Structural evidence for intermediates during O2 formation in photosystem II

In natural photosynthesis, the light-driven splitting of water into electrons, protons and molecular oxygen forms the first step of the solar-to-chemical energy conversion process. The reaction takes place in photosystem II, where the Mn4CaO5 cluster first stores four oxidizing equivalents, the S0 to S4 intermediate states in the Kok cycle, sequentially generated by photochemical charge separations in the reaction center and then catalyzes the O–O bond formation chemistry1–3. Here, we report room temperature snapshots by serial femtosecond X-ray crystallography to provide structural insights into the final reaction step of Kok’s photosynthetic water oxidation cycle, the S3→[S4]→S0 transition where O2 is formed and Kok’s water oxidation clock is reset. Our data reveal a complex sequence of events, which occur over micro- to milliseconds, comprising changes at the Mn4CaO5 cluster, its ligands and water pathways as well as controlled proton release through the hydrogen-bonding network of the Cl1 channel. Importantly, the extra O atom Ox, which was introduced as a bridging ligand between Ca and Mn1 during the S2→S3 transition4–6, disappears or relocates in parallel with Yz reduction starting at approximately 700 μs after the third flash. The onset of O2 evolution, as indicated by the shortening of the Mn1–Mn4 distance, occurs at around 1,200 μs, signifying the presence of a reduced intermediate, possibly a bound peroxide.

In natural photosynthesis, the light-driven splitting of water into electrons, protons and molecular oxygen forms the first step of the solar-to-chemical energy conversion process.The reaction takes place in photosystem II, where the Mn 4 CaO 5 cluster first stores four oxidizing equivalents, the S 0 to S 4 intermediate states in the Kok cycle, sequentially generated by photochemical charge separations in the reaction center and then catalyzes the O-O bond formation chemistry [1][2][3] .Here, we report room temperature snapshots by serial femtosecond X-ray crystallography to provide structural insights into the final reaction step of Kok's photosynthetic water oxidation cycle, the S 3 →[S 4 ]→S 0 transition where O 2 is formed and Kok's water oxidation clock is reset.Our data reveal a complex sequence of events, which occur over micro-to milliseconds, comprising changes at the Mn 4 CaO 5 cluster, its ligands and water pathways as well as controlled proton release through the hydrogen-bonding network of the Cl1 channel.Importantly, the extra O atom O x , which was introduced as a bridging ligand between Ca and Mn1 during the S 2 →S 3 transition [4][5][6] , disappears or relocates in parallel with Y z reduction starting at approximately 700 μs after the third flash.The onset of O 2 evolution, as indicated by the shortening of the Mn1-Mn4 distance, occurs at around 1,200 μs, signifying the presence of a reduced intermediate, possibly a bound peroxide.
Serial femtosecond X-ray crystallography at X-ray free electron lasers (XFELs) 7 enabled us to collect crystallography data of photosystem II (PS II) in real time as the reaction progresses at physiological temperature.The four photon-induced water-oxidation reaction in PS II (Fig. 1a,b) was initiated with multiple visible laser flashes.Using this capability, intermediate S-state structures (S 0 , S 1 , S 2 and S 3 ) have been studied 4,5,[8][9][10] that revealed the structural changes of the oxygen evolving complex (OEC) of PS II, which is a functional unit composed of the Mn 4 CaO 5 cluster and its water-ligand environment (Fig. 1c,d) 2,11 .Recently, we collected snapshot data at several time points during the S 2 →S 3 transition, the step in which one substrate water is introduced into the cluster.The study 6 suggested the sequence of Mn oxidation, incorporation of an extra oxygen bridge (O x or O6 in Suga et al. 10 ) between the open coordination site at Mn1 and Ca (forming Mn 4 CaO 5 -O x in S 3 ), the potential entry path for substrate water and the proton release with its gating mechanism 6,12 .The XFEL studies also clearly established that the electronic and geometric structure of the OEC obtained by these measurements is unaffected by X-ray photoelectrons under the conditions used 6,8,13 .
In the current study, we investigate the oxygen evolving step of Kok's water oxidation cycle (Fig. 1b), the S 3 →[S 4 ]→S 0 transition (Fig. 1c,d).In dark-adapted PS II samples, this transition is initiated by the third visible laser flash.The OEC is oxidized in this step from the all-Mn(IV) S 3 state to the proposed highly reactive S 4 state with formal oxidation states of Mn(IV) 4 O ⦁ or Mn(IV) 3 (V).This initiates O-O bond formation and O 2 release, and the now vacant binding site is filled by a new water substrate forming the lowest oxidation state of the cluster (S 0 ).This multistep process, which also involves the release of two protons, has the longest time constant among the S-state transitions, and its kinetics depend on the species and sample preparation [14][15][16][17][18][19] .
To provide structural insight into this complex reaction step and specifically, the important interplay between the Mn 4 CaO 5 cluster and its protein-water environment (Fig. 1c,d), we collected room temperature crystallography data of PS II at seven different time points during the S 3 →S 0 transition, ranging from 50 μs to 4 ms after initiating this transition (Fig. 1b).All datasets have resolutions between 2.00 and 2.16 Å (Extended Data Tables 1 and 2).
The S 3 state was populated by illumination of dark-adapted PS II microcrystals with two in situ visible nanosecond laser flashes (2F in Fig. 1b) 4 , in which the interval between flashes was 200 ms to account for acceptor quinone Q A and Q B kinetics and efficiently drive S-state transitions.The time points between the S 3 and S 0 states were generated by giving the third visible pump laser flash at various delay times (Δt) before the crystals were exposed to the femtosecond XFEL pulse (the time labels in Fig. 1b).While our illumination protocol achieves the highest possible populations of particular S states, there is a higher mixing of S-state populations with increasing flash number due to intrinsic PS II-specific inefficiencies ('misses') (Methods) 20 .We model this distribution in a multicomponent model during structural refinement, with the 'primary' component being the centers that advance from the S 3 to S 0 state.The 'secondary' and 'tertiary' components are the known starting and end points: for example, the decreasing S 3 population and at longer delay times, the increasing S 0 population from centers that have completed the transition (Methods and Extended Data Table 3 have details).We note that the primary component at each time point may consist of a mixture of multiple structures, which are intermediates between S 3 and S 0 .All results discussed below correspond to these refined primary components from monomer I (chains annotated as uppercase in the deposited structures).

