Rapid evolution of the Photosystem II electronic structure during water splitting

Abstract

Photosynthetic water oxidation is a fundamental process that sustains the biosphere. A Mn₄Ca cluster embedded in the photosystem II protein environment is responsible for the production of atmospheric oxygen. Here, time-resolved x-ray emission spectroscopy (XES) was used to observe the process of oxygen formation in real time. These experiments reveal that the oxygen evolution step, initiated by three sequential laser flashes, is accompanied by rapid (within 50 μs) changes to the Mn Kβ XES spectrum. However, no oxidation of the Mn₄Ca core above the all MnIV state was detected to precede O−O bond formation, and the observed changes were therefore assigned to O−O bond formation dynamics. We propose that O−O bond formation occurs prior to the transfer of the final (4th) electron from the Mn₄Ca cluster to the oxidized tyrosine Y_Z residue. This model resolves the kinetic limitations associated with O−O bond formation, and suggests an evolutionary adaptation to avoid releasing of harmful peroxide species.

Enzymes function as nature’s catalysts, facilitating virtually all the reactions necessary for life. By carefully coordinating electron dynamics and atomic rearrangements within a predefined energy landscape, they enable a broad range of efficient and highly selective transformations, many of which have proven challenging for chemists. Among these, the reaction catalyzed by the Mn₄Ca cluster of photosystem II (PS II) during photosynthesis holds a special place, as the ability to biosynthesize O₂ from H₂O occurred only once during evolution 3. The development of this process dramatically altered our planet by generating the oxygen-rich atmosphere we live in today. The catalytic activity and quantum efficiency of the oxygen evolving complex (OEC) remain unmatched by synthetic systems developed for artificial photosynthesis `. Despite its far-reaching consequences, the underlying mechanism of PS II remains debated.

In 1970, Bessel Kok et al. described a potential water splitting mechanism in which the OEC cycles through five states (S₀-S₄), corresponding to the oxidation states of manganese, following sequential visible light absorptions, Figure 1A [1]. Antenna pigments from the surrounding protein matrix absorb these photons and funnel energy towards P₆₈₀, the chlorophyll a special pair responsible for charge separation. Within nanoseconds, the tyrosine Y_Z residue (Tyr_Z) located between P₆₈₀ and the OEC is oxidized by the special pair. Tyr_Z• is subsequently reduced by the OEC on a microsecond timescale. This process drives the water splitting reaction [2].

Figure 1. Schematic outline of the TR-XES experimental setup.

(A) Current model of the Kok cycle. Depiction of sequential incident visible light photon absorptions triggering electron/proton release [13]. The dashed region is based on previous analysis of the S₃ to S₀ transition in which the S₄ state was proposed [13,49]. (B) Electronic transitions, reflected in the Kβ main lines, are influenced by the spin state of the Mn ion. (C) A spectral comparison of Mn oxides depicts the effect of changing oxidation state on the Kβ emission lines. (D) Nanosecond laser pulses (1, 2 or 3) are used to advance the Kok cycle in the protein, Table S1. The pump/probe delay time, Δt, measured from the final laser flash to the center of the x-ray pulse, is set dependent on the desired S-state. X-ray fluorescence from the sample is reflected by 10 flat analyzer crystals onto a 2D-position sensitive detector. Kβ emission spectra are extracted to form snapshots of the electronic structure in time. Smoothed emission spectra are presented for 2F (majority S₃) and two time points during the S₃ to S₀ transition. (E) The proposed reaction scheme shows the early evolution of the OEC during the S₃ to S₀ transition, providing an interpretation of spectroscopic results.

