Bicarbonate-controlled reduction of oxygen by the Q_A semiquinone in Photosystem II in membranes

Significance

In Photosystem II (PSII), O₂ reduction by ${\text{Q}}_{\text{A}}^{•-}$ is often discussed but has not been demonstrated. Here, we show in PSII membranes that ${\text{Q}}_{\text{A}}^{•-}$ can reduce O₂ to superoxide, but only when bicarbonate is absent from its binding site on the nonheme Fe²⁺. Bicarbonate’s role in PSII was recently shown to involve a regulatory/protective redox-tuning mechanism linking PSII function to CO₂ concentration. A key aspect is the presence of stable ${\text{Q}}_{\text{A}}^{•-}$ causing release of bicarbonate from its site on Fe²⁺. Here, we show that under these conditions, O₂ binds to the empty site on the Fe²⁺ and is reduced by ${\text{Q}}_{\text{A}}^{•-}$. This unexpected reaction may be a further indication of cross-talk between the regulation of PSII and CO₂ fixation.

Abstract

Photosystem II (PSII), the water/plastoquinone photo-oxidoreductase, plays a key energy input role in the biosphere. ${\text{Q}}_{\text{A}}^{•-}$, the reduced semiquinone form of the nonexchangeable quinone, is often considered capable of a side reaction with O₂, forming superoxide, but this reaction has not yet been demonstrated experimentally. Here, using chlorophyll fluorescence in plant PSII membranes, we show that O₂ does oxidize ${\text{Q}}_{\text{A}}^{•-}$ at physiological O₂ concentrations with a t_1/2 of 10 s. Superoxide is formed stoichiometrically, and the reaction kinetics are controlled by the accessibility of O₂ to a binding site near ${\text{Q}}_{\text{A}}^{•-}$, with an apparent dissociation constant of 70 ± 20 µM. Unexpectedly, ${\text{Q}}_{\text{A}}^{•-}$ could only reduce O₂ when bicarbonate was absent from its binding site on the nonheme iron (Fe²⁺) and the addition of bicarbonate or formate blocked the O₂-dependant decay of ${\text{Q}}_{\text{A}}^{•-}$. These results, together with molecular dynamics simulations and hybrid quantum mechanics/molecular mechanics calculations, indicate that electron transfer from ${\text{Q}}_{\text{A}}^{•-}$ to O₂ occurs when the O₂ is bound to the empty bicarbonate site on Fe²⁺. A protective role for bicarbonate in PSII was recently reported, involving long-lived ${\text{Q}}_{\text{A}}^{•-}$ triggering bicarbonate dissociation from Fe²⁺ [Brinkert et al., Proc. Natl. Acad. Sci. U.S.A. 113, 12144–12149 (2016)]. The present findings extend this mechanism by showing that bicarbonate release allows O₂ to bind to Fe²⁺ and to oxidize ${\text{Q}}_{\text{A}}^{•-}$. This could be beneficial by oxidizing ${\text{Q}}_{\text{A}}^{•-}$ and by producing superoxide, a chemical signal for the overreduced state of the electron transfer chain.

Photosystem II (PSII) is the water/plastoquinone photo-oxidoreductase that uses the energy of red light to drive the oxidation of water and the reduction of plastoquinone (PQ). This process leads to the formation of a radical pair on the chlorophyll P_D1^•+ and pheophytin Pheo_D1^•−. The electron hole is transferred from the chlorophyll cation radical, P_D1^•+, via a redox-active tyrosine (Tyr_Z) to the Mn₄O₅Ca cluster, where, after four sequential photochemical turnovers, two water molecules are oxidized on the luminal side of the membrane. The electron is transferred from the pheophytin anion radical (Pheo_D1^•−) via a nonexchangeable plastoquinone (Q_A), which acts as a one-electron relay, to an exchangeable plastoquinone (Q_B), the terminal electron acceptor (1–3). Q_B accepts two electrons and takes up two protons from the aqueous phase to form Q_BH₂ before it is released as a PQH₂ into the PQ/PQH₂ pool in the membrane (4). From here, the PQH₂ delivers electrons to cytochrome b₆f complex, the subsequent enzyme of the photosynthetic electron transfer chain.

As photochemical charge separation is intrinsically a one-photon-per-electron process, two photochemical turnovers are required to form PQH₂, with the one-electron reduced semiquinone, ${\text{Q}}_{\text{B}}^{•-}$_, formed as an intermediate. Semiquinones can be very reactive, but ${\text{Q}}_{\text{B}}^{•-}$ is rendered thermodynamically stable by the environment provided by its binding site on the D1 protein (4). The electron on ${\text{Q}}_{\text{B}}^{•-}$ is still able to back-react via ${\text{Q}}_{\text{A}}^{•-}$ with the S₂ and S₃ states, the two semistable intermediates of the water oxidizing enzyme (5). This back-reaction occurs by thermal repopulation of the intermediate radical pairs between ${\text{S}}_{2/3}{\text{Q}}_{\text{B}}^{•-}$ and ${\text{P}}_{\text{D1}}{}^{•+}{\text{Pheo}}_{\text{D1}}{}^{•-}$. ${\text{P}}_{\text{D1}}{}^{•+}{\text{Pheo}}_{\text{D1}}{}^{•-}$ can either repopulate P*, which can decay radiatively (5, 6), or it undergoes direct charge recombination, forming the chlorophyll triplet state, ³P₆₈₀ (7–9). As expected for a long-lived chlorophyll triplet state, it reacts with dioxygen to form a highly reactive singlet oxygen species, ¹O₂, which causes photodamage (10, 11).

