Structure-based mechanism of photosynthetic water oxidation

The recently-published 3.5 Å resolution X-ray crystal structure of a cyanobacterial photosystem II (PDB entry 1S5L) provides a detailed architecture of the oxygen-evolving complex (OEC) and the surrounding amino-acids [K. N. Ferreira, T. M. Iverson, K. Maghlaoui, J. Barber and S. Iwata, Science, 2004, 203, 1831–1838].

The revealed geometry of the OEC lends weight to certain hypothesized mechanisms for water-splitting, including the one propounded by this group, in which a calcium-bound water acts as a nucleophile to attack the oxygen of a MnVO group in the crucial O–O bond-forming step [J. S. Vrettos, J. Limburg and G. W. Brudvig, Biochim. Biophys. Acta, 2001, 1503, 229–245].

Here we re-examine this mechanism in the light of the new crystallographic information and make detailed suggestions concerning the mechanistic functions (especially the redox and proton-transfer roles) of calcium, chloride and certain amino-acid residues in and around the OEC.

In particular, we propose an important role for an arginine residue, CP43–Arg357, in abstracting protons from a substrate water molecule during the water-splitting reaction.


Photosystem II (PSII) is a dimeric, multi-subunit, transmembrane protein complex, of molecular weight ca. 650 kDa, that is found in the thylakoid membrane of plant chloroplasts and in cyanobacteria.

It catalyzes photosynthetic water oxidation and is therefore responsible for the presence of oxygen in the earth’s atmosphere.

PSII’s primary role in the photosynthetic chain is to use energy obtained from sunlight to produce a charge separation, leading to reduction of the membrane quinone pool.

This reducing power, augmented by further photonic energy in photosystem I, is used to assimilate carbon dioxide in the dark reactions of photosynthesis.

The ultimate source of these low-potential electrons is water, which is oxidized to dioxygen in the oxygen-evolving complex (OEC) of PSII according to the four-electron reaction given in eqn. (1):
2H2O → O2 + 4 H+ + 4 e

The electrons produced by this water-splitting reaction reduce the previously photo-oxidized chlorophyll P680, the site of primary charge separation in PSII, from which electrons are passed through a series of cofactors to a membrane-mobile plastoquinone.

P680+˙ is the strongest known biological oxidant, having a reduction potential recently estimated to be around +1.3 V.1

The OEC is incrementally oxidized by P680+˙ in one-electron steps through a five-state redox cycle, called the Kok or S-state cycle (the states being labeled S0 to S4).

Oxidizing power is accumulated in the OEC until, upon reaching the transiently-stable S4 (most oxidized) state, two molecules of water are oxidized to a single molecule of dioxygen, and the OEC returns from S4 to S0, its most reduced state.

The structure of the OEC, and the mechanism by which it catalyzes water oxidation, have together been the subject of prolonged investigation using many different spectroscopic and biochemical techniques (see refs. 2–5 for some recent reviews).

It has been established that the OEC consists of four Mn ions, Ca2+, and a redox-active tyrosine residue which shuttles electrons from the OEC to P680+˙.

The presence of one or more Cl ions has also been strongly suspected.

Until very recently, however, detailed structural information has been scarce.

The first X-ray crystal structure of a cyanobacterial PSII, obtained to 3.8 Å resolution (R factor 59%) in 2001, revealed the positions of the transmembrane α-helices, but left unresolved the precise locations of the cofactors, the identities of the individual amino-acids and the arrangement of the atoms in the OEC.6

In contrast, the recently-published structure of PSII at 3.5 Å resolution (R factor 30%) assigns for the first time a precise structure to the OEC, and identifies its ligating amino-acids.7

The complex is seen to take the form of a cuboidal Mn3CaO4 cluster, with a fourth Mn ion ‘dangling’ off one of the corner oxygen atoms.

We will discuss in detail this structure and its implications later in the paper.

First, however, we shall consider some concepts and experimental results that are important in understanding the water-splitting mechanism, and summarize our existing mechanistic proposal in these terms.

Then we will compare our proposal with the new crystal structure, and describe in what ways our proposal is supported or contradicted by this crystallographic information.

Our mechanism is largely corroborated by the crystal structure, but the structure is incompatible with our incorporation of Babcock’s proposal that the nearby tyrosine, YZ, abstracts protons as well as electrons from the OEC.8

We propose instead that an arginine residue close to the OEC plays the basic role in the water-splitting reaction, and offer an electrostatic explanation of how this occurs.

The role of proton-coupled electron transfer (PCET) in S-state advancement

The OEC catalyzes the oxidation of water to dioxygen and protons.

It follows, therefore, that progression through its S-state cycle must involve the movement of protons as well as electrons within the OEC.

The existence of proton-coupled electron transfer (PCET) in certain S-state transitions is suggested by several strands of experimental evidence, namely, (1) pH-dependences that correspond to the protonation/deprotonation of nearby residues, (2) H/D kinetic isotope effects (reviewed in refs. 9,10), and (3) temperature dependences that reveal high activation energies in those steps that are proposed to involve proton transfer.

In addition, constituents of the OEC (Mn, Ca2+ and Cl) have been removed or replaced, and nearby amino-acid residues mutated, to alter proton transfer in a revealing fashion.

The results of some of these experiments are summarized below.

The OEC exhibits a pronounced pH-dependence in its steady-state oxygen-evolving activity, with a maximum around pH 6.0–7..011,12

The pH-dependence of each individual S-state transition has also recently been obtained.13

Only the S1 → S2 transition was found to be independent of pH between 4.1 and 8.4.

This transition was proposed to involve electron transfer (ET) alone, with the result that a positive charge accumulates on the oxidized cluster.