The past forty years have yielded new insights into the structure of PS II [3–12], as well as the nature and timing of the S-state transitions that form the Kok cycle, Figure 1A [13–17]. Nonetheless, the critical step during which the O−O bond is formed remains poorly characterized and thus cannot be implemented in artificial systems. O−O bond formation likely occurs on a microsecond timescale during the S₃ to S₀ transient step of the catalytic cycle, culminating in O₂ evolution. Direct monitoring of this transient process however, has proven challenging and details remain elusive. A pre-eminent report by Babcock et al., as well as recent studies by Nilsson et al., support a rate of Tyr_Z• reduction with t_1/2 ~1 ms following three flashes, and associate this rate constant with the formation of the S₄-state, subsequently capable of fast O₂ evolution [18,19]. It was pointed out later that such hypotheses face a serious kinetic challenge in that the timescale of molecular oxygen release, following the formation of the S₄-state, is very short for the associated redox chemistry and bond formation dynamics [20]. As an alternative hypothesis, S₃-state peroxide formation was proposed [20], but this has not yet been confirmed experimentally.

Given the lack of definitive spectroscopic results, computational simulations have been performed to model the O−O bond formation path [21–24]. Many of these imply oxidation of the OEC past Mn₄IV via the formation of a MnVMn₃IV state, also presented as Mn₄IV−O• (oxyl radical). This oxidized configuration is currently associated with S₄ and would precede O−O bond formation [21,25]. Experimental proof for a MnVMn₃IV intermediate state is currently lacking, and our data rule out its formation. Here, we examine the earliest dynamic in the S₃ to S₀ transition via time-resolved x-ray emission spectroscopy (TR-XES) utilizing dispersive detection to aid our understanding of this critical biological process [26,27]. In 2015, we proposed a new mechanistic model in which O−O bond formation occurs prior to transfer of the final (4th) electron from the Mn₄Ca cluster to explain our preliminary spectroscopic results [28]. An x-ray crystallographic study of the S₃-state [10] recently confirmed our DFT-based proposal, producing a virtually indistinguishable S₃-state model, within the experimental resolution of x-ray diffraction [29].

Here, we deliver an extensive statistical analysis of these initial datasets in conjunction with those collected subsequently, each consistently composed of almost a half million repetitive interrogations of the OEC electronic structure. The power of large statistical data sets has long been realized in science, often enabling the development of new research tools. A single, blurry electron microscopy image of a complex biomolecule, for instance, provides little insight into its structure, while several thousand images can now deliver atomically-resolved models [30]. In a similar fashion, repetitive measurement of a spectroscopic response has allowed us to solidify the rapid electronic structure evolution in the S₃ to S₀ transition, a result which is required to complete the description of O−O bond formation in natural photosynthesis.

Mn Kβ spectral emission lines reflect the number of unpaired 3d electrons and, thus, provide information about the oxidation and/or spin states for a given Mn ion, inaccessible via structural methods such as x-ray crystallography [31]. The exchange interaction between the 3p hole and 3d valence electrons in the final state causes multiplet splitting that results in separate Kβ_1,3 and Kβ′ peaks, Figure 1B, C. This coupling is directly linked to the electronic state of Mn such that an increase in the oxidation state results in decreased splitting between the Kβ spectral lines. This effect is most apparent in the Kβ_1,3 peak position shift to lower energies with increasing oxidation. XES also allows for dispersive detection, in which the full emission spectrum is recorded during a single, intense, polychromatic x-ray pulse, Figure 1D [26]. The temporal resolution of such a setup is limited in practice only by the time structure of the x-ray source. In our experiments, we utilized multi-bunch x-ray exposures of 22 μs to 44 μs duration to match the microsecond kinetics of the OEC [13], Table 1 & Figure S1A. We previously determined exposures of up to 66 μs under these conditions to be undamaging to PS II [26,32]. Data collection was performed using a von Hamos style miniature x-ray emission spectrometer (miniXES – Figure S1B) [26,33], and a non-jet open-air sample delivery system (see SI and Figure S1C) was used to supply fresh, unrecycled PS II for each measurement at a controlled repetition rate. Samples were excited given a defined number (0–3) of laser flashes (F) and probed at a time (Δt) after the final laser flash by a single x-ray pulse, Figures 1D & S1A, Table S1. For clarity, we note that 0F, corresponding to no flashes, corresponds to majority state S₁, 1F corresponds to majority state S₂, etc.

Table 1.