On the electron acceptor-side of PSII (Fig. 1), the reduced forms of the electron transfer cofactors can potentially reduce oxygen, forming superoxide radical (${\text{O}}_2^{•-}$) by a one-electron transfer to O₂. This would be a wasteful leak of electrons, and the ${\text{O}}_2^{•-}$ formed could be damaging and act as redox signal within the cell. During forward electron transfer, ${\text{Q}}_{\text{A}}^{•-}$ has a half-time of ∼1 ms (1) and is unlikely to react with O₂. However, when, for example, Q_B and the PQ pool are reduced, ${\text{Q}}_{\text{A}}^{•-}$ is longer lived, and it is then more likely to reduce oxygen. Oxygen reduction by ${\text{Q}}_{\text{A}}^{•-}$ is often discussed in the literature (12–16), but clear experimental evidence for this reaction has not been reported. Here, using PSII membranes, we have directly tested for electron transfer from ${\text{Q}}_{\text{A}}^{-}$ to O₂ and for formation of superoxide. We show the reaction does occur, and we characterized the reaction in terms of the O₂ binding site using biochemical and computational approaches. The results indicate the reaction constitutes an unexpected regulatory mechanism involving bicarbonate.

Fig. 1.

The quinone-iron complex in PSII. The nonheme iron shown as a red sphere and the bicarbonate in pink. The amino acid side chains involved in the hydrogen-bonding network are shown, green if belonging to D1 and cyan if belonging to D2. Data from the 1.9-Å crystal structure (PDB ID: 3WU2) (41).

Results

${\text{Q}}_{\text{A}}^{•-}$ Oxidation by O₂.

Fig. 2A (orange trace) shows the stability of ${\text{Q}}_{\text{A}}^{•-}$ generated in low-PQ PSII membranes (SI Appendix, Fig. S1) under anaerobic conditions and with the concentration of ${\text{Q}}_{\text{A}}^{•-}$ monitored using chlorophyll fluorescence (17, 18). To generate this state, a dark-adapted sample was illuminated with three saturating flashes in the presence of the exogenous electron donor, NH₂OH, at 250 µM, a concentration that was sufficient to donate electrons to the Mn cluster but low enough to avoid overreduction and loss of the Mn cluster in the majority of centers (SI Appendix, Materials and Methods, and, for example, ref. 19). This treatment resulted in the trapping of stable ${\text{Q}}_{\text{A}}^{•-}$ in ∼50% of the centers, based on a comparison to maximum fluorescence. The decay of ${\text{Q}}_{\text{A}}^{•-}$ in the other centers is mainly due to 1) forward electron transfer in the millisecond timescale (18) as a result of the incomplete removal of the PQ pool and 2) charge recombination (${\text{S}}_2{\text{Q}}_{\text{A}}^{•-}$ and ${\text{S}}_3{\text{Q}}_{\text{A}}^{•-}$) in the seconds timescale (20) in centers where electron donation from NH₂OH was insufficient to eliminate these back reactions. In the absence of O₂, the stable fluorescence signal from ${\text{Q}}_{\text{A}}^{•-}$ showed a slow decay with a half-time of ∼600 s. A decay rate for ${\text{Q}}_{\text{A}}^{•-}$ with similar kinetics has previously been correlated with the decay of Tyr_D• and attributed to charge recombination between these two states (21), which could explain the slow decay of ${\text{Q}}_{\text{A}}^{•-}$ observed here.

Fig. 2.

Kinetics of ${\mathrm{Q}}_{\mathrm{A}}^{•–}$ reoxidation by O₂. Fraction of ${\text{Q}}_{\text{A}}^{•-}$ obtained after three saturating flashes of a degassed sample of PSII membrane fragments (60 nM PSII cores) in the presence of 250 µM hydroxylamine in a 40 mM MES buffer, 5 mM MgCl, pH 6.5. The data in A, C, and D were fitted to Eq. 1. (A) Orange symbols represent the decay in the absence of added oxygen, and the data were fitted with a linear decay. Other symbols represent the decay in the presence of oxygen in the concentration range from 40 ± 3 µM (black symbols) to 155 ± 6 µM (blue symbols). Data were fitted with Eq. 1. (B) Observed kinetic rate constants obtained from the fitting of the kinetic data with Eq. 1 plotted in function of the concentration of added oxygen. Data were fitted with the hyperbole of Eq. 2. (C) Effect of the addition of Q_B site inhibitors. Control experiments with (open black symbols) and without (closed orange symbols) oxygen are shown for comparison; 10 µM DCMU (open blue symbols) or 100 µM bromoxynil (open red symbols) was added prior to the degassing step and the kinetics were measured upon the addition of 110 ± 5 µM oxygen. (D) Effect of the addition of 1 mM bicarbonate (open circles) at the degassing step prior to the addition of oxygen is compared with a control kinetics in the absence of bicarbonate (open squares). Kinetics were recorded upon addition of 110 ± 5 µM oxygen.

The decay of the trapped ${\text{Q}}_{\text{A}}^{•-}$ state was significantly accelerated when O₂ was added back to the medium (Fig. 2A), which provided clear experimental evidence for the reduction of O₂ by ${\text{Q}}_{\text{A}}^{•-}$. Fig. 2A also shows that the rate of ${\text{Q}}_{\text{A}}^{•-}$ oxidation accelerated as the O₂ concentration increased. The observed fluorescence decays were found to be biphasic, with a fast exponential decay rate that depended on the O₂ concentration and a slow decay rate that was independent of the O₂ concentration. The slower phase corresponded to that observed in the absence of oxygen. The kinetic data were fitted using a linear combination of the fast exponential phase and a slow, linear decay,

$$f=e^{(k_{\text{obs}} t)}+ ((b\times t)+c)$$ [1]

where k_obs is the pseudo first-order rate constant for ${\text{Q}}_{\text{A}}^{•-}$ reoxidation at a fixed O₂ concentration, b is the slope, and c is the intercept of the linear decay observed in the absence of O₂. This biphasic behavior indicated heterogeneity in the preparation, where most of the centers were susceptible to reactions with O₂, but a smaller fraction remained unreactive. We demonstrate in the following that this heterogeneity appears to be related to bicarbonate binding to the nonheme iron. We investigated whether the effect of “connectivity” influenced the kinetics observed here in Fig. 2. Connectivity is a phenomenon seen in fluorescence when PSII shares an extended antenna, leading to a situation where there is a significant probability that excitation energy can visit more than one closed center before finding an open PSII. This results in a nonlinear relationship between fluorescence intensity and ${\text{Q}}_{\text{A}}^{•-}$ concentration (22, 23). We found that connectivity has a negligible effect on the present experiment (SI Appendix, Fig. S2).

