Ammonia binding to the oxygen-evolving complex of photosystem II identifies the solvent-exchangeable oxygen bridge (μ-oxo) of the manganese tetramer

Abstract

The assignment of the two substrate water sites of the tetra-manganese penta-oxygen calcium (Mn₄O₅Ca) cluster of photosystem II is essential for the elucidation of the mechanism of biological O-O bond formation and the subsequent design of bio-inspired water-splitting catalysts. We recently demonstrated using pulsed EPR spectroscopy that one of the five oxygen bridges (μ-oxo) exchanges unusually rapidly with bulk water and is thus a likely candidate for one of the substrates. Ammonia, a water analog, was previously shown to bind to the Mn₄O₅Ca cluster, potentially displacing a water/substrate ligand [Britt RD, et al. (1989) J Am Chem Soc 111(10):3522–3532]. Here we show by a combination of EPR and time-resolved membrane inlet mass spectrometry that the binding of ammonia perturbs the exchangeable μ-oxo bridge without drastically altering the binding/exchange kinetics of the two substrates. In combination with broken-symmetry density functional theory, our results show that (i) the exchangable μ-oxo bridge is O5 {using the labeling of the current crystal structure [Umena Y, et al. (2011) Nature 473(7345):55–60]}; (ii) ammonia displaces a water ligand to the outer manganese (Mn_A4-W1); and (iii) as W1 is trans to O5, ammonia binding elongates the Mn_A4-O5 bond, leading to the perturbation of the μ-oxo bridge resonance and to a small change in the water exchange rates. These experimental results support O-O bond formation between O5 and possibly an oxyl radical as proposed by Siegbahn and exclude W1 as the second substrate water.

In oxygenic photosynthesis, light-driven water splitting is catalyzed by the oxygen-evolving complex (OEC) of the membrane bound, pigment-protein complex photosystem II (PSII). The OEC consists of an inorganic tetra-manganese penta-oxygen calcium (Mn₄O₅Ca) cluster (1–3) and the nearby redox-active tyrosine residue Y_Z (D1-Tyr161) that couples electron transfer from the Mn₄O₅Ca cluster to P680, the photo-oxidant of PSII. The cluster resembles a “distorted chair”, where the base is formed by an oxygen-bridged (μ-oxo) cuboidal Mn₃O₄Ca unit (1) (Fig. 1A). The fourth Mn (Mn_A4) is located outside of the cuboidal unit and is linked via a μ-oxo–bridged ligation (O4) to one of its corners (Mn_B3). A second linkage between the outer Mn and the cube is provided by a fifth oxygen O5. The Mn₄O₅Ca cluster is also held together by six carboxylate ligands and has only one directly coordinating nitrogen ligand, D1-His332 (Fig. 1B).

Fig. 1.

(A) The S-state cycle of the OEC. The crystal structure of the manganese tetramer is also shown, indicating the unusual ligation of O5, equidistant between Mn_A4 and Mn_D1 (1). (B) A representative, unified DFT model of the OEC in the S₂ state (10). The oxygen ligands W1 (pink), W2 (cyan), and O4/O5 (green) were assigned as sites exchangeable with solvent water in the S₁ state (17). (C) The effect of ammonia on the CW-EPR (multiline) signal of the S₂ state [11,29] (Fig. S1 and Table S1).

The OEC cycles through a series of five intermediate states that are known as S states (4) (Fig. 1A): S₀, S₁ (dark stable), S₂, S₃, and S₄ (not yet isolated), where the subscript refers to the number of oxidizing equivalents stored in the OEC through successive electron withdrawals by Y_Z^•. In the 1.9-Å resolution structure, the S state of the cluster was assigned to be S₁ (1). However, this is unlikely as all Mn-Mn, Mn-Ca, and Mn-O/N distances of the crystal structure are ∼0.1 Å longer compared with those determined by extended X-ray absorption fine structure (EXAFS) spectroscopy (5–7). Moreover, the central O5 has unusually long bonds to three Mn ions and to the Ca ion, outside the range seen for model complexes. All these structural details suggest that the Mn ions of the cluster were photoreduced during X-ray data collection, and as such, the X-ray structure represents a nonphysiological, overreduced S state (8, 9). This structural ambiguity can be eliminated by combining the X-ray data with spectroscopic constraints and the introduction of computational modeling. In these unified models, O5 is generally considered to be a μ-oxo bridge between Mn_A4 and Mn_B3 in the S₁ and S₂ states, rendering this unit bis–μ-oxo bridged, and Mn_D1 as five coordinate (10–13) (Fig. 1B).