Experimental characteristics of pulsed x-ray source

Characteristics	BioCARS
Excitation Energy	Peak energy 7.85 keV, FWHM ~500 eV
X-ray Spot Size	~45 × 100 μm2
Pulse Length	44 μs (22 μs dataset 5)
Photon Flux	3 × 1011 photons/pulse
Dose Delivered per Pulse	~7 × 107 photons/μm2

A total of seven beamtimes were accomplished, with two devoted to methodology development and five to data collection (note that dataset #5 was recorded after a beamline upgrade and is analyzed separately), ultimately accumulating over two million x-ray pulses to measure different S-states, Table 2. Time-resolved spectra of the majority S-states S₁, S₂, and S₃ were collected following zero, one and two laser flashes with spacing between consecutive laser flashes of 100 ms, corresponding to a laser frequency of 10 Hz. Samples were probed with the x-ray beam following a Δt=500 μs time delay from the final laser flash to allow for the full reduction of Tyr_Z• with limited decay of the formed S-state, Table S1 [13,34]. Given the multiplet character of the spectra and the noise inherent for such a dilute biological sample, previous studies recommend the use of the statistical first moment, $\frac{\sum {}_j{E}_j{I}_j}{\sum {}_j{I}_j}$, surrounding the Kβ_1,3 peak, to determine any changes to the electronic structure, Figures 2 & 3 [35]. It has been reported, and should be emphasized here, that any data manipulation such as background subtraction and smoothing can affect first moment magnitudes, leading to risk of misinterpretation of small spectral changes [17]. To avoid uncertainties due to these processing methods, we provide the first ever analysis of primary photosystem II emission data. These data sets were subject to no manipulation beyond extraction of spectra via energy calibration of the detector. The statistical significance of the observed spectral changes is then determined using one-way ANOVA (ANalysis Of VAriance), a simple statistical method employed broadly across many scientific disciplines. Resultant p-values represent the probability of falsely rejecting the null hypothesis, i.e. that a difference in the first moment between the two states listed is a random statistical variation. Thus, the lower the p-value, the stronger the evidence for changes in the spectra between compared data sets. In this study we take the 95% confidence interval to be significant. Here, background subtraction and smoothing is only used to produce figures for comparison with previous XES PS II studies, which all present significantly processed data.

Table 2.

Approximate number of x-ray pulses per state, per beamtime.

Majority S-State	Flash	Data Sets
		1	2	3		4	5**
S0	3F40ms	43,250	60,000	55,333		107,222	97,800
S1	0	62,267	72,800	60,233		n/a	100,000
S2	1	42,083	61,556	55,333		n/a	88,900
S3	2	n/a	62,000	57,178	48,889	107,222	95,600
S4a	3F50μs	n/a	n/a	57,178	48,889	111,667	93,300
S4b	3F200μs	n/a	60,889	57,178	48,889	110,556	86,700
S′4	3F500μs	43,750	60,889	57,400		n/a	91,100
Total		191,350	322,334	546,500		436,667	653,400

^*Additional statistics were collected for S₃, S_4a and S_4b for data set 3. The columns are split to reflect the additional x-ray pulses for these states.

^**This dataset was collected after 2015 upgrades to the optics at the BioCARS beamline which made impossible the use of the old experimental conditions and required twice shorter (1/2 intensity) pulses to minimize heat load on new optics components.

Figure 2. Analysis of statistical significance for S₁→S₂→S₃ transitions serves as a proof of concept.

