Rapid Loss of Structural Motifs in the Manganese Complex of Oxygenic Photosynthesis by X-ray Irradiation at 10–300 K*

Structural changes upon photoreduction caused by x-ray irradiation of the water-oxidizing tetramanganese complex of photosystem II were investigated by x-ray absorption spectroscopy at the manganese K-edge. Photoreduction was directly proportional to the x-ray dose. It was faster in the higher oxidized S2 state than in S1; seemingly the oxidizing potential of the metal site governs the rate. X-ray irradiation of the S1 state at 15 K initially caused single-electron reduction to S0* accompanied by the conversion of one di-μ-oxo bridge between manganese atoms, previously separated by ∼2.7 Å, to a mono-μ-oxo motif. Thereafter, manganese photoreduction was 100 times slower, and the biphasic increase in its rate between 10 and 300 K with a breakpoint at ∼200 K suggests that protein dynamics is rate-limiting the radical chemistry. For photoreduction at similar x-ray doses as applied in protein crystallography, halfway to the final MnII4 state the complete loss of inter-manganese distances <3Å was observed, even at 10 K, because of the destruction of μ-oxo bridges between manganese ions. These results put into question some structural attributions from recent protein crystallography data on photosystem II. It is proposed to employ controlled x-ray photoreduction in metalloprotein research for: (i) population of distinct reduced states, (ii) estimating the redox potential of buried metal centers, and (iii) research on protein dynamics.

. The crystallographic results on PSII represent a long awaited breakthrough. With respect to the manganese complex, however, the question has emerged regarding to what extent the obtained structural information is invalidated by modifications caused by the numerous radicals that are inevitably created by x-ray irradiation (11)(12)(13). In all four structures (8 -11) manganese ions were found; however, there were inconsistencies in their probable number and position with respect to the protein matrix. In addition, surprisingly few amino acid side chains were at a reasonably short distance to provide ligands ( Fig. 1). Water molecules, which were invisible at the best resolution of 3.2 Å (11), or disorder effects may explain the void around manganese. As previously estimated (12), recently shown by XAS on PSII crystals (14), and substantiated in this work, all crystallographic data likely refer to a metal site where x-ray exposure resulted in reduction of the initially high valent manganese ions to the Mn II level.
In protein crystallography, which is mostly done at 100 K at high flux third generation synchrotron radiation sources, prolonged x-ray irradiation of crystals eventually leads to a decrease in the diffracted intensity (15)(16)(17). Frequently, data collection is continued until this "radiation damage" becomes too pronounced for compensation by scaling of mosaicity and temperature factors. This approach becomes critical, if distinct modifications occur already prior to the loss in diffraction quality. Indeed, specific structural changes result from x-ray irradiation at ϳ13 keV (0.95 Å) even at low temperatures, e.g. breaking of disulfide bonds, modification of carboxylates and histidines, and tyrosine dehydroxylation, before the global fading out of diffraction intensity (18 -20). Reduction of active site transition metal ions may precede modifications of the protein matrix (21,22).
In the present study, the influence of ionizing x-ray irradiation on the manganese complex of PSII is investigated systematically by XAS at the manganese K-edge (23,24). As opposed to crystallography, XAS is particularly sensitive to the oxidation state and local structure of proteinbound metal centers (6). The position on the energy scale of the K-edge in the XANES region of XAS spectra is indicative of the formal oxidation state of the metal, at least in the case of the manganese complex of PSII (see Ref. 6 and the references therein), and thus was employed to monitor the time course of the photoreduction process. The EXAFS spectral region, which carries information on the metal-ligand distances, on the number and chemical nature of ligands, and on the geometry of the site, e.g. as maintained by well known bridging motifs such as di-and mono--oxo bridges between the metal ions, was scrutinized in detail to detect changes in the structure of the manganese complex.
The relations between x-ray flux and exposure time, temperature, and the rate and extent of manganese photoreduction and the associated structural changes were quantified. The implications of the results for interpretation of the crystallographic data are discussed. The obtained insights are of likely relevance for protein-bound metal centers in general and for the use of x-ray photoreduction as a specific tool to study intermediates of the catalytic cycle.