The S₂ state is readily observed using EPR spectroscopy and related techniques. In this state, the four Mn ions of the OEC are coupled together, resulting in a ground electronic state with one unpaired electron, i.e., effective spin S_eff = 1/2 (14). A distinctive “multiline” EPR spectrum is observed at liquid helium temperature, where the line splittings reflect the coupling of the four ⁵⁵Mn magnetic nuclei to the unpaired electron spin (hyperfine interaction) (Fig. 1C). The unpaired electron of the Mn₄CaO₅ cluster also couples to other magnetic nuclei in the vicinity of the OEC (e.g., ¹⁷O, ¹⁴N/¹⁵N, ¹H/²H), such as those that coordinate the Mn ions, e.g., ¹⁷O, ¹⁴N/¹⁵N, ¹H/²H. These hyperfine couplings are sufficiently small so that the interactions are not directly observed by continuous wave (CW)-EPR spectroscopy. Such interactions can instead be detected using pulse magnetic resonance techniques that probe NMR transitions (15). Such techniques include electron spin echo envelope modulation (ESEEM), electron nuclear double resonance (ENDOR), and electron–electron double-resonance–detected NMR (EDNMR). Each technique is suited to probe specific electron–nuclear interactions of the OEC. For example, exchangeable oxygen sites of the OEC, which are potential substrate sites (16), have been recently studied with W-band EDNMR using ¹⁷O isotopic labeling (17). This methodology is particularly useful as it allows all water-exchangeable sites, including fully deprotonated Mn–μ-oxo bridges, to be observed. It is known from time-resolved membrane inlet mass spectrometry (TR-MIMS) that at least one substrate is bound in all S states and exchanges with bulk water on a seconds timescale (16, 18–20). In the equivalent EDNMR experiment performed in the S₁ state, rapid mixing of PSII with ¹⁷O-labeled water led to the uptake of the ¹⁷O label at three different Mn-ligand sites: (i) as one μ-oxo bridge, most likely O4 or O5; (ii) as a terminal hydroxide ligand, most likely W2, a ligand of Mn_A4; and (iii) as a terminal water ligand, most likely W1, also a ligand of Mn_A4 (17). This latter species dominates the weakly coupled “matrix” envelope, which also has contributions from the Ca-bound waters (W3/W4) and second coordination shell H₂O ligands. These assignments are based on comparison with model compounds and the recent 1.9-Å resolution PSII crystal structure in conjunction with density functional theory (DFT) models (10).

Enzymological studies have indicated that there are at least two independent ammonia-binding sites, SYI and SYII (21, 22) in PSII. Ammonia binding at the SYI site is chloride-concentration dependent, is S-state independent, and results in the inhibition of oxygen-evolving activity (21–23), indicating that SYI likely represents one of the chloride sites identified in the crystal structure (1). In contrast, ammonia binding to SYII is independent of the chloride concentration (21, 22, 24, 25) and does not reduce O₂ evolution. It binds only upon formation of the S₂ state, to be subsequently released at some later point during the S state cycle (after S₃), such that it is not bound upon return to the S₁ state (26). SYII exhibits steric selectivity for small Lewis bases and appears to be only accessible to ammonia.

The ESEEM study by Britt et al. (25) demonstrated that SYII represents a Mn coordination site. Interestingly, the bound ¹⁴NH₃ species displayed a large, rhombic quadrupole coupling (e²Qq/h) of 1.61 MHz, with η = 0.59. From comparison with model compounds, it was suggested that the ammonia is taken up as an amido bridge between either two Mn ions or one Mn ion and the Ca ion, i.e., replacing or modifying one μ-oxo bridge of the complex. Low-frequency FTIR spectroscopy supports this basic hypothesis, identifying a putative Mn-μO-Mn or Mn-μO-Ca vibrational mode (27) lost upon ammonia addition (28).

