What Are the Oxidation States of Manganese Required To Catalyze Photosynthetic Water Oxidation?

Abstract

Photosynthetic O₂ production from water is catalyzed by a cluster of four manganese ions and a tyrosine residue that comprise the redox-active components of the water-oxidizing complex (WOC) of photosystem II (PSII) in all known oxygenic phototrophs. Knowledge of the oxidation states is indispensable for understanding the fundamental principles of catalysis by PSII and the catalytic mechanism of the WOC. Previous spectroscopic studies and redox titrations predicted the net oxidation state of the S₀ state to be (Mn^III)₃Mn^IV. We have refined a previously developed photoassembly procedure that directly determines the number of oxidizing equivalents needed to assemble the Mn₄Ca core of WOC during photoassembly, starting from free Mn^II and the Mn-depleted apo-WOC complex. This experiment entails counting the number of light flashes required to produce the first O₂ molecules during photoassembly. Unlike spectroscopic methods, this process does not require reference to synthetic model complexes. We find the number of photoassembly intermediates required to reach the lowest oxidation state of the WOC, S₀, to be three, indicating a net oxidation state three equivalents above four Mn^II, formally (Mn^III)₃Mn^II, whereas the O₂ releasing state, S₄, corresponds formally to (Mn^IV)₃Mn^III. The results from this study have major implications for proposed mechanisms of photosynthetic water oxidation.

Introduction

Oxygenic photosynthesis extracts electrons and protons from water via the water-oxidizing complex (WOC) found within photosystem II (PSII). The WOC is comprised of an inorganic core (Mn₄CaO₅) that cycles through a series of four intermediate states, called S states, before releasing a molecule of O₂. These states are denoted as S₀, S₁ (dark stable), S₂, and S₃, with the subscript referring to the number of oxidizing equivalents that are stored through successive electron removal. Despite extensive studies, neither the individual oxidation states of the Mn ions nor their mean oxidation levels during the catalytic process have been conclusively determined in any of the S states. Identifying the oxidation states of Mn operating in the cluster is essential to understanding the catalytic potential and the molecular mechanism of this pervasive biological process, as well as the synthesis of functional water-splitting catalysts modeled on its principles.

The four Mn ions of the WOC are spin coupled, which creates a ground electronic state with a single value for the electronic spin. In the S₀ and S₂ intermediates, distinctive multiline electron paramagnetic resonance (EPR) spectra are seen at low temperature, and have been assigned to a spin state with one unpaired electron (total spin S = 1/2 (1,2)). This multiline structure in S₂ fit simulations based on ⁵⁵Mn hyperfine interactions with all four Mn ions (3), which constrained the allowed individual Mn spin states to values consistent with this total spin. On this basis, the S₂ intermediate is widely accepted to contain only Mn^III (S = 2) and Mn^IV (S = 3/2) ions, but the number of each (one or three Mn being formally Mn^IV) has not been determined and is not agreed upon (4–7).

Three general types of experiments have been applied to identify the Mn oxidation states: 1), spectroscopic techniques applied to the intact WOC (4,8–11); 2), biogenesis and reassembly of disassembled apo-WOC-PSII complexes from the free inorganic cofactors (12); and 3), chemical titrations using reductants to selectively reduce the PSII-WOC to Mn^II, which is subsequently quantified after it is released from the PSII protein (13–17). Debate over the Mn oxidation states originates from uncertainties in interpretation and calibration of these methods. For example, oxidation-state assignments derived from spectroscopic techniques are semi-empirical, relying upon comparisons with data from synthetic complexes. These databases are invariably incomplete and generally different for each technique. They are further subject to ambiguity owing to differences in the coordination number and local environment (4,18).