Sample Preparation and S-State
Population-PSII-enriched membrane particles were prepared from market spinach using betaine as a stabilizer in all media as described in Refs. 25 and 26 and stored at Ϫ80°C until use. Their oxygen evolution activity under saturating white light illumination was 1200 -1400 mol of O 2 (mg chlorophyll ϫ h) Ϫ1 at 28°C. PSII membrane particles were dissolved at 1 mg/ml of chlorophyll in a medium containing 15 mM NaCl, 5 mM MgCl 2 , 5 mM CaCl 2 , 1 M betaine, and 25 mM MES, pH 6.2, and partially dehydrated membrane multilayer samples on Kapton foil were prepared by the previously described centrifugation technique (25) and enriched in the S 1 state using a preflash protocol (7,25). S 2 state samples were prepared by continuous white light illumination (440 nm Ͻ Ͻ 750 nm) of S 1 samples for 2 min at the 200 K of a dry ice/ethanol bath (7,25). Quantitative population of the S 2 state was verified by EPR measurements (not shown) of its well known multiline signal (27).
XAS at Low Flux, Bending Magnet Beam Line-XAS measurements at the manganese K-edge were performed at the EMBL bending magnet beam line D2 at Hamburg Synchrotron Laboratory (Deutsches Elektronen-Synchrotron, Hamburg, Germany) using an energy-resolving solid state 13-element germanium detector (Canberra) for fluorescence detection (25). The Si111 crystal monochromator was detuned to 70% of maximal flux. The samples were kept in a helium cryostat at 20 K under ϳ200 mbar of helium gas. The x-ray spot size on the samples was ϳ4 ϫ 1 mm. The scan duration (6400 -7100 eV) was ϳ20 min.
XAS at High Flux, Undulator Beam Line-XAS measurements at cryogenic temperatures as well as at room temperature were carried out at the undulator beam line ID26 of the European Synchrotron Radiation Facility (Grenoble, France). Manganese XAS spectra were measured by monitoring the excited x-ray fluorescence perpendicularly to the incident beam by a PIN photodiode (3.8-cm 2 active area; Eurisys Meassures), as in Refs. 6 and 28. In the room temperature experiments, the samples were exposed to plain air of the climatized (18 Ϯ 2°C) experimental hutch of the beam line. In the low temperature experiments, the samples were kept in a helium cryostat at ϳ200 mbar of helium gas. Rapid scan XAS spectra were collected by simultaneous scanning of a Si220 crystal monochromator (scan range, 6500 -7100 eV; x-ray spot size on the sample, ϳ1 ϫ 1 mm) and the undulator gap in the rapid scan mode of ID26 (29) as described in Refs. 28 and 30. At both synchrotron radiation sources, the angle between the electric field vector of the incident x-ray beam and the sample normal was 55°( magic angle; see Ref. 31). Energy calibration of each XAS spectrum was facilitated by simultaneous measurements and Gaussian simulation of the narrow pre-edge peak absorption centered at 6543.3 eV of a KMnO 4 powder sample mounted in front of a detector at the end of the beam line. The accuracy of the energy calibration procedure is better than Ϯ 0.1 eV (25,28).
Data Evaluation-XAS spectra were normalized, and EXAFS oscillations were extracted and transformed to a wave vector scale (k scale) using an E 0 value of 6540 eV as described in Ref. 6. Edge energies were determined by the "integral method" (6). EXAFS oscillations were weighted by k 3 to compensate for the fall-off of amplitudes with increasing energy. Spectra were simulated by a least squares procedure using the in-house software SimX (32). Complex phase functions for different elements in the various shells of backscatterers were calculated using FEFF-7 (33). The amplitude reduction factor, S 0 2 , was 0.85. In the simulations a value for E 0 of 6547 eV was consistently used. The fit quality was judged by calculation of the Fourier-filtered R-factor (R F ) (28) subsequently to the curve fit itself, which involved only unfiltered EXAFS oscillations (6). Fourier transforms of EXAFS oscillations were calculated for energies between 20 and 500 eV above E 0 using fractional cosine windows extending over 10% at high and low ends of the k range.

RESULTS
X-ray Photoreduction of the Manganese Complex at 10 -20 K-A series of consecutive XANES spectra was collected on one single spot on the PSII samples. These spectra reflect the radiation-induced modifications of the manganese complex, which gradually develop with increasing x-ray exposure time. To resolve the initial steps of the x-ray photoreduction of manganese at 20 K, spectra were collected at a low flux bending magnet beam line (D2, Hamburg Synchrotron Laboratory, Deutsches Elektronen-Synchrotron, Hamburg, Germany). The slower photoreduction steps became experimentally accessible only at an undulator beam line (ID26, European Synchrotron Radiation Facility, Grenoble, France), where the x-ray flux was ϳ3 orders of magnitude higher and photoreduction was accelerated accordingly.