Here we investigate the binding of ammonia to the OEC, using multiple-pulse EPR techniques and TR-MIMS. It is shown that, although ammonia significantly perturbs all exchangeable Mn-O ligand signals, it only moderately affects the exchange rates of both substrate waters. Instead of it displacing a μ-oxo bridge, our data support a mechanism in which ammonia modifies the μ-oxo bridge by displacing a water ligand trans to the bridge position, specifically the water ligand W1 trans to the μ-oxo bridge O5. Broken symmetry (BS)-DFT calculations, which model this displacement, quantitatively reproduce all spectroscopic observables. Together, our data show that W1 is not a substrate binding site, but instead favor O5 as one of the two substrate waters.

Results and Discussion

Ammonia Binds to the OEC Without Significantly Changing Its Electronic Structure.

PSII isolated from the thermophilic cyanobacteria Thermosynechococcus elongatus was used throughout this study. Ammonia was added to PSII samples in the S₁ state, which was advanced before the EPR measurements to the S₂ state by low-temperature (180 K) illumination with visible light. In agreement with the literature, this resulted in an unperturbed S₂-state multiline EPR signal similar to the “no NH₃” spectrum shown in Fig. 1C (26). Subsequent annealing of the sample to 260 K for 30 s led to the induction of the ammonia-modified multiline form (NH₃ spectrum, Fig. 1C) (26). No change was observed in the background cytochrome c550/b559 signals upon annealing the sample at 260 K. The ammonia-modified S₂-state multiline signal is also centered about g ≈ 2.0, spread over the 250- to 430-mT field range and characteristically contains more hyperfine peaks than the control sample (at least 24 vs. 20; see Fig. 1C) (24, 29). Simulations of the EPR and ⁵⁵Mn-ENDOR spectra using the spin Hamiltonian formalism are given in Fig. S1 and Table S1.

Nitrogen ligands of the OEC can be readily detected using ESEEM. In this type of pulse EPR experiment, the EPR signal intensity (spin echo) is recorded as a function of the time intervals between the successive microwave pulses. Signal intensity modulations arise from the weak coupling of the electron spin with nearby magnetic nuclei such as ¹⁴N [I(¹⁴N) = 1]. Both native ¹⁴N-PSII and universally labeled ¹⁵N-PSII [I(¹⁵N) = 1/2] were measured. X-band three-pulse ESEEM experiments of ¹⁴NH₃-containing PSII illuminated at 180 K and subsequently annealed at 260 K (see above) are shown in Fig. 2 A and B. A new modulation, consistent with ¹⁴NH₃ binding to the Mn₄O₅Ca cluster, is observed in the light-minus-dark difference spectra only after the 260-K annealing step (25). The bound ¹⁴NH₃ species displays three sharp nuclear-quadrupole lines (N.Q.L.) at 0.5, 1.0, and 1.5 MHz in the Fourier-transformed spectrum (Fig. 2B). Spin Hamiltonian simulations of the lineshape are shown in Fig. 2 A and B as dashed red lines and the fitted parameters (Table 1, Fig. S2, and Table S2) for T. elongatus are similar to those reported in the earlier higher-plant study (25). The relatively small magnitude of the hyperfine coupling supports the assignment of ¹⁴NH₃ as a ligand to one of the Mn^IV ions as opposed to the Mn^III ion of the S₂ state. This is because the Mn^IV ions carry a lower spin density (spin projection) than the Mn^III ion and thus their ligands are expected to display smaller effective hyperfine couplings (30).

Fig. 2.

(A) X-band three-pulse ESEEM traces measured at the center of the S₂-state multiline signal (Fig. 1, B₀ = 333 mT, microwave frequency = 9.4 GHz). The data represent annealed-minus-dark difference traces collected on ammonia (¹⁴NH₃)-treated ¹⁴N-PSII (black) and (¹⁴NH₃)-treated ¹⁵N-PSII (blue). The traces shown in A were measured with an interpulse spacing τ of 136 ns. Additional data traces using the τ-values 152 ns, 168 ns, and 184 ns are shown in Fig. S2. (B) Fourier transform (FT) of the X-band time domain data. N.Q.L. identifies the nuclear-quadrupole lines caused by the coupling of the OEC with the added ¹⁴N (I = 1). The spectrum shown represents the sum of the FT of the four ESEEM traces measured using different τ-values (136–184 ns) to minimize spectral artifacts. (C) Q-band three-pulse ESEEM traces measured at the center of the S₂-state multiline signal (B₀ = 1.22 T, microwave frequency = 34.0 GHz). The data represent light-minus-dark and annealed-minus-dark difference spectra of native ¹⁴N-PSII (black) and ammonia (¹⁴NH₃)-treated ¹⁴N-PSII (blue) respectively. The time domain data were measured using an interpulse spacing τ of 260 ns. Additional data traces using τ-values of 240 ns and 300 ns are shown in Fig. S2. (D) Corresponding FT of the data traces presented in C. S.Q. and D.Q. identify single-quantum and double-quantum transition lines from the coupling with ¹⁴N-His332. The red dashed lines superimposing the data represent a simulation using the spin Hamiltonian formalism (SI EPR Theory/Simulations, Fig. S2, and Table S2). The label N.Q.L. identifies the quadrupole lines observed in the X-band ¹⁴N-ESEEM spectrum.