(A) Spectral shifts occurring as a result of 0F→1F and 1F→2F transitions are characterized using 2-D heatmaps, illustrating the p-value (see Table S2 for n) for changes to first moments calculated over the ranges defined by the x and y axes, i.e. start and end energy, respectively. Contours appear in plots comparing 0F→1F data indicate statistically significant difference. By contrast, limited low p-value regions are observed for the 1F→2F transition, suggesting a smaller change. The directionality and magnitude of spectral shifts are shown in the final column. These 2-D heatmaps, which graphically illustrate the change in first moments (ΔFM=FM_post-flash-FM_pre-flash) calculated over the ranges defined by the x and y axes. 0F→1F and 1F→2F transitions are dominated by positive, or oxidative, shifts. An example of statistical noise is presented by randomly dividing a 0F dataset and performing the same analysis, e.g. 0F→0F. Relevant data from datasets 1–4 were merged to generate the final columns. Note that 0F and 1F data were not collected for dataset 4, while dataset 5 was collected independently following beamline upgrade and is therefore analyzed separately, see Fig. S4. (B) Wavelet-transform smoothed via wavelet transforms and background subtracted emission spectra for merged 0–2F data. The region (6.485–6.495 keV) over which the first moment was calculated is highlighted and a magnified inset as well as difference spectra are presented. Difference spectra were smoothed with a rolling average calculated over 14 points (~ ±0.7 eV). (C) (left) Average first moments from unprocessed (color) and processed (grey) spectra. Errors are presented as SEM with n given in Table 2. Those moments with a statistically significant difference (p<0.05) from 0F data are indicated with an asterisk. (right) Dot plot of first moments from raw data. Each dot represents a calculated first moment from a thread collected during the beamtime corresponding to its color in the legend. Dashed lines represent the average first moment while solid bars are the 95% confidence interval.

Figure 3. Rapid onset of spectral changes observed during the S₃ to S₀ transition.

(A) Spectral shifts occurring as a result of the 2F→3F transition are characterized using 2-D heatmaps colored by p-value (see Table S2 for n), for changes to first moments calculated over the ranges defined by the x and y axes, i.e. start and end energy, respectively. Strong contours are consistently observed indicative of a statistically significant spectral change. The directionality and magnitude of spectral shifts occurring with these transitions are also shown as 2-D heatmaps, which graphically illustrate the change in first moments (ΔFM=FM_post-flash-FM_pre-flash) calculated over the ranges defined by the x and y axes. All 2F→3F comparisons yield primarily negative, or reductive, shifts. Determined p-values ~0.02 (Table 3) for the 200 μs time point are further supported by the above heatmaps, however, little information can be gleaned from these plots of dataset 2, likely due to much lower x-ray shot count, see Table 2. Although the 500 μs time point is equally compelling, the traditional energy range (6.485–6.495 keV) selected for reporting first moments does not deliver statistical significance, Table 3. (B) Wavelet transform smoothed and background subtracted emission spectra for merged 2–3F data, calculated as in Fig. 2B. (C) (left) Average first moments from unprocessed (color) and processed (grey) spectra. Errors are presented as SEM with n given in Table 2. Those moments with a statistically significant difference (p<0.05) from 2F data are indicated with an asterisk. (right) Dot plot of first moments from raw data, presented as in Fig. 2C. See Fig. S3 for 2-D plots of the 40 ms (majority S₀ state) time point.

While it is common to analyze first moments calculated over 6485–6495 eV, this approach fails to convey the full information content of XES spectra. To show that our trends are not dependent on a particular range chosen for analysis of the 1st moment, p-contours over variable energy integration ranges are shown in Figures 2, 3, & S3. It is evident that merged data are dominated by statistically significant shifts, with the exception of 1F to 2F, in agreement with earlier work [35]. Note that corners of the contour maps, which either encompass integration ranges outside the peak position, or incomplete portions of the peak, may be influenced more significantly by noise. This is especially true for contours maps of individual beamtimes.

S₁ to S₂-state transition.

The obtained S₁ state spectral shape and peak position are in good agreement with previous RT PS II measurements [26]. A comparison of 0F and 1F first moments evaluated by one-way ANOVA shows a reproducible, statistically significant shift of the Kβ_1,3 peak to lower energies, Table 3 & Figures 2 & S4. Cryogenic measurements previously reported a −0.059 eV shift following spectral smoothing and background subtraction procedures [35]. Analysis of three combined data sets (#1, 2, and 3) following similar data processing yields a first moment shift of −0.16 eV for our 0F → 1F transition (note that dataset #4 was measured without the S₂-state, and #5 was measured after beamline update is analyzed separately). The observed shift of the Kβ_1,3 peak to lower energies following a single laser flash is likewise consistent with previous cryogenic XANES [8,36] and recently published room-temperature XES [17] results for the S₁ to S₂ state transition, both of which indicate Mn-centered oxidation. X-ray free electron laser (xFEL)-based TR-XES experiments have struggled to reproduce this result, likely due to a combination of lower spectrometer resolution and data processing methods [27]. Although some xFEL studies [37,38] have observed spectral shifts for x-ray measurements made following a different number of laser flashes, they have not yet provided any new mechanistic insights. Overall, we consider our results for the 0F → 1F transition robust and in good agreement with previous characterization of the S₁ to S₂ transition in which one Mn center is oxidized from MnIII to MnIV.