Reduction of manganese typically results in a shift of its K-edge position to lower energies (6). Fig. 2 shows edge spectra of the manganese complex not previously exposed to X-rays for the dark-stable S 1 state and for the S 2 state, both measured at the low flux beam line at 20 K. The oxidation state of the Mn 4 complex in these states most likely is  Mn III 2 Mn IV 2 and Mn III Mn IV 3 , respectively (Ref. 7 and the references therein). Accordingly, the light-induced S 1 3 S 2 transition is associated with a one-electron oxidation of the Mn 4 complex, which causes a shift of the K-edge by 0.5-0.9 eV to higher energies (7,25,34). Using the integral method (6), a similar upshift by ϳ0.7 eV as previously found (7) was obtained here (Fig. 2).
For samples in the more oxidized S 2 state, ϳ6 h of x-ray irradiation at the low flux beam line yielded a XANES spectrum, which resembles that of S 1 with respect to edge energy and shape (Fig. 2). The K-edge position at 20 K in the S 1 and S 2 state samples as a function of the x-ray exposure time is shown in Fig. 3. For samples initially in S 2 (triangles), it was confirmed that ϳ6 h of irradiation caused reduction by 1 equivalent of the Mn 4 complex as judged by the downshift of the K-edge energy by ϳ0.7 eV. On the other hand, for samples initially in S 1 (circles), ϳ18 h of irradiation yielded a downshift by ϳ0.7 eV (Fig. 3) to a state that resembled the native S 0 (7) with respect to its edge energy (Fig. 2). Thus, the initial photoreduction of the manganese complex in its higher oxidized S 2 state was more than two times faster than in S 1 . We conclude that the velocity of x-ray photoreduction critically depends on the oxidizing power of the metal center, in line with previous results (7).
The slower steps of x-ray photoreduction were monitored by collecting successive XAS spectra at 10 K at the high flux beam line on S 1 state samples (Fig. 4B). A clearly biphasic decrease of the edge energy was observed. Its bi-exponential simulation yielded for the faster edge shift by approximately Ϫ0.7 eV a time constant of fast ϭ 1/k fast ϭ 1.1 min. Comparison of the initial phase of photoreduction in Figs. 3 and 4 revealed its 600-fold acceleration at the high flux beam line, which is by a similar factor as the increase in the x-ray flux. The second main reduction phase accounting for most of the edge shift was described by main ϭ 1/k main ϭ 110 min. The rate constants are summarized in Table 1.
The edge energies in Fig. 4B were obtained from a series of complete XANES spectra. In Fig. 4C the edge energy was monitored by an alternative approach, namely the time scan technique (35). The x-ray fluorescence intensity was recorded at a fixed excitation energy of 6549.5 eV. At this energy, the x-ray absorption and thus the excited fluorescence increases for increasing reduction of the manganese complex (Fig. 4A). Fig. 4 (B and C) reveals similar kinetic behavior of the K-edge energy and of the increase in the x-ray fluorescence. Thus, the time scan approach can be used to monitor the K-edge energy of the metal and hence the level of x-ray photoreduction in an experimentally efficient way, which may also be employed at crystallography beam lines.
The pronouncedly biphasic downshift of the K-edge suggests the rapid formation of a discrete intermediate state upon the one-electron reduction of the manganese complex initially in S 1 . In the water oxidation cycle, the light-driven oxidation of S 0 leads to S 1 . Thus, a single reduction of S 1 leads to a formal S 0 state. Accordingly, the one-electron reduction of S 2 results in a formal S 1 state. We denote the intermediates created by x-ray photoreduction as S 0 * and S 1 *, respectively. The similar XANES spectra suggest a close correspondence between the native states (S 0 (7) and S 1 ) and the x-ray-induced intermediates (S 0 * and S 1 *). Notably, the more reduced states beyond S 0 * created by prolonged x-ray irradiation have no precedent in the native catalytic cycle.
Temperature Dependence of Photoreduction- Fig. 5 shows the results of x-ray photoreduction experiments at room temperature (290 K). Again, the irradiation time dependence of the edge shift did not exhibit a single-exponential behavior. Rather, as opposed to the rapid reduction phase observed at 10 -20 K, at room temperature an initial lag phase was observed (Fig. 5A) as in Refs. 7 and 30. Accordingly, at room temperature the initial rate of photoreduction is close to zero, whereas it is particularly high at low temperatures. In the following, the initial phases were not considered but only the subsequent apparently monophasic reduction process, which accounts for the major fraction of the irradi-  ation-induced edge shift. In comparison with the 10 -20 K data, at room temperature the rate k main of this photoreduction phase was accelerated by ϳ2 orders of magnitude (Table 1). Fig. 5B reveals that also at room temperature the time scan technique can be employed to monitor x-ray photoreduction of the manganese complex properly. By measuring such fluorescence time scans in the range from 10 to 300 K and their single-exponential simulation, the temperature dependence of k main was determined (Fig. 6). It is pronouncedly biphasic and reveals a sharp breakpoint around 200 K. Description of the data by a single activation energy was not possible as apparent from the Arrhenius-plot (ln k main versus T Ϫ1 ) (not shown). Such a plot yielded an activation energy of 45 meV for temperatures ranging from 200 to 300 K but a value of only 0.2 meV between 10 and 50 K.