OEC and the Y z region
Figure 2 shows the omit map density of selected atoms at the OEC and its surroundings of the refined population at the time points (Δt = 250, 500, 730, 1,200 and 2,000 μs) after the third flash (we use The relevant channels for water and proton transfer (O1, O4 and Cl1) are indicated as red, blue and green shaded areas, respectively.The dotted circles mark structural differences between the S 3 and S 0 states.the nomenclature of 3F(Δt μs)).To visualize the sequence of events during the S 3 →[S 4 ]→S 0 transition, we follow the changes in three areas: Y z and D1-H190 (Fig. 2a,b) and O x and O5 (Fig. 2c), as well as the water ligands W1-W4 (Fig. 2d) and two carboxylate ligands of the Mn 4 CaO 5 / Mn 4 CaO 5 -O x cluster that bridge between Mn and Ca (Fig. 2e).Additionally, selected atomic distance changes are shown in Fig. 3, including the earlier (50 μs) and later (4,000 μs) time point data.
The distance between Y z and D1-H190 has been established previously to be an indicator of the oxidation state of Y z 6 .In the reduced state, a strong hydrogen bond between Y z and D1-H190 leads to a short distance of about 2.6 Å. Upon oxidation of Y z by P680 + , the phenolic proton of Y z is transferred to D1-H190, the distance increases to 2.8 Å and a movement of the His ring plane is observed (Fig. 2b).Our data show that the distance already increases between the S 3 data and the first time point (50 μs), indicating that Y z is fully oxidized by this time (Fig. 3).Thereafter, this distance remains constant until 500 μs and returns to the base level between 730 and 1,200 μs.Thus, the data demonstrate that Y z ox reduction by the Mn 4 CaO 5 -O x cluster starts only after around 500 μs (Extended Data Table 4) 16,21,22 and appears to be complete by the 1,200-μs time point.Additional distance changes between Y z and His190 are observed at 2,000 and 4,000 μs, which may be due to the rearrangement of the hydrogen bonding network related to the last proton release but are not well understood currently.Interestingly, a shift in the position of a Tyr residue next to the special pair Chl upon light excitation was previously noted in time-resolved crystallography data of the purple bacterial reaction center and interpreted as originating from a change in the hydrogen-bonding interactions of this Tyr upon deprotonation 23 .The extra oxygen O x , present in the S 3 state, is lost upon S 0 formation, indicating that O x may participate in the O-O bond formation, and changes in its density probably inform on the onset of this process.For tracing the O x population, O x was eliminated from the OEC model, and its omit map density is shown in Fig. 2c.O5 was separately omitted, allowing comparison of their densities (see also Extended Data Fig. 1).The O x omit density becomes asymmetric starting at 250 μs, with a clear reduction in intensity after 500 μs and dropping to the noise level between 1,200 and 2,000 μs, and O x can only be modeled with a population of less than 20% in the 2,000 μs time point.The O x intensity Additional markers for the presence of O x in the cluster are Mn-Mn distances.As reported previously, the Mn1-Mn4 distance increases during the S 2 →S 3 transition due to the insertion of O x (ref.6). Figure 3 shows that the Mn1-Mn4 distance remains elongated in the 3F structures until 1,200 μs (Methods has a more detailed analysis) and then declines over the next 3 ms to attain the same value as seen in the S 1 and S 2 states.A similar trend is also seen in the Mn1-Mn3 distance.Thus, there is a delay between the onset of O-O bond formation (500-730 μs based on the changes observed for Y z and O x ) and the time when the Mn-Mn distances start to decrease (1,200 μs).The finding indicates that during this period, a water oxidation intermediate likely exists before the release of O 2 .