Table 3.

p- and F-values from ANOVA for state-to-state comparisons from data sets 1–4 combined.

Majority S-State (n)		S0	S1	S2	S3	S4a	S4b	S’4
	Flash	3F40ms	0F	1F	2F	3F50μs	3F200μs	3F500μs
S0 (117)	3F40ms	1	0.08 (3.06)	0.06 (2.19)	0.06 (3.60)	0.17 (1.82)	0.63 (0.23)	0.66 (0.19)
S1 (78)	0F		1	<0.01 (9.00)	<0.01 (13.56)	0.68 (0.17)	0.19 (1.67)	0.05 (3.80)
S2 (81)	1F			1	0.85 (0.04)	<0.01 (7.27)	0.02 (3.68)	0.98 (0.82)
S3 (96)	2F				1	<0.01 (11.32)	0.02 (5.87)	0.22 (1.50)
S4a (75)	3F50μs					1	0.11 (0.80)	0.18 (2.60)
S4b (98)	3F200μs						1	0.39 (0.73)
S’4 (82)	3F500μs							1

^*p-values are based on the first moments over the range 6.485 – 6.495 keV. The number of “samples” (i.e. threads) is shown in parentheses for each state. These values are broken down by beamtime in Table S2. See Table 2 for additional comparisons between the states based on the number of x-ray pulses per state per beamtime and Figures 2C & 3C for dotplots of all first moments used for p-value calculations. F values are given in parentheses and, as each value represents comparison between only two groups, can be taken as F(1,n₁+n₂-2)=F with n_e, n₂ being the number of measurements in the table. Analysis of data set 5 is shown in Fig. S4.

S₂ to S₃ state transition.

For the ensuing S-state transition, S₂ → S₃, previous cryogenic measurements determined that changes to the Mn Kβ emission spectrum are minimal, and a small (−0.02 eV) shift, on the order of our systematic error (0.02 eV, see SI for more details), was reported [35]. A comparison of our 1F and 2F first moments, analyzed with one-way ANOVA indicates a lack of statistical significance over most energy ranges, Table 3 & Figures 2 & S4. Any associated spectral differences are likely too small to reach statistical significance under our experimental conditions. In contrast to both these measurements and earlier XAS studies [8,32,33], Zaharieva et al. recently observed a shift in the room temperature emission spectrum equivalent to that observed during the S₁ → S₂ transition [17]. Although the proposed Mn oxidation to form Mn₄IV in the S₃ state [39–43] Figure 1A, at its most basic, suggests a comparable XES shift for S₁ → S₂ and S₂ → S₃ transitions, we now know that the OEC undergoes a major structural change during the S₂ → S₃ transition from both extended x-ray fine structure (EXAFS) [36], femtosecond (fs) x-ray crystallography [9–11], and electron paramagnetic resonance (EPR) [42]. Studies on model compounds indicate that changes to the local spin densities associated with structural rearrangements, such as changes to the ligand environment, could obscure a Mn-centered oxidative shift [29,44,45]. Minimal changes to the emission spectra, observed upon the S₂ → S₃ transition in previous studies, were attributed to ligand-centered oxidation [35]. Low S₁ to S₂ state conversion following the first laser flash could produce equivalent changes in the S₁ to S₂ and S₂ to S₃ transitions observed by Zaharieva et al. [16], however, the origin of such discrepancies between reports has yet to be elucidated.

S₀ forms after O₂ evolution.