Three different sets of oxidation patterns for the Mn ions through the water-splitting cycle have been proposed, with each set starting from an assumed oxidation distribution for the least oxidized intermediate, S₀. Subsequent S-state intermediates are derived from these sets by progressive electron withdrawals. In S₀, these proposed individual oxidation states are (Mn^III)₃Mn^II, (Mn^III)₃Mn^IV, and (Mn^IV)₂Mn^IIIMn^II. These assignments are based on electronic spectroscopy methods, including EPR (4), ⁵⁵Mn-ENDOR (8), Mn-K-edge absorption (XANES) (9), NMR relaxation (10) and Mn-Kβ emission (11). The latter two possibilities, (Mn^III)₃Mn^IV and (Mn^IV)₂Mn^IIIMn^II, can be considered equivalent for our purposes because they have lost the same number of electrons relative to the initial Mn^II ions from which they formed during the light-dependent assembly process (also called photoactivation in earlier literature). Thus, five one-electron photo-oxidation steps are required to reach S₀ for this assignment of oxidation states. By contrast, if S₀ has the (Mn^III)₃Mn^II oxidation states, S₀ is reached after three one-electron photo-oxidation steps starting from four Mn^II. Consequently, with the inclusion of the four one-electron steps of the S-state water oxidation cycle, the minimum number of photons required to produce the S₄ state is expected to be either 7 or 9 for the above two oxidation-state assignments. Here, we call these proposed paradigms the low oxidation (LO)-state and high oxidation (HO)-state paradigms, respectively.

Measurements of the photoassembly process involve detecting O₂ after short saturating flashes of light and, to date, favor the LO-state paradigm (19), whereas XAS and ENDOR spectroscopic data have generally been interpreted within the HO-state paradigm (8,9). Photoassembly measures the yield of O₂ as a function of flash delivery rate and flash duration starting from the cofactor-extracted apo-WOC-PSII. This method has identified two sequential light-induced intermediates (IM₁ and IM₂) below the native S₀ state (19,20), which have also been characterized by EPR spectroscopy (21–23). These experiments clearly show that only seven flashes are required to advance to the O₂ evolving state, using millisecond-duration pulses (10–30 ms). These pulses are long compared with the flash requirements for single oxidation state turnover in the intact WOC, but are the optimum duration for the lowest quantum yield photoassembly step in apo-WOC-PSII (QY ∼ 1%). Single turnover pulses used to advance native S states in the holoenzyme are shorter (<500 μs) (24). As a result, skepticism has prevailed about how to interpret the oxidation state assignments based on photoassembly results, because they contradict the prevailing spectroscopic interpretation.

The chemical-reduction approach for counting the number of oxidation equivalents above the Mn^II level that are needed to achieve functional turnover relies on titrations using reductants such as NH₂OH, NH₂NH₂, and hydroquinone to form new reduced S states (S_-1, S_-2, etc.). These states have been invoked to rationalize the delayed patterns of flash O₂ yield (25–27) seen on reduction, and proposed to be transiently populated during photoassembly (28). Alternatively, the number of released Mn^II can be measured (14). Such methods all require a large excess of reductant to observe a response, which then must be fully washed away before assaying for O₂ by flashing. Photoassembly studies have revealed that excess chemical reductants invariably result in carryover that cannot be removed by washing (29), resulting in an increase in the number of flashes required to reoxidize the WOC-PSII and produce O₂. By contrast, photoassembly in vivo is done without an exogenous reductant except for the native substrate, water.

Here we present an improved photoassembly method that uses a series of much shorter, single-turnover laser flashes of variable duration to eliminate the above uncertainties, coupled with higher-sensitivity detection of O₂. This approach reaffirms the previously determined LO-state paradigm as deduced on the basis of photoassembly with long pulses.

Materials and Methods

Preparation of PSII membranes and apo-WOC-PSII

Apo-WOC-PSII samples depleted of Mn^II, Ca²⁺, and three extrinsic proteins (PsbO, PsbP, and PsbQ) were prepared as previously described (30), starting from BBY PSII membrane particles (31) with minor modifications. Briefly, the samples were incubated for 40–90 s in an aminosulfonic buffer (20 mM CHES/NaOH, pH 9.4) and divalent cations (200 mM MgCl₂) at room temperature, followed by dilution with 50 mM MES, 35 mM NaOH, and 400 mM sucrose buffer (pH 6.0) to stop the reaction. This method reversibly inactivates O₂ evolution (32–34) and results in >98% yield of apo-WOC-PSII membranes that are devoid of the three extrinsic WOC proteins as determined by SDS-PAGE (30), exhibit no residual O₂ evolution activity (below ∼5 × 10⁻¹⁵ mol O₂), and efficiently photoassemble the water-splitting system of PSII with high yield equal to 60–70% as measured by the O₂ production rate observed with extrinsic protein-depleted PSII (30).