Table 1.

Experimentally determined ESEEM and EDNMR spin Hamiltonian parameters: Comparison with calculated magnetic resonance parameters from DFT

	Hyperfine couplings |Aiso|/MHz
		Exchangable ligands, 14N/17O
Experiment/Theory	His332*, 14N*	W1 17O/NH3 14N*	W2 17O*	O5 17O†
DFT
Native	4.8	1.7	5.2	17.4
+NH3‡	5.2	1.5	4.3	12.2
Δ§	0.4	—	−0.9	−5.2
Δ/%§	8.3	—	−17	−30
Experiment
Native	7.2	1.4	4.5	9.7
+NH3*	7.2	2.4	3.1	6.5
Δ§	0.0	—	1.4	3.2
Δ/%§	0.0	—	−31	−28

^*Calculated (projected) BS-DFT hyperfine values directly comparable to experiment (/MHz).

^†Calculated (raw) BS-DFT hyperfine values are not directly comparable to experiment; the percentage change (Δ) due to ammonia binding can, however, be compared.

^‡NH₃ replacing W1.

^§Δ = difference between native and +NH₃ samples.

In contrast to the X-band measurements, at Q-band, no difference is seen between the control and the ¹⁴NH₃-treated sample (Figs. 2C and 2D). Instead, the observed ESEEM modulation is dominated by a ¹⁴N hyperfine coupling assigned to the D1-His332 ligand of Mn_D1 (31). At Q-band the histidine ¹⁴N signal is at or near the cancellation condition and as such displays a maximal ESEEM response (30, 31). As a consequence of the D1-His332 ¹⁴N coupling matching the cancellation condition, the signal of the bound ammonia in comparison is suppressed at Q-band and no direct information on its binding site can be obtained in this way. However, the ¹⁴N histidine ESEEM signal, which resolves multiple spectral lines at 0.6, 2.0, and 7.3 MHz, and 14.8 MHz representing single-quantum (SQ) and double-quantum (DQ) transitions, respectively, can be used as spin probe reporting on the electronic structure and the oxidation state of Mn_D1. Spin Hamiltonian simulations of the lineshape of this signal are shown in Fig. 2 C and D (dashed red lines). The parameters used (Table 1, Fig. S2, and Table S2) are similar to those reported earlier by Stich et al. (31) for PSII purified from Synechocystis sp. 6803. The relatively large magnitude of the D1-His332 ¹⁴N coupling suggests that it is ligated to the only Mn^III ion in the S₂ state, i.e., to the Mn ion that carries the largest spin density/spin projection (11, 30, 31). As the D1-His332 signal does not change upon the addition of ammonia, the oxidation state and ligand field of the Mn_D1 ion cannot change. Thus, the binding site of ammonia at the manganese tetramer is unlikely to be proximal to the Mn_D1 but instead is distal to it, consistent with NH₃ binding to a Mn^IV ion. It is also noted that protons in the vicinity of the OEC can be readily detected using Q-band ¹H-ENDOR (SI EPR Theory/Simulations and Fig. S3A). The addition of NH₃ does not change the width of the signal envelope, which has been assigned to the protonated oxygen ligands on Mn_A4 (32). The absence of a large proton coupling suggests ammonia does not replace one of the μ-oxo bridges of the OEC, excluding this previous suggestion for its binding site (25).

Ammonia Perturbs All Exchangeable Oxygen Ligands of the Manganese Tetramer.

