pre-edge “white-line” at 6543.30 eV. The X-ray photon flux on the samples was kept below a maximum flux density of 7 × 10⁷ photons s⁻¹ mm⁻² of PS II sample. We have examined the effect on the various S states, and this is the maximum photon density, in the energy region between 6500 and 7100 eV, that one can safely use, within the conditions of temperature, pressure, and cryo-protectant concentration as described above.

EXAFS Data Analysis

EXAFS data were analyzed as reported previously.^11,26 A brief outline follows. The EXAFS modulation of the absorption χ(k) is described by the equation

$$\chi (k)={S_0}^2\sum _i{N}_i{B}_i(k)\frac{f_{\text{eff}}(\pi ,k,R_i)}{{{kR}_i}^2}exp(-2{\sigma }^2{k}^2)exp\left(\frac{-2R_i}{\lambda (k)}\right)\times sin[2{kR}_i+2{{\delta }_i}^{\mathrm{c}}(k)+{\varphi }_i(k)]$$ (1)

where for each shell i, N_i is the number of scatterers at a distance R_i, S₀² is the many-body amplitude reduction factor, B_i(k) is an amplitude reduction factor caused by inelastic forces in the central atom, f_eff is the effective backscattering amplitude of the scattering atom, δ_i^c and ϕ_i are the phase shifts for the absorber and backscatterer, respectively, σ² is the Debye–Waller term describes the attenuating effect resulting from thermal and static disorder, and λ(k) is the mean free-path of the photoelectron. The basic structure used for calculation of FEFF5.05 fitting functions (f_eff, δ, ϕ, λ) was a di-μ-oxo-bridged binuclear structure linked to a similar moiety by a mono-μ-oxo bridge. k is the magnitude of the photoelectron wave vector given by k = (2π/h)[2m_e(E − E₀)]^1/2, where m_e is the electron mass, E is the X-ray photon energy, E₀ is the threshold photoionization energy, and h is Planck’s constant. All spectra were weighted by k³, and Fourier transforms were obtained from k-space spectra in the range of ~3.5–11.5 Å⁻¹. A three- or four-domain cubic spline was applied to the k³-weighted data to minimize background contributions. For some data sets, a two-domain spline was applied in energy space to remove the background contribution. For such data sets, no further background removal was applied to the k-space data. A window function was applied to isolate Fourier peaks individually and in pairs to minimize the distorting effects due to Fourier isolation. These isolated Fourier peak(s) were back-transformed into k-space data and used for curve fitting. Further details of the EXAFS analysis have been reported previously.^20,23,27

The Fourier isolates were fit with theoretical values for the backscattering amplitude and phase shifts for the scattering–absorbing pairs, based on values calculated using the FEFF5.05 program.^28–30 Details of parameters using the FEFF5.05 program have been reported previously.²⁶ Coordination numbers N_i are calculated on a per Mn basis and are interpreted here in the context of a total of four Mn atoms per PS II. This means that a single Mn–scatterer interaction in the complex would result in a coordination number of 0.25. More scatterers would appear in multiples of 0.25. However, in the case of Mn–Mn interactions, one Mn–Mn interaction would give a minimum coordination number of 0.5, since both of the Mn atoms would “see” each other at the same distance. More Mn–Mn interactions at the same distance would give multiples of 0.5 for the coordination number. In this work the Debye–Waller parameters were allowed to vary over a range comparable with those derived from synthetic multinuclear Mn compounds. The energy of the absorption edge (E₀) was varied in simulations to allow better phase matching to the experimental waves. The values of the average distance R_i, the numbers of scattering atoms N_i at distance R_i, the Debye–Waller factor σ², and the threshold energy E₀ were simultaneously fit with a nonlinear least-squares fitting program.^20,27,31

The normalized error sum (Φ) used to verify the quality of the fits is given by the equation

$$\mathrm{\Phi }=\sum _i^N{\left(\frac{1}{s_i}\right)}^2{[{\chi }^{\text{expt}}(k_i)-{\chi }^{\text{calc}}(k_i)]}^2$$ (2)

where N is the number of data points, and χ^expt(k_i) and χ^calc(k_i) are the experimental and calculated EXAFS. The normalization factor s_i is defined as

$$\frac{1}{s_i}={k_i}^3/\sum _j^N{k_j}^3\mid {{\chi }_j}^{\text{expt}}(k_j)\mid$$ (3)

The ε² error accounts for the number of variable fit parameters (p), and the number of independent data points (N_ind) is estimated to be equal to 2ΔkΔR/π, where Δk is the k range of the data used and ΔR is the width of the Fourier-filtered peak.^32,33

$${\epsilon }^2=[N_{\text{ind}}/(N_{\text{ind}}-p)]N^{-1}\mathrm{\Phi }$$ (4)

A negative value of ε² indicates that the fit is underdetermined, and the fit solution is considered not unique. Generally, including more shells (which increases the value of p) results in a decrease in Φ. The ε² value therefore indicates whether fits were under- or overdetermined and whether the quality of fit is improved upon inclusion of extra shells.

Results

Synchronization of the S State Formation

The PS II samples were treated with PPBQ and given two preflashes to prevent loss of the S₂ and S₃ states by two processes of recombination. One path is decay of the S₂ and S₃ states to the S₁ state by recombination with the electrons from plastoquinone Q_B⁻.^34,35 In the presence of exogenous electron acceptor PPBQ, the half-time for the S₂ and S₃ state recombinations increased to ~3.5 min.³⁶ The flash interval in our experiment was typically 1.5 s, and the PS II samples were frozen within 1 s in liquid nitrogen after the last flash. In the presence of PPBQ, the loss of S₂ and S₃ owing to recombination after one or two flashes was therefore negligible.

A second path for S₂ and S₃ state loss is through fast recombination with reduced Y_D in the dark.^37–40 Because these dark recombination reactions happen in seconds, the interval between the Xe lamp or Nd:YAG laser flashes (~1.5 s) in our experiments cannot be ignored. The preflash treatment oxidizes most of the Y_D to Y_D^ox, as shown by a 30–50% increase in the amplitude of the Y_D^ox EPR signal.¹⁶ Therefore, the PS II centers are synchronized to the S₁Y_D^ox state before the samples receive the subsequent flashes, thus minimizing Y_D recombination with higher S states. The application of 0–5 flashes after the preflash protocol did not lead to a further increase of the Y_D^ox signal, indicating that, at that point, the recombination between reduced Y_D and the higher S states did not occur to any significant extent.