Charge build-up in this step is indicated by electrochromic shifts seen in UV-visible absorbance spectra obtained from nearby chlorophylls,14–17 although it has been suggested that these observed Stark effects may be explained by invoking a change in the electric dipole moment of the OEC instead of an increase in its overall charge.18

All other transitions are blocked at low pH by a process with a pKa ≈ 4.8, which was ascribed to the protonation of a carboxylic acid involved in the proton expulsion pathway.13

At high pH, the S3 → [S4] → S0 transition is deactivated with a pKa ≈ 8.0, and the S2 → S3 transition with a pKa ≈ 9..413

This high-pH deactivation was interpreted as being the result of YZ having decreased in reduction potential, rendering it unable to oxidize the higher S-states.

The importance of the changing protonation state of YZ has been emphasized by a number of researchers.

The large discrepancy between the pKa of the oxidized and reduced form of the residue (pKox ≈ –2, pKred ≈ 9) has been used to argue that, energetically, it is very likely that its redox reactions are coupled to proton transfer.18

YZ has thus been suggested to exist in its oxidized state principally as the neutral, oxygen-centered radical, and in its reduced state as the neutral, protonated, phenolic form.

That the oxidized form is the neutral radical is supported by evidence from electron paramagnetic resonance (EPR) experiments19–22 and from Fourier-transform infra-red (FTIR) spectroscopy.23

FTIR spectroscopy also suggests that the reduced form of YZ is protonated at physiological pHs.23

Furthermore, site-directed mutagenesis of D1–His190, and subsequent chemical rescue studies, have shown that this base is required for effective oxidation of YZ˙.24–28

These results suggest that each time the OEC reduces YZ–O˙, ET is accompanied by proton transfer to form YZ–OH.

Equally, each oxidation of YZ–OH by P680+˙ is accompanied by the loss of a proton from YZ–OH to D1–His190, re-forming the neutral tyrosyl radical.

There is disagreement, however, on the ultimate source and destination of these protons.

Some groups (including Babcock and co-workers) have proposed that the protons are exchanged to and from the bulk lumenal solution,18 and other groups (including Junge and co-workers) claim evidence that the same proton stays near YZ and exerts an electrostatic effect on its surroundings.29

The important ramifications of this debate will be explored later in the paper.

Kinetic deuterium isotope experiments have been conducted which reveal that the rate of oxidation of YZ has a kH/kD = 2.9–3.4 at pL 6.5 in manganese-depleted core complexes, but that YZ oxidation is almost H/D insensitive in oxygen-evolving centers, at least on the nanosecond timescale.30–35

This suggests that oxidation of YZ in the intact OEC requires only minimal proton movement, which has been explained by positing a hydrogen-bonding network, stabilized by the complete cluster, which facilitates proton movement around YZ.23,36,37

Furthermore, calcium in particular is necessary for maintaining this hydrogen-bonding network.38

Acetate, a well-known inhibitor of the OEC, binds competitively with chloride39 and may exert its inhibitory effect by disrupting proton transfer around the OEC.40

The different temperature-dependences of the S-state transitions also offer clues about proton transfer.

In catalytically competent PSII, the S1 → S2 transition proceeds at temperatures as low as 140 K,41 whereas the S0 → S1 and S2 → S3 transitions advance only at temperatures above 220 K42,43 and have correspondingly greater activation energies (59.4 kJ mol−1 for S0 → S1, 26.8 kJ mol−1 for S2 → S3, and 9.6 kJ mol−1 for S1 → S2).44

The S2 → S3 transition is also shown by EPR to be inhibited in samples depleted of calcium or chloride22,45 or inhibited by acetate;46 the Kok cycle stalls under illumination at the S2YZ˙ state.

In other words, the OEC cluster remains in the S2-state, and the YZ˙ species, which would normally oxidize the OEC cluster, is unable to do so, perhaps because the reduction potential of the modified S3 cluster state is higher than that of YZ˙.

This stalling has been interpreted as indicating that, in order to reach the S3-state of the OEC, a well-ordered hydrogen-bonding network must exist through which protons or hydrogen atoms are able to move around the OEC, lowering the reduction potential of S3 by coupling its formation to proton-abstraction10.

Concerted electron transfer/proton transfer (ETPT) in the OEC

Some years after first identifying YZ as the OEC’s immediate oxidant,47 Babcock and co-workers suggested that YZ intimately combined its ET and proton transfer (PT) roles by abstracting hydrogen atoms from the OEC.8,48

Hydrogen atom transfer is a special instance of concerted electron and proton transfer (ETPT).

ETPT is itself a type of PCET and is defined by the simultaneous (rather than consecutive) transfer of electrons and protons.49

On the basis of spectroscopic information, Babcock et al.

noted that YZ was sufficiently close to the catalytic cluster to abstract hydrogen atoms on the observed timescale of the reaction, and that the low dielectric of the environment made favorable such a charge-neutral process.

Furthermore, the strong coupling of YZ’s pKa to its redox state (see above) was considered to provide essential thermodynamic driving force for the oxidation of the higher S-states of the OEC.

This aspect of the hydrogen atom transfer theory was influenced by the work of Krishtalik, who had considered water-splitting from a theoretical standpoint some years previously50 and whose thinking had already persuaded some researchers that ETPT was likely to be important in the reaction.51

The Babcock group proposed that the reduced tyrosine passed its acquired electron to P680+˙ and its proton to the hydrogen-bonded D1–His190 residue, from whence the proton passed through a channel of charged residues to the protein surface.

As mentioned above, some researchers (notably Junge) found evidence that the YZ proton always stayed close to the residue, and considered this observation fatal to the proposal of hydrogen atom transfer.29

Babcock and co-workers supposed that YZ abstracted hydrogen atoms from the OEC cluster in every one of the S-state transitions.48

While in general adopting the idea of hydrogen atom transfer, we (along with several other groups) pointed out that there is reason to believe that the tyrosyl radical only abstracts hydrogen atoms from the cluster in the later S-state transitions.10

This theory is consistent with a considerable body of evidence suggesting that there is a discontinuity in the means of S-state advancement beyond the S2 state.

Firstly, as mentioned above and considered in detail later, it is the transitions beyond S2 that are blocked by the absence of calcium or chloride, and by the presence of acetate.