EDNMR (32), a pump–probe technique, which employs two independent microwave pulses, has recently been shown to be the magnetic resonance method of choice for the detection of (¹⁷O) ligands of metallocofactors, such as the Mn₄O₅Ca cluster of the OEC. In this experiment, the EPR signal is monitored at a fixed microwave frequency matched to the resonator (probe pulse). Before the detection sequence, a microwave pulse of varying frequency, termed the high turning-angle (HTA) pulse, is applied (pump pulse). The pumping (HTA) pulse drives spin-forbidden transitions where both the electron spin and the nuclear spin state change (|Δm_s| = 1, |Δm_I| = 1). Magnetic nuclei appear as doublets centered about their characteristic (Larmor) frequencies; i.e., ν_N(¹⁴N) = 10.46 MHz and ν_N(¹⁷O) = 19.6 MHz at 3.4 T. As described in Rapatskiy et al. (17), in S₂-state PSII samples resuspended in H₂¹⁷O-containing buffer, two structured signal envelopes are observed centered at the Larmor frequency and at twice the Larmor frequency of ¹⁷O. These two signal envelopes correspond to SQ and DQ transitions of exchangeable oxygen ligands of the manganese tetramer (Fig. 3A). Three components were identified (17): (i) a large coupling, assigned to a μ-oxo bridge from comparison with model complexes, most likely O4 or O5; (ii) an intermediate coupling, assigned to the terminal oxygen ligand of Mn_A4 (W2); and (iii) a weak coupling (unsplit matrix line), representing the second terminal oxygen ligand of Mn_A4 (W1) but also including contributions from W3 and W4. The couplings of W1 and W2 are proposed to differ due to their protonation state. In DFT models, W2 is preferentially a hydroxo ligand in the S₂ state, whereas W1 represents a water ligand (10, 13). In comparison with the hydroxo ligand (W2), the water ligand (W1) is expected to have a much smaller coupling, owing to its additional covalent bond to hydrogen, which weakens its bond to the Mn^IV ion.

Fig. 3.

(A) W-band ¹⁷O-EDNMR spectra of native and ¹⁴NH₃-treated ¹⁴N-PSII samples (17). The black line represents the data; the red dashed line represents the total simulation. Fitted isotropic hyperfine values are listed in Table 1. A complete list of parameters is given in Table S3. The colored traces represent the four components of the fit: the ¹⁴N of D1-His332, blue; the strongly coupled ¹⁷O species, green; the intermediately coupled ¹⁷O species, orange; and the weakly coupled ¹⁷O species, pink. (B) TR-MIMS traces monitoring substrate exchange in the S₂ state at pH 7.6 in the presence of either 100 mM NH₄Cl (red triangles) or 100 mM NaCl (black squares). The lines represent biexponential (³⁴O₂, Left) and monoexponential (³⁶O₂, Right) fits. NH₄Cl: k_f = 52 s⁻¹, k_s = 2 s⁻¹. NaCl: k_f = 38 s⁻¹, k_s = 3 s⁻¹.

Ammonia binding to the OEC modifies the ¹⁷O signal profile (17) (Fig. 3A and Fig. S4). The widths of the ¹⁷O single- and double-quantum envelopes narrow by ∼30%, and the splitting of the two outer single-quantum satellite peaks, which corresponds to the large coupling (μ-oxo bridge), becomes unresolved. Additionally, the sharp central matrix line (W1) appears to be of lower intensity.

The intermediate coupling is best resolved in the double-quantum region, owing to spectral congestion in the single-quantum region. Ammonia binding modifies the intermediate-coupling feature, narrowing it by ∼1–2 MHz. Furthermore, the whole double-quantum region becomes more symmetric compared with the spectra of the control sample (Fig. 3A, Fig. S4, and Table S3); this asymmetry was previously thought to be due to the matrix signal (17). The reduced asymmetry in the double-quantum region is taken as additional evidence that the matrix component is reduced by ammonia binding to the OEC, which is further supported by considering the power dependence of the EDNMR signal (for further details see SI EPR Theory/Simulations and Fig. S4). Thus, ammonia likely displaces W1, perturbing W2 and the μ-oxo bridge signal. It is also noted that the water ligands of the Ca²⁺ ion (W3, W4) were measured independently using ¹⁷O-Mims ENDOR, and no change was observed; ergo, W3 and W4 are not displaced by ammonia (Fig. S3B).