Extraction of S₃ EXAFS

Sets of PS II samples were synchronized in the S₁ state before being given a series of 0–5 flashes. The S₂ multiline EPR signal (MLS), generated after one flash, oscillates with a period of four, and can thus be used to precisely quantitate the S state distribution for each flash number. The actual S state distributions in samples given a particular number of flashes were obtained from the best fit of the flash-induced multiline EPR signal (MLS) oscillation pattern to the Kok model.⁶ The first flash induced the maximal amount of MLS. The MLS oscillation pattern exhibits a periodicity of four, with maxima after the first and fifth flash, as expected.⁴¹

Assuming that after the preflash protocol the samples have a 100% S₁ population, the Kok model cast in terms of only two parameters (misses and double hits) describes the damped period-four MLS oscillation well. The best fit was found with 12% misses and 5% double hits for Xe flash lamp illumination and 14% misses and 0% double hits when a Nd:YAG laser was used for flash illumination.¹⁶ Using these fit values, the S state distributions in samples given two flashes was calculated. This independent information on the S state distributions allows us to extract reliably the EXAFS spectra of the pure S₃ state from the experimentally acquired spectra of samples given two flashes. The best fit to the MLS oscillation pattern usually led to ~70% (for the Xe flash lamp illumination) and ~75% (for laser flash illumination) yield of the S₃ state after two flashes. In samples given two Xe lamp flashes the composition was 8% S₀, 2% S₁, 20% S₂, and 70% S₃, and from the Nd:YAG laser the composition was 0% S₀, 2% S₁, 24% S₂, and 74% S₃. The S₀ contribution was neglected for samples made using Xe flash lamp illumination. Because the EXAFS spectra of S₁ and S₂ states are nearly identical,¹¹ we used either 22% (Xe flash) or 26% (YAG laser) S₂ spectra. The S₂ state spectrum used for subtractions was obtained using a concentrated sample that was continuously illuminated at 190 K to produce the S₂ state characterized by the multiline EPR signal. This protocol was followed to minimize the contribution of noise that is introduced as a consequence of subtraction procedures. There were no discernible differences in the S₃ state spectra obtained by the two flash protocols.

The S₂ state contribution to the spectrum of a sample given two flashes, determined using MLS data as described above, was subtracted from the total spectrum and renormalized to yield the S₃ state EXAFS spectrum. There are no significant differences in the EXAFS spectra generated by subtraction in E-space or k-space. We present only the S₃ spectra generated by using the raw k-space data. Fourier transformation and isolation of individual Fourier peaks for the S₃ state was carried out on EXAFS spectra calculated using raw k-space data generated as described above.

Nine independent S₃ spectra collected from five X-ray runs were constructed and are presented as six data sets (S₃A–F) in this report. Each of S₃A, S₃B, and S₃C was the average of two S₃ sample spectra.

EXAFS of the S₃ State

The k³-weighted EXAFS spectra of representative S₃ and S₂ states are presented in Figure 1. Relative to the S₂ state, the spectra of the S₃ state samples show an overall difference in phase and amplitude within the data range of 3.5 to 11.5 Å⁻¹. We have previously observed changes in the modulations in the S₂ g = 4.1 and annealed ammonia-treated S₂ state samples.^42,43 In both cases, a Mn–Mn distance increase in one of the two di-μ-oxo dimanganese units was observed. Figure 1 clearly shows that there is an increase in the frequency of the oscillations in the S₃ state compared to the S₂ state as shown by the progressive change in phase of the oscillations toward higher k values between the two states. This result is indicative of an increase in distance for one or more of the scattering interactions. Figure 2 shows the Fourier-filtered back-transform of peaks I, II, and III (0–4 Å) of Figure 3 for the S₃ and S₂ states, where the differences in phase and amplitude between the two states are clearly evident. Interestingly, the EXAFS spectrum from a sample given two flashes is very similar to the calculated S₃ spectrum shown in Figure 1 (data not shown). This is not surprising because the two-flash sample is predominantly S₃ in composition (70–75%). However, this is important as it shows that subtraction used to generate the pure S₃ state spectrum does not introduce artifacts into the data.

Figure 1.

Background-subtracted k-space EXAFS data from S₂ (dashed line) and S₃ (solid line) states of PS II samples. The S₃ state spectrum was calculated from the spectrum obtained from a sample illuminated by two flashes by subtracting the residual S₂ state contribution. The amount of S₂ spectrum subtracted was determined by quantitating the amount of multiline EPR signal in samples given two flashes. The details of the quantitation are described in the Results section. These data have been multiplied by k³. Differences in the phase and frequency of the EXAFS modulations and in the amplitude extend over the entire range of the spectra.

Figure 2.

Fourier-filtered k-space EXAFS data from S₂ (dashed line) and S₃ (solid line) states of PS II samples (see Figure 1 caption and text for details of calculation of the S₃ state EXAFS spectrum). The Fourier window extended over the three Fourier peaks labeled I, II, and III (0–4 Å) shown in Figure 3. The differences in phase, frequency, and amplitudes between the raw S₂ and S₃ state EXAFS spectra (Figure 1) become very obvious in these Fourier-filtered spectra, where the high-frequency noise has been removed.

Figure 3.

Fourier transform power spectra of S₂ (dashed line) and S₃ (solid line) states of PS II. The Fourier transform of the S₃ state was obtained from the calculated S₃ state EXAFS shown in Figure 1 (see text for details). The FTs are of the k³-weighted EXAFS data from 3.5 to 11.5 Å⁻¹ shown in Figure 1. The major Fourier peaks are labeled I, II, and III. The spectra are clearly different between the S₂ (dashed line) and S₃ states (solid line). The arrow points to a shoulder between peaks I and II at an apparent distance of 2.0 Å. There is a reduction in amplitude in all three peaks in the S₃ state compared to the S₂ state. More importantly, peaks II and III are at a greater apparent distance R′ for the S₃ state compared to the S₂ state as shown. The upper part of the figure shows an expanded version of the Fourier transform from 0 to 4 Å to show the differences in amplitude and apparent distance.

Fourier transforms (FTs) of EXAFS data for pure S₂ and S₃ states are shown in Figure 3. The positions of the three peaks labeled I–III correspond to the shells of scatterers at different “apparent” distances from the Mn absorbers. These apparent distances, R′, are shorter than the actual distances, R, due to an average phase shift induced by the potential of the given absorber–scatterer pair on the photoelectron.

In the S₂ state, peak I has been assigned to two N/O ligand atoms at a distance of ~1.8 Å and ~2–4 O/N ligand atoms at distances between 1.95 and 2.15 Å. The main contribution to peak II is from the backscattering of 1.2 ± 0.2 Mn neighbors at 2.7 Å from the Mn absorber. Peak III has been assigned to contributions from both Mn and Ca backscatterers at about 3.3 and 3.4 Å, respectively.^10,26,44 Studies from other groups prefer the assignment of peak III to backscattering from Mn or Ca at 3.3 Å,⁴⁵ or from Ca at 3.7 Å.⁴⁶

Presented in the lower part of Figure 3 are Fourier transforms from 0 to 10 Å. The data at R′ > 4 Å are a measure of the noise in the data showing the quality of the data obtained for the S₂ and S₃ states. The FTs from the six S₃ data sets all show peaks I and II that are significantly smaller in amplitude than the corresponding peaks of the S₂ spectrum. For the S₃ state, the amplitude of peak II is consistently smaller than that of peak I in all of the data sets. The vertical lines in Figure 3 emphasize that peaks II and III of the S₃ spectra are shifted to longer distances compared to those observed in the S₂ data. This is more evident in the expanded Fourier transform from 0 to 4 Å shown in the upper part of Figure 3. The distance increase is consistent with the shorter period in the oscillation pattern observed in the k³-weighted EXAFS spectra shown in Figure 1 and discussed above. Fits to the EXAFS equation (discussed below) confirm this as a lengthening of Mn–Mn distances. The pattern of FT peaks in all of the six S₃ data sets remains the same under various background-removal procedures (in k-space or energy-space) and are therefore unlikely to be affected by data-processing artifacts.