Photoassembly conditions

Freshly prepared apo-WOC-PSII membranes were diluted to a final concentration of 1 μM apo-WOC-PSII in buffer A (pH 6.0) and 2 mM K₃Fe(CN)₆. Optimal concentrations of inorganic cofactors were added (100 μM MnCl₂, 100 mM CaCl₂, and 10 mM NaHCO₃) (30,35,36), mixed for ∼30 s, and loaded into the photoassembly cell. Samples were incubated in the dark for 10 min before the start of the illumination. Double hits were suppressed by lowering the Mn^II/PSII ratio and by using a 1000-fold higher concentration of CaCl₂, which competes directly with Mn^II for binding to the photo-oxidizable site on apo-WOC-PSII (K_D = 40–50 μM, at low Ca²⁺ (10 mM)) (22,31), whereas all other sites are considerably weaker (36,37). At the Mn^II and Ca²⁺ concentrations used here, there is <1 Mn^II bound per PSII. This produces the fastest photoassembly kinetics, requiring the fewest total flashes to reach the endpoint (n = 400). By contrast, using 10× higher MnCl₂ (1 mM) and lower CaCl₂ (10–20 mM) concentrations results in a considerably slower rate of photoassembly, requiring 1000–3000 flashes and proceeding via a different pathway involving a 150-ms dark process (20,38,39).

O₂ electrode/photoassembly cell

O₂ was measured amperometrically using an in-house-built Clark-type electrode described previously (12,29), employing a thin microcell configuration (sample cavity 0.25 mm thick, 5 μL volume) covered with a thin, low-density polyethylene membrane that allowed a rapid response time (∼100 ms). Integration of the O₂ pulse (for 1.6 s) was done for improved sensitivity (∼1 nM O₂). For stable illumination, a GaAs laser diode was used (665 nm, optical power 500 mW in continuous regime that generates rectangular pulses from 1 μs to 100 ms via an in-house-made current driver).

Standard Kok models

The data were modeled using a standard Kok model (40,41) implemented with a Markov chain (42). The PSII-WOC cycles through five flash-induced intermediates (S₀, S₁ (dark stable), S₂, and S₃, with O₂ formation occurring during the S₃-to-S₀ transition via a transient S₄ state). A standard Kok model includes four possible outcomes of photon excitation (transitions):

• γ (hit), single-turnover advancement to the next stable S state.

• α (miss), no S-state transition occurs.

• β (double hit), double-turnover advancement.

• ε (inactivation), enzyme inactivates, no further S-state transitions occur.

The model assigns identical transition probabilities for all S_i states (average transition probabilities), allowing a unique determination of the system (27,42). In the real PSII, the flash response is S-state dependent (43–47).

Modeling photoassembly

We used both two-step and four-step models (denoted LO and HO, respectively) described by the Markov chain in Fig. 1 (IM_i intermediates) with either the same model with two photoassembly intermediates (IM₁^∗ and IM₂) as in Baranov et al. (30), or four photoassembly intermediates (IM₁^∗, IM₂, IM₃, and IM₄).

Figure 1.

Markov models describing photoassembly processes. Photoassembly (IM_i) and Kok (S_i) intermediates. Parameters of the LO (solid line), and HO (dotted line) models are indicated. Previously identified kinetic steps leading to formation of IM₁^∗ are included (k₁ and k₂) (30) but precede the Markov modeling chain used herein. See Materials and Methods for definitions.

Results and Discussion

Optimizing the Kok parameters

Fig. 2, A and B, display the O₂ oscillation pattern observed in intact PSII membranes that were dark-adapted in media with FeCN for 5 min and 20 min, respectively. The simulated trace offset from the data represents a least-squares fitting using a standard Kok model as described above. Factors contributing to the intensity of O₂ yield oscillations were systematically investigated, and conditions were optimized for each of the following parameters: flash duration, dark preincubation period, and [FeCN]. A full description follows.

Figure 2.

Flash O₂ yields observed for intact PSII membranes after dark incubation in 2 mM FeCN/buffer for (A) 4 min and (B) 20 min. (C) Reassembled PSII (reconstituted apo-WOC). Laser flash length: 60 μs; repetition rate: 0.5 Hz. Simulated traces offset from data use a standard Kok model, with parameters indicated.