The Site of Ammonia Binding: A Mechanism for the Perturbation of the μ-oxo Bridge.

The binding of ammonia as a terminal ligand to Mn_A4 instead of W1 could potentially modify the hyperfine coupling of the μ-oxo bridge O5 via the trans effect. To test whether this rationale can quantitatively explain the observed spectral changes, DFT calculations were performed using previously reported S₂-state OEC models consistent with geometric, thermodynamic, and spectroscopic parameters (10, 17). Calculated EPR parameters (33, 34) of both the W1- and the NH₃-containing structure are shown in Table 1 and Tables S2–S4. This single-ligand substitution quantitatively reproduces all experimental observables, including the ¹⁴N hyperfine and quadrupole couplings of the bound ammonia, the ∼1-MHz decrease in the ¹⁷O hyperfine coupling of the terminal hydroxide (W2), and the ¹⁴N-His332 hyperfine coupling and its insensitivity to ammonia addition. Although it is currently not possible to reliably calculate projected hyperfine coupling constants for bridging ligands, as this is yet to be calibrated in model systems, it is possible to compare the raw BS-DFT values to ascertain the effect of the ammonia ligand. The calculations show that ammonia binding at the W1 site selectively perturbs the O5 μ-oxo bridge. The observed change in coupling is again quantitatively reproduced, with a decrease in the hyperfine coupling of O5 by 30%, the same as seen for the μ-oxo bridge species using EDNMR. All other μ-oxo bridge couplings are calculated as being very similar for the H₂O- and the NH₃-containing structure, including the O4, which actually increases upon NH₃ binding, excluding it as the exchangeable bridge. The only exception is O1, where the calculated raw BS-DFT hyperfine has a large percentage change; however, the absolute magnitude of the O1 hyperfine coupling is small and the absolute change is only 0.55 MHz (Table S4).

From this, we can confidently assign the site of NH₃ binding to the W1 coordination site of Mn_A4. A comparison of the different geometries of the two BS-DFT structures (with and without NH₃) shows a small elongation of the Mn_A4-O5 bond of 0.02 Å upon NH₃ substitution, as expected. This bond lengthening reduces the Mn_A4 to O5 spin polarization and consequently the overall spin density on O5, resulting in the 30% decrease in the observed ¹⁷O hyperfine value. This change should also modify the vibrational mode of the O5 bridge, consistent with low-frequency IR spectroscopic results reported in ref. 28. Indeed, vibrational frequencies computed for the optimized structures of the two models indicate that a Mn_A4–O5 stretching mode along the Mn_A1-Mn_D1 vector at 644 cm⁻¹ shifts upon NH₃ binding to 617 cm⁻¹ with concomitant ∼50% loss in intensity, consistent with experimental observations.

W1 Is Not a Substrate Water.

TR-MIMS, a mass spectrometric pump–probe technique, employing H₂¹⁸O labeling, provides important information regarding the binding of the substrate to the catalyst during the S-state cycle (16). This experiment involves poising the OEC in the desired S state with light flashes and the subsequent rapid injection (t_1/2 = 3 ms) of isotopically labeled water (H₂¹⁸O), followed by successive light flashes to release the product O₂. By varying the incubation time of the sample in labeled water, the extent to which ¹⁸O is incorporated into the product O₂ is varied, allowing the determination of substrate water exchange rates with the bulk solvent. These experiments have established that the two substrate waters exchange with different rates that also vary independently with the S states. Thus, the two substrates bind at chemically distinct sites. The slowly exchanging substrate (W_s) is bound throughout the S-state cycle, whereas the fast-exchanging substrate (W_f) is bound latest in the S₂ state (16, 18, 20, 35, 36).

TR-MIMS data monitoring the fast and slow substrate exchange in the S₂ state at pH 7.6 in the presence of 100 mM NH₄Cl (red) or 100 mM NaCl (black) are shown in Fig. 3B. If ammonia displaces a substrate, a major slowing or even abolishment of one exchange rate is expected. This is not observed experimentally: The exchange rates of W_s and W_f with bulk water lie within factors of 1.5 in the presence and absence of NH₄Cl. This demonstrates that ammonia does not displace a substrate water, but instead slightly modifies exchange rates by binding in their vicinity. Thus, the combined EPR and TR-MIMS data exclude W1 as a substrate site. Importantly, these results exclude O-O bond mechanisms that involve both terminal Mn oxygen ligands on Mn_A4, i.e., the Kusunoki-type mechanism (37).