Interestingly, a small Fourier feature is observed between peaks I and II, indicated by an arrow in Figure 3. It is possible that this small shoulder is also present in the S₂ state but becomes more evident in the S₃ state because of the shift of peak II to a longer apparent distance. Recent XAS studies in our laboratory on oriented PS II membranes in the S₃ state show that this peak is also clearly present, that it is differently dichroic from peak II, and that it increases in amplitude when the e-vector of the X-rays is parallel to the membrane normal. Preliminary analysis suggests that this peak arises from a Mn–Cl interaction.⁴⁷ EXAFS spectra by Evans and co-workers of Ca-depleted samples prepared by NaCl treatment have also been modeled with a Cl contribution at ~2.4 Å.²¹ Recent studies have indicated that the presence of Cl is necessary only for the S₂→S₃ and S₃→S₀ transitions of the OEC, while the earlier steps of the cycle can proceed in its absence.^48,49

The amplitude and distance changes of Fourier peak III in the S₃ state are less obvious. Peak III appears substantially above the noise level in three of the six S₃ FTs and is readily distinguishable from the noise in two others. Furthermore, all discernible third peaks in the data sets are shifted to a greater “apparent distance” than that observed in the S₂ spectrum. However, the amplitude of peak III is susceptible to noise components in the data, and the difference in the amplitude of this peak between the S₂ and S₃ spectra is harder to determine. Therefore, the changes of the amplitude of peak III in the S₃ spectra are less definite.

Curve Fitting of Mn EXAFS Data

Fourier isolation is commonly used to help simplify curve fitting of EXAFS data, allowing analysis of the contribution of single peaks. The isolation process, however, can introduce distortions in the extracted k-space data, especially at the start and end of the spectrum.⁵⁰ Closely spaced Fourier peaks can also be distorted if isolated individually. For analysis of the data presented here, Fourier peaks were isolated individually and in pairs to help simplify the spectra and to minimize the effects of distortions due to Fourier isolation.⁵¹ Data sets from samples that allowed a clean isolation of the Fourier peaks of interest, I+II, II, II+III, or III, were used for fitting. Fourier peaks I and II were best isolated from samples A–D, peak II from samples B–F, peaks II and III from samples B–D, and peak III from samples A and D. Curve-fitting results of EXAFS data from representative S₃ and S₂ data sets (S₂ data from DeRose and co-workers¹¹ and Liang and co-workers⁴²) are shown in Tables 1–4.

Table 1.

Two-Shell and Four-Shell Simulations of Fourier Peak I and Peak II of the Native S₃ State Samples^a

(A) Two-Shell Simulations
sampleb	Mn–O/N interaction			Mn–Mn interaction			ΔE0	Φ (×102)d	ε2 (×104)d
	R (Å)	N	σ2 (Å2)	R (Å)	N	σ2 (Å2)
S3A	1.85	1.90	0.0060	2.78	0.81	0.0045	−17.23	1.38	1.46
S3B	1.86	2.08	0.0083	2.80	0.92	0.0072	−14.22	1.56	1.15
S3C	1.86	2.19	0.0098	2.82	0.97	0.0046	−9.98	1.38	0.94
S3D	1.88	1.55	0.0065	2.79	1.23	0.0093	−14.09	1.43	2.10
〈S3〉	1.86	1.93	0.0077	2.80	0.98	0.0064
S2 e	1.86	2.50	0.0050	2.73	1.10	0.0045

(B) Four-Shell Simulations
sampleb	two Mn–O/N interactions			two Mn–Mn interactions			ΔE0	Φ (×102)d	ε2 (×104)d
	R (Å)	N	σ2 (Å2)	R (Å)	N	σ2 (Å2)
S3A	1.85	1.50	0.0020c	2.82	0.75	0.0020c	−8.81	0.96	1.47
	2.02	0.94	0.0020c	2.96	0.27	0.0020c
S3B	1.85	1.34	0.0025	2.81	0.56	0.0030	−7.69	1.33	0.81
	2.01	0.77	0.0011	2.93	0.36	0.0043
S3C	1.87	1.56	0.0030	2.84	0.73	0.0015	−4.61	0.48	0.42
	2.05	1.09	0.0011	2.97	0.40	0.0087
S3D	1.86	0.82	0.0010	2.79	0.75	0.0028	−8.62	1.28	0.95
	1.99	0.76	0.0019	2.94	0.56	0.0037
〈S3〉	1.86	1.31	0.0021	2.82	0.70	0.0023
	2.02	0.89	0.0015	2.95	0.40	0.0047
S2	1.80	1.40	0.0010	2.73	1.20	0.0025
	1.98	0.80	0.0010

^aFit parameters are defined in the text. Data are fit from k = 3.5 to 11.5 Å⁻¹.

^bFour individual S₃ data sets.

^cParameter is fixed in the fit.

^dQuality of fit parameters Φ and ε² are defined in Materials and Methods.

^eThe S₂ data sets are from refs 11 and 42. 〈S₃〉 is the average of the individual fits presented. Three shell fits were also performed (see text for details).

Table 4.

Fit Parameters for Fourier Peak III of the Native S₃ State Samples^a

One-Shell Simulation
sampleb	Mn–Mn interaction			ΔE0 (eV)	Φ (×102)c	ε2 (×104)c
	R (Å)	N	σ 2 (Å2)
S3A	3.42	0.65	0.0056	−5.74	0.16	0.84
S3D	3.41	0.82	0.0076	−6.93	0.39	0.21
〈S3〉	3.42	0.74	0.0066
S2	3.36	0.30	0.0030

Two-Shell Simulation
Mn–Mn Interaction Only
sampleb	Mn–Mn interaction			Mn–Mn interaction			ΔE0 (eV)	Φ (×102)c	ε2 (×104)c
	R (Å)	N	σ2 (Å2)	R (Å)	N	σ2 (Å2)
S3A	3.43	0.52	0.0020d	3.58	0.23	0.0020d	−2.29	0.15	0.77
S3D	3.41	0.57	0.0026	3.57	0.37	0.0031	−2.39	0.30	0.36
〈S3〉	3.42	0.55	0.0023	3.58	0.30

Mn–Mn and Mn–Ca Interactions
sampleb	Mn–Mn interaction			Mn–Ca interaction			ΔE0 (eV)	Φ (×102)c	ε2 (×104)c
	R (Å)	N	σ2 (Å2)	R (Å)	N	σ2 (Å2)
S3A	3.43	0.48	0.0020d	3.64	0.27	0.0020d	−1.43	0.10	0.53
S3D	3.40	0.48	0.0023	3.60	0.45	0.0026	−2.50	0.17	0.12
〈S3〉	3.42	0.48	0.0020	3.62	0.36	0.0025
S2	3.38	0.30	0.0020	3.39	0.30	0.0070

^aFit parameters are defined in the text. Data are fit from k = 3.5 to 11.5 Å⁻¹.

^bFive individual S₃ data sets.

^cQuality of fit parameters Φ and ε² are defined in Materials and Methods.

^dParameter is fixed in the fit. 〈S₃〉 is the average of the individual fits presented.

We were concerned about introducing artifacts because of the subtraction procedure used to generate the S₃ state EXAFS spectrum. To address this concern we performed fits on EXAFS data from a two-flash sample that is predominantly S₃ in composition. The data for 70–75% S₃-state composition are similar to those of the S₃ state spectrum. The fit parameters are also very similar to those reported below for the calculated pure S₃ state EXAFS data (data not shown).