Relation between X-ray Flux and Photoreduction Rate-The relation between the rate of photoreduction and the x-ray flux was investigated at the high flux beam line for PSII samples initially in the S 1 state by the time scan technique (Fig. 7). The time scans carried out at 20 K with maximal x-ray flux (Fig. 7A) or using an x-ray flux that was attenuated to 10% of the maximum (Fig. 7B) revealed identical biphasic behaviour, but a 10 times slower photoreduction in the latter case. Thus, a linear dependence of the reduction rate on the x-ray flux may be anticipated. To confirm this conjecture, fluorescence time scans at increasing flux were measured and simulated by Equation 1, taking into account the biphasic photoreduction at low temperatures, ϳ0.1 ϫ 10 12 0.5 ND 20 ϳ1.0 ϫ 10 12 4.9 ND 20 ϳ0.5 ϫ 10 12 0.9 0.009 ϳ20,000 10 ϳ1 ϫ 10 12 ND 0.020 ϳ10,000 10 ND 0.038 ϳ5,000 100 Lag phase 0.9 ϳ200 290 Crystallography c ϳ1 ϫ 10 14 (ϳ13 keV) ϳ100 ϳ1.0 ϳ200 100 S 2 D2 ϳ1 ϫ 10 9 0.0040 ND (Weeks) 20 a k fast accounts for the formation of S 0 . b Estimated irradiation time for 90% reduction of the manganese complex to Mn II 4 with rate k main . On the basis of our data at 100 K, full Mn-reduction in crystallography is estimated to occur after application of a dose of ϳ2 ϫ 10 16 photons mm Ϫ2 . The applied dose in crystallography was Ն2 ϫ 10 16 photons mm Ϫ2 (Ref. 14 and references therein); hence, most manganese ions likely were reduced to Mn II . c Estimates for crystallography beam lines refer to fluxes reported in (Ref. 14) and to the considerations under "Discussion." ND, not determined.
where F 0 denotes the starting fluorescence level, A fast denotes the amplitude (or shift in K-edge energy), and k fast denotes the rate constant of the initial phase. The slow phase was approximated by a linear decrease in edge energy with slope r main because it was not completed in the measuring time interval. The resulting values for k fast and R main ϭ r main /A fast are shown in Fig. 8A.
The rates of both photoreduction phases at 20 K rise linearly with increasing x-ray flux. As shown above, the x-ray photoreduction process at room temperature revealed an initial lag phase, after that monoexponential manganese reduction occurred until the final Mn II 4 level was reached. Also the rate constant of photoreduction at room temperature was directly proportional to the x-ray flux (Fig. 8B); the lag phase became shorter with increasing x-ray flux (not shown). Thus, at both temperatures the extent of manganese photoreduction depends, for otherwise unchanged conditions, exclusively on the integrated number (dose) of absorbed x-ray photons.
Radiation-induced Structural Changes of the Manganese Complex-Successive EXAFS scans at 10 K on a single spot of PSII samples during continuous x-ray irradiation at the high flux beam line yielded a series of nine spectra that was analyzed to unravel the accompanying structural changes of the manganese complex. Selected k 3 -weighted EXAFS oscillations obtained after distinct x-ray exposure times and their Fourier transforms (FTs) are shown in Fig. 9.
The Fourier-transformed spectrum of the native manganese complex in its S 1 state (Fig. 9B, 0 min) reveals two major peaks at 1.35 and 2.2 Å of reduced distance corresponding to backscattering atoms ϳ1.85 and 2.7 Å, respectively. Peak I stems from the oxygen and nitrogen atoms in the first coordination sphere of manganese, whereas peak III is dominated by short manganese-manganese vectors within the Mn 4 complex. In our hands, there is compelling evidence that the ϳ2.7-Å distances reflect the presence of two di--oxo-bridged manganese pairs in the S 1 state (Refs. 6, 7, and 12 and the references therein). (Deviating interpretations have been reported in Refs. 14 and 36.) Already 1.6 min of x-ray exposure prior to the EXAFS scan resulted in a significant decrease in the magnitude of peak III, suggesting a reduction in the number of the ϳ2.7-Å manganese-manganese vectors. After 5.5 min irradiation, peak III was diminished by ϳ50%. More prolonged x-ray irradiation resulted in the complete disappearance of peak III. In parallel, the formation of a new FT peak at ϳ1.8 Å (peak II) indicated the presence of manganese-ligand distances of ϳ2.2 Å. Such distances are typical for manganese-oxygen bonds in Mn II complexes (37). After very long x-ray irradiation, the peak II attributed to Mn II -O interactions became the only detectable one (not shown) so that the spectrum resembled that of [Mn II (H 2 O) 6 ] 2ϩ in solution (14,38).