This model also provides a simple rationale for ammonia binding/release during the S-state cycle (24, 26). In the lower S states (S₀, S₁), Mn_A4 is usually considered to be in the Mn^III oxidation state and is thus potentially five-coordinate, with W1 being only a weakly associated ligand. It is noted that DFT calculations support assigning the Jahn–Teller axis of Mn_A4^III along the W1/O5 axis (13, 38, 39). As such, ammonia does not bind in these S states as its nominal binding site is preferentially unoccupied. Upon formation of the S₂ state, the Mn_A4 is oxidized to +IV and is required to be six-coordinate, thus allowing ammonia to bind to the OEC. As NH₃ is a better (more tightly bound) ligand to Mn^IV than water in the S₃ and presumably the S₄ states, ammonia is unlikely to be released until after the O-O bond formation step, at which point Mn_A4 returns to its +III oxidation state and is again five-coordinate.

O5 Represents a Substrate Site.

The slow rate of exchange of W_s and the observation that the rate is S-state (i.e., Mn oxidation state) dependent suggest that W_s represents a Mn–oxygen ligand (16, 18, 20, 35). In Rapatskiy et al. (17), three exchangeable Mn-O ligands were identified, and thus, all three potentially represent W_S: W1, W2, and a μ-oxo bridge, either O4 or O5. As described above, the ammonia effect excludes W1 and demonstrates that O5 (and not O4) represents the exchangeable bridge. Thus, we can now reduce the number of possible candidates for W_s to only two: W2 and O5.

A series of studies are converging with regard to the role of O5 instead of W2 as the W_s substrate site. Critical to this assignment has been the recent demonstration that one of the μ-oxo bridges (shown here to be O5) exchanges rapidly with bulk water (17), with an exchange rate consistent with mass spectrometry measurements (16, 18–20) and over 1,000 times faster than that seen in synthetic model systems (40). A rationale for this enhanced exchange rate was recently provided by the theoretical study of Pantazis et al. (13), where it was shown that O5 has a flexible coordination, acting as either a μ-oxo linkage to the outer Mn (Mn_A4) or a vertex of the cuboidal unit proper. Similarly, the OEC appears to contain several pathways for internal oxygen exchange between terminal water ligands to Ca or Mn, which may allow a calcium-ligated bridge such as O5 to exchange rapidly (41).

Site-selective perturbations such as protein mutagenesis provide further support for the assignment of O5 over W2 as W_s. The replacement of Ca with Sr strongly enhances the exchange rate of W_s (36). As O5 (not W2 or O4)) is a ligand to Ca/Sr (1), this result is readily understood (36, 41). Similarly, the mutation of the D1-Glu189 (bridge between Mn_D1 and Ca), the D1-Asp170 (bridge between Mn_A4 and Ca), and the CP43-Glu354 (bridge between Mn_B3 and Mn_C2) all enhance the rate of W_s exchange (20, 42, 43). As O5 is a ligand to Mn_A4, Mn_B3, and Mn_D1 (owing to its two isoenergetic forms in the S₂ state and potentially the S₃ state) (1, 13), the observed perturbation in the exchange rate seen in these mutants is again readily explained.

An O-O Bond Formation Mechanism Involving O5.

The O-O bond reaction can proceed via either (i) a nucleophilic attack of O5 by a nearby substrate, i.e., between the μ-oxo bridge (O5) and a terminal hydroxide/Ca²⁺-bound water (W3), or (ii) an oxo/oxyl radical coupling of O5 and an as yet unidentified water (possibly previously bound to Ca/Mn_A4) that is located proximal to O5 in the S₃/S₄ states, as proposed by Siegbahn (12) (see also refs. 41, 44).

Of the two pathways to O-O bond formation, only the nucleophilic attack mechanism has been previously observed in Mn model systems, albeit with a much slower rate than seen for the OEC (45, 46). In contrast, the radical coupling mechanism has no precedence in Mn model chemistry, but has been demonstrated as an efficient O-O bond formation pathway in second-row transition metal catalysts; see, for example, the ruthenium (Ru-Hbpp) dimer complex (47). This latter mechanistic route has been demonstrated in silico by Siegbahn as the most efficient O-O bond formation pathway (12).