Fourier Peaks I and II

Peaks I and II were isolated together and fit as a combination of two subshells (one Mn–O/N, one Mn–Mn) or four subshells (two Mn–O/N, two Mn–Mn) of scatterers. The corresponding parameters from the curve-fitting results are summarized in Table 1. Fitting results for the S₂ state agree well with our previous results on PS II samples.^10,11,42,43

Results from curve-fitting peak I+II with two subshells indicate an increase in the Mn–Mn distance from 2.73 to ~2.80 Å upon the S₂ → S₃ transition. Furthermore, the Debye–Waller factor (σ²) for the Mn–Mn shell is larger for the S₃ data. These results are reminiscent of our previous studies from the S₂ g = 4.1 and the ammonium-treated S₂ samples; in both cases an amplitude reduction of peak II was observed and shown to result from a larger disorder in the Mn–Mn shell.^42,43

For the Mn–O/N subshell, fits to peak I+II of the S₃ state resulted in a small change in the Mn–O/N distance relative to the S₂ sample, but larger Debye–Waller factors were required for the best fits. The results were suggestive of a more heterogeneous environment of these two subshells in the S₃ state (Table 1A). Best fits are obtained when two Mn–O/N subshells are included (Table 1B). There is little change between the S₂ and S₃ states in the Mn–O/N distances that are assigned to Mn-terminal ligands at ~2.0 Å. However, there is a lengthening of the Mn–O/N subshell from 1.80 Å in the S₂ state to 1.86 Å in the S₃ state. The shorter 1.8 Å Mn–O distances are assigned to the Mn–O μ-oxo bridging ligands.

To test whether an increase of heterogeneity in both peaks I and II could account for the amplitude reduction in the S₃ state, the data were also fit under the assumption that peak II contains two Mn–Mn subshells. As shown in Table 1B, this fit leads to a significant decrease of the normalized error sum (see Φ value, Table 1) in each of the S₃ samples. The best fits for peak II were found at distances of ~2.82 (N = 0.56–0.75) and 2.95 Å (N = 0.27–0.56) for the Mn–Mn shell. The fact that ε² remained positive after peaks I+II were fit with four subshells suggests that the fitting results were not underdetermined, because the number of free parameters does not exceed the number of independent data points. To test whether we have obtained a unique solution set at minimum Φ, we decreased number of the degrees of freedom by assuming the disorders for the individual subshells to be equal (σ² = 0.002 Ų). The unique solution under this assumption was found to be similar to those fits without restriction. One such fit is listed for sample S₃A in Table 1. Peak I of the S₂ state was better fit with two subshells of N/O atoms at distances of 1.80 and 1.98 Å. Peak II of the S₂ state was also well fit with a single shell of 1.2 Mn at ~2.7 Å. An additional shell of Mn does not improve the overall fit quality of the S₂ sample (Table 1B).^11,42,43

Three-shell fits were also performed that included two Mn–O/N distances and one Mn–Mn distance, or one Mn–O/N distance and two Mn–Mn distances (data not shown). When only one Mn–Mn distance is included, the Debye–Waller factors are large and the distance is ~2.85 Å, and when two Mn–Mn distances are included the Debye–Waller factors decrease and one can obtain fits with two distances at ~2.82 and 2.96 Å. The inclusion of one or two Mn–O/N shells has little bearing on the fits to the Mn–Mn distances. The results are not different from what is presently shown in the tables for two- and four-shell fits.

Fourier Peak II

The isolates of Fourier peak II from the S₂ or S₃ states are shown in Figure 4. It is obvious that there is a significant difference in the contribution to this peak between the S₂ and S₃ states. The change in frequency of the modulations, as shown by the progressive change in phase toward higher values of k, and differences in the amplitude of the modulations are clearly evident. Fitting parameters for peak II alone for five S₃ data sets are given in Table 2. Fitting of this isolate using a single shell confirmed an elongation of the Mn–Mn distance to ~2.85 Å for the S₃ state relative to the S₂ state (2.73 Å). The Debye–Waller factors for the S₃ samples were larger than the values for S₂ samples, indicating a large disorder in peak II for the S₃ state. As shown in Table 2B, peak II was better fit with two Mn–Mn subshells at 2.82 (N = 0.58–0.73) and 2.96 Å (N = 0.47–0.62) for the S₃ state. We could force a two-distance fit for the S₂ state that yielded distances of 2.71 and 2.81 Å. However, the fits are physically unreasonable because the Debye–Waller factors for a one-shell fit increased for a two-shell fit.

Figure 4.

EXAFS contributions from Fourier peak II from S₂ (dashed line) and S₃ (solid line) states of PS II. The Fourier transform of the S₃ state was obtained from the calculated S₃ state EXAFS shown in Figure 1 (see text for details). Fourier peak II is assigned to Mn–Mn interactions at 2.7 Å in the S₁ and S₂ states. The increase in frequency evident in this figure in the S₃ state compared to that in the S₂ state is because of the increase in the Mn–Mn distance in the S₃ state. The sin(2kR + α(k)) contribution to the EXAFS (eq 1) dictates that an increase in R will be reflected in an increase in the frequency, as shown in the S₃ data compared to the S₂ state spectra. EXAFS techniques are most sensitive to such changes in frequency because the changes are manifest over the entire range of the spectra and hence less susceptible to contributions from noise in the spectra.

Table 2.

One-Shell and Two-Shell Simulations of Fourier Peak II of the Native S₃ State Samples^a

(A) One-Shell Simulation
sampleb	Mn–Mn interaction			ΔE0 (eV)	Φ (×102)c	ε2 (×104)c
	R (Å)	N	σ2 (Å2)
S3B	2.85	1.27	0.0096	−5.67	0.30	0.50
S3C	2.84	1.33	0.0067	−5.48	0.28	1.04
S3D	2.80	1.29	0.0094	−12.05	0.39	1.17
S3E	2.88	1.05	0.0058	−1.70	0.29	0.57
S3F	2.87	1.14	0.0099	−2.69	0.68	0.55
〈S3〉	2.85	1.22	0.0083
S2	2.73	1.20	0.0025

(B) Two-Shell Simulation
sampleb	Mn–Mn interaction			Mn–Mn interaction			ΔE0 (eV)	Φ (×102)c	ε2 (×104)c
	R (Å)	N	σ2 (Å2)	R (Å)	N	σ2 (Å2)
S3B	2.83	0.64	0.0021	2.97	0.47	0.0021	−3.04	0.13	0.54
S3C	2.82	0.73	0.0010	2.96	0.55	0.0024	−2.24	0.12	0.11
S3D	2.80	0.69	0.0018	2.95	0.62	0.0032	−5.30	0.22	0.65
S3E	2.84	0.58	0.0022	2.95	0.53	0.0026	−0.71	0.23	0.27
S3F	2.83	0.62	0.0046	2.97	0.58	0.0050	−0.64	0.30	0.34
〈S3〉	2.82	0.65	0.0023	2.96	0.55	0.0031
S2	2.71	0.60	0.0010	2.81	0.60	0.0060

^aFit parameters are defined in the text. Data are fit from k = 3.5 to 11.5 Å⁻¹.

^bFive individual S₃ data sets.

^cQuality of fit parameters Φ and ε² are defined in Materials and Methods. 〈S₃〉 is the average of the individual fits presented. One Mn–Mn shell fit is better than a Mn–C fit by a factor of 3–4, and the two-shell Mn–Mn fits are better than one Mn–Mn and one Mn–C fits by a factor of 2–3 (see text for details).