Flash duration

Titration of the flash length was performed to optimize the O₂ yield of intact PSII samples. A 60 μs flash was sufficient to saturate the O₂ yield while limiting the double turnovers to 9% (see Figs. 3 A and 3 A; also see optimization in Fig. S1, Fig. S2, Fig. S3, and Table S1 in the Supporting Material). This relatively high value of double turnovers arises in part from the use of FeCN as an electron acceptor during a long dark adaptation, which oxidizes the nonheme in PSII (48). In turn, it serves as an endogenous electron acceptor, causing a higher double-turnover probability (49).

Figure 3.

(A) Laser flash length dependence of O₂ yield from intact spinach PS II membranes in the presence of 2 mM FeCN and flash lengths of 30–500 μs. (B–D) Flash length dependence of the Kok (B) miss parameter (α), (C) single-hit parameter (γ), and (D) double-hit parameter (β). Samples were dark-adapted for 4 min. Flash rate = 0.5 Hz. Parameters for the simple Kok model were described previously (40). All data traces in Figs. 1–7 are an average of three samples.

Preincubation dark time

The intact PSII enzyme requires a dark-adaptation time to allow the S states to relax to their dark-adapted populations and the plastoquinone pool to reoxidize. The S₂ and S₃ states decay to S₁, whereas S₀ remains unaltered on our timescales (41). The decay time required to maximally populate S₁+S₀ corresponds to that required to maximize the O₂ yield on flashes 3 and 4. Here we observed the decay to be first order with a t_1/2 of ∼120 s in the presence of FeCN, in agreement with the literature (50). Consequently, the initial preincubation dark time was typically set to 5 min.

Electron acceptor contribution

FeCN is the preferred electron acceptor for photoassembly because it does not interfere with formation of the photoassembly intermediates and produces the fastest rate and highest yield of photoassembly of all tested electron acceptors (29,35,36), including exogenous quinones (30). The optimum concentration of FeCN for photoassembly experiments was previously determined to be 1–2 mM based on O₂ yield (29,35,36). Our studies confirm that outcome and reveal that this concentration directly increases the initial dark S₂ population, as evidenced by an increase in the double-hit parameter on the first 60-μs flash only. This outcome was identified as being due to oxidation of the nonheme iron (normally ferrous) (48), and was quantified as described next and verified by EPR spectroscopy (Fig. S4).

The effects of two different dark preincubation times on the flash pattern are compared in Fig. 2, A and B. Simulations are shown assuming 0% and 17% population of the initial dark state containing the oxidized nonheme iron (S₁Q_AFe³⁺). In the simulations presented in Fig. 2, the dark state is assumed to be an admixture of centers with lower and higher probability of first-flash double hits, corresponding to S₁Q_AFe²⁺ and S₁Q_AFe³⁺, respectively. From the perspective of the O₂ simulations, the S₁Q_AFe³⁺ state is equivalent to the S₂Q_A state when using saturating flashes. Both states require only one 60-μs flash to progress to the S₃ state. Simulations of the Kok S-state cycle, in which all parameters were varied, showed that the initial dark S₂ population was specifically increased by dark incubation in FeCN on the timescale of oxidation of the nonheme iron (Fig. 2, A and B, and Fig. S1). This effect was included in the simulations.

O₂ oscillation pattern in photoassembled PSII

The required use of excess Mn^II for fast photoassembly kinetics leads to competition for electron donation at other sites without reconstitution of O₂ evolution, and limits the reconstitution yield to ∼70% of active centers (30). This, together with the absence of the extrinsic PSII subunits, results in less efficient turnover (i.e., the miss parameter doubles) (51), which dampens the oscillation pattern of flash-induced O₂ yield compared with intact PSII (Fig. 2 C). A significant O₂ yield is observed on the second flash, and peaks on both the third and fourth flashes. No oscillations are observed beyond the sixth flash even though the steady-state yield is quite high (60–70%). No unique fitting of these data could be achieved using the Kok model employed in this study. It is reasonable to assume that the double-hit probability (β) in these samples should be no higher than in native samples, because β is controlled by the intrinsic optical cross section of PSII and the pulse duration. A fit assuming the same β-value as in native samples (9%) is shown in Fig. 2 C (fitting results are given in Fig. 1).