Interestingly, there was a pronounced FT peak at ϳ3 Å that increased in size during the early phase of photoreduction and disappeared after longer irradiation times (Fig. 9B, asterisks). The transient increase of this peak likely reflects the appearance of a longer manganese-manganese vector of ϳ3.5 Å, presumably because of the formation of a mono-oxo-bridged manganese pair in S 0 *. Thereafter, all manganese-manganese vectors gradually became elongated to Ͼ3.5 Å and thereby EXAFSinvisible during photoreduction to the final Mn II 4 state. To extract quantitative structural information, the k 3 -weighted EXAFS spectra (Fig. 9A) were simulated as previously described (6). We focused on the manganese-oxygen/nitrogen and manganese-manganese distances in the first and second coordination spheres. Therefore, all backscatterers at distances above 3 Å were not considered in the curve fitting. This distance range has been shown to comprise manganese-manganese and manganese-calcium vectors of ϳ3.1-3.4 Å length (4 -7, 14, 39). Their neglect may result in a minor increase of the apparent coordination number (from ϳ1.0 to ϳ1.25) of the ϳ2.7 Å shell assignable to di--oxo-bridged manganese pairs (6,38). Otherwise, however, the simulation results for the manganese-backscatterer shells below 3 Å remained essentially unaffected. The simulation curves obtained for a joint fit (38) of the nine spectra, including two shorter manganese-oxygen interactions (backscatter shells I and II) to account for the complex distance distribution function of the manganese-oxygen/nitrogen bonds (6) and one ϳ2.7 Å manganese-manganese vector  . Structural changes of the manganese complex during x-ray photoreduction at 10 K revealed by EXAFS analysis. The spectrum before x-ray irradiation (0 min) corresponds to the manganese complex in its S 1 state. A, k 3 -weighted EXAFS oscillations and simulation curves (thick lines). B, the respective FTs for increasing irradiation time periods. The arrows in B mark the FT peak III stemming from ϳ2.7 Å manganese-manganese vectors; asterisks mark contributions from longer manganese-manganese vectors.
(shell III), are shown in Fig. 9A (smooth lines); the corresponding fit parameters are depicted in Fig. 10.
The coordination number N III of the ϳ2.7-Å manganese-manganese vector exhibited a clearly biphasic decrease for increasing x-ray irradiation time. A rapid phase essentially completed after 5 min was followed by a slower decrease; already after ϳ60 min of irradiation the ϳ2.7 Å vectors became practically undetectable (Fig. 10A). Concomitantly, the mean manganese-oxygen/nitrogen distance ϽR IϩII Ͼ rapidly increased by ϳ0.04 Å, which was followed by a much slower and almost linear increase not completed within 90 min (Fig. 10B). The manganese-oxygen/nitrogen bond length is expected to increase upon manganese reduction. Typical bond lengths for manganese coordinated by six oxygens are ϳ2.2 Å for Mn II , 2.0 Å for Mn III , and 1.9 Å for Mn IV (37). That the average bond length of the Mn III 2 Mn IV 2 complex in the S 1 state was shorter than the ϳ1.96 Å suggested by the above numbers is explainable by the presence of several short 2 -O bonds in the Mn 4 complex and presumably by the presence of penta-coordinated Mn III (6,40). The initial increase of the average manganese-oxygen/nitrogen bond length is compatible with a single Mn IV 3 Mn III reduction, whereby the respective manganese-oxygen/nitrogen bonds are elongated by ϳ0.15 Å. This result is in accord with the edge energy analysis, suggesting the rapid formation of a singly reduced S 0 * state from S 1 .
The results in Fig. 10 were obtained by simulation of the 2.7 Å manganese-manganese interaction using a common value for the Debye-Waller parameter ( III ) for all nine spectra. In alternative simulations involving a variable III , its irradiation time dependence was found to be insignificant. In particular, no increase in III during the initial 5-min reduction phase was observed. As shown in Fig. 10B, also the ϳ2.7-Å manganese-manganese distance (R III ) changed only marginally, but its apparent coordination number (N III ) was reduced by ϳ50%. If it is assumed that there are two 2.7-Å manganese-manganese vectors in the S 1 state assignable to di--oxo-bridged manganese pairs, seemingly after only ϳ5 min of irradiation one bridge was lost or modified. Dissolution of the second di--oxo bridge during photoreduction was clearly slower but essentially completed after ϳ60 min. Then, as judged by the K-edge energy (Fig. 4) and by the coordination numbers of the manganese-oxygen/nitrogen vectors (Fig. 10A), the manganese complex was reduced by ϳ4 equivalents, presumably to the Mn II 2 Mn III 2 state. Thus, already after approximately half of the irradiation time required to reduce all four manganese ions to Mn II , di--oxo bridges seem to be absent in the complex.