To probe samples enriched with S₀, XES spectra were collected after a 40 ms delay following a third laser flash, (3F_40ms). The first moment of 3F_40ms is consistently shifted to higher energies (Figure 3C, S3 & S4), supporting the expected reduction of Mn. Overall, the RT TR-XES results are in good agreement with previous Mn Kβ emission data [17,35]. Having validated the experimental technique, we investigated the elusive transient S₃ to S₀ process also initiated by three laser flashes (3F).

O-O bond formation step.

Figure 3 depicts the earliest evolution of the OEC electronic structure following three laser flashes and the associated first moment changes of Mn Kβ_1,3 measured at Δt ~50 μs (3F_50μs) and Δt ~200 μs (3F_200μs). The trend of increasing first moment is robust and statistically significant, Table 3 & Figures 3 & S4. It is important to note that this statistically significant change occurs after S₂ to S₃ transition, for which our analysis did not detect a statistically significant change in the spectra, Table 3 & Figure 2. Furthermore, the observed increase in the first moment cannot be explained by mixing of S-states due to poor protein advancement, as this would produce an oxidative (decrease in the first moment) trend or no change, see the SI for details regarding laser excitation. Based on the results of our statistical analysis, there is less than a 5% chance that this trend of increasing first moment is merely noise. It is extremely unlikely that we would repeatedly observe p-values less than 0.05 if we were prevented from detecting small spectral changes due to inherent limitations of our spectroscopic setup. Random reassignment of S-state labels, representative of datasets limited by spectrometer resolution, for example, yields p-values less than 0.05 in only 5% of the random assignments, which is the expected false positive rate. We are therefore confident that these trends are not resolution limited. To further probe the robustness of the shifts, we repeated our analysis on randomly divided halves of the data and repeated our ANOVA analysis. Recovery of statistical significance after this procedure demonstrates that there is, in fact, a (5%)2 = 0.25% chance that the reported effect is due to noise.

This statistically significant spectral shift suggests that during the S₃ to S₀ transition the OEC undergoes a significant transformation at short timescales. Changes observed at 50 μs and 200 μs are likely due to a short-lived isoform of the S₃ state (S₃OO), in which the O-O bond has been formed and Mn centers have been reduced from the (MnIV)₄ state [46]. The accumulation of this species is controlled by its rate of formation (k₁), and rate of oxidation due to electron transfer to Tyr_Z• (k₂), which produces a one electron more oxidized S₄OO state, Figure 1E. The first moment of the Mn Kβ_1,3 line is commonly correlated with the nominal spin value of the Mn center, Figure 4. We observe that the nominal spin of 3F_50μs and 3F_200μs deviates from the all MnIV assignment reported for S₃ and represented on our plot by MnO₂ oxide as well as two model compounds containing a single MnIV center. Intriguingly, the shift between 2F and 3F_200μs is greater than that observed between 2F and 3F_500μs, for which statistical significance was only visible on some plots (Figure 3) but was not strictly demonstrated in the 6.485–6.495 keV range. It is currently unclear whether this effect is rooted in statistical uncertainty, or is simply a weak spectral change due to transient oxidation of the OEC by electron transfer to Tyr_Z• (k₂) Figure 1E. Transient oxidation would likely temporarily pause or reverse the reductive (increasing) trend in first moments.

Figure 4. Analysis of Mn Kβ first moments.

Placement of fully processed 2F/3F data along a linear fit to a series of Mn oxide first moments empirically predicts the average spin state of Mn centers in the OEC. For reference, relevant Mn coordination complexes are placed along the line by using the nominal spins reported by Davis et al. and Jensen et al. [29,44]. Note that S_4a and S_4b correspond to 3F states with Δt between the final laser flash and x-ray pulse of 50 μs and 200 μs, respectively. Compounds 1–3 are formally mixed valence MnIII/MnIV complexes. 1 is a di-μ-oxo dimer, [Mn₂O₂L′₄](ClO₄)₃, while 2 and 3 are two examples from the Mn cubane family, Mn₄O₄L₆. Compounds 4 and 5, by contrast, are mononuclear Mn complexes [MnIV(OH)₂(Me₂EBC)]2+ and [MnIV(O)(OH)- (Me₂EBC)]+, respectively.