We had shown for the S₂ state that the fit for peak II with at least two Mn–Mn interactions at 2.7 Å is better than that with Mn–C alone.¹¹ We used a similar approach for fitting the S₃ state data, and a similar conclusion was drawn. We compared C fits to Mn fits under very similar fit conditions, and the Mn–Mn fits are better by a factor of 3–4. We also compared the two Mn–Mn distance fits (at 2.82 and 2.96 Å) with one Mn–Mn and Mn–C interactions, and again in this case the fits with two Mn–Mn distances are better than those with one Mn–Mn and Mn–C interactions by a factor of 2–3 (data not shown).

Fourier Peaks II and III

Fitting parameters for Fourier peaks II and III, using two subshells or four subshells, are listed in Table 3. Fits were performed on three of the six pure S₃ data sets that showed a peak III that was substantially above the noise level in the FTs. In the top parts of Table 3A and B we consider a simple model of one Mn–Mn shell accounting for peak II and one Mn–Mn or Mn–Ca shell for peak III. Again, the fitting results indicate a larger Mn–Mn distance at ~2.82 Å, with a much larger Debye–Waller factor observed in two of the three S₃ data sets. Furthermore, the Mn–Mn or Mn–Ca interaction at ~3.3 Å also increases by ~0.06 (for a single Mn–Mn shell) or ~0.08 Å (for a single Mn–Ca shell).

Table 3.

Two-Shell and Four-Shell Simulations of Fourier Peaks II and III of the native S₃ State Samples^a

(A) Mn–Mn Interaction Only
Two-Shell Simulation
sampleb	Mn–Mn interaction			Mn–Mn interaction			ΔE0 (eV)	Φ (×102)c	ε2 (×104)c
	R (Å)	N	σ2 (Å2)	R (Å)	N	σ2 (Å2)
S3B	2.84	1.24	0.0095	3.39	0.33	0.0044	−6.04	0.28	0.39
S3C	2.84	1.07	0.0048	3.43	0.61	0.0065	−6.17	0.52	0.35
S3D	2.80	1.28	0.0094	3.39	0.93	0.0078	−9.82	0.37	0.33
〈S3〉	2.83	1.20	0.0079	3.40	0.62	0.0062
S2	2.71	1.20	0.0025	3.34	0.30	0.0050

Four-Shell Simulation
sampleb	Mn–Mn interactions			Mn–Mn interactions			ΔE0 (eV)	Φ (×102)c	ε2 (×104)c
	R (Å)	N	σ2 (Å2)	R (Å)	N	σ2 (Å2)
S3B	2.82	0.72	0.0042	3.42	0.28	0.0043	−3.20	0.15	0.21
	2.96	0.55	0.0052	3.58	0.10	0.0090
S3C	2.83	0.73	0.0011	3.43	0.49	0.0056	−2.29	0.17	0.11
	2.96	0.55	0.0029	3.56	0.36	0.0018
S3D	2.80	0.68	0.0015	3.41	0.58	0.0024	−4.84	0.16	0.14
	2.95	0.61	0.0023	3.56	0.38	0.0034
〈S3〉	2.82	0.71	0.0022	3.42	0.45	0.0041
	2.96	0.57	0.0035	3.57	0.28	0.0047

(B) Mn–Mn and Mn–Ca Interactions
Two-Shell Simulation
sampleb	Mn–Mn interaction			Mn–Ca interaction			ΔE0 (eV)	Φ (×102)c	ε2 (×104)c
	R (Å)	N	σ2 (Å2)	R (Å)	N	σ2 (Å2)
S3B	2.84	1.04	0.0082	3.45	0.43	0.0045	−6.65	0.21	0.17
S3C	2.84	0.92	0.0040	3.51	0.63	0.0037	−5.61	0.52	0.42
S3D	2.81	1.08	0.0070	3.44	1.07	0.0067	−9.70	0.48	0.42
〈S3〉	2.83	1.01	0.0064	3.47	0.71	0.0050
S2	2.71	1.20	0.0025	3.39	0.40	0.0050

Four-Shell Simulation
sampleb	Mn–Mn interactions				Mn–Mn and Mn–Ca interaction				ΔE0 (eV)	Φ (×102)c	ε2 (×104)c
		R (Å)	N	σ2 (Å2)		R (Å)	N	σ2 (Å2)
S3B	Mn–Mn	2.83	0.70	0.0039	Mn–Mn	3.41	0.29	0.0045	−2.78	0.13	0.18
	Mn–Mn	2.96	0.56	0.0051	Mn–Ca	3.63	0.16	0.0066
S3C	Mn–Mn	2.83	0.73	0.0010	Mn–Mn	3.43	0.49	0.0052	−2.40	0.16	0.11
	Mn–Mn	2.96	0.55	0.0030	Mn–Ca	3.61	0.36	0.0011
S3D	Mn–Mn	2.80	0.68	0.0017	Mn–Mn	3.40	0.48	0.0021	−4.73	0.15	0.13
	Mn–Mn	2.95	0.61	0.0027	Mn–Ca	3.60	0.44	0.0026
〈S3〉	Mn–Mn	2.82	0.70	0.0022	Mn–Mn	3.41	0.42	0.0039
	Mn–Mn	2.96	0.57	0.0036	Mn–Ca	3.61	0.32	0.0034

^aFit parameters are defined in the text. Data are fit from k = 3.5 to 11.5 Å⁻¹.

^bThree individual S₃ data sets.

^cQuality of fit parameters Φ and ε² are defined in Materials and Methods. 〈S₃〉 is the average of the individual fits presented.

The best fits for the S₃ state were obtained by including four subshells (bottoms of Table 3A and B). The normalized error sum for each data set decreased by a factor of 2–3, and the Debye–Waller factors were also smaller compared to the single-shell fits. Peak II was again better fit with two Mn–Mn interactions at distances (2.82 and 2.96 Å) similar to the results from fitting peak I+II. Peak III can be fit reasonably well to two subshells of Mn–Mn (~3.42 and ~3.57 Å) or one Mn–Mn (~3.41 Å) plus one Mn–Ca interaction (~3.61 Å).

Fourier Peak III

Table 4 presents fitting parameters of peak III alone on data sets S₃A and S₃D. Fits with a single shell of Mn–Mn were worse than those with two subshells. Also, Mn–Mn fits are clearly better when compared to Mn–C fits. A higher degree of disorder was observed for the Mn–Mn fits, which suggested the need for an extra subshell. However, as shown above, the Mn–Mn distance was longer relative to that in the S₂ state in the one-shell fits. Best fits were obtained in fits with two subshells in which one 3.62 Å Mn–Ca interaction was included. For fits of peak II+III or peak III only, the distance obtained from each subshell of the native S₃ state is consistently longer (0.04–0.23 Å) than those from the native S₂ state.

Fits to peak III alone are difficult for the S₃ samples. First, the quality of fits is poorer owing to the noise level in the data. Also, these samples are very dilute (~4–5 mg/mL of Chl) to guarantee a complete turnover by single laser or Xe lamp flash. In addition, the contribution from peak III is small compared with the total EXAFS, and multiple contributions from other low-Z elements are possible at longer distances from the absorber. Finally, distortions can arise by Fourier isolation of such a small feature, especially because it is near a much larger peak II. Considering the limitations, we chose to perform peak III simulations on the two best S₃ data sets. The fact that the fitting results are similar to those from peak II+III (Table 3) indicates the fitting parameters obtained are valid.