In conclusion, even at temperatures as low as 10 K, x-ray-induced photoreduction of the manganese complex is coupled to the loss of most or all -oxo-bridging motifs between manganese ions. On the other hand, controlled photoreduction may be employed to investigate distinct intermediates as the S 0 * and S 1 * state, which likely are related to the native structure of the complex.

X-ray Photoreduction: Mechanistic Considerations-
The PSII manganese complex initially is more than twice as rapidly reduced when starting from the S 2 state than from S 1 (Ref. 7 and this work). But also the S 1 state is reduced faster than metal sites of other enzymes (see below). At low temperatures, the S 1 state first undergoes a rapid one-electron reduction resulting in S 0 *. Further reduction to the Mn II 4 level is 100 times slower.
These results are in accord with the estimated redox potential of the manganese complex. In S 1 it is highly positive; approximately ϩ1 V (41). A potential increase upon the S 1 3 S 2 transition and a decrease in the S 0 * state is anticipated (41,42); the potential of the Mn II 4 state presumably is comparably low (Յ0 V) (43,44). That electrophilic centers are the primary target for reduction by electrons liberated because of x-ray exposure is also supported by investigations on protein crystals where disulfide bridges first became reduced at 100 K (18,19) and by EPR results (Ref. 45 and the references therein). We conclude that the high susceptibility of the PSII manganese complex to photoreduction is governed by its extraordinary positive potential.
Whereas the oxidation state of the metal center frequently can be determined from the energy of the metal K-edge, the redox potential of buried protein-bound metal centers mostly remains obscure. Quantification of the relation between electrophilicity and photoreduction rate could provide an estimate of the redox potential also for deeply buried metal sites.
Interestingly, the temperature dependence of the main rate of photoreduction of the manganese complex showed a sharp breakpoint at ϳ200 K. A similar breakpoint was observed in recent measurements of the mean square displacements, ⌬x 2 , of the atoms in PSII membrane particles by quasi-elastic neutron scattering (46). Also in other proteins where ͗⌬x 2 ͘ was studied for bound iron (47,48), hydrogen atoms (49,50), methyl-bearing side chains (51), and specific amino acid residues and associated water molecules (52), a breakpoint consistently was found at ϳ200 K. Similar breakpoint behavior was also observed for certain electron transfer steps in photosystem I (53) and in the bacterial photosynthetic reaction center (54).
The sharp increase of ͗⌬x 2 ͘ above 200 K has been attributed to the onset of inharmonic, protein-specific dynamics. The breakpoint temperature may reflect the transition between two regimes of distinctly different dynamic behavior; occasionally the term "protein-glass transi- FIGURE 10. Changes in simulation parameters of EXAFS spectra of the manganese complex with increasing x-ray irradiation time. A, coordination numbers N. B, manganese-backscatterer distances R. The following restraints were used in the joint fit (38) of nine spectra employing three shells of backscatterers. The sum of the coordination numbers N I and N II was fixed to 5.5 (12), 2 2 I and 2 2 II were coupled to yield equal values for all nine spectra, R II was coupled to yield equal values for all nine spectra, and 2 2 III was fixed to a value of 0.002. FEBRUARY 24, 2006 • VOLUME 281 • NUMBER 8 tion" is used (55,56). The here observed breakpoint in the temperature rate relation of x-ray photoreduction of the manganese complex suggests that protein-specific dynamics affect the chemistry of irradiationinduced radicals, presumably by influencing the electron transfer steps that lead to reduction of the metal site. Such motions also may be of relevance for the electron transfer kinetics involving the native manganese complex.

X-ray Photoreduction of the Manganese Complex of PSII
We observed a linear relation between x-ray flux and the photoreduction rate of the manganese complex, suggesting that the x-ray dose determines the extent of photoreduction. Good evidence for a linear relation between flux and rate of radiation damage relation also has been obtained for protein crystals irradiated at 100 K (57), but also deviations from linear behavior were reported (58). Linearity in case of the PSII manganese complex implies that it is straightforward, at least in this system, to adjust either the x-ray flux or the irradiation time so that a crystallographic structure of the unperturbed manganese complex in its S 1 state may be reachable.