Changes in shape and intensities are observed for the omit map density of the water ligands W1-W4, which are displayed in Fig. 2d.Specifically, a slight elongation of the O5 density toward W2 and a slight elongation of the W4 density toward W3 are observed at 250 μs.This latter trend continues, and at 500 μs, an overlap of the W3 and W4 densities is seen, with an mF obs − DF calc peak (F obs and F calc are the experimental and model structure factors respectively, while m and D are weighting factors) at 2.5σ between W3 and W4 (Extended Data Fig. 2).At the same time point, the W1 density becomes extended toward D1-D61 and W19, indicating a higher mobility of W1.We speculate that all these motions are related to the deprotonation of the OEC and proton transfer toward the Cl1 channel (see the next section and Fig. 4).
At 730 μs, the O5, W2 and W3 densities become anisotropic, all pointing toward a region between these three oxygen atoms (Fig. 2d), which indicates increased mobility of these ligands.At the same time, the O x density is also highly anisotropic.This movement of all four oxygen atoms is likely related to the formation of the water oxidation intermediate.We note that accurate modeling of the O x position will require higher-resolution data.At 1,200 μs, the densities for all the terminal water ligands (W1-W4) and the bridging O5 become weakest.A comparison of these omit map densities within the OEC with that of the O2 atom, which is believed to not play a prominent role in the S 3 →S 0 step, shows that the reduction of the density is specific to these five atoms (W1-W4 and O5) and O x (Extended Data Fig. 1).The overlap of the O5, W2 and W3 densities is no longer observed at 1,200 μs.
At 2,000 μs after the third flash, the O5 omit map density is restored considerably compared with the S 3 and S 0 states, but omit map densities of the waters W1-W4 have not yet reached a similar level and remain elongated.We note a similar overlap of W1/W2 as observed at 500 μs.This could indicate the onset of the second proton release, known to occur in the S 3 →S 0 transition, after the binding of a water that refills the vacant site formed by O 2 release 21,24,25 .Interestingly, the elongated shape of the W3 density persists even in the S 0 state, which we modeled previously with two possible positions of W3 (ref.4).
Figure 2e shows that the D1-D170 and D1-E189 ligands, which both bridge between an Mn and Ca, change their conformation during O 2 formation and release.Consistent with the high mobility observed for the W1 and W2 water ligands, the connection between D1-D170 and Ca appears to be weakened between 730 and 2,000 μs and is only fully restored at the S 0 state (3F(200 ms)).
From 1,200 to 4,000 μs, several structural changes occur, which are reversed upon formation of the stable S 0 state (3F(200 ms)).These include the increase of the Y z -D1-H190, Ca-D1-E189 and Mn4-O5 distances, as well as a decrease of the Mn1-Mn4 and Mn1-Mn3 distances.Most of these changes are indicative of O 2 release and/or water insertion via the Ca ion 26,27 , possibly from the O1 channel (see below).This indicates that O 2 release and refilling of the cluster by bulk water and resetting of the catalytic center occur over an extended timescale.