Discussion

Comparison of the Structure of Mn Cluster in the S₁, S₂, and S₃ States

By comparison with the structural motifs present in multinuclear Mn complexes (for details see refs 52 and 53, and refs 1b, 9, 43, and 44 in ref 5), we proposed that the Mn cluster in PS II consists of a pair of di-μ-oxo-bridged Mn binuclear clusters linked through a mono-μ-oxo bridge. The 2.7 Å Mn–Mn distance is characteristic of di-μ-oxo-bridged models, and the 3.3 Å Mn–Mn distance is characteristic of mono-μ-oxo-bridged Mn–Mn distances. The dimer-of-dimers model was proposed on consideration of the above data.¹⁰ We described in a previous publication the various topological alternatives that are compatible with our EXAFS data,¹¹ and in this paper we discuss the structure of the Mn cluster in the S₃ state with the simplest of options that is compatible with our EXAFS data.

The changes in Mn–Mn distances occurring during the S₁→S₂ and →S₃ transitions within the context of the dimer-of-dimer model described above is shown in Figure 5.

Figure 5.

Summary of the changes in Mn–Mn distances in the native S₁, native S₂, modified S₂, S₃′ (S₂Y_Z^•), and native S₃ states as determined by XAS. The increase in Mn–Mn distances in the S₃ state argues against the presence of di-μ-oxo bridges between Mn atoms. It is possible that the di-μ-oxo motif is unchanged in the other states presented in the schematic.

The first step in the scheme is the conversion of the dark, stable S₁ state to the S₂ state that is characterized by the multiline EPR signal. Little change is observed in the fitting parameters for Fourier peaks II and III when the Mn cluster undergoes this transition. This step involves a Mn-centered oxidation, as demonstrated by the appearance of the MLS during this transition and by our Mn K-edge and Mn Kβ emission studies.^16,54 Fourier peak II, corresponding to the averaged Mn–Mn distance in each di-μ-oxo dimanganese unit, is best fit to ~2.7 Å for both the S₁ and S₂ states.¹¹ EXAFS studies on oriented PS II in the S₁ and S₂ states show that there is heterogeneity in the 2.7 Å vectors, which suggests that these two binuclear species are not completely equivalent to each other.^55,56

The inequivalence of the two di-μ-oxo-bridged Mn units becomes more evident in the S₂ state that is prepared by illumination at 130 K and characterized by the g = 4.1 EPR signal.⁴² Similar results were obtained with NH₃-treated and annealed S₂ state samples⁴³ and in F⁻-treated samples.⁵⁷ The scheme in Figure 5 shows that there are paths from the MLS S₂ state that lead to two states, the S₂ g = 4.1 state inhibited by NH₃/F⁻ or the S₂Y_Z^• states; neither of these states can proceed to the physiologically relevant S₃ state. These states are depicted in Figure 5 as branching away from the normal pathway leading to the S₃ state by photon absorption. However, the S₂ g = 4.1 state generated by 820 nm illumination at 130 K⁵⁸ can proceed to an S₃ state.59 In the modified S₂ states, one of the Mn–Mn distances increases to ~2.85 Å, whereas there is very little change in the other 2.7 Å or the 3.3 Å Mn–Mn distance. With the degeneracy of the 2.7 Å distance lifted, we were able, by studying the dichroism of the Fourier peak, to assign the relative orientation of the 2.7 and 2.85 Å vectors.⁴³

The NH₃-treated and F⁻-treated samples cannot advance beyond the modified S₂ state (S₂Y_Z^• state). It is likely that F⁻or NH₃ prevents the structural changes from occurring that are necessary for the formation of the S₃ state. In the case of NH₃, it is probably due to an amido group (NH₂⁻) replacement of the oxo bridge involved in oxidation. The modification of the MLS spectra upon addition of NH₃ ⁶⁰ and ESEEM studies using ¹⁴NH₃ and ¹⁵NH₃ demonstrated that NH₃ becomes a ligand of Mn.⁶¹ The asymmetry parameter derived from ESEEM results suggested that the amido group is likely to be a bridging ligand.⁶¹

The other unproductive state generated from the S₂-MLS state is denoted in Figure 5 as the S₂Y_Z^• state generated in Ca-depleted samples. The Ca-depleted samples are inactive in O₂ evolution, while a broad g = 2 EPR signal has been observed in the “S₃′ state” of such samples. In such samples the g = 2 broadened EPR signal has been confirmed to arise from the tyrosine Y_Z radical,^62,63 and it is proposed that the signal is broadened by interaction with the spin on the Mn cluster.⁶⁴

We recently reported the XAS analysis from calcium-depleted S₃′ state samples (S₂Y_Z^•).²⁰ The position of peak II in the FT of S₃′ state samples (S₂Y_Z^•) was invariant relative to that of the native S₂ state sample (Figure 6). The EXAFS fits showed that the 2.7 Å Mn–Mn distances did not lengthen as observed in the native S₃ state samples and are essentially unchanged from those of the native S₂ state. This finding is surprising because the Mn K-edges from these calcium-depleted samples showed a behavior similar to that of the native PS II in that little or no shift was observed in the S₂′→S₃′ (or S₂Y_Z^•) transition. This difference between the native S₃ and calcium-depleted S₃′ states indicates that the core di-μ-oxo-bridged structure is probably dissimilar in the native S₃ and the S₃′ (S₂Y_Z^•) states, with the structure in the S₂Y_Z^• state resembling the native S₂ state structure. It is reasonable to question how the transfer of one electron from the Mn cluster to Y_Z results in major changes in the Mn–Mn distances, as is observed in the native S₃ state. The roles of the Tyr_z and the Mn cluster in the process of water oxidation are clearly delineated by the comparison of the S₃ and the S₃′ (or S₂Y_Z^•) states. Instead of the Mn cluster just providing a scaffolding for water oxidation, this comparison shows that the structural change in the Mn cluster, initiated by the transfer of an electron from the Mn cluster to Tyr_z during the S₂→S₃ transition, might provide the trigger to the chemistry of the formation of the O–O bond. The results show that the Mn cluster is involved in a much more intimate manner in the catalysis than just providing the framework. The implications to the mechanism are significant (see below).

Figure 6.

Fourier transform power spectra of S₂ (dashed line) and Ca-depleted S₃′ or S₂Y_Z^• (solid line) states of PS II.²⁰ The FTs are of k³-weighted EXAFS data from 3.5 to 11.5 Å⁻¹. There is a small difference in amplitudes, but there is no shift in apparent distance, in marked contrast to the Fourier transforms shown in Figure 3. The expanded version from 0 to 4 Å shown in the upper part of the figure makes the lack of a shift obvious.

In Ca-depleted systems the tyrosine Y_Z radical is stabilized, and the oxidation of the Mn–OEC and the concomitant changes in Mn–Mn distances are prevented. In the S₃ state, with Ca present, the radical resides on the Mn cluster; however, in the absence of Ca, the radical resides on tyrosine Z. Thus, Ca is proposed to play a crucial role in controlling the redox potential and thus the course of the mechanism of water oxidation.