Choosing an optimal flash regime for photoassembly

Fig. 1 describes the sequence of photoassembly steps at low Mn^II loading of the high-affinity site (12,30). IM₀ represents all the dark precursor states in which only the Mn^II oxidation state is present. IM₁ and IM₁^∗, the first light-induced intermediate, has one Mn^III bound to the high-affinity site and forms reversibly. IM₁^∗ forms from IM₁ as a result of a dark rearrangement process (k₂) that increases the quantum efficiency for advancement upon the next turnover. IM₂, the second light-induced intermediate, can only form after this slow, dark rearrangement step (36,37). IM₁ and IM₁^∗ have been characterized by EPR (22,52,53), whereas IM₂ and other intermediates (IM₃, IM₄, etc.) have not yet been detected by EPR. The dark step (k₂) is rate-limiting and determines the low overall quantum yield (typically <1%) as the back reaction out competes forward assembly (19,35). The lifetime of the k₂ dark step exceeds 10 s in spinach PSII and is instrument limited (19,30). Accurate measurement of k₂ is described next.

k₂ is intrinsically slow

Light flashes of variable duration (60–500 μs), such as those used to progress the intact PSII through the S-state cycle, do not efficiently assemble the apo-WOC. This is due to the low quantum yield of the initial Mn^II photooxidation step that forms IM₁, and competition between the back reaction and the forward k₂ step (19,20). The consequence of this competition on O₂ yield is shown in Fig. 4. Using single-turnover flashes of 60 μs duration (spaced 2–3 s apart), we did not observe O₂ until approximately the 20th flash. The flash number for observation of O₂ rapidly decreased to two flashes as the flash length increased. A significant O₂ signal was not observed on the first flash for any flash duration, consistent with the k₂ step being intrinsically slow.

Figure 4.

Photoassembly of apo-WOC-PSII using long laser flashes of variable length. (A) Flash sequences used in this experiment. Each flash train is comprised of 20 flashes. Each flash is comprised of a series of n shorter flashlettes (500 μs each). (B) The resulting O₂ signal for the 0–50 s range. (C) Enlarged O₂ signal for the 0–20 s range. Flash length in units of 500 μs (and O₂ activation flash number minus the number of flashes before O₂ is observed) for each trace: (a) 1 (14), (b) 2 (8), (c) 4 (6), (d) 8 (4), (e) 16 (2), (f) 32 (2), (g) 64 (2), and (h) 128 (2). Traces are scaled for clarity.

Determination of k₂

We performed a titration of the dark time between the first and second flashes to determine k₂ (Fig. 5) (19,30). The flash sequence used for this titration is shown in Fig. 5 A. Here, a long preflash of 1 s duration (100 × 60 μs flashlettes) was used to form IM₁. Subsequent flashes were clusters of five (single turnover) 60-μs flashlettes separated by 100 ms (10 Hz). We chose to use five flashlettes per cluster to obtain a stronger signal (we obtained the same dark time regardless of the number of flashlettes). These clusters were repeated at a fixed interval of 3 s. Each cluster of five flashlettes is denoted as a flash in Fig. 5, and produced an O₂ signal that was used to monitor the course of photoassembly (Fig. 5 B). Fig. 5 C plots the integrated O₂ yields for flashes 3–8 as a function of the dark time between the long preflash and the second flash. This figure demonstrates that the O₂ yield increased with increasing dark time and approached saturation within the 500 s measurement interval. A simple monoexponential expression gave a good fit with lifetime ∼70 s (parameter b in Y = Y₀ + a(1 − e^−bt)). Parameter b is equivalent to k₂ for the dark rearrangement process. This lifetime provides a more accurate determination than our earlier, instrumentally limited estimate (>10 s) (19).

Figure 5.

Determination of k₂ by titration of the dark time between the first and second flashes used to photoassemble the apo-WOC. (A) Flash sequence used in this experiment. (B) Baseline corrected O₂ data. (C) Integrated O₂ signal (from B) for flashes 3–8 as a function of the dark time length. (Refer to Fig. 4 for sample.)