Water and proton channels
PS II has several hydrophilic channels that extend from the OEC to the lumenal side of the thylakoid membrane 11,[28][29][30][31] , and some of these are proposed to play a critical role in transporting protons and substrate waters during the catalytic cycle (O1, O4 and Cl1 channels are shown in Fig. 1) 24,32,33 .In our recent study of the S 2 →S 3 transition 12 , we assigned the O1 channel, which extends from the O1 and Ca of the OEC to the bulk, to be a substrate water channel and the Cl1 channel, which extends from W1 and W2 of the OEC to the bulk, to be a proton release channel during the S 2 →S 3 transition 6 .The S 3 →S 0 transition also involves the insertion of one substrate water into the OEC and the release of two protons to the bulk; it has been suggested that one proton is released before the O-O bond formation and the other after the rebinding of a water molecule to the OEC 24,25,34 .
Figure 4 shows the time point data for the O1 and the Cl1/O4 channels near the OEC during the S 3 →S 0 transition.The electron density of water molecules in the O1 channel in the vicinity of the OEC changes substantially as shown in the 2mF obs − DF calc maps (Fig. 4a), similar to what was observed during the S 2 →S 3 transition 6 .Waters W27, W28 and W32 (Supplementary Table 1 has water numbering) have low electron density and high B factors (about 50 Å 2 ) (Extended Data Fig. 3) relative to the more stable waters, such as W29 (B factor of about 37 Å 2 ), in particular at 1,200 μs.We interpret this as an indication of the high mobility of these waters and hypothesize that this region could serve as the inlet for the substrate water that refills the OEC after the release of molecular oxygen 12 .We, therefore, propose that PS II uses the O1 channel for the substrate intake in both the S 2 →S 3 and S 3 →S 0 transitions.Among the group of five waters (W26-30) ('water wheel' in ref. 6), W26 shows high electron density throughout the transition, with a substantially elevated density at 500 μs.As W26 is within hydrogen-bonding distance to O1 of the OEC, this interaction may be important for balancing the charge on the cluster, when the OEC advances through the last oxidation step (that is, S 4 state formation) and the subsequent four-electron reduction to form the S 0 state.
Changes are also observed in the Cl1 channel.At 250 μs, the D1-E65 residue rotates by 19° toward W40, resulting in shortening the distance between D1-E65 and W40 by 0.3 Å (Fig. 4b).This is the time point when the O x density starts to become asymmetric, which is even more pronounced at 500 μs (Fig. 2c).At 500 μs, the W1 omit map density also becomes elongated toward the region of D1-D61 and W19 (Fig. 4b).This coincides with a decrease of the D1-D61 carboxylate oxygen density that is within hydrogen bond distance of W40 and a shortening of the distance between W40 and D1-E65 by 0.5 Å due to rotation (25°) of the side chain.Consequently, a continuous hydrogen bond network is formed that connects the OEC to the D1-E65/D2-E312 region.We speculate that the changes are related to the first proton transfer from the OEC toward the Cl1 channel.These changes are reversed by the 730-μs time point (Fig. 4b).An early deprotonation event has also been suggested by other studies using different methods, with time constants that range from 50 to 300 μs (that is, before the last oxidation event; the transient S 4 state formation) [15][16][17]21,24,34,35 .
At 1,200 μs, the D1-E65 residue rotates toward W40 for the second time during this transition, accompanied by a shortening of the W42-D1-E65 distance by 0.4 Å.Thus, a hydrogen bond network from the OEC to the D1-E65/D2-E312 region, similar to what we observed at 500 μs, is reformed at this time point.The D1-E65/D2-E312 distance elongates from roughly 2.6 to roughly 3.2 Å, which points to a substantial weakening of the interaction, likely forming a configuration that can accept the next proton.This change coincides with the time at which the O x electron density decreases below the detection level in the omit map (Fig. 2), and the W28, W27 and W32 densities decrease in the O1 channel (Fig. 4a).We interpret this series of changes to be related to the onset of the recovery process of the Mn 4 CaO 5 cluster (that is, initiated by the insertion of water into the OEC along with a deprotonation).At 2,000 and 4,000 μs, the side chain of D1-E65 is rotated almost 40° from its position in the S 3 structure toward W119 (Fig. 4b).The distance D1-E65-W119 is around 2.5 Å, suggesting a shared proton or very tight interaction between these two groups.D1-R334 also moves by 20°, forming a hydrogen-bonding interaction with W41.
The changes around the D1-E65/D2-E312 region are indicative of the proton release to the bulk since they are reminiscent of what we observed for proton release during the S 2 →S 3 transition 12 .In the S 3 →S 0 transition, we hypothesize that the earlier changes of this region (250-730 μs) are related to the first proton transfer and that the later changes (1,200-4,000 μs) are related to the second proton transfer from the OEC to the bulk, through D1-D61 36,37 via the rotation of D1-E65.Thus, the current result suggests that the D1-E65/D2-E312 region functions as a gate for proton release twice during the S 3 →S 0 transition.