Fig. 2B shows the observed rate constants (k_obs) for the fast, exponential phase of the ${\text{Q}}_{\text{A}}^{•-}$ decay, measured in pseudo first-order conditions and plotted as a function of the added concentration of O₂. The plot shows saturation behavior characteristic of the presence of a discrete binding site. This binding site becomes saturated above ∼150 µM O₂ (SI Appendix, Fig. S3). The data were fitted with a hyperbolic curve (Eq. 2) from which the apparent K_d was calculated to be 70 ± 20 µM and a rate constant, at physiological O₂ concentrations, to be 0.07 s⁻¹, corresponding to a half-time of ∼10 s (10.6 ± 0.8 s). The pseudo first-order rate constant k_obs for ${\text{Q}}_{\text{A}}^{•-}$ reoxidation at a fixed oxygen concentration was estimated as

$$k_{\text{obs}}=\frac{k_{\text{sat}}}{(K_d^{\text{app}}+[{\text{O}}_2\left]\right)}$$ [2]

where k_sat is the rate constant when the binding site is fully occupied, and $K_{\text{d}}{}^{\text{app}}$ is the apparent dissociation constant. The second-order rate constant, calculated from the linear fit, was found to be 340 ± 83 M⁻¹ s⁻¹.

Influence of Herbicides on the Reactivity to Oxygen.

To test if the driving force determines the rate of the electron transfer, kinetics was measured by modulating the reduction potential of ${\text{Q}}_{\text{A}}/{\text{Q}}_{\text{A}}^{•-}$ by binding herbicides into the Q_B site.

Different classes of herbicides shift the potential in different directions: by +50 mV (−144 mV to −94 mV) with 3-(3,4-dichlorophenyl)-1,1-dimethylurea (DCMU), and by −50 mV (−144 mV to −194 mV) with bromoxynil (24). These herbicide-induced shifts in the ${\text{Q}}_{\text{A}}{\text{/Q}}_{\text{A}}^{•-}$ E_m result in differential effects on the kinetics of ${\text{Tyr}}_{\text{D}}{}^{•}{\text{Q}}_{\text{A}}^{•-}$ recombination (SI Appendix, Fig. S4), thereby confirming herbicide binding under our experimental conditions. Fig. 2C shows the kinetics of ${\text{Q}}_{\text{A}}^{•-}$ reoxidation at 110 ± 5 μM O₂ concentration. With DCMU, the reaction was essentially blocked, whereas with bromoxynil, the reaction was virtually unaffected. Similar behavior was observed when the experiment was performed at lower O₂ concentration (40 ± 3 µM) (SI Appendix, Fig. S5). The effect of DCMU appears to be too drastic to be due to the small change in the ${\text{Q}}_{\text{A}}{\text{/Q}}_{\text{A}}^{•-}$ reduction potential, and it is thus attributed to a structural effect of DCMU binding rather than a redox effect on the quinone (in the following, we discuss DCMU binding having a redox effect on the nonheme iron and on the binding of bicarbonate as viable options to explain its effect). The lack of a bromoxynil effect, on the other hand, suggests that the rate of O₂ reduction by ${\text{Q}}_{\text{A}}^{•-}$ is not determined by the thermodynamic driving force between ${\text{Q}}_{\text{A}}{\text{/Q}}_{\text{A}}^{•-}$ and ${\text{O}}_2/{\text{O}}_2{}^{•–}$. As bromoxynil occupies the Q_B site (bromoxynil binding data on flash 1 is provided in SI Appendix, Fig. S6B), this result eliminates the possibility that ${\text{Q}}_{\text{B}}^{•-}$ or Q_BH₂ donates electrons to O₂, allowing the ${\text{Q}}_{\text{A}}^{•-}$ to decay by forward electron transfer.

Influence of Bicarbonate on ${\text{Q}}_{\text{A}}^{•-}$ Reoxidation Kinetics.

The degassing used to make the PSII anaerobic is also expected to remove CO₂ and, at this pH (pH 6.5), bicarbonate from solution. The degassing is thus expected to result in partial loss of bicarbonate from PSII. Fig. 2D shows that addition of bicarbonate after degassing the sample, eliminated electron transfer from ${\text{Q}}_{\text{A}}^{•-}$ to O₂. A control experiment using 1 mM NaCl instead of NaHCO₃⁻ had no effect on the kinetics of ${\text{Q}}_{\text{A}}^{•-}$ oxidation by O₂ (SI Appendix, Fig. S7A). Addition of bicarbonate also blocked ${\text{Q}}_{\text{A}}^{•-}$ reoxidation by O₂ in the presence of the herbicides (Fig. 2B and SI Appendix, Fig. S8).

Fig. 2D thus suggests that bicarbonate binding to the nonheme Fe²⁺ controls electron transfer from ${\text{Q}}_{\text{A}}^{•-}$ to O₂. To test the specificity of this effect, sodium formate, which is known to bind in the bicarbonate site on the nonheme Fe²⁺, was added (SI Appendix, Fig. S7B). Formate, at 100 mM, a concentration known to be competitive with bicarbonate (25, 26), behaved like bicarbonate: it prevented ${\text{Q}}_{\text{A}}^{•-}$ from reducing O₂ (SI Appendix, Fig. S7B). Because the effect is not specific to bicarbonate, this suggests electron transfer from ${\text{Q}}_{\text{A}}^{•-}$ to O₂ requires that nonheme Fe²⁺ lacks the carboxylic acid and thus that the O₂ is reduced when bound to the Fe²⁺.

Superoxide as the Product of the Reaction.

The electron paramagnetic resonance (EPR) spin trap 5-(diisopropoxyphosphoryl)-5-methyl-1-pyrroline-N-oxide (DIPPMPO) was used to demonstrate ${\text{O}}_2^{•-}$ formation under continuous illumination (Fig. 3A). This spin trap forms trapped radical adducts with either ${\text{O}}_2^{•-}$ or the more reactive OH^•, but these can be distinguished by their different EPR spectra (27). The experiments in Fig. 3A were done in the presence of catalase, which removes peroxide (formed by dismutation of the ${\text{O}}_2^{•-}$), which could give rise to OH^• by Fenton chemistry. The presence of residual OH^• forms a spectroscopically distinct adduct with DIPPMPO (28); it can be deconvoluted from the ${\text{O}}_2^{•-}$ be spin adduct and quantified as being less than 20% of the OH^• adduct (SI Appendix, Figs. S9 and S10 EPR deconvolution).