Optimal flash sequence

From the results in Figs. 4 and 5, we can design an optimal flash sequence for photoassembly of the apo-WOC-PSII. This sequence achieves the fastest rate of assembly given the intrinsic kinetic limitations of the process. The sequence, depicted in Fig. 6 A, includes two key changes from previous photoassembly studies: 1) a long initial preflash of 1 s duration consisting of 100 60-μs flashlettes with interflashlette separation of 10 ms; and 2), a fixed dark time of 240 s before the next (second) flash, of 60-μs duration (single-turnover duration). This long dark interval allows conversion of IM₁ to form IM₁^∗. The 1-s duration preflash does not allow a significant proportion of the sample to advance past IM₁^∗, because the required dark process is much slower (k₂ ∼ 70 s; Fig. 5). We estimate that no more than 1% of the total number of centers can progress past IM₁^∗ during the 1-s-long preflash. This experimental approach is designed so that the system must end up in either of the following two measurable states: 1), centers progress to IM₁^∗; or 2), centers progress past IM₁^∗ to form completely assembled cluster (in S₀ or S₁) due to the higher quantum efficiency of the later photoassembly steps (see next paragraph). Centers that do not progress beyond IM₀ or decay back to IM₀ from IM₁ or IM₁^∗ are not measurable until about the 40th flash (Fig. 6 B) because only inefficient photoassembly takes place during the single 60-μs flashes due to the much slower dark process (k₂ ∼ 70 s).

Figure 6.

Photoassembly of the apo-WOC using the optimal laser flash sequence. (A) Flash sequence showing expected positions for first O₂ release in fully assembled (intact) PSII (photo-assembly LO and HO models from Fig. 1). (B) Comparison of the photoassembly kinetics using the standard flash sequence (60 μs laser flashes spaced by 3 s) or the optimal flash sequence with long preflash (see panel A). The data shown correspond to the integrated O₂ signal/flash.

Few centers progress past IM₁^∗

Data from Tamura et al. (20,54) and others (30,35) established that subsequent photoassembly intermediates (IM₂, and potentially IM₃ and IM₄) form with much higher quantum efficiencies than IM₁^∗. As a consequence, <1% of the centers that do progress past IM₁^∗ during the first preflash are expected to be fully assembled by flash 2. Because there is a long dark time between preflash 1 and flash 2 (240 s), the few centers that assembled during the preflash are expected to decay back to the dark stable S states, S₁ or S₀. This population is accounted for in the simulations.

Photoassembly using the optimal flash sequence

The integrated O₂ signal during photoassembly is shown in Fig. 6 B, using the optimal flash sequence and 1-s-long preflash to populate IM₁ (Fig. 6 A). The data in Fig. 6 B show that the population of fast photoassembling centers (those that are in IM₁^∗ or later) is ∼24% of all PSII centers (compared with the steady-state O₂ yield of intact PS II). After 20 flashes, a sharp decrease in the slope of O₂ recovery occurs, which is an order of magnitude slower than the initial slope (for flashes 20–100). The O₂ recovery of apo-WOC-PSII without the long preflash, but otherwise using the same flash sequence (60-μs standard), is also given in Fig. 6 B. It shows only the slow recovery phase due to centers starting from IM₀. Thus the rapid phase (∼24% of centers) corresponds to centers that efficiently assembled IM₁ (during the preflash) and advanced to IM₁^∗ (in the dark following the 240-s adaption period) and beyond to fully assemble.

Lag in O₂ production during photoassembly

The first 20 flashes (i.e., the first phase) of the photoassembly curve further resolve into two temporally distinct phases when expanded (Fig. 7 A). An initial lag phase producing no O₂ occurs on the first two flashes, followed by a shallow increase on flashes 3–5, followed by a sharp increase on flashes 6–9. This small signal observed on flashes 3–5 cannot be readily interpreted within any plausible photoassembly model. An O₂ response on the third flash would require both double and triple hits to occur to fully assemble the WOC and evolve O₂ from centers starting from IM₁^∗. Because a double hit for the 60-μs flash duration already has a low probability (9%), the probability of multiple double hits is vanishingly small. By contrast, an initial low amplitude phase is expected due to the low leakage of centers past IM₁^∗ during the 1-s-long preflash, as described above. Due to the highly cooperative nature of photoassembly, the majority of such centers are expected to be fully assembled and thus generate an O₂ signal beginning on the third flash and continuing thereafter.