Earlier XAS data from the S₃ state samples produced by a cryogenic double-turnover method indicated increased disorder in the peak at 2.7 Å, indicative of the presence of two nonequivalent di-μ-oxo-bridged clusters in the S₃ state.¹² The principal result of that work was that a small but significant structural change was found to accompany the S₂→S₃ transition. Specifically, the second Fourier peak, representing Mn scatterers at about 2.7 Å, was decreased in amplitude by about 40–50% compared to the Fourier peak for the S₂ state and was better simulated by distances differing by ~0.15 Å. We note that the S₃ state in that work was generated by cryogenic turnover at 240 K of chemically treated PS II. The earlier data were collected at 190 K using a single-element Si(Li) detector, and the S/N was poor compared to that of our present data that were collected at 10 K using a 13-element Ge detector. The quality and smaller range of data, and the method of generation of the S₃ state, might be responsible for the differences. However, it was obvious even with the earlier data that a structural change that involved the lengthening of the Mn–Mn distance was occurring between the S₂ and S₃ states and was absent during the S₁→S₂ state transition.

The EXAFS results presented in this paper represent the first detailed analysis of data from native S₃ state samples created under physiological conditions with saturating actinic flash illumination (preliminary results from our group, Liang and coworkers,¹⁹ have appeared in conference proceedings). During the S₂→S₃ transition there is a significant change in the 2.7 Å Mn–Mn distances that characterize the di-μ-oxo-bridged Mn–Mn units. One of the main conclusions from this study is that there is an increase in both the short Mn–Mn distances from 2.7 Å to ~2.8 and ~3.0 Å. The results for the native S₃ state are shown in the scheme in Figure 5.

The k³-weighted EXAFS spectra of the native S₃ state show an overall amplitude reduction with the most pronounced difference in the region k = 7.5–10 Å⁻¹, and a distinct difference in the phase and frequency of the modulations (Figures 1, 2, and 4). The perturbation in the Mn–Mn distances is different from that observed in the modified S₂ states. Each of these altered S₂ states gives a two-shell Mn–Mn fit that is consistent with the alteration of one of the two di-μ-oxo-bridged Mn–Mn distances from ~2.72 to 2.85 Å. Unlike the case in the modified S₂ states, the position of peak II from the native S₃ state clearly shifts to a longer distance in the FTs.

Parameters from the best fits for peak II derived from isolates of peak II only, peak I+II, or peak II+III all converge to Mn–Mn distances of 2.82 and 2.95 Å. These results underscore the point that during the S₂→S₃ transition under physiological conditions, both of the di-μ-oxo-bridged Mn–Mn units undergo structural changes that lead to an increase in the Mn–Mn distance. The structural changes are different between the two units. One is probably similar to those observed for the modified S₂ states, with a final Mn–Mn separation of 2.82 Å. For the other unit, a more significant change is observed that results in the Mn–Mn distance increasing to ~3.0 Å. A similar change in one of the Mn–Mn distances from 2.7 to 3.0 Å was observed in samples that had been depleted of Ca by NaCl treatment.²¹

The best fits for the Mn–O/N shell in Table 1B also show interesting differences between the S₂ and S₃ states. The fits converge to a distance of 1.80 and 1.98 Å for the S₂ state and to 1.86 and 2.02 Å for the S₃ state. The shorter Mn–O distance is characteristic of Mn–O bridging distances, and it increases from 1.80 Å in the S₂ state to 1.86 Å in the S₃ state. This is a significant change, and it provides additional evidence for the involvement of the bridging oxygen atoms during the S₂→S₃ transition.

The results from isotropic S₃ samples are supported by polarized EXAFS studies on oriented PS II in the native S₃ state. The data confirm that the two di-μ-oxo-bridged Mn–Mn dimer units are not equivalent. The polarized EXAFS data are different from those observed in the S₂ state.^43,56 Fourier peak II of S₃ is dichroic and is readily resolved to Mn–Mn distances of ~2.8 and ~3.0 Å, each with its own distinct projection on the membrane normal.⁴⁷ Recently, Dau and co-workers have reported results on the oriented S₃ state, showing an increase in the amplitude of Fourier peak II that is interpreted as an increase in the number of 2.7 Å Mn–Mn interactions in the S₃ state.⁶⁵ However, an examination of their FTs shows that Fourier peak II occurs at a longer distance and is broader compared to that in the S₂ state.

EXAFS analyses of PS II samples depleted of calcium and reconstituted by either calcium or strontium favor a model in which both manganese and calcium scatterers contribute to peak III in the S₁ state.²⁶ More evidence for the proximity of calcium to the Mn cluster in the S₁ state has been obtained from Sr EXAFS studies.⁴⁴ In the present study of the S₃ state, it is seen that Fourier peak III occurs at a greater apparent distance than that of the S₂ state. Simulation of peak II+III and of peak III alone indicates a lengthening of the two Mn–Mn distances or the Mn–Mn and Mn–Ca distances compared to the S₂ state (Tables 3 and 4).

Implication to the Mechanism of Water Oxidation

Two essential factors are considered to construct possible structural arrangements of the Mn cluster in the S₃ state. First, our Mn K-edge inflection-point data better support the interpretation that Mn is not oxidized during the S₂→S₃ transition. These data are reinforced by our recent Mn Kβ emission studies of the various S states.⁵⁴ Second, the Mn–Mn distance in both of the di-μ-oxo-bridged units increases from 2.72 to 2.82 and 2.95 Å, respectively, upon the formation of the S₃ state. Furthermore, the Mn–Mn distance of 3.3 Å increases by 0.04 Å, and the Mn–Ca (or Mn–Mn) distance also increases by 0.2 Å during the S₂→S₃ transition (Table 4). These changes imply a significant structural change in the Mn cluster as it proceeds to the S₃ state. It is hard to rationalize such changes in Mn–Mn distances as arising purely from Mn oxidation. We propose that substrate/water oxidation chemistry is occurring at this transition, leading to the significant structural changes. This hypothesis is supported by various experiments reported in the literature.

First, marked differences of about 10–20-fold have been observed in the efficiency of exogenous reductants such as NH₂OH and NH₂NH₂ toward the S₂ and S₃ states, with S₃ being more resistant to reduction than S₂.⁶⁶ These large differences in reactivity between the S₂ and S₃ states have been interpreted as indicating that a significant structural change in electronic and nuclear geometry occurs during the S₂→S₃ transition, which is consistent with our proposal.