Article
W20, which forms a tight hydrogen bonding network with O4 via W19 in the O4 channel, disappears during the S 1 →S 2 transition and reappears in the S 0 state 4 .Its return during the S 3 →S 0 transition is therefore an indicator for the full recovery of the S 0 state.The first clear indication of the return of W20 is found in the F obs (4,000 μs) − F obs (2F) difference map (orange density in Fig. 4b), and it was modeled at 40% occupancy in the 4,000-μs refined component.This implies that the W20 restoration happens in the later stage of the OEC recovery.

Sequence of events during S 3 →S 0
The snapshots of the structures of PS II during the S 3 →S 0 transition show the sequence and the progression of each of the events at multiple locations with different time constants.The structural changes can be broadly grouped into four sections with different onset times and kinetics, as shown in Fig. 5; they are the redox state changes of Y z , first deprotonation, OEC oxidation and O 2 formation and the complete recovery and resetting of the Kok clock in the S 0 state.
In the S 3 state, all four Mn are formally in the (+IV) oxidation state.We note that oxidation of ligand instead of Mn (formation of oxyl or an oxo-oxyl bond) during the S 2 →S 3 transition has been suggested in the literature 5 , but this is not in line with the room temperature X-ray emission spectroscopy (XES) data, which show the oxidation of Mn 6,13 .Upon the third flash, the oxidation of Y z occurs by donation of an electron to P680 + after the charge separation at the reaction center chlorophylls.A distance increase between Y z and D1-H190 is observed in comparison with the S 3 state in 3F(50 μs) (Figs. 3 and 5(i)).This change is assigned to the Y z ox formation, which is known to occur within 30 μs after photoexcitation 38 , and related proton translocation between Y z ox and D1-H190.
The Y z ox formation triggers the first deprotonation event likely during the time period of 200-500 μs [15][16][17]24,34,39 . We obsered the start of the rotation of D1-E65, which is proposed to be part of the proton gate, and the formation of the hydrogen bond pathway from the OEC to this region at 250 μs.Along with these changes, the electron densities at W1 and D61 become more prominent at 500 μs (Figs.2d and 4b), potentially related to a proton release from the OEC to the proton gate residues D1-E65 and D2-E312 (Fig. 5(ii)).
In the early stage of the 500-to 1,200-μs period, the last oxidation event (transient S 4 state formation, with Mn(IV) 4 O ⦁ or Mn(IV) 3 (V)) occurs, and subsequently, the reduction of Mn takes place (Fig. 5(iii)).The O-O bond formation should be triggered by this final oxidation event of the OEC to the potentially short-lived S 4 state.The change in distance we observe for Y z -D1-H190 between 500 and 730 μs suggests that the reduction of Y z ox takes place during this time, through the electron transfer from the OEC to Y z .After the transient formation of S 4 , the four-electron reduction may proceed in one step with the O-O bond formation and immediate release of O 2 or in two steps with the presence of an intermediate before the release of molecular oxygen from the OEC.In the latter case, a peroxo species formed by an initial two-electron reduction appears most likely as an intermediate.Our data show that there is a delay between the onset of O-O bond formation (500-730 μs) as indicated by the Y z -D1-H190 distance/rotation and the decrease of the O x electron density and the onset of the O 2 release supported by the Mn1-Mn4 distance contraction (1,200 μs).This onset time for O 2 release is also in line with studies of O 2 evolution 22,40 .The delay indicates that there is an intermediate state, possibly a peroxide-like species, pointing toward the two-step electron reduction mechanism.
Several O-O bond formation sites have been proposed in the literature based on theoretical studies (Fig. 5) 3,[41][42][43][44][45][46][47][48][49] .Among these, O5-O x best account for our data because of their proximity and the reduced occupancy of O5 around 1,200 μs (Fig. 5, model a).However, two other possibilities in which O5 reacts with either W2 or W3 and O x replaces O5 cannot be excluded at this time (Fig. 5, model b).While other mechanisms that do not involve O5 cannot be ruled out, there is no clear evidence to support those in the current data.At 1,200 μs, the O x omit map density is below the 2.5σ threshold, indicating that a predominant fraction of O x has shifted from its original position in the cluster.The contraction of the Mn1-Mn4 and Mn1-Mn3 distances starting at this time point suggests that the onset of O 2 release happens around this time.Once O 2 is released, refilling of the cluster with a new substrate water seems to occur immediately.This is based on the observation that there is no missing oxygen density besides O x , although the omit map densities of all the terminal waters (W1-W4) and bridging O5 are weakened at 1,200 μs.The data support that the O 2 release and refilling of the site are highly coordinated and occur likely via a terminal water already ligated to the OEC.
At both 2,000 and 4,000 μs (Fig. 5(iv)), O x density is within the noise level, which implies that Mn1 becomes predominantly five coordinate.Mn4 is six coordinate, although the Mn4-O5 interaction is weak (2.2-2.3Å), suggesting that O5 may be a hydroxide.The Ca-D1-E189 distance is still more elongated at these time points than in the S 0 state.Other slow recoveries are observed in the Y z region (Fig. 2b), the 'water wheel' (Fig. 4a) and the proton gate regions (D1-E65/D2-E312) (Fig. 4b).We hypothesize that a water from the 'water wheel' region in the O1 channel, similar to the S 2 →S 3 transition 6,12 , replaces the terminal water ligand of the OEC.The changes in the proton gate region might indicate the deprotonation of the newly inserted water 36 .Concomitantly, the amino acid coordination environment, the hydrogen bonding network around the OEC and the waters in the channels reset to the S 0 state.This includes recovery of W20 in the O4 channel, which is proposed to be involved in proton release during the S 0 →S 1 transition 4,50 .
In the current study, room temperature snapshots of PS II structures through the final step of Kok's clock (S 3 →[S 4 ]→S 0 ) reveal details of the molecular processes for photosynthetic water oxidation.Until now, these processes were interpreted largely based on kinetic studies.Importantly, the results reported here provide experimental support for a two-step reduction mechanism of the Mn 4 CaO 5 -O x cluster upon the O-O bond formation and O 2 release with a transient intermediate, most likely a bound peroxide.This is a major step forward toward understanding the chemistry of the water oxidation reaction.The results also show how biological catalysts, such as the OEC in PS II, enable multielectron/multiproton reactions through the interplay between the metal center, the protein environment and the water network.The active role of the microenvironment in natural enzymes provides inspiration for how to control such reactions in artificial photosynthetic systems that can be made from earth-abundant elements.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-023-06038-z.