Figure 7.

Initial kinetic phases of integrated O₂ flash yields recovered during photoassembly using the optimal laser flash sequence (see Fig. 6). (A) a: During photoassembly of apo-WOC-PSII; b: after photo-assembly is complete. (B) a: The same as panel A (a), corrected photoassembly curve equal to a – b. (C) Model simulations of the corrected photoassembly curve using the LO-state model (2 IM) and the HO-state model (4 IM). Parameters used for the LO model simulation: α₁ ∼ 20%, α₂ ∼ 70%, IM₁^∗ = 100% (fixed), IM₂ = 0% (fixed). Parameters used for the HO model (unconstrained) simulation: α₁ ∼ 87%, α₂ ∼ 0%, α₃ ∼ 0%, α₄ ∼ 66%, IM₁^∗ = 31%, IM₂ = 21%, IM₃ = 25%, IM₄ = 24%. Parameters used for the corresponding constrained HO model simulation: α₁ ∼ 0%, α₂ ∼ 0%, α₃ ∼ 0%, α₄ ∼ 66%, IM₁^∗ = 100% (fixed), IM₂ = 0% (fixed), IM₃ = 0% (fixed), IM₄ = 0% (fixed). Shared parameters for the LO and both HO simulations: α = 34%, γ = 56%, β = 9% (fixed), ε = 0.02% (fixed), S₀ (Q_AFe²⁺) = 49%, S₁ (Q_AFe²⁺) = 38%, S₁ (Q_AFe³⁺) = 13%.

O₂ from photoassembled centers is produced on the sixth flash

As expected, the magnitude of these leaked centers is small and is ∼2% of the total number of centers. This compares well with our above estimate of ∼1% leaked centers based on the value of k₂ ∼70 s. The calculated O₂ signal arising from this fraction of preassembled centers is shown in Fig. 7 A (denoted the fast-assembled population) using a scaled subtraction of the O₂ oscillation pattern seen for reassembled PSII (Fig. 2 C). This signal was subtracted from the photoassembly curve. The amount subtracted was such that the corrected photoassembly curve had a zero third flash O₂ response (∼2% of the total number of PSII centers, or 8% of the fraction of PSII centers that assembled rapidly). This scaling is not arbitrary because no plausible photoassembly model, based on current understanding, can rationalize a third flash O₂ response of significant magnitude. The corrected photoassembly O₂ curve is given in Fig. 7 B and reveals a sharp O₂ increase beginning on the sixth flash. This number of flashes should therefore correspond to the number of electrons transferred to PSII that are required to make O₂, starting from the state IM₁^∗. A smaller O₂ response is observed on the fifth flash (∼10% of the sixth flash signal). This residual peak is expected due to the finite probability of a double-turnover event for a single flash (i.e., ∼9%) and represents the typical offset in O₂ response seen for O₂ oscillation patterns of the fully assembled oxygen-evolving complex (OEC; see Fig. 2 and below).

Simulation of photoassembly data

Photoassembly steps preceding normal S-state transitions were fitted with the use of transmission parameters, γ_I = 1− (α_i + β) (defined in Fig. 1), whereas S-state transitions used the identical Kok parameters (α and β) obtained from fitting the O₂ oscillation pattern of the reassembled PSII (see Fig. 2 C). The values of all parameters used to simulate the photoassembly curves, using the Markov model shown in Fig. 2, are given in the legend to Fig. 7. Details of the fitting are given in Fig. S6, Fig. S7, and Table S2.

Correction for nonheme iron oxidation

Because the dark time between the first preflash and the second flash is long (∼240 s), the nonheme iron partially reoxidizes during this time. This effect is included in the simulations, effectively as a larger double hit (∼10%) on the second flash (first 60-μs flash). For simulation of the initial phase of the corrected O₂ photoassembly curve (Fig. 7 B), only centers that had progressed to IM₁^∗ after the preflash were considered, because only a small fraction of centers had leaked beyond this point, and these were removed as described above. The states IM₂, IM₃, and IM₄ are expected to have vanishingly small initial populations.