Secondly, ET between the OEC and P680+˙ is significantly slower (and biphasic) in the S-state transitions beyond S2 than in the early S-states.52

Thirdly, the reorganization energies of the two higher S-state transitions have been found to be significantly higher than those of the two previous transitions.53

Finally, the sub-microsecond ET rates from the cluster to P680+˙ (via YZ) vary much more with temperature in post-S2 transitions than they do in the pre-S2 steps.54

We rationalized these findings by positing a switch in the mode of proton and electron transfer at the S2 stage of the cycle, a change activated by the positive charge developed on the OEC in the S1 → S2 transition.10

In the S0 → S1 transition we inferred the action of PCET, in which an electron and a proton move consecutively, and in the transitions after S2 we inferred the action of ETPT, in which the two particles move together and simultaneously.

The same conclusion has been independently reached,9,55 largely on the basis of proton release measurements in the different S-states.51,56

It might be hoped that PCET and ETPT would be distinguishable by measuring the effect of H/D exchange on the ET kinetics of the different transitions.49

Unfortunately, the results of such experiments have so far proven equivocal and from them no consensus has yet emerged32,34,57,58.

Mechanisms of water oxidation

A great number of specific mechanisms for the catalytic action of the OEC have been proposed over the years (for a list of some of these see .ref. 59)

In accordance with the results of experiments conducted on various PSII preparations and on model inorganic compounds, the oxidizing power of the OEC is now presumed to accumulate largely, if not entirely, on the four Mn ions.

The exception to this consensus has been the S2 → S3 step, which has been suggested by some to involve manganese oxidation8,10 and by others the oxidation of a species near to the cluster,60,61 perhaps a manganese-bridging oxo group.62

Manganese-bridging oxo ligands have also often been proposed to react with one another in the O–O bond forming step.63,64

Alternatively, one of the constituent atoms of dioxygen has been proposed to come from a manganese-bound terminal water molecule, perhaps deprotonated by basic μ-oxo ligands.65

Our proposal is amongst those that hypothesizes the role in catalysis of a unique, highly-oxidized MnVO species, along with a specific mechanistic role for calcium.

This type of mechanism is also favored by Pecoraro and co-workers.66

Siegbahn, using density functional theory calculations, has proposed a related mechanism in which a Mn-oxyl species is involved in the O–O bond-forming step67,68.

The Brudvig group proposal for the water oxidation mechanism

Fig. 1 outlines the S-state cycle previously propounded by this group.10

Experimental and theoretical justification of our assigned manganese oxidation states have been published previously.10

A recent EPR study of a di-manganese model compound, replicating the S1 multiline signal, supports our assignment of the manganese oxidation states in the S1 state.69

The distinctive aspects of our mechanism are as follows:

(1) The Ca2+ ion plays a precise, mechanistic role in the reaction by coordinating one of the substrate water molecules.

This ligation both aligns the water properly for the O–O bond-forming reaction and optimally tunes the pKa of this substrate water.

This water nucleophilically attacks an oxo oxygen derived from the second substrate water molecule.

(2) This second substrate water is coordinated to a Mn ion.

During the course of the S-state cycle, this Mn ion is successively oxidized, while at the same time its ligated water is deprotonated.

By the S4 state, a MnVO group has been achieved, whose oxo group is electrophilic enough to engage in an SN2-type reaction with the calcium-bound substrate water.

(3) The manganese-bound substrate water was proposed to be deprotonated by YZ, acting not only as an oxidant but also as a base throughout the reaction.

As discussed above, we have suggested that, in the S0 → S1 transition, ET is coupled to PT but the transfers are consecutive rather than simultaneous, and that the S1 → S2 transition involves only ET.

In contrast, we have proposed that in the final two cluster redox transitions (S2 → S3 and S3 → S4), YZ performs a concerted electron and proton transfer, abstracting a hydrogen atom from the OEC.

This aspect of our mechanism is a modification of Babcock’s proposal of hydrogen atom transfer in all S-state transitions.8

The switch from PCET to ETPT, activated by charge build-up in the S1 → S2 transition, was proposed to explain the apparent mechanistic discontinuity at the S2 → S3 transition (see above).

The role of calcium in the water-splitting reaction

Calcium has been revealed in the new crystal structure to be integrated into the OEC cluster, which agrees with the absolute requirement for Ca2+ in O2-evolution70 and, specifically, to its mechanistic role in water oxidation.

Recent 18O-isotope exchange experiments71 and pulsed EPR studies72,73 have corroborated calcium’s mechanistic role, as well as our proposal that the ion binds water early in the S-state cycle.

Several metals have been substituted into the Ca2+-binding site in the OEC, but only Sr2+ confers any catalytic activity, and that only partially.74

Other (inhibitory) metal ions, which are similar in their size and charge to Ca2+ and bind strongly to the site, have aqua-ion pKas significantly lower than that of Ca2+.74

Their inability to play a part in water oxidation suggests that they bind OH in the active site, which may be too strongly bound to the metal, and too strongly charged, to act as an effective nucleophile towards the MnVO oxygen atom.67

We have argued that calcium provides a bound water molecule, rather then a bound hydroxide ion, as the nucleophile in the O–O bond-forming reaction.10

This interpretation is supported by experiments in which trivalent cations have been substituted, with inhibitory effect, into the Ca2+-binding site.74,75

The strongly size-selective metal-binding site does not discriminate between di- and trivalent cations, indicating that the trivalent ions, which are strong Lewis acids, bind a hydroxide ligand and so are incorporated into the cluster with a net +2 charge.5

A similar proposal has been made to account for the observation by X-ray absorption spectroscopy (XAS) that only very small structural effects are induced by Dy3+/Ca2+ substitution in the OEC.76

However, the Dy3+-substituted OEC is unable to attain the positively-charged S2 state,77 indicating that the charge of the metal in this site does play a part in the cluster’s reactivity.