Sample injection and illumination
Acoustic droplet ejection 54 was used in combination with the Drop-on-Tape sample delivery method 55 .For capturing the stable intermediates S 2 , S 3 and S 0 , each droplet of the crystal suspension was illuminated by 120-ns laser pulses at 527 nm using an Nd:YLF (yttrium lithium fluoride) laser (Evolution, Coherent) at Linac Coherent Light Source (LCLS) or by 8-ns laser pulses at 532 nm using a combination of two Nd:YAG (yttrium aluminium garnet) lasers (Minilite, Continuum) at Spring-8 Angstrom Compact free electron Laser (SACLA) via three fiber-coupled outputs with a delay time of 200 ms between each illumination and of 200 ms between the last illumination and the X-ray probe, similar to what was used previously to accommodate the acceptor quinone Q A and Q B kinetics and efficiently drive S-state transitions 4,6,55 .We implemented a feedback control system of the belt speed and deposition delay, and the flashing delay and droplet phase were adjusted accordingly 55 .To achieve time delays shorter than 200 ms between illumination and the X-ray probe, a fourth 'free space' laser was utilized.This was either an Opolette 355 LD laser (Opotek, 530-nm wavelength, 7-ns pulse width) at the macromolecular femtosecond crystallography/LCLS instrument or an NT230 OPO laser system (530-nm wavelength, 5-ns pulse width, EKSPLA Co.) at SACLA.This free space laser was triggered to be synchronized with the X-ray pulse with an adjustable delay that was set between 50 and 4,000 μs for this study.The laser was guided with optics to the X-ray interaction spot, and its position was fine-tuned for each delay time to ensure that the laser spot position coincides with the position of the sample droplet at the selected delay timing.At the XFELs, a light intensity of 120 ± 10 mJ per cm 2 was applied as O 2 evolution was found to be saturated at 70 mJ per cm 2 for the dimensions and concentrations of samples used in our experiments 4 .A light intensity of 120 mJ per cm 2 corresponds to about 140 photons absorbed per PS II monomer in the front 5-μm layer of the crystal and approximately 9 photons per PS II monomer when assuming a 60-μm thickness of the crystal (which is the upper size limit of the crystals used in this study) for the back 5-μm layer.This photon density ensures saturation over the entire crystal volume, even in the case of two crystals stacked on top of each other in the laser beam.Given a minimum pulse length of 5 ns and 35 Chl per PS II monomer, the light intensity used averages to 0.8 photons per (Chl and nanosecond) for the front and 0.05 photons per (Chl and nanosecond) for the back part of the crystal.If a PS II center is undergoing charge separation, additional photons absorbed by the internal antenna Chl are rapidly dissipated in the form of fluorescence with an average fluorescence lifetime of around 0.5-1 ns, hence preventing any overexcitation of the reaction center or causing any heating artefacts.

X-ray data collection
The crystallography data were collected at various facilities, and details are listed in Supplementary Table 2.The experimental beam conditions and detector configurations used to collect each dataset are also tabulated.The sample was delivered into the X-ray interaction region using the previously described Drop-on-Tape setup 55 .Illumination conditions for populating different S states are detailed in ref. 4.

X-ray diffraction data processing
The data collected for the different illumination states were processed using the program dials.stills_process with a target unit cell of a = 117.0Å, b = 221.0Å, c = 309.0Å, α = β = γ = 90° and the space group P2 1 2 1 2 1 .Bragg spots were integrated to the edge of the detector.A Kapton absorption correction due to the conveyor belt of our sample delivery system was applied to each integrated Bragg spot, taking into account the droplet size, tape thickness, tape angle and the position of the diffraction spots on the detector with respect to the crystal position.Before integration, we also performed ensemble refinement of the crystal and detector parameters using the program cctbx.xfel.stripe_experiment,which has been shown to narrow the unit cell distribution and improve the final isomorphous difference maps 56 .Finally, the intensities were merged using the program cctbx.xfel.merge,which applies a per-image resolution cutoff and filtering of the lattices using a unit cell threshold of 1% from the reference model.To merge the reflections, we use the best practices described in ref. 57.The unit cells and number of lattices merged for each dataset are tabulated in Extended Data Tables 1 and 2.