In the used PSII samples there are ϳ10 5 light atoms (carbon, nitrogen, and oxygen) per manganese complex (hydrogen atoms can be neglected because of their small absorption cross-section). At energies above the manganese K-edge (e.g. 6700 eV), their x-ray absorption cross-section, C,N,O is roughly 40 times smaller than Mn (59). Thus, less than 0.2% of the x-ray photons are absorbed by manganese. Also in absolute terms, direct absorption by manganese is a rare event as even at the sample surface; only ϳ1.5% of the manganese ions are hit during 1 h of irradiation at 10 12 photons s Ϫ1 mm Ϫ2 . Thus, the x-ray absorption by manganese ions is excluded as the primary cause of radiation-induced modifications.
Below ϳ15 keV the dominant primary event after x-ray absorption is photoionization. Mostly core electrons of carbon, nitrogen, and oxygen are liberated, causing a cascade of secondary ionization events, which results in the formation of electron-deficient cation radicals, electron surplus anion radicals, and population of excited states (45,60). Assuming that less than 25 eV is required per ionization event, we estimate that more than 250 radical pairs are initially formed per absorbed x-ray photon in PSII. A significant fraction of these radicals is stable at temperatures Ͻ100 K (61).
The photoreduction is proportional to the accumulated dose but independent of the dose rate. This shows that thermally activated radical migration or chemistry proceeding on the seconds to minutes time scale is not rate-limiting for photoreduction. We propose that the initially formed excited radical states represent, prior to thermalization, the source for electron transfer to the manganese complex (electron sink). The electron transfer rate is determined by their distance and potential difference to the manganese complex and by the temperature (62). Electron transfer and possibly also hydrogen atom transfer and more complex chemistry competes with thermal and radiative (luminescence) excited state decay. Thus, thermal activation behavior concomitant to dose-rate independence of x-ray photoreduction is predicted.
The S 0 * State Created by X-ray Irradiation-X-ray irradiation of proteins has been used to reduce their metal site so that putative intermediates of the catalytic cycle became accessible for structure determination (21,63). Underlying was the assumption that controlled irradiation generates distinct redox states of the metal before significant radiation damage to the protein occurs. We obtained evidence that this rationale also holds true for the PSII manganese complex.
Starting in the S 1 state, a rapid one-electron photoreduction leads to the formation of S 0 *. Accompanying is a reduction in the number of ϳ2.7 Å manganese-manganese vectors by ϳ50%, suggesting that one of the two di--oxo-bridged manganese pairs is modified or lost. Modification might result from protonation of a bridging oxide so that an increase of the original ϳ2.7 Å manganese-manganese distance in the Mn(-O) 2 Mn pair to ϳ2.85 Å in the Mn(-O)(-OH)Mn unit is expected (64). Such a modified manganese complex should show a mean manganese-manganese distance of ϳ2.77 Å instead of ϳ2.7 Å. However, in the EXAFS analysis, such a distance increase was not observed, and formation of a -OH bridge in S 0 * thus is unlikely. On the other hand, loss of a -O bridge, e.g. by change of the bridging to a terminal oxygen is compatible with the EXAFS data. In this case a lengthening of the manganese-manganese distance to Ͼ3 Å is expected (7). Indeed, the S 0 * state EXAFS suggested an additional ϳ3.5 Å manganese-manganese vector. We conclude that the S 1 3 S 0 * reduction is associated with the conversion of a di--oxo bridge to a mono--oxo bridge.
Interestingly, also for the native S 0 state, a decrease in the apparent number of ϳ2.7 Å manganese-manganese vectors relative to S 1 has been observed, but in this case individual manganese-manganese distances of ϳ2.7 and 2.8 -2.9 Å were resolvable (4,7,36). Thus, the native S 0 as created by light flash application at room temperature may contain one Mn(-O) 2 Mn and one Mn(-O)(-OH)Mn unit (4,7). At room temperature the S 0 3 S 1 transition is electroneutral (65,66); probably the manganese oxidation is coupled to deprotonation of a -OH bridge (7,36,42). At the cryogenic temperatures where the S 1 3 S 0 * step occurred, charge-compensating long-range proton movements are unlikely so that protonation is impaired, and instead the -O bridge is broken.