Similarly, calcium is found to be released from the cluster more readily in the higher, positively-charged S-states than in the lower S-states.78

Calcium also appears to have a role in maintaining the hydrogen-bond network necessary for proton transfer around the OEC.22,38

The inhibitory effect of calcium-depletion is exerted by blocking the S2 → S3 transition,79–81 an observation consistent with the necessity for proton transfer through a hydrogen-bonded network in this step.

These results point to three roles for calcium in the OEC.

The first is as a Lewis acid, binding and activating one substrate water as a nucleophile in the O–O bond-forming reaction.

The second is in maintaining the hydrogen-bonding network around the OEC, which appears to be necessary for progression beyond the S2 state.

The third is a general electrostatic role, in which the metal modulates the charge distribution, and thus the reduction potential, of the cluster76.

The role of chloride in the water-splitting reaction

Chloride, like calcium, appears specifically to be required for the S2 → S3 transition,45 and has also been implicated in facilitating proton movement around the OEC.82,83

However, the OEC’s chloride requirement is certainly less stringent than its calcium requirement.

It seems that chloride-depleted samples may be reconstituted to a catalytically active form by a variety of anions, including bromide and nitrate.84,85

Chloride’s necessity in O2 evolution has recently been seriously called into question by Andreasson and co-workers.82

These authors claim that observed chloride-depletion effects are largely artefactual, caused by protein alteration linked to the sulfate treatments employed in removing chloride from the complex.

Nevertheless, the balance of evidence suggests that one Cl ion is normally associated with the functional OEC.86

The proximity of the chloride ion to the OEC is unknown, with some groups hypothesizing that it is ligated to the tetramanganese cluster10,87 and others postulating that it is further removed.84

Recent work in assigning an extended X-ray absorption fine structure (EXAFS) signature to chloride bound to Mn-oxo model clusters will help to resolve this question.88

In our previous proposal, chloride was present as a bridging ligand between Ca2+ and the substrate-binding Mn ion (see Fig. 1).

The new X-ray crystal structure of the OEC7

The new crystal structure of the OEC is shown in panel A of Fig. 2.

It appears that the catalytic site takes the form of a metal-oxo distorted cube, comprising four metals and four bridging oxygen ions, with one of these corner oxygens (a μ4 ligand) binding a fifth metal on the cuboid’s exterior.

Using X-ray anomalous difference maps at different wavelengths, the cuboid is proposed to contain three Mn ions and one Ca2+, making a Mn3CaO4 cluster, and the exterior ‘dangler’ metal ion is taken to be the fourth manganese.

Although this exact arrangement had never been previously proposed on the basis of spectroscopic investigations, the Mn ions had indeed been proposed to exist as a ‘3+1 tetramer’, largely on the basis of EPR spectroscopies and EXAFS measurements.89

D1–Tyr161 (‘YZ’), which has long been established as the immediate oxidant of the OEC in the cluster’s catalytic redox cycle,47 is 5.0 Å from Ca2+, the nearest cluster atom.

This short distance is compatible with fast electron transfer from the cluster to YZ, and also, potentially, with proton transfer.

D1–His190, which has been previously identified as YZ’s hydrogen-bonding and proton-exchange partner (see above), is revealed in the structure to be excellently positioned for this function.7

Several other hydrophilic residues have been found close to the OEC, whose possible roles in hydrogen-bonding and proton transfer will be discussed below.

Of particular importance is the side-chain of CP43–Arg357, which is shown to be very close to our proposed active site of the cluster.

Finally, crystallographic analysis revealed non-protein electron density in the space above Ca2+ and Mn(4), which was best fitted as a bicarbonate, carbonate or nitrate ion.

Because nitrate was not expected to be present, this space in the structure has been tentatively occupied with a bicarbonate ion7.

Comparison of previously proposed mechanism to new crystal structure

It is important to emphasize that the recent X-ray structure provides an imperfect image of the OEC in an indeterminate redox state that cannot do more than suggest a catalytic mechanism in vivo.

However, the active site of the metal cluster may be guessed at with some confidence from the crystal structure.

There are three principal reasons for suspecting that amongst the metals, the ‘dangler’ Mn(4) ion and the Ca2+ ion play key roles, binding unseen reactant water molecules between this pair of metal ions.

Firstly, the distance and angle between the two metals make the O–O bond-forming reaction between two such bound waters geometrically plausible (see below).

Secondly, Mn(4) lies at one end of a row of charged amino-acid residues (beginning with D1–Asp61) which leads to the lumenal protein exterior, an arrangement which stands out in a generally hydrophobic area of the protein.7

These residues would be well-suited to moving protons out of the active site only if Mn(4) were intimately involved in the catalytic reaction.

Thirdly, Mn(4) and Ca2+ are together the closest of the metal ions to YZ, which acts in its radical form as the immediate oxidant of the OEC.

Taking these telling points together, the positions, proximity and alignment of Ca2+ and Mn(4) favors those proposed mechanisms (such as ours) that have envisaged a direct role for Ca2+ and one of the manganese ions in coordinating the substrate water molecules.

The coordination of the metal atoms in the OEC

The new crystal structure shows that the four Mn ions have between them, besides the four bridging oxo groups of the cuboidal cluster, only six ligating amino-acid residues.

These residues had all previously been identified as integral to the OEC on the basis of mutational studies,90,91 except for Glu354 of the CP43 subunit, which was unexpectedly found in the new crystal structure to ligate Mn(3).7

Manganese in proteins is almost invariably five- or six-coordinate, and the metal generally has a preference for octahedral geometry, especially in the higher oxidation states.92

Calcium is typically six- to eight-coordinate.93

This implies that there are probably coordination sites on each of the metals that are occupied (at least in some oxidation states) by small, unseen molecules.

In the absence of evidence to the contrary, it may be assumed that most of these small molecules are waters or hydroxides.7

The two manganese ions on the opposite side of the cluster from our proposed active site, Mn(1) and Mn(2), are structurally important but not directly involved in our proposed mechanism.

They are assigned to the two manganese ions that remain in the +4 oxidation state throughout our proposed cycle, and the details of their coordination need not concern us here.