In addition, Renger and co-workers have shown in a series of experiments that the activation energies for the S₁→S₂ and S₂→S₃ transitions are markedly different in PS II preparations from a thermophilic cyanobacterium and from either PS II core or membrane fragments from spinach.^67–69 The activation energy increases by a factor of about 3 for S₂→S₃ compared to that for the S₁→S₂ transition. The authors attribute this marked difference in activation energy to the involvement of a significant structural change in the S₂→S₃ transition that is absent in the S₁→S₂ transition. Moreover, they have deduced an increase in the reorganization energy for the S₂→S₃ transition compared to that for the S₁→S₂ transition. Renger notes that these results are consistent with the assumption that a significant structural change accompanies the S₂→S₃ transition,⁷⁰ perhaps reflecting the onset of water oxidation chemistry.⁷¹

Protonation of oxo bridges in the S₃ state is another mechanism by which one can rationalize the lengthening of Mn–Mn distances. Successive protonation of di-μ-oxo bridges in synthetic binuclear Mn complexes has been shown to result in increases in the Mn–Mn distance successively from 2.7 to 2.8 and to 2.9 Å.^72,73 The distance increase seen for the S₃ state might be due to such protonation. However, it is generally agreed that deprotonation at the higher S states is more likely, and mechanisms based on deprotonation or H atom abstraction of the substrate, OH⁻ or H₂O, linked to the tyrosine Z moiety have been proposed.^62,74,75

Alternatively, it has been proposed that formation of peroxide might follow protonation of a bridging μ-oxo ligand in the S₃ state and subsequent release of a dioxygen molecule.^76,77 Whether a peroxo-bridged Mn structure forms in the S₃ state is still under debate. The Mn–Mn distance in peroxo-bridged complexes is likely to be greater than 3 Å, assuming that the peroxo group replaces one of the di-μ-oxo bridges. Another option would be for the peroxo bridge to form between the two di-μ-oxo units. Calculations based on density functional theory support a structural arrangement in which the peroxo bridge forms between Mn atoms with accessible ferromagnetic states, as is the case for the Mn atoms that are linked by the mono-μ-oxo bridge.⁷⁸ Replacement of both of the di-μ-oxo bridges in one binuclear unit with a peroxo bridge results in a reduction of Mn. This arrangement is considered less likely because the Mn XANES spectra do not support such a reduction during the S₂→S₃ transition.¹⁶

However, Renger has proposed that there could be equilibration between two species: one in which there is peroxide bond formation between two Mn(III) species, and one involving two Mn(IV) species with bound hydroxide or water.⁷⁰ Such a scenario would lead to no net change in Mn oxidation state and would be compatible with the Mn K-edge and Kβ emission results observed by our group.^16,54

In addition to oxo-bridging ligands, Mn-terminal oxo ligands have also been proposed to be involved in dioxygen formation.^79,80 It has been proposed that Mn is ligated to activated O (Mn=O) produced by the abstraction of protons from Mn-bound water by a nearby tyrosine,^62,74,81 and the O–O bond is proposed to form between the oxo ligands of two adjacent Mn atoms.⁸² However, it is hard to understand how changes in terminal ligation can generate such a profound change on the Mn–Mn distances in the S₃ state as reported here. Replacement of terminal ligands in di-μ-oxo-bridged model compounds has a minimal effect on the Mn–Mn distance of 2.7 Å that is characteristic of such di-μ-oxo-bridged Mn compounds.^53,83

Limburg and co-workers⁸⁴ have recently reported a di-μ-oxo Mn complex that oxidizes water. They proposed that the O–O bond formation can be explained by a nucleophilic attack of a Mn=O or Mn–O radical species on a OH⁻ or H₂O that is either free or ligated to another Mn atom. On the basis of the Mn K-edge and EXAFS results, it is less likely that a Mn=O exists at the S₃ state of the OEC. However, a radical species is supported by the lack of any significant shift in the Mn K-edge inflection point energy and also by the Mn Kβ emission studies.⁵⁴

The case for the involvement of the bridging oxygen atoms during the S₂→S₃ transition and in the mechanism of oxygen evolution is manyfold. It is easy to rationalize increases in Mn–Mn distances as being due to changes in the bridging structure. As described above, the Mn–O bridging distances increase from 1.80 to 1.86 Å during the S₂→S₃ transition. NH₃, an inhibitor of water oxidation and an analogue of the substrate/water, has been shown by ESEEM studies⁶¹ to be coordinated to Mn, probably in a bridging position. XAS studies showed that one of the 2.7 Å Mn–Mn vectors is perturbed in NH₃-treated samples, with the distance increasing to 2.85 Å, an increase easily rationalized if an amido group were replacing an oxo group. Replacing a terminal O/N ligand with an NH₃ ligand cannot easily justify such an increase in the Mn–Mn distance.

We proposed earlier a model for the S₃ state in which an oxyl radical is generated on one of the μ-oxo bridges, which results in an increase of one of the Mn–Mn distances to ~2.95 Å (Figure 7). Consistent with our XANES results, this structure implies that the oxidative equivalent is not stored on the Mn atoms per se during the S₂→S₃ transition but is delocalized with significant charge and spin density on the bridging oxo ligand. Junge and co-workers have proposed a mechanism in which Mn is oxidized during the S₀→S₁ and S₁→S₂ transitions but involves the oxidation of bound substrate, OH⁻, to a hydroxide radical during the S₂→S₃ transition, with a small fraction already present as a peroxide intermediate.^85,86

Figure 7.

Scheme showing the two different pathways by which an oxyl radical formed in the S₃ state could form an O–O bond in the transient S₄ state before release of dioxygen. One path involves the formation of an O–O bond between the two oxo groups of one binuclear unit. The other path for the formation of the O–O bond is the reaction of the oxyl radical formed with OH⁻or H₂O that is a ligand of Mn or in the outer sphere of the Mn cluster. The Mn–Mn distances in the transient S₄ state are unknown at present.

The presence of an unpaired electron on the oxygen changes the spin state of the Mn complex; therefore, the MLS signal disappears, as observed from the results of EPR spectroscopy. Parallel-mode EPR studies by Matsukawa and co-workers have shown the presence of EPR resonances at g = 12 and 8 that are best simulated by an S = 1 excited-state spin species in the S₃ state.⁸⁷ The authors rationalize these results as arising from a coupling between the S = ½ complex, characterized by the MLS EPR signal, interacting with an S = ½ radical produced during the S₃ transition that leads to the disappearance of the MLS and generation of an integer spin state.

Siegbahn and Crabtree have proposed a new mechanism on the basis of quantum chemical studies where they have tried to reconcile the various biophysical data.⁸⁸ Spin-state considerations led them to propose a Mn–O oxyl intermediate in the S₃ state, with radical character on the terminal oxygen ligand, with the formation of the O–O bond proposed to occur between the oxyl radical and an outer-sphere water molecule. Recent calculations by Siegbahn have produced an energy minimum for the formation of the oxyl radical on the bridging O atom.⁸⁹ The Siegbahn and Crabtree mechanism involves only one Mn in the oxidation process and involves a radical species in the S₃ state, as proposed by our group.

We propose that the O–Mn–O angle in one of the di-μ-oxo-bridged Mn–Mn cores decreases in the S₃ state and, as a consequence, draws the two bridging oxo ligands closer for the imminent formation of an O–O bond before O₂ release. Consequently, the distance between the two Mn atoms lengthens. We previously proposed that in the S₄ state a dioxyl radical is produced which spontaneously converts to a peroxo species with the formation of an O–O bond (Figure 7).⁵ This proposal results from the study of a synthetic inorganic di-μ-oxo-bridged di-Cu compound that has been shown to convert into an isomer with the formation of an O–O bond between the two bridged oxygens.⁹⁰ Alternatively, the formation of the oxyl radical on one dimer in the S₃ state and its reaction with OH⁻ or H₂O either on the other Mn dimer or in the outer sphere during the subsequent S state conversion can lead to the formation of the O–O bond (Figure 7). The results reported by Messinger and co-workers support the presence of two nonequivalent exchangeable sites in the S₃ state as required by this proposal.^80,91

Our proposed mechanism avoids the formation of the O–O bond until the most oxidized state (S₄) is reached. This precludes the formation and release of peroxide or other oxidation products of water in the earlier S states, thus preventing the system from “short circuiting” and avoiding the risk of damaging the polypeptides of Photosystem II.