Recent crystallographic data suggested that one specific manganese ion (Mn 1 in Fig. 1), presumably coordinated by D1-Asp 170 , is the first to become disordered after prolonged x-ray exposure of PSII. 3 It is tempting to speculate that Mn 1 is connected to a neighboring manganese by a di--oxo bridge in the S 1 state but by a mono--oxo bridge in the S 0 *. Proton release from the -OH bridge upon the native S 0 3 S 1 transition may then occur via the proton channel close to Asp 170 , which has been 3 J. Kern and J. Biesiadka, personal communication.  Fig. 1 the view is rotated by ϳ90°around the horizontal axis (1). The S 1 3 S 0 * transition results from one-electron reduction of Mn 1 , which leads to conversion of a bridging oxygen to a terminal one; consequently the Mn 1 -Mn distance is lengthened to ϳ3.5 Å (dark arrows) (2). Reduction by five electrons to the Mn II 4 state is accompanied by the loss of all -O motifs and thus an increase in the manganese-manganese distances to Ͼ3.5 Å; the increased manganese-manganese distances likely result in extension of the apparent electron density of manganese in the crystallographic data (light arrows); changes in the positions of neighboring amino acids are probable. At present it is still unclear which of the manganese ions contribute to the second di--oxo motif present in the intact complex and lost during the slower phase of x-ray photoreduction.
tentatively assigned in the PSII structure (10). Upon its reduction to Mn II , the mono--oxo bridge may be lost so that Mn 1 becomes increasingly disordered. We summarize these attributions in Fig. 11.
Implications of Manganese Photoreduction for Crystallography-In protein crystallography at synchrotron radiation sources, data frequently are collected using x-rays of an energy E of ϳ13 keV. The x-ray absorption coefficient, (E), of carbon, nitrogen, and oxygen matrix atoms to a first approximation is proportional to 1/E 3 , whereas the deposited energy per photon increases linearly with energy. Thus, for identical photon flux the photoreduction rate should decrease quadratically with increasing x-ray energy. Therefore, for crystallographic data collection at ϳ13 keV, the rate of photoreduction may be approximately four times lower than that reported here for irradiation at ϳ6.7 keV. This estimate is in line with previous experimental results (14).
In the present study, photoreduction of the manganese complex was investigated for a specific type of PSII preparation, namely multilayers of membrane particles in which the water content was reduced to ϳ50%. 4 An increased water content (use of frozen solutions) or a decreased lipid content (in PSII core particle preparations used for crystallization) (14) and also stabilizers such as betaine and cryoprotectants (glycerol and sucrose) may slightly affect the photoreduction rate. In any event, on the basis of the results of the present study (Table 1), our previous estimates (12), and recent XAS experiments on PSII crystals (14), we conclude that likely all manganese ions became reduced to the Mn II level in crystallography at 100 K and using X-rays of ϳ13 keV.
To what extent is the crystallographic structure of the manganese complex affected by x-ray photoreduction? In crystallography at 100 K involving x-ray fluxes of ϳ10 14 s Ϫ1 mm Ϫ2 (12,14), the S 0 * state certainly is rapidly reached (within a few seconds). In particular this rapid initial phase of photoreduction may easily be overlooked and indeed has not been noticed previously (14). Already in the S 0 * state where only a single manganese ion is reduced by one equivalent, one di--oxo bridge between manganese ions is lost, and the respective ϳ2.7 Å manganesemanganese distance is elongated to ϳ3.5 Å. More prolonged irradiation results in progressive Mn II formation and in the total loss of -oxo bridges between manganese ions so that all manganese-manganese distances become elongated even at 10 K. It is important to note that the loss of di--oxo bridges was already completed halfway to reduction to the final Mn II 4 state. Thus, in the crystallographic electron densities, manganese-manganese distances Ͻ3 Å presumably are no longer present, and the arrangement of the Mn II ions is pronouncedly different from the intact complex.
It is an unexpected result of this and previous (14) studies that the final Mn II 4 state created at low temperatures spectroscopically resembles the [Mn II (H 2 O) 6 ] 2ϩ complex in solution. A straightforward interpretation of this observation is that not only -oxo bonds are broken but that also rearrangements of the manganese ions and of their ligands from amino acids occur so that additional H 2 O molecules may become bound directly to Mn II . At the current crystallographic resolution of 3.2 Å (11), the displacement of manganese ions and amino acid side chains by Ͻ1 Å may be hard to detect. However, such structural changes could explain the unassigned space around manganese ions and the inconsistencies with respect to side chain orientations among the crystal structures (8 -11). A distinct structural model including a Mn 3 Ca( 3 -O) 4 cubane cluster plus a nearby "monomeric" manganese ion has been proposed (10). This model represents a stimulating hypothesis, but its specific features are likely to be revised if photoreduction during crystallography is decreased by more optimized experimental strategies. Preliminary evidence for a different position of the Mn 1 ion (Fig. 11) in crystal structures obtained at 100 and 20 K has been derived. 3 Specifically, we propose that the apparently monomeric manganese ion may originally be connected to the complex by a di--oxo bridge in the S 1 state but by a mono--oxo bridge in the irradiation-induced S 0 * state.