The catalytically important metals in our mechanism are Mn(4), Mn(3) and Ca2+, and so we must closely examine their probable coordination.

They comprise what we take to be the ‘active face’ of the OEC cluster.

Our proposal for the coordination (in the S0 state) of these metals and nearby amino-acid residues is shown, superimposed on the crystallographic data, in panel B of Fig. 2.

Mn(4) is seen to have three ligands in the crystal structure: the μ4 cluster oxygen atom, D1–Asp170 and D1–Glu333, these last two each binding via their carboxylate side-chains in monodentate mode.

The three ligands are aligned 90 ± 5° to each other in the same plane, and it seems reasonable to take them as three of the metal’s equatorial ligands, constituting part of a near-regular octahedral coordination.

Maintaining this geometry, the fourth equatorial ligand, trans to D1–Glu333, we take as one of the substrate water molecules.

The angle, relative to the Mn3CaO4 cluster, of the bond joining Mn(4) to the μ4-O atom is such that the putative substrate water on Mn(4) is tilted towards the coordination sphere of the Ca2+ ion.

We assume that there is a non-substrate water ligand in one of the Mn(4) axial positions (pointing towards the viewer in panel B of Fig. 2), which is well positioned to hydrogen-bond to D1–Asp61, the first residue of the apparent hydrophilic proton exit pathway.7

The sixth and final supposed ligand of Mn(4), the second axial ligand, points away from the area of chemical activity and therefore does not concern us.

Mn(3) is coordinated to three oxygen atoms within the cuboidal cluster, and by CP43–Glu354 in a seemingly bidentate fashion.7

The alignment of these ligands again indicates octahedral geometry, with at least one (axial) ligand unseen in the crystal structure.

This axial position, which points towards the CP43–Arg357 side-chain on the active face of the cluster, we have tentatively occupied with a non-substrate hydroxide ligand.

This is because in our proposed mechanism this Mn ion is always in the +3 or +4 oxidation state, and would be expected to prefer to coordinate hydroxide rather than water.94

The only ligands of Ca2+ revealed in the crystal structure are the two μ3- and the single μ4-oxo atoms of the OEC cuboidal cluster.

No proteinaceous ligands are explicitly identified, although the carboxyl terminus of the D1 subunit, D1–Ala344, is found to be very close (2.01 Å) to the metal.7

There is good evidence to suggest that the OEC is ligated by this terminus95 and we therefore assume that D1–Ala344 does indeed ligate Ca2+ in a mono- or bi-dentate fashion.

Judging by a recent survey of carboxylate binding to Ca2+ in proteins, the former mode is a little more likely.96

Assuming D1–Ala344 carboxylate-terminal coordination of Ca2+, this metal presumably also has bound to it a certain number of unseen, small molecules: a minimum of one (bidentate carboxylate coordination, total metal coordination number 6) and a maximum of four (monodentate carboxylate coordination, total metal coordination number 8).

As described above, there is good evidence to suggest that one of these small molecules is the second substrate water molecule.

We have added this on the active face of the OEC cluster, that is, the face comprising Ca2+, Mn(3) and Mn(4), and near to CP43–Arg357.

In our previous proposal, Cl was present as a bridging ligand between the substrate-binding Mn ion and Ca2+, a position occupied in the new structure by the μ4-O atom of the cuboidal cluster.

There is no evidence for Cl around the OEC in the new crystal structure, although chloride is a weak X-ray scatterer compared to the metal ions of the cluster and might easily be overlooked.

Acetate, which competes with chloride for the same binding site in the OEC,39 binds close to YZ˙.97,98

Calcium depletion and chloride depletion both cause the reaction cycle to stall at the same stage in the cycle, giving similar stable EPR S2YZ˙ signals, which suggests that these ions are functionally connected in the OEC.

We therefore propose as previously that chloride is present in the OEC bound to calcium, but instead of a μ-Cl we propose a terminal chloride situated between the metal and YZ.

This would place the Cl ion roughly 2.5 Å from YZ, in accordance with recent pulsed EPR data which estimated a distance of ca. 3 Å between YZ and OEC-bound acetate.97

Both calcium and chloride are important, we suggest, in maintaining a hydrogen-bonded network around the OEC, and this aspect of their function will be dealt with later.

The identity of the catalytic base in the water-splitting reaction

Despite the generally good agreement between the structural requirements of our proposed mechanism and the new crystallographic information, a problem arises in the identity of the catalytic base and the subsequent exit of protons from the active site.

We had hypothesized that YZ acts as both an oxidant and as a base in the course of the Kok cycle (see above).

Protons derived from the substrate waters were thought to pass from YZ to D1–His190, YZ’s well-established hydrogen-bonding partner, and from this residue through a series of hydrophilic amino-acid residues to the protein exterior.

The new crystal structure shows that D1–His190 is indeed ideally placed to hydrogen-bond to the YZ side-chain.7

However, there are no further hydrophilic species in the structure positioned so that they might accept a proton from the imidazole ring and eventually pass it to the protein surface.

The apparent proton-exit pathway, comprising a chain of promisingly positioned hydrophilic residues, begins at D1–Asp61 (on the other side of the cluster’s active face from YZ) and leads directly away from YZ.7

If we take the crystal structure at face value we are therefore forced to one of two conclusions.

The protein in this region may be sufficiently plastic that certain amino-acid residues move, at one time or another in the catalytic cycle, from the arrangement seen in the crystal structure to one that establishes a proton-transporting wire leading from D1–His190 to the protein surface.

Alternatively, YZ might in fact accept neither protons nor hydrogen atoms from the OEC.

Instead, as the Kok-cycle proceeds, YZ might repeatedly exchange the same proton with D1–His190, receiving the proton as YZ is reduced, and donating it as YZ is oxidized.