Model building and map calculation
Each dataset was refined using a high-resolution PS II structure (1.89 Å) that was published in a previous work (PDB ID code 7RF1) 12 as the starting point using the program phenix.refine 58.The refinement is done in several stages.First, the B factors of the starting model are set to 30, and all waters and the atoms of the OEC are removed.An initial rigid body refinement coupled with refinement of xyz coordinates and isotropic B factors was done for 15 cycles to adjust the model into the unit cell.Next, the OEC atoms are added back and refined with custom bonding restraints for several cycles.We also use custom bonding restraints for chlorophyll-a (to allow correct placement of the Mg relative to the plane of the porphyrin ring) and unknown lipid-like ligands (steric acid) in the refinement.After initial refinement of the OEC + protein complex, waters were added to the model using the phenix.refinewater picking protocol as well as manual placement of waters via coot 59 and doing multiple cycles of refinement.
At this stage, we split the model in the vicinity of the OEC and the OEC itself (only protein and OEC atoms) into multiple components (Extended Data Table 3).The split was done only in parts of chains A/a, C/c and D/d.The rationale and population of the components in each time point used are described in the section Estimating population distribution in each time point.In each dataset, the primary conformer (defined as the intermediate that is advancing from S 3 to S 0 ) is refined using a strategy of reciprocal xyz + isotropic B-factor refinement.For the secondary/ tertiary components (whose structures are known as they are either in the S 3 or S 0 state), only the group B factors are adjusted (group_adp strategy in phenix.refine) to adjust them to the resolution of the dataset.For the remaining part of the model that is not split, regular reciprocal xyz refinement and isotropic B-factor refinement are performed in tandem for multiple cycles.All waters (except for the terminal waters ligated to the OEC; that is, W1-W4) were refined as a single component.
The refinement of the OEC in the primary component of the multicomponent model was done using custom restraints that were used to model the S 3 state.However, for all the time points, we used slightly looser estimated s.d.values for the restraints (0.1 Å for bonds, 10° for angles) to allow the OEC atoms during refinement to move toward where the electron density is optimally modeled and reduce strain in the refinement while at the same time maintaining the overall shape of the cluster.The OEC in the S 0 state was modeled with restraints used for our previously published S 0 state structure.The restraints used to model the OEC atoms in the time points have been provided as text files (schemes 1-3 in Supplementary Data).

Fig. 1 |
Fig. 1 | An overview of PS II and the electron donor site where water oxidation takes place.a, The structure of PS II with the membrane-embedded helices and the membrane extrinsic regions on the lumenal side of PS II shown in gray.The main electron transfer components are shown in colour, which include the reaction center chlorophylls (P680), pheophytins, acceptor quinones Q A and Q B , redox-active tyrosine Y z and the catalytic Mn 4 CaO 5 cluster.The Y z and Mn 4 CaO 5 cluster are the cofactors of the electron donor site.b, Kok cycle of the water oxidation reaction taking place at the donor site that is sequentially driven by charge separations in the reaction center P680 induced by the absorption of photons (nanosecond light flashes, 1F-4F) in the antenna system of PS II.Room temperature X-ray crystallography data were collected at the time points indicated during the S 3 →S 0 transition.c,d, The structure of the OEC in the S 3 (c) and S 0 (d) states and the sequence of events occurring between them.Mn, purple; Ca 2+ , green; O, red.W1, -2, -3 and -4 are water ligands of Mn4 and Ca.The relevant channels for water and proton transfer (O1, O4 and Cl1) are indicated as red, blue and green shaded areas, respectively.The dotted circles mark structural differences between the S 3 and S 0 states.

Fig. 2 |
Fig. 2 | mF obs − DF calc electron density omit map of key components of the redox active donor site of PS II at five time points along the S 3 →S 0 transition as well as the S 3 and S 0 states.a, Residues D1-Y161 (Y z ) and D1-H190.The omit map from the S 3 state reference is shown in light brown for comparison with the time point data (blue).b, A simplified representation of the structural changes observed at the Y z region.c, Omit density of atoms O5 and O x of the OEC.d, Omit density of atoms O5 and the terminal water ligands W1, W2, W3 and W4 of the OEC.e, Omit density of carboxylate oxygen atoms of D1-E189 and D1-D170.All omit maps shown in a and c-e were generated by omitting the atom or residue of interest individually, and only the primary component (that is, the state that is advancing to S 0 ) was used.Notable features are highlighted with red arrows and black dashed circle.All omit maps shown are contoured at 2.5σ, 3σ and 4σ using the colour scheme annotated in d for easier visualization.See also Supplementary Information Video 1.

Fig. 3 |
Fig.3| Distance changes between selected atoms/residues in the OEC during the S 3 →S 0 transition.All distances are taken from the refined component of each time point (that is, the state that is advancing to the S 0 state).Error bars are calculated from the end/rapid approach described in Methods and are an upper limit.Data here are shown as mean values ± standard deviation.The error bars for each time point were obtained from n = 100 independent END/RAPID refinements.More details about the END/RAPID procedure can be found in Methods.Dashed arrows in the schematics of the OEC on the right indicate the location of the individual distances.Mn is shown as purple spheres, and O is shown as red spheres.