Fig. 3.

EPR characterization of ROS. (A) Generation of ${\text{O}}_2{}^{•–}$ produced in the absence (Top) and in the presence (Bottom) of 10 mM bicarbonate measured by EPR using the spin probe DIPPMPO. PSII membrane fragments (600 nM PSII cores) in 40 mM MES buffer, 5 mM MgCl, pH 6.5. The sample was illuminated with 50 µM photons m⁻² s⁻¹ red light (590 nm longpass filter) for 20 min. (B) ${\text{O}}_2{}^{•–}$ quantified as reduced cyt c in the absence (blue bar) and in the presence (red bar) of 1 mM bicarbonate. A total of 40 to 50% ${\text{Q}}_{\text{A}}^{•-}$ was obtained after three saturating flashes of a degassed sample of PSII membrane fragments (60 nM PSII cores) in the presence of 250 µM hydroxylamine in a 40 mM MES buffer, 5 mM MgCl, pH 6.5.

Fig. 3A shows that the concentration of ${\text{O}}_2^{•-}$ formed under these conditions was diminished by ∼30% when the experiment was done in the presence of 12 mM bicarbonate. The higher bicarbonate concentration was used to maintain the bicarbonate ratio to PSII in the EPR experiment, in which 12 times more PSII is required to obtain appropriate signal to noise.

The spin-trapping EPR experiment showing the formation of ${\text{O}}_2^{•-}$ was insufficiently sensitive to allow its exact quantification. A more sensitive approach is to monitor the reduction of cytochrome c (cyt c) by ${\text{O}}_2^{•-}$ by its absorption change at 550 nm (29). Fig. 3B shows the amount of reduced cyt c from the reaction of ${\text{Q}}_{\text{A}}^{•-}$ with O₂, in a sample containing 60 nM PSII. The sample was treated as described for the fluorescence kinetics measurements (Materials and Methods). The experiment was performed in the presence and absence of 1 mM bicarbonate, and the amount of reduced cyt c, which is equivalent to the amount of ${\text{O}}_2^{•-}$ generated, was found to be 32 ± 16 nM and 68 ± 16 nM respectively. The bicarbonate-dependent decrease in the amount of reduced cyt c corresponds to ∼50% of the concentration of PSII and thus approximately the same amount of ${\text{Q}}_{\text{A}}^{•-}$ present prior to O₂ addition. This confirms that ${\text{Q}}_{\text{A}}^{•-}$ decay upon the addition of O₂ is due to stoichiometric ${\text{O}}_2^{•-}$ formation. Control experiments showed that the direct reduction of cyt c by PSII was negligible, consistent with previous reports (e.g., ref. 30). Interference by H₂O₂ and OH^• on the cyt c reduction was prevented by the presence of catalase (Materials and Methods).

The residual cyt c reduction occurring after the addition of bicarbonate does not arise from the ${\text{Q}}_{\text{A}}^{•-}$ present prior to O₂ addition, as all of that is accounted for by the bicarbonate-sensitive ${\text{O}}_2^{•-}$. Other electron transfer components have been suggested as possible reductants of O₂ forming ${\text{O}}_2^{•-}$, including cyt b₅₅₉ (31) and Q_BH₂ (14) as well as the PQH₂ pool (15, 32), which could be present as a residual in our low PQ preparation. Furthermore, small amounts of contaminant Photosystem I (PSI) in the PSII-enriched thylakoid membrane (BBY)-prep (Materials and Methods) could represent another possible source of ${\text{O}}_2^{•-}$. While such effects would be expected to occur to a minor extent in the dark, they could also be driven by an actinic effect of the measuring flashes. Despite their weak intensity, their frequency might result in sufficient photochemistry during the time course of the experiment to account for the observed residual ${\text{O}}_2^{•-}$-mediated cyt c reduction (SI Appendix, Fig. S11). This would be consistent with the observation that experiments run in the presence of DCMU show a further decrease in the amount of reduced cyt c.

${\text{O}}_2^{•-}$ Production When ${\text{Q}}_{\text{A}}^{•-}$ Is Generated by Continuous Illumination.

The ${\text{O}}_2^{•-}$ accumulation was also studied by cyt c reduction under constant illumination (SI Appendix, Fig. S12). After 2 min of illumination (50 μmol photons m⁻² s⁻¹ red light), an average of 736 ± 10 nM reduced cyt c was generated, and when 1 mM HCO₃⁻ was present, this was diminished by approximately half (351 ± 60 nM). This is consistent with what we observed in the single flash stoichiometric experiments. When the same experiment was done by using 20 min of illumination without HCO₃⁻, an average of 2,076 ± 10 nM reduced cyt c was generated, while in the presence of HCO₃⁻, 1,457 ± 60 nM reduced cyt c was formed. The effect of HCO₃⁻ on decreasing the formation of reduced cyt c seemed to diminish at longer illumination periods, suggesting that other routes for production of ${\text{O}}_2^{•-}$, such as cytochrome b₅₅₉ and PQH₂ and perhaps involving ¹O₂, might play a role upon prolonged illumination. Furthermore, the prolonged illumination leads to an increased probability of ${\text{Q}}_{\text{A}}^{•-}$ being formed that in turn favors bicarbonate dissociation even in the presence of 1 mM bicarbonate in solution (33, 34).

When the DCMU was present in the absence of added bicarbonate, 2 min of illumination resulted in hardly detectable amounts of reduced cyt c, while after 20 min of illumination, the amount was 626 ± 70 nM (SI Appendix, Fig. S12). These results differ from those shown in Fig. 2B, where DCMU was found to be less effective than bicarbonate for inhibiting the ${\text{Q}}_{\text{A}}^{•-}$ reduction of O₂ in the dark. The more pronounced effect of DCMU under continuous light could result from other reactions generating reactive oxygen species (ROS), such as long-lived Pheo_D1^•− and/or ³Chl-mediated ¹O₂ formation.