Such ‘proton-rocking’ between YZ and D1–His190 has indeed been proposed on the basis of the temperature-dependence of ET kinetics99 and by Junge and co-workers on the basis of extensive proton release measurements.29

This mode of PCET is also widely believed to take place at the homologous tyrosine in the D2 subunit, YD.100

The new crystal structure has shifted the balance of evidence, although not definitively, towards the proton-rocking hypothesis for YZ.

Nevertheless, acknowledging the structure’s imperfect resolution and bearing in mind the evidence for conformational change around the OEC during the catalytic cycle101 it remains possible to argue that YZ is able to abstract protons or hydrogen atoms from our proposed active site, a mere ca. 5 Å distant.

As for the fate of the abstracted proton, YZ and D1–His190 are both less than 4 Å from D1–Gln165, which, in turn, is less than 4 Å from CP43–Arg357, a residue plausibly linked to the proton exit pathway beginning at D1–Asp61 (see below).

Future structures obtained in definite oxidation states might show that YZ is indeed linked via these residues to the lumenal surface.

In the meantime, our attention is focused on identifying another species capable of playing the basic role previously allotted to YZ in the water-splitting mechanism, and which is linked to the crystallographically observed proton exit pathway.

Moreover, in so adapting our proposal we improve its mechanistic plausibility and its agreement with experimental results.

For the first time we are able to propose a precise electrostatic mechanism to explain the mechanistic discontinuity proposed to exist in the Kok cycle at the S2 state.

The role of CP43–Arg357 in the water-splitting reaction

We propose that the nearby CP43–Arg357 side-chain is central in organizing the flow of protons from the OEC during the later part of the catalytic cycle.

It is only with the publication of the recent crystal structure that this residue’s proximity to the OEC has emerged: the terminal nitrogens of the CP43–Arg357 side-chain are each a little over 4 Å from the active face of the cluster.

The residue’s functional indispensability had already been established by observing that the O2-evolving activity of PSII from Synechocystis sp.

PCC 6803 is abolished if one mutates the homologous CP43 arginine (Arg342) to serine.102

Furthermore, the residue is conserved in the same genetic position in all known CP43 sequences.7

We propose that CP43–Arg357 is necessary for two, interconnected reasons.

First, it helps to organize the hydrogen-bonding network of metal-ligated waters and hydroxides around the OEC cluster.

Second, in its neutral form it abstracts protons from the manganese-bound substrate water in both S-state progressions beyond the S2 state.

These protons then leave to the lumenal bulk solution through the pre-organized hydrogen-bonding network.

The arginine sidechain is a well-known, important and versatile component of hydrogen-bonding networks in proteins.

Cytochrome c oxidase, for instance, contains a hydrogen-bonded network incorporating two highly-conserved arginines which are important in optimizing both electron and proton transfer around the active site.103,104

In the new structure of the PSII OEC, the guanidinium side-chain of CP43–Arg357, straddling the active face of the metal-oxo cluster, is positioned to hydrogen-bond not only to the two putative substrate waters (ligated respectively to Ca2+ and Mn(4)), but also to the putative non-substrate ligands (probably water or hydroxide) of Mn(3) and Mn(4).

Because the positions and nature of these ligands are unknown, we have not attempted to propose detailed hydrogen-bond distances and angles.

Nevertheless, a network of hydrogen-bonds similar to that shown in Fig. 4, to be discussed later, seems eminently plausible.

It may be that not all of these bonds are extant at any one time, but that the components of the network move a little through the course of the reaction cycle, making some connections and breaking others.

Both of the substrate water ligands, bound respectively to Ca2+ and Mn(4), may plausibly be fitted into the structure within a few angstroms of the terminal nitrogens of the arginine side-chain, which points almost directly at these spaces in the structure.

Arginine’s postulated hydrogen-bonding interactions with the substrate waters are significant because our mechanism implies that these waters are prevented from hydrogen-bonding to one another.

Such a bond would fix a hydrogen between the two oxygen atoms, preventing their close approach in the O–O bond-forming step.

CP43–Arg357’s ability to hydrogen-bond with the other, non-substrate ligands of Mn(3) and Mn(4) presents a plausible route through which protons may travel from the substrate waters to D1–Asp61, the beginning of the proton exit pathway (see Fig. 4, later).

As the Kok cycle progresses beyond the S2 state, therefore, concerted ETPT may be imagined to take place in a bifurcated manner: electrons pass from the cluster to oxidized YZ˙, and protons, abstracted by basic CP43–Arg357, pass to the lumen through a hydrogen-bonded network via D1–Asp61.

While arginine’s ability to hydrogen-bond is not in doubt, it is not well-known as a general acid/base catalyst in protein active sites.

However, there are now well-established examples of it serving in such a fashion, namely in the fumarate reductases105,106 and in two families of polysaccharide lyases.107,108

Simply on the basis of comparing the pKa of the side chain in the free amino-acid (ca. 12)109 with the pH of the chloroplast lumen (as low as 5)110 one would not anticipate the existence of a neutral, deprotonated arginine at the OEC.

However, the electrostatic environment anticipated at the OEC in its higher oxidation states makes the momentary existence of the neutral arginine highly plausible.

It is known that the pKa of an amino-acid residue may be dramatically altered by the effect of nearby electrostatic charges (see, for instance, .ref. 111)

Specifically, we propose that the increase in positive charge of the OEC during the S1 → S2 transition, along with the positive charge of the D1–His190 residue present in the S2YZ˙ state, is sufficient in catalytically-competent PSII preparations to deprotonate the side-chain of CP43–Arg357 and poise it to abstract protons from the substrate water in subsequent steps.

This proposition is described schematically in Fig. 3, which shows the OEC cluster surrounded by four important amino-acid residues: YZ, the immediate oxidant of the cluster; D1–His190, YZ’s hydrogen-bonding and proton transfer partner; CP43–Arg357, which we propose acts as a base in the later S-states; and D1–Asp61, which constitutes the beginning of the crystallographically revealed proton exit pathway.

The simple S1 → S2 transition is illustrated in the left hand side of Fig. 3.

The S1 state of the cluster is, we assume, uncharged.