While density functional theory (DFT) modeling has proven inconclusive regarding the formation of a MnV=O species [22,24,47,48], we were unable to observe oxidation of the OEC, which would likely be associated with lower values of the first moment, at any observation time point following the third flash and measured over multiple beamtimes. This lack of evidence for oxidation past 2F (majority state S₃) was consistently observed (Figures 3, S3 & S4) thereby excluding formation of a MnVMn₃IV state kinetic intermediate. Current DFT models [21–24] also do not resolve the previously identified kinetic challenge [20]. UV-Vis difference spectra show that Tyr_Z• is reduced quite slowly, ~1 ms after the third flash. Given that this time constant is comparable to the rate of O₂ evolution, only a very short ~50 μs time window remains for all bond formation dynamics and product/substrate exchange to occur.

Discussion

The TR-XES results detailed above cannot be explained by most DFT mechanistic models. Those which propose a Mn₄IV−O• (oxyl radical) in place of MnV=O, for example, predict significant activation barriers for O−O bond formation [21,25], in disagreement with spectroscopic results that show early onset reduction in both XES and XAS. Our results are, however, in agreement with the only other published TR studies probing the electronic structure evolution of the Mn centers via x-ray absorption spectroscopy (XAS) [13,16,49]. TR-XAS detected no oxidation and instead, suggests the gradual (milliseconds) reduction of the OEC initiated 250 μs into the S₃ to S₀ transition. In contrast to the XAS study, where only two energy points along the Mn K-edge were analyzed, we collected full spectra with high multiplicity, representing the complete electronic structure of the OEC at Δt ~50 μs and Δt ~200 μs. In addition, XAS and XES probe different electronic transitions. It is therefore possible that some early spectral changes between t = 0–200 μs may have previously escaped detection, or are less pronounced in XAS due to the transient nature of the early intermediate. A more recent XES study performed at the Linac Coherent Light Source indicates no changes to the Mn Kβ spectra 250 μs after the third laser excitation [50]. We attribute this discrepancy to the differences in experimental conditions that will likely be clarified in the future.

To overcome the kinetic challenge, we proposed that the O−O formation step takes place in the S₃ to S₀ transition prior to the reduction of TyrY_Z•, Figure 1E [28,29]. Rapid evolution of the OEC during the S₃ to S₀ transition has long been a primary target of PS II research. Based on UV-Vis difference spectroscopy [15,51] and TR infrared spectroscopy [14] a deprotonation event has been proposed to occur early (0–300 μs) during this transition. Our results do not explicitly exclude a deprotonation event, but necessitate significant changes to the electronic structure of the 3d Mn frontier orbitals to explain the observed spectroscopic effect. The results presented here can be better rationalized if the formation of the (TyrY_Z•)S₃ state, occurring on the order of 100 ns [14], triggers a sequence of events resulting in significant redox or structural changes to the OEC, such as the formation of the O−O bond. The most recent isotope exchange studies show that substrate exchange stops early (0–300 μs) in the transition [19] hinting at such a possibility. The highest activation energies have also been noted in this early time window [52]. Likewise, the only molecularly defined system for water oxidation functioning with a comparative reaction rate, [Ru(bda)(isoq)₂], is hypothesized to work via a radical coupling mechanism, which results in a peroxo intermediate. During the final catalytic step of this artificial system, the peroxo intermediate is further oxidized, and release of O₂ follows [53]. These results suggest that the same catalytic mechanism engenders rapid O₂ evolution in both biological and bio-mimetic systems, while at the same time preventing peroxide release due to the transient presence of the peroxo isoform.

In summary, to analyze the evolution of the photosystem II electronic structure, we observed intermediate states of photosynthetic O₂ production via microsecond resolution time-resolved x-ray emission spectroscopy at a synchrotron source. Consistent with the obtained full spectra is a mechanism involving the O−O bond formation in the S₃ to S₀ transition prior to TyrY_Z• reduction. This mechanism resolves the previously highlighted kinetic problems. Parallel advancements in the development of molecular catalysts for artificial photosynthesis strongly support O-O bond formation prior to the final oxidation step of such peroxo intermediates and provide further support for the mechanistic proposal detailed herein.