Mechanism of ${\text{Q}}_{\text{A}}^{•-}$ Reduction of O₂.

To probe possible molecular mechanisms of O₂ reduction, we performed quantum chemical density functional theory (DFT) calculations and hybrid quantum mechanics/molecular mechanics (QM/MM) molecular dynamics (MD) simulations to assess the electronic structure of the nonheme iron site and its reactivity with molecular oxygen (Fig. 4 and SI Appendix, Fig. S13). In our DFT models, we first optimized the structure of the nonheme iron site with either ${\text{HCO}}_3^{–}$ or an O₂ ligand bound to the Fe²⁺ ion or in the absence of either ligand. The DFT models suggested that the O₂ can bind with a high affinity (∼−4 kcal mol⁻¹) to the nonheme Fe²⁺, forming a 2.4-Å covalent bond between the Fe²⁺ and the O₂. We next added an electron on Q_A and reoptimized the models. The DFT calculations suggested that the electron was localized on Q_A, forming ${\text{Q}}_{\text{A}}^{•-}$ in the ${\text{HCO}}_3^{–}$-bound form and when no exchangeable ligand on the iron was modeled (Fig. 4A). In contrast, in the dioxygen-bound form, the electron instantly moved from ${\text{Q}}_{\text{A}}^{•-}$ to O₂ during the calculations, leading to formation of an ${\text{Fe}}^{\text{II}}-{\text{O}}_2{}^{•–}$ species (Fig. 4 B–D). By varying the Fe-O₂ distance, we observed that the electron transfer from ${\text{Q}}_{\text{A}}^{•-}$ was triggered at around 2.6 Å, when a chemical bond forms between the dioxygen and iron (SI Appendix, Fig. S14). The findings thus suggest that the nonheme iron catalyzes O₂ reduction/superoxide formation (Fig. 4 B and D). Our calculations also suggested that the presence of ${\text{Q}}_{\text{A}}^{•-}$ led to weakening of the ${\text{HCO}}_3^{–}$ affinity for the iron due to electrostatic repulsion, in accordance with experimental evidence (33). The calculations also support that the Fe³⁺ form of the nonheme iron can still bind ${\text{HCO}}_3^{–}$, whereas both O₂ and ${\text{O}}_2^{•-}$ favor the Fe²⁺ form (SI Appendix, Fig. S15).

Fig. 4.

O₂ reduction mechanism, bicarbonate binding, and accessibility of the nonheme Fe site of PSII. Structure and spin densities from DFT models of the (A) HCO₃⁻-bound and (C) O₂-bound forms of the nonheme iron (the state without HCO₃⁻ is shown in SI Appendix, Fig. S13F). The figure shows the spin density difference (red/blue sphere ±0.001e for alpha/beta spin) after adding an electron into the system, with only the Q_A and iron/histidine residues shown for clarity (SI Appendix, Fig. S13). (B) Spin population during the QM/MM MD simulations for HCO₃⁻-bound (in blue) and without HCO₃⁻-bound (in orange) and the O₂-bound forms (in red) on the Fe²⁺ (solid lines), Q_A (dotted lines), and O₂ (dashed line). (D) The O-O bond length during electron transfer from Q_A. (E and F) Snapshot at 150 ns of the water-filled tunnels formed around the Q_A site during MD simulations in (E) with the HCO₃⁻-bound form and (F) without the HCO₃⁻. (G) Histogram of water molecules in tunnels 1 and 2 connecting to the iron from the bulk solvent. The bicarbonate blocks water entry to the nonheme iron site.

Hybrid QM/MM MD simulations were also done to study the dynamics of the electron transfer process in more detail, including the effects of temperature (Fig. 4 B and D, SI Appendix, Fig. S14, Materials and Methods). This was done by initiating the simulations from a state in which ${\text{Q}}_{\text{A}}^{•-}$ was relaxed in classical molecular dynamics (MD) simulations. The reduction of Q_A tightened the local hydrogen bonds between the plastoquinone and His214/Phe261 relative to the neutral Q_A state, which stabilized the reduced form by further subtle conformational changes in the surrounding helices. Consistent with our results obtained from the DFT models, we found that Q_A remained reduced throughout the 2 ps QM/MM MD simulation when HCO₃⁻ was bound or when no exchangeable ligand on the iron was modeled. In contrast, the electron was rapidly transferred during the initial 0.2 ps to the Fe-bound O₂, forming a ${\text{Fe}}^{\text{II}}-{\text{O}}_2^{•-}$ species.

To probe structural effects linked to dissociation of HCO₃⁻ on longer nano-to-microsecond timescales, we performed classical atomistic MD simulations of PSII embedded in a lipid membrane, with the Q_A site modeled either as oxidized or as semiquinone, ${\text{Q}}_{\text{A}}^{•-}$. During the 200 ns MD simulations, we observed two water-filled tunnels, 1 and 2, that connect the stromal side of the membrane to the nonheme Fe center, while the Q_A site remained sequestered from water molecules (Fig. 4 E–G, SI Appendix, Fig. S15). In simulations without an exchangeable ligand modeled on the nonheme iron, we observed a significant increase in the hydration state of the tunnels compared to the simulations performed when the ${\text{HCO}}_3^{–}$ was bound (Fig. 4G). Conformational changes around Lys264 seem to regulate the accessibility of the tunnel to the nonheme iron site (Fig. 4 E and F, SI Appendix, Fig. S15 A and B). We note that the water-filled tunnels are large enough for dioxygen diffusion into the site. These findings fit with the model in which occupation of the binding site on the nonheme iron prevents O₂ from binding there.

Discussion

Electron Donation from ${\text{Q}}_{\text{A}}^{•-}$ to O₂.