The side-chain of CP43–Arg357 we take here to be positively charged, in accordance with its normal high pKa, and D1–Asp61 we always take to be in its negatively-charged carboxylate form.

First of all, YZ–OH reduces P680+˙ and, in a coupled reaction, transfers its tyrosyl proton along a hydrogen-bond to the imidazole ring of D1–His190, making this side-chain positively charged.

The newly-formed YZ–O˙ now oxidizes the OEC cluster from the S1 to the S2 state.

In this S-state transition, no proton transfer is associated with oxidation of the cluster, and so the cluster acquires a positive charge.

As YZ–O˙ is reduced, the proton it recently lost to D1–His190 rocks back along the hydrogen-bond to bind again to the YZ side-chain.

The bottom left-hand panel of Fig. 3 shows a neutral YZ–OH residue alongside the cluster and the CP43–Arg357 side-chain, both of which are now positively charged.

We start with this S2 state on the right-hand side of Fig. 3.

As before, the first step in the S-state transition is the oxidation of YZ, coupled to proton transfer to D1–His190.

Now, however, the mechanism differs from the S1 → S2 case.

In the S2YZ˙ state, both D1–His190 and the OEC cluster are positively charged.

The positively-charged CP43–Arg357 side-chain is ca. 4 Å from the cluster and less than 9 Å from the imidazolium ring of D1–His190.

Furthermore, these two repulsive electrostatic interactions occur in a site that is deeply buried from bulk solvent and is therefore presumably a low-dielectric environment.112

We propose that, in order to diminish these unfavorable electrostatic interactions, the guanidinium group of CP43–Arg357 deprotonates to the lumenal bulk solution via D1–Asp61, to which arginine is connected through hydrogen-bonded non-substrate ligands of Mn(3) and Mn(4) (see above).

Having deprotonated, the neutral arginine side-chain constitutes a powerful base close to the Mn(4)-bound substrate water ligand.

This redox-active Mn(4) is now able simultaneously to lose an electron to YZ–O˙ and a proton to CP43–Arg357 in a concerted ETPT reaction.

The reduction of YZ–O˙ is, as always, coupled to proton transfer from D1–His190.

The final panel of the figure shows the S3 state, which is electrostatically poised to undergo a similar ETPT reaction in its next transition, S3 → S4.

The cluster begins with a positive charge in both of the transitions after S1 → S2, and each Sn>1YZ˙ state provides sufficient positive charge to create transiently the guanidine base required for electroneutral oxidation of the cluster.

The ET-coupled change in pKa of the arginine side-chain (from a value below the ambient pH in the protonated form to a value higher than the ambient pH in the deprotonated form) admirably satisfies Krishtalik’s energetic requirement for water oxidation, which was proposed by Babcock to be fulfilled by the changing pKa of YZ.18,50

We similarly suggest that oxidation of the cluster in the higher S-states is thermodynamically coupled to the simultaneous protonation of a strong, nearby base, but that the base is the guanidine side-chain of CP43–Arg357 rather than the tyrosinate side-chain of YZ.

Although there is no unambiguous experimental support for our proposed role of CP43–Arg357 in the Kok cycle, ET kinetic data have been interpreted in terms of a species near YZ (CP43–Arg357 is a little over 7 Å from YZ) shifting its pKa upon YZ oxidation from a value ≥9 to around 6, and thus deprotonating to the protein exterior.113

It is possible that this species is CP43–Arg357.

Also in agreement with our proposal are ET and proton release kinetic data which have shown correspondence between a lag phase of ET from the metal–oxo cluster to YZ˙ and a rapid phase of local electrochromism occurring in the time range of rapid deprotonation.51,114

Lavergne and co-workers have interpreted these results as indicating that YZ˙ reduction by the OEC cluster is controlled by the deprotonation of a group closer to the cluster than it is to YZ, so that, by acting as a base, this group increases the driving force of the reduction of YZ˙.

This interpretation perfectly describes our proposal for the role of CP43–Arg357, which is shown in the crystal structure to be ca. 3 Å closer to the cluster than it is to YZ.

It is important to note that these ideas, which concern the coupling of YZ–OH oxidation to the alteration of the pKa of a nearby species, require that the proton lost by YZ–OH remain in the OEC in order to exert its electrostatic effect.29

Such ideas are given credibility by the new crystal structure, which shows no clear way for protons to escape from the hydrogen-bonded pair of residues, YZ and D1–His190.

The hydrogen-bonding network around the OEC is stapled in place by the CP43–Arg357 side-chain, organizing the flow of protons from the substrate water to D1–Asp61 as electrons flow to YZ in the later S-state transitions.

We suggest that it is the disruption of this network by chloride depletion, calcium depletion or acetate inhibition that prevents the OEC from being oxidized beyond the S2 state.

Such disruption means that CP43–Arg357 is unable easily to deprotonate to bulk upon attainment of the S2YZ˙ state, with the result that proton abstraction cannot be coupled to the further oxidation of the cluster.

The reduction potential of YZ˙ is then not high enough, at least at normal pH values, to continue the Kok cycle.

The proton lost in the S0 → S1 transition evidently does not require such an ordered hydrogen-bonding network, and so we propose that this proton is lost from a Mn(4)-ligand that is directly connected to the proton exit route, rendering the hydrogen-bond network less important in this step.

The role of calcium and chloride in hydrogen-bonding around the OEC

In our proposal the substrate water and chloride bound to Ca2+ are linked into the hydrogen-bonded network as described above and indicated in Fig. 4.

Metal-bound chloride is an established hydrogen-bond acceptor115 and, consonant with its proposed role in proton translocation (see above), chloride in our proposed position would be able to accept a hydrogen-bond from the substrate water bound to the dangler Mn(4).

Both bromide and iodide could partially fulfill this function, although metal-bound fluoride, in line with its inhibitory effect, exhibits markedly different hydrogen-bonding behavior from that of its heavier congeners.116

It is conceivable that acetate bound at the OEC would break up the hydrogen-bonding network by binding to calcium in place of chloride (possibly also displacing a substrate water) and by presenting its methyl group to neighboring species.