A unique feature of the oxo/oxyl mechanism proposed by Siegbahn is that the second, fast-exchanging water substrate (W_f) binds to the OEC late in the S-state cycle, a conclusion supported by FTIR difference spectroscopy (48). This additional substrate from the bulk binds to the open coordination site of Mn_D1 as a water/hydroxide ion in the S₃ state, forming an oxyl radical in the S₄ state (Fig. 4) (12). Superficially, this appears to be in disagreement with TR-MIMS measurements, which suggest that W_f has a similar affinity in the S₂ state to that in the S₃ state, requiring it to be in a chemically similar environment in both states. The inherent structural flexibility of the OEC provides a rationale for this problem, suggesting a second binding sequence for W_f, reconciling the oxo/oxyl mechanism with the observation that W_f is already bound in the S₂ state. Instead, of binding directly to Mn_D1, the second substrate could bind to the solvent-accessible outer Mn_A4 ion, as the open coordination site of the complex can exist at either Mn_A4 or Mn_D1 via the facile movement of the O5/W_s bridge. In this instance, the terminal hydroxide ligands of Mn_A4 in the S₃ state (W2 and W_f) would be indistinguishable, owing to rapid interchange, and could be considered to represent the same species. O5/W_s, which upon proton movement from W_f returns to the putative S₃ state proposed by Siegbahn, represents in this tautomeric structure a terminal hydroxide ion bound to a Mn^IV ion. This ligand motif is considered to exchange with bulk solvent on a seconds timescale in Mn model complexes. The Mn_D1-bound oxygen is, however, within a more hydrophobic pocket compared with the Mn_A4-bound oxygen, which explains why two exchange rates are still observed for the two putative Mn^IV-O(H) substrate ligands in the S₃ state. The hydrophobic region about Mn_D1 potentially acts to stabilize the subsequent ligand oxidation of the Mn_D1-bound oxygen to an oxyl radical upon advancement to the S₄ state.

Fig. 4.

(Left) Site for NH₃ binding to the OEC poised in the S₂ state. NH₃ displaces W1, a water ligand of the outer Mn_A4 (a Mn^IV ion in the S₂ state), which slightly affects the binding strength of the oxo-bridge O5, which is trans to this position. (Right) O-O bond formation mechanisms consistent with this study (see main text): (I) a nucleophilic attack of O5 by a nearby substrate; (II) an oxo/oxyl radical coupling of O5 and an as yet unidentified additional water marked W_f (possibly W2). Mn, purple; Ca, yellow; N, blue; O, red; and substrate O, green.

Thus, a concerted tetramer mechanism involving O5, which uses the unique geometry of the Mn₄O₅Ca cluster to bind and position the two substrates, provides a rationale for the substrate exchange phenomenology described in the literature. The sequential uptake of the two substrates ensures that simultaneous binding of both substrates does not occur in the resting states (S₀, S₁) of the catalyst, which is likely critical for efficient (high turnover frequency) and highly selective O₂ product formation.

Materials and Methods

¹⁴N- and ¹⁵N-PSII core complex preparations from T. elongatus were isolated as described earlier (49, 50) with modifications described in SI Materials and Methods. The S₂ state was generated by short, white-light illumination (5 s) with a tungsten lamp at 185–200 K.

EPR measurements were performed at X-band using Bruker ELEXSYS 500 and 580 spectrometers, at Q-band using a Bruker ELEXSYS E580 spectrometer, and at W-band using a Bruker ELEXSYS E680 spectrometer. X-band CW and pulse EPR measurements were performed at 8.6 K and 4.2 K, respectively. Q- and W-band pulsed EPR measurements were performed at 4.8–5.2 K. Experimental settings were as reported in refs. 11 and 17 and in Figs. S1–S3.

TR-MIMS experiments were performed at 20 °C using a modified membrane-inlet cell connected to a magnetic sector field isotope ratio mass spectrometer. Further details regarding experimental procedures and data analysis are described in SI Materials and Methods and refs. 16, 19, and 35.

Density functional theory calculations of geometries, exchange coupling constants, vibrational frequencies, and EPR parameters were performed similarly to those described in refs. 10 and 17. Computational details and Cartesian coordinates of the optimized structures are given in SI Materials and Methods and Table S5, respectively.