Here, we trapped the ${\text{Q}}_{\text{A}}^{•-}$ state in PSII membranes with a depleted PQ pool by illuminating in the presence of low concentrations of the electron donor, NH₂OH, under anaerobic conditions. The ${\text{Q}}_{\text{A}}^{•-}$ is stable in a significant fraction of centers because forward electron transfer is blocked due to Q_B being reduced or absent in the PQ-depleted PSII. In addition, ${\text{Q}}_{\text{A}}^{•-}$ cannot recombine with the stable S₀ or S₁ states on the minutes timescale of the experiment (5). The presence of ${\text{Q}}_{\text{A}}^{•-}$ was monitored using chlorophyll fluorescence (17, 35). This “trapped” ${\text{Q}}_{\text{A}}^{•-}$ undergoes a slow decay (t_1/2 = 5 min) that may correspond to the recombination of the electron on ${\text{Q}}_{\text{A}}^{•-}$ with a relatively stable oxidized species, such as Tyr_D^• (21) (Fig. 2A).

We used this experimental system to test if the trapped ${\text{Q}}_{\text{A}}^{•-}$ would react with added O₂. Upon addition of O₂, the lifetime of the trapped ${\text{Q}}_{\text{A}}^{•-}$ was found to decrease, with the reaction rate increasing with increased O₂ concentration and saturating above ∼150 µM O₂ (SI Appendix, Fig. S3). The kinetics of the O₂-induced ${\text{Q}}_{\text{A}}^{•-}$ decay showed a single exponential curve, consistent with a pseudo first-order reaction at a fixed ${\text{Q}}_{\text{A}}^{•-}$ concentration, with a transition to a zero-order reaction above 150 µM O₂, thus resulting in a half-time of ∼10 s under physiological O₂ concentrations (270 µM) (Fig. 2B).

The reaction rate did not appear to be affected by the thermodynamic driving force, as judged by the lack of effect of the phenolic herbicide, bromoxynil, which is known to lower the E_m of the ${\text{Q}}_{\text{A}}{\text{/Q}}_{\text{A}}^{•-}$ redox couple (24) (Fig. 2C and SI Appendix, Fig. S4 Tyr_D ${\text{Q}}_{\text{A}}^{•-}$ recomb). DCMU binding is expected to increase the E_m of the ${\text{Q}}_{\text{A}}{\text{/Q}}_{\text{A}}^{•-}$ couple and thus potentially slow down the O₂ reduction. However, DCMU binding strongly inhibited the electron transfer from ${\text{Q}}_{\text{A}}^{•-}$ to O₂ (Fig. 2C). The DCMU inhibition was so significant that it seems unlikely to result from the predicted +50 mV change in reduction potential. Possible explanations for the DCMU-induced inhibition of O₂ reduction by ${\text{Q}}_{\text{A}}^{•-}$ include minor structural/conformational effects, redox effects on the nonheme iron, which may affect bicarbonate binding (see The Nonheme Iron as the Binding Site for O₂).

Bicarbonate and Formate Inhibition of O₂ Reduction.

Under the conditions of the fluorescence measurements, the bicarbonate was expected to be depleted from its binding site on the nonheme Fe²⁺ because 1) the degassing of the sample to make it anaerobic is expected to lower the concentration of CO₂ and thus of the bicarbonate; and 2) the presence of ${\text{Q}}_{\text{A}}^{•-}$ greatly decreases the affinity of the bicarbonate for its binding site on the nonheme iron (33). Therefore, we tested the effect of adding back bicarbonate prior to the addition of O₂. Bicarbonate readdition is expected to lower the E_m of ${\text{Q}}_{\text{A}}{\text{/Q}}_{\text{A}}^{•-}$ and thus to increase the driving force for O₂ reduction. Instead, we found that bicarbonate addition completely blocked electron transfer from ${\text{Q}}_{\text{A}}^{•-}$ to O₂ (Fig. 2D). This finding suggests that O₂ reduction occurs when it is bound to the nonheme iron in the empty bicarbonate site.

Formate had the same effect as bicarbonate. This lack of formate/bicarbonate specificity contrasts with the specific roles of bicarbonate in optimizing proton-coupled electron transfer (26, 34, 36) and in redox tuning for regulation and photoprotection (33), roles that are not shared by formate. This lack of specificity of bicarbonate-versus-formate binding is consistent with the inhibition of O₂ reduction being a steric effect, with either bicarbonate or formate binding occluding the O₂ binding site on the nonheme iron. The occlusion of the nonheme iron binding site by bicarbonate was also indicated by the MD simulations.

Formation of Superoxide.

EPR spin trapping and cyt c reduction confirmed that ${\text{O}}_2^{•-}$ was the product of the reaction of ${\text{Q}}_{\text{A}}^{•-}$ with O₂ (Fig. 3) and that this reaction was blocked by bicarbonate addition. The cytochrome reduction method allowed the ${\text{O}}_2^{•-}$ to be quantified, and its concentration was found to be close to that of the ${\text{Q}}_{\text{A}}^{•-}$. A second fraction of ${\text{O}}_2^{•-}$ was formed in the dark in a reaction that was not inhibited by bicarbonate addition and was unrelated to the oxygen-dependent decay of ${\text{Q}}_{\text{A}}^{•-}$. We have not characterized the electron source of this fraction of ${\text{O}}_2^{•-}$ formation, but we note that there are several possible candidates, including PQH₂ bound to PSII or in the membrane, cyt b₅₅₉, and PSI (14, 15, 31, 32).

Physiological Significance: Rates and Conditions.

(In Redox and Mechanistic Considerations, we discuss the possibility of faster rates of O₂ reduction by ${\text{Q}}_{\text{A}}^{•-}$, but in this section, we discuss the slower rate measured using the experimental approach used here.) The rate of electron transfer from ${\text{Q}}_{\text{A}}^{•-}$ to O₂ (t_1/2 ∼ 10 s) implies that under conditions where the Q_B and quinone pool are oxidized, the reduction of O₂ is too slow to compete with the forward electron transfer rate of t_1/2 ∼ 1 ms (35) even when bicarbonate is absent (34). When the quinone pool is fully reduced, a large proportion of ${\text{S}}_2{\text{Q}}_{\text{A}}^{•-}$ and ${\text{S}}_3{\text{Q}}_{\text{A}}^{•-}$ will recombine with a t_1/2 ∼ 1.5 s (20), i.e., much faster than the rate of electron transfer of the ${\text{Q}}_{\text{A}}^{•-}$ to O₂ reaction reported here. However, when forward electron transfer from ${\text{Q}}_{\text{A}}^{•-}$ is blocked in centers where either S₁ or S₀ are present, the reaction with O₂ could be the dominant ${\text{Q}}_{\text{A}}^{•-}$ reoxidation pathway, provided the bicarbonate is absent from its binding site on the nonheme iron. Such conditions are likely to occur when CO₂ levels are limiting, as previously discussed (33).