The role of bicarbonate in the water-splitting reaction

There is currently considerable debate concerning the role of bicarbonate in the donor-side reactions of PSII.

The debate stretches back to Otto Warburg, who proposed that bicarbonate was a substrate in the oxygen-forming reaction, and the ion’s role in the OEC has since been intermittently investigated.117

However, isotope-exchange experiments show that water molecules, rather than carbonate or bicarbonate, are the substrates in the oxygen-forming reaction.118

There is evidence, notably from Dismukes and co-workers, that bicarbonate is involved in the assembly of the OEC.119

Investigations into the effects of bicarbonate on the OEC are complicated by the diverse effects of bicarbonate on PSII, particularly on the acceptor-side.120

The appearance in the new crystal structure of electron density that may be modeled as bicarbonate, carbonate or nitrate near to the Mn3CaO4–Mn cluster is certain to provoke renewed discussion of bicarbonate’s role in the OEC.

Given that, in the structure, the modeled bicarbonate occupies the apparent active site of the OEC, and that bicarbonate is not a substrate in the oxygen-evolving reaction, it is hard to see what mechanistic conclusions may be drawn from its presence.

We treat it as an adventitious anion binding in an inactive form of the OEC and have therefore neglected it in our mechanism.

We do not rule out a role for bicarbonate in the operation of the OEC, perhaps in regulating proton currents by forming part of the hydrogen-bonding network, but there is no evidence for its direct involvement in the water-splitting reaction.

Mechanism of water oxidation

We show in Fig. 4 the Mn3CaO4 cluster and some surrounding amino-acid residues as the OEC cycles through our proposed mechanism of water oxidation.

The changing redox state of YZ, which is alternately reduced by the cluster and oxidized by P680+˙, is indicated alongside the reaction arrows.

Note that the S0–3 states are drawn in the SnYZ˙ state, which means that the YZ˙ radical is poised to oxidize the cluster.

The transiently-stable S4 state never co-exists with the YZ˙ radical and so is shown in the S4YZ state.

The two Mn ions on the bottom face of the cuboidal cluster are not redox-active and remain in the +4 oxidation state throughout the cycle.

Mn(3) and Mn(4) change their oxidation states as described previously in Fig. 1.

The first transition, S0 → S1, involves the oxidation of Mn(4) and the concomitant loss of a proton from its non-substrate water ligand.

This water is directly hydrogen-bonded to the D1–Asp61, the start of the proton exit pathway, and so this PCET is able to occur in the absence of a normal hydrogen-bonded network around the rest of the OEC (Cl depletion or acetate inhibition) and even in the absence of a normally-constructed cluster (Ca2+-depletion).

The second oxidation of the cluster, to reach the S2YZ state, involves ET only and so the cluster develops a positive charge.

This reaction, involving no proton transfer at all, occurs particularly easily.

Oxidation of YZ to reach the S2YZ˙ state is swiftly followed by deprotonation of CP43–Arg357, allowing concerted, bifurcated ETPT now to take place.

YZ accepts an electron from Mn(4), while at the same time neutral arginine abstracts (along a hydrogen-bond) a proton from the metal’s bound substrate water.

Oxidation of YZ by P680+˙ forms the S3YZ˙ state of the OEC.

Owing to the positive charge on the cluster, the next transition follows a similar mechanism.

In this S3 → S4 step, transiently-neutral CP43–Arg357 abstracts the second and final proton from Mn(4)’s substrate water, forming MnVO.

The subsequent S4 → S4′ transition is the crucial O–O bond-forming step, and involves the formation of a peroxo intermediate at Mn(4), which is concomitantly reduced from MnV to MnIII.

We assume that the arginine side-chain remains protonated in this step, because YZ stays reduced until the formation of S0 and therefore does not exert any electrostatic effect on CP43–Arg357.

Therefore the proton that is lost in the spontaneous S4 → S4′ step we take to move to bulk, without the intercession of any specific base, through a hydrogen-bonded proton-relay network via D1–Asp61.

The nearby D1–Gln165 residue may play a role in this step by accepting a hydrogen-bond from the calcium-bound substrate water as it nucleophilically attacks the MnVO group.

Finally, the unstable MnIII-bound peroxo intermediate S4′ decays to produce dioxygen and the OEC is reset to the S0 state.

Because Mn(4) in this step is reduced all the way to weakly-acidic MnII, it is plausible that the proton lost from the peroxo intermediate simply moves to the Mn(4) hydroxo ligand next to D1–Asp61, reforming the metal’s non-substrate water ligand.


In this paper we have considered the mechanism of water-oxidation by the OEC with reference to the newly-published crystal structure of PSII.7

The structure generally agrees well with our existing mechanistic proposal,10 and is consistent with the involvement of two substrate water molecules bound respectively to Ca2+ and Mn(4).

However, the structure does not reveal a proton-exit route from YZ, which casts doubt on this residue’s ability to act as a catalytic base.

CP43–Arg357 is apparently well-suited to play this role instead.

The arginine side-chain is close to the active face of the metal-oxo cluster, and its protonation state is plausibly coupled to the redox state of YZ in the higher S-state transitions, during which the metal-oxo cluster is positively charged.

We propose that these transitions (S2 → S3 and S3 → S4) proceed by bifurcated, concerted electron/proton transfer (ETPT), with an electron passing from Mn(4) to YZ˙, and a proton simultaneously passing from the Mn(4)-bound substrate water to the transiently-neutral CP43–Arg357 side-chain.

Such highly-coupled electron and proton transfer requires an organized hydrogen-bonding network around the OEC, linking the substrate waters both to CP43–Arg357 and to D1–Asp61, the start of the crystallographically-resolved proton-exit pathway.

Proton-coupled electron transfers in other S-state transitions do not involve neutral CP43–Arg357 as a base and have less stringent organizational requirements, in line with experimental data.