LO versus HO model

Using these values for the initial populations and fit parameters, we simulated the O₂ photoassembly curve using the LO and HO models (Fig. 7 C). Only the two-intermediate model is consistent with the observed data. It fits the threshold, the two limiting slopes of the corrected O₂ photoassembly curve (initial rise and plateau), and the small fifth flash contribution arising from a double hit. It correctly models the first significant O₂ burst on the sixth flash and the decrease in slope without oscillations beyond flash 9. By contrast, the same is not true for the four-intermediate model. In this case, IM₂, IM₃, and IM₄ all need to be significantly populated initially to remotely approximate the photoassembly curve. If fitting to the four-intermediate model is constrained such that the initial IM₁^∗ population dominates, the simulated curve does not agree well with the data (see Fig. 7 C). The simulated curve is offset by two flashes, and large oscillations in O₂ are predicted because the calculated miss parameters must be very small (α ∼ 0). As a consequence, our analysis indicates that the two-intermediate model is the best model for interpreting the photoassembly data. The four-intermediate model would require a significant revision of the current photoassembly literature.

Conclusions

The two-intermediate photoassembly model suggested here requires the Mn oxidation level in S₀ to be (Mn^III)₃Mn^II, with all subsequent S states determined by successive one-electron removals. This result supports the validity of our earlier photoassembly data obtained using long, 30-ms flashes and a constant flash separation for photoassembly (19). In those experiments, we had to obtain seven flashes to observe the first O₂ signal starting from IM₀. This outcome is the expected result because the dominant factor that determines the flash number for initial O₂ detection is the low (∼1%) quantum efficiency of the first photo-oxidation step from IM₀. The use of long, 30-ms pulses overcame this limitation by advancing a sufficient population of centers to IM₁. Additionally, the use of a relatively short interpulse delay in our earlier studies (much shorter than the intrinsic dark rearrangement time, t_1/2 ∼ 70 s) did not allow a significant number of these centers to advance past IM₁^∗ during the long pulses. The ability of the two-intermediate photoassembly model to explain both the present and earlier results measured under very different flash conditions further supports the validity of the underlying model.

The number of holes captured during photoassembly represents the formal Mn oxidation states and is completely independent of reference standards. As noted in the Introduction, this is not the case for some spectroscopic techniques (such as XAS and ENDOR), the results of which have been interpreted as favoring the HO-state paradigm. To illustrate this ambiguity, a recent XANES study assigned the lower oxidation state model to the OEC (55). When properly structured to accommodate the differential kinetics and inefficiencies of the various steps, which may be independently determined, the photoassembly method relies only on the ability to produce one-electron photo-oxidation steps and to count flashes. However, the method does require the ability to remove all Mn ions from the WOC in PSII and to detect O₂ with very high sensitivity. Both of these conditions have been fully met and demonstrated in multiple previous publications and reviews (12,29). Our conclusion that only three photoassembly intermediates can be populated below S₀ in apo-WOC-PSII (IM₂, IM₁, and IM₀ differing by one electron each) is in agreement with the conclusion reached by Messinger et al. (25), i.e., that intact PSII membranes treated with hydroxylamine also produce three sequentially reduced states (S_-3, S_-2, and S_-1). However, the latter treatment produces PSII with all extrinsic subunits present and normal probabilities of misses and double hits, unlike the case with apo-WOC-PSII.

The two-intermediate model for photoassembly determined here specifies that the S₄ state, which is a precursor to O₂ release, should be assigned a formal (Mn^IV)₃Mn^III oxidation level. It implies that the mean oxidation state of the four Mn progresses from 2.75 in S₀ to 3.75 in S₄. No Mn or ligand oxidation to potentially dangerous levels beyond this is required in the natural system.

This clearly has important implications for the catalytic mechanism and suggests, among other things, that a Mn^V oxidation state (invoked in some mechanistic proposals for water oxidation chemistry) is not needed. These results should have important consequences for mechanistic proposals that rely on the possible chemistry of Mn^IV and Mn^III ions bound within the topological heterocubane geometry suggested by EXAFS and crystallographic data. It also reinforces efforts to synthesize bioinspired catalysts based on the principles of a photosynthetic WOC to target the lower Mn oxidation state found herein.