Physiological conditions other than those where the PQ pool is reduced are likely to exist in which electrons are trapped on ${\text{Q}}_{\text{A}}^{•-}$ long enough to allow bicarbonate release and reduction of O₂ to be a relevant reaction. These circumstances could include those associated with assembly and photoactivation of PSII, photoinhibition, and repair, in which ${\text{Q}}_{\text{A}}^{•-}$ is longer lived (9, 37–39). An intermediate state in PSII assembly is structurally modified by the binding of assembly factors (Psb27, Psb28, and Psb34) that cause the bicarbonate site on the nonheme iron to be occupied instead by a glutamate (39). This iron coordination mirrors the situation that exists in the purple bacterial reaction centers. From the present results, we expect such assembly intermediates of PSII to be unable to reduce O₂ from ${\text{Q}}_{\text{A}}^{•-}$. As the Q_B site is significantly modified in this assembly intermediate, it seems likely that ${\text{Q}}_{\text{A}}^{•-}$ will recombine with Tyr_Z^•(H⁺) via a direct tunneling step between P⁺ and ${\text{Q}}_{\text{A}}^{•-}$, given the high potential of the ${\text{Q}}_{\text{A}}{\text{/Q}}_{\text{A}}^{•-}$ couple prior to photo-assembly of the Mn₄CaO₅ cluster (9, 39).

The Nonheme Iron as the Binding Site for O₂.

The single saturable site for O₂ reduction and its complete inhibition by bicarbonate indicate that the nonheme Fe²⁺ is the O₂ binding and reduction site. DFT and QM/MM calculations reported here further support this assignment. Our calculations showed that reduction of O₂ by ${\text{Q}}_{\text{A}}^{•-}$ is favorable when oxygen binds to the Fe²⁺ in the absence of bicarbonate binding to that site. This result contrasts with a previously proposed mechanism involving direct oxidation of ${\text{Q}}_{\text{A}}^{•-}$ by O₂ (12–14), which is expected to require close contact between the oxygen and the semiquinone (40). The crystal structure (41) and the MD simulations indicate that this is unlikely because the Q_A is not exposed to the solvent.

In enzymes, transition metals and Fe²⁺ in particular often activate O₂ for reduction, overcoming the intrinsic spin-transition that makes O₂, a ground state triplet, anomalously stable. It seems that this is also the case in PSII. This mechanism seems reasonable given that bicarbonate dissociates from the iron when ${\text{Q}}_{\text{A}}^{•-}$ is long-lived (33).

The DCMU inhibition of O₂ reduction may also be taken as an indication that the nonheme Fe²⁺ is the O₂ binding site. DCMU binding shifts the redox potential of the Fe³⁺/Fe²⁺ couple 120 mV to higher values, while other herbicides/inhibitors (atrazine and o-phenanthroline) induced much smaller shifts (42–44). Whether the DCMU inhibition of O₂ reduction reflects a perturbation of the electronic structure of the Fe²⁺ (as manifest by the redox shift), minor structural changes, or both of these is unclear. These effects could be responsible for DCMU inhibiting bicarbonate binding and dissociation (45, 46). Given the conditions of our experiment, it is possible that DCMU binding prevents bicarbonate loss, leaving the nonheme Fe²⁺ site occupied. This would explain the lack of O₂ reduction (Fig. 2). The observation that addition of bicarbonate to a DCMU-treated sample eliminates the residual O₂ reduction (SI Appendix, Fig. S10) is consistent with bicarbonate depletion is very limited when DCMU is present.

The suggestion that the nonheme Fe²⁺ could play an oxygen reduction role was already discussed (47), although superoxide was suggested to bind to the Fe²⁺, and then to undergo further reduction to form OH^•. In this mechanism, the superoxide was suggested to arise from oxygen reduction by ${\text{Q}}_{\text{A}}^{•-}$ directly, a reaction that seems unlikely based on the present work. The direct catalytic role for the nonheme iron in oxygen reduction and its control by the bicarbonate, reported here, have not been discussed previously, as far as we are aware.

Redox and Mechanistic Considerations.

Nominally, the electron transfer from ${\text{Q}}_{\text{A}}^{•-}$ to O₂ is thermodynamically unfavorable, based on the E_m of ${\text{Q}}_{\text{A}}{\text{/Q}}_{\text{A}}^{•-}$ = −144 mV, a value that is shifted to −70 mV in the absence of bicarbonate (33), and thus, it is more oxidizing than E_m = −160 mV for ${\text{O}}_2{\text{/O}}_2{}^{•–}$ (48). However, given the very low concentration of ${\text{O}}_2^{•-}$ compared to O₂, its functional potential is likely to be closer to ∼0 mV (13), rendering the overall process thermodynamically favorable. In addition, it seems likely that the binding of the O₂ to the iron will change the redox potential of the ${\text{O}}_2{\text{/O}}_2^{•-}$ couple.

Given the thermodynamic driving force and the ∼8 Å distance between ${\text{Q}}_{\text{A}}^{•-}$ and O₂, rapid (∼ps) reaction rates are expected, as indeed observed in the QM/MM MD simulation, when O₂ was already bound to the iron. The measured reaction rates are, however, much slower (t_1/2 ∼10 s), and this could be due to the reaction rates being limited by O₂ diffusion along the constricted channels to the iron binding site (Fig. 4 E and F and SI Appendix, Fig. S14). This diffusional restriction could limit the measured rate in the experiments performed, i.e., when an anaerobic sample is mixed with oxygenated buffer. The possibility arises that under aerobic conditions, O₂ may be already within the access channel and thus have more rapid access to the nonheme iron. This raises the possibility that in vivo the reduction of O₂ by ${\text{Q}}_{\text{A}}^{•-}$ could be faster than reported here.