Fig. 4 |
Fig. 4 | Structural changes in select regions of the water and proton channels of PS II during the S 3 →S 0 transition.a, The terminus of the O1 channel near the OEC that includes the group of five waters (W26-W30) in this region.Overlaid is the 2mF obs − DF calc electron density map contoured at 0.8σ, 1.0σ and 1.5σ.b, The O4 and Cl1 channels (branch A) that include the D1-D61 and D1-E65/D2-E312 region that is suggested to function as a proton gate.Overlaid are mF obs − DF calc omit maps for W1, W2 and D61 shown at 2.5σ, 3.0σ and 4.0σ.Also shown is the F obs (time point) − F obs (2F) difference density map within a 1.5-Å radius of W19/ W20/W48 in the O4 channel at 3σ (orange map).The observed rotation angle at the side chain of E65 at particular time points is calculated with respect to the corresponding side chain position at the 2F state.Major changes are highlighted with an arrow or dashed circle.All waters are coloured by their B factors according to the diverging colour scheme shown in the figure.Important hydrogen-bond interactions are shown with a binary colour scheme to indicate strength (distance < 2.8 Å is red and 2.8-3.2Å is gray).

Fig. 5 |
Fig. 5 | Schematic of the S 3 →S 0 transition and proposed mechanism for O-O formation.The sequence of events (i-iv) leading to the first deprotonation event, the molecular oxygen release, the water insertion and the second deprotonation event.The OEC atoms are shown in purple (Mn), green (Ca) and red (O).The O1 channel is shown in red, the O4 channel is in blue and the Cl1 channel in green.The ligands of the OEC and the residues forming the water-proton channels are coloured based on the subunit they belong to (D1, blue; D2, green).Possible pathways for proton (cyan arrow), water (red dashed arrow), oxygen (red solid arrow) and electron (green arrow) transfer are depicted.Notable features are highlighted with black arrows.The right tan box shows the suggested models (model a and model b) for O-O bond formation.Oxygen highlighted with magenta indicates the candidate atoms for O-O formation.

Extended Data Fig. 1 |
mF obs − DF calc electron density omit map of relevant atoms in the OEC for time points along the S 3 →S 0 reaction.For comparison of the peak height level, the omit map of the oxygen atom, O2, is also shown.Omit maps for each atom were generated by individually omitting the atom only in the primary conformer in each dataset.The maps are shown at contour levels of 2.5, 3 and 4σ.(a) Omit maps of O X , O5 and O2 (b) Omit maps of W1, W2, W3, W4 and O2.A clear reduction in omit map peak height of O X , O5, W1, W2, W3, W4 is observed with respect to the reference O2 omit map between 730-2000 μs.Extended Data Fig. 2 | mF obs − DF calc electron density map near the OEC region in the 3F(500 µs) time point.The map is shown at a contour level of 2.5σ(in yellow).Electron density is observed at this level in-between W3 and W4 (see red arrow), possibly indicating a transient water motion that was also corroborated by the overlap of the omit map densities.For comparison, the individual omit maps of W3 and W4 are also overlaid at 2.5, 3 and 4σ (color scheme same as in Fig.2in main text).Stronger electron density is also observed around W1/D61 possibly related to motion involving a proton transfer as discussed in the main text related to Fig.2.Electron density around the O1 is possibly related to changes observed in the W26-W30 (referred to as the 'water wheel' region) in this time point (discussed in main text related to Fig.4).Extended DataFig. 3 | Comparing B-factors of key waters in the water wheel region of the O1 channel in the S 3 →S 0 transition (W27, W28, W32) against a reference water (W29) for time points shown in Fig. 4 in the main text.The B-factor distribution of all the channel waters are overlaid for each time point for comparison.The y-axis values of the marker points for the 4 waters are just placeholders.There is a stark shift in the relative B-factor values of W27, W28 and W32 in the 3F(1200 μs) time point which coincides with the disappearance of O X .Extended Data Table 1 | Data collection, merging and refinement statistics (2F, 2F-alternate, 3F(50 µs), 3F(250 µs), 3F(500 µs)) Note that all reference to the 2F dataset or the S 3 model in the paper uses the 2.09 Å model.The 2F-alternate model is only provided here for the isomorphous difference maps shown in Fig. 4b.Extended Data Table 2 | Data collection, merging and refinement statistics (3F(730 µs), 3F(1200 µs), 3F(2000 µs), 3F(4000 µs), 3F(200 ms)) Extended Data Table 3 | The estimated population distribution and identity of each of the conformers used in the multi-component modeling of the datasets Refer to the discussion Estimating population distribution in each timepoint in Methods for more details regarding the conformers.Extended Data Table 4 | Source data for the plots shown in Fig. 3 of the main text