The Fe³⁺/Fe²⁺ couple has an E_m of +430 mV at this pH (44). The Fe²⁺ is located between the quinones Q_A and Q_B, but there is no evidence of a distinct redox role of the metal in the electron transfer process between the two quinones (49). Similarly, while the DFT and QM/MM calculations indicate that both Fe³⁺ and Fe²⁺ are stable with HCO₃⁻ bound, no Fe³⁺ is formed during the electron transfer process with O₂ bound, suggesting that the Fe²⁺ has a catalytic rather than redox role in enabling the electron transfer to O₂ (Fig. 4 and SI Appendix, Fig. S14).

Bicarbonate Regulatory Mechanism.

Recently, it was shown that the dissociation of bicarbonate leads to an increase in the reduction potential of ${\text{Q}}_{\text{A}}{\text{/Q}}_{\text{A}}^{•-}$, and consequently, the presence of ${\text{Q}}_{\text{A}}^{•-}$ decreased the bicarbonate affinity for its binding site on the nonheme iron (33). The redox-tuning/bicarbonate-binding relationship suggested the following photoregulation/protection model: 1) when the intracellular CO₂ concentration is low and PSII is exposed to light, limitations in CO₂ fixation result in the overreduction of the electron transfer chain, leading to the formation of a long-lived ${\text{Q}}_{\text{A}}^{•-}$; 2) the long-lived ${\text{Q}}_{\text{A}}^{•-}$ triggers the dissociation of the bicarbonate by lowering its affinity for the nonheme iron; 3) the loss of the bicarbonate raises the E_m of ${\text{Q}}_{\text{A}}{\text{/Q}}_{\text{A}}^{•-}$, increasing the energy gap between the ${\text{Q}}_{\text{A}}{\text{/Q}}_{\text{A}}^{•-}$ and Pheo_D1/Pheo_D1^•− redox couples (33); 4) this increased energy gap disfavors the back-reaction, preventing the formation of P^+•Pheo^•−, the precursor of the chlorophyll triplet state (7) that reacts with O₂ to form damaging ¹O₂ (9).

Our current findings that ${\text{Q}}_{\text{A}}^{•-}$ can reduce O₂ and that bicarbonate binding prevents this reaction suggest another layer of complexity on the regulatory and protective role of bicarbonate in PSII (Fig. 5). Under normal functional conditions (light, high intracellular CO₂, stomata open), the ${\text{Q}}_{\text{A}}^{•-}$ lifetime is expected to be short, the bicarbonate is bound, and ${\text{O}}_2^{•-}$ formation is blocked (Fig. 5A), leading to minimal electron leaks. However, when the intracellular CO₂ concentration is low, the electron transfer chain is reduced, and long-lived ${\text{Q}}_{\text{A}}^{•-}$ is formed, which favors bicarbonate release (Fig. 5B). Under these conditions, O₂ can bind to the Fe²⁺ and then be reduced by ${\text{Q}}_{\text{A}}^{•-}$, forming Q_A and ${\text{O}}_2^{•-}$ (Fig. 5C), which is then released (Fig. 5D).

Fig. 5.

Structural scheme showing the working model for O₂ reduction by ${\text{Q}}_{\text{A}}^{•-}$ in PSII. (A) When photosynthesis is limited by low CO₂, illumination results in the reduction of the PQ pool, Q_B and Q_A. (B) The presence of long-lived ${\text{Q}}_{\text{A}}^{•-}$ changes the dissociation constant of the bicarbonate ligand on the nonheme Fe²⁺, leading to its release and allowing O₂ to enter the channel and to bind to the Fe²⁺. (C) When the O₂ is bound to the Fe²⁺, it is rapidly reduced by the electron coming from ${\text{Q}}_{\text{A}}^{•-}$. (D) The superoxide formed is released from the Fe²⁺ and diffuses away.

The slow rate (t_1/2 = 10 s) for reoxidation of ${\text{Q}}_{\text{A}}^{•-}$ will only have small effects on relieving the electron transfer block. It is debatable whether this will constitute a significant benefit to the system. As mentioned previously, this slow rate appears to represent the diffusional limit when O₂ is added to a fully anaerobic system, and it seems quite possible that faster rates may occur in equilibrated aerobic conditions. The possibility of faster O₂ reduction notwithstanding, low concentrations of superoxide can act as a signal, either directly or by dismutation to form hydrogen peroxide (11, 50, 51). Superoxide formation is stoichiometric and could accumulate, and a signaling role does not necessarily require a high concentration. PSI is usually considered the major source of superoxide (11, 15, 50). However, the different locations of the two photosystems in the thylakoid membrane and the specific localization of superoxide dismutases near PSI could allow the ${\text{O}}_2^{•-}$ produced by PSII to play a role in regulation.

Relevance to Previous Studies.

The phosphorylation of the D1 protein in PSII, which is associated with the migration of damaged PSII from the grana to the stroma lamellae during PSII repair cycle, was shown to decrease ${\text{O}}_2^{•-}$ production (52). This was confirmed in STN8 kinase knockout mutant strains in rice that showed enhanced ${\text{O}}_2^{•-}$ production under high light (53). In both cases, it was suggested that a conformational change induced by the phosphorylation (or the lack of it in the absence of the kinase in the knockout mutant) was responsible for the formation of ${\text{O}}_2^{•-}$ due to the reduction of oxygen by Q_B^•−. Considering 1) that the experiments in ref. 52 were performed in high light, 2) the improved understanding of the redox properties of Q_B (4), and 3) the findings in the present work, it seems more likely that ${\text{Q}}_{\text{A}}^{•-}$ is responsible for the oxygen reduction and that the phosphorylation restricts O₂ accessibility to the nonheme iron, perhaps by favoring bicarbonate binding.

PsbS knockout mutants in rice were shown to produce more ${\text{O}}_2^{•-}$ due to a proposed decrease in the ${\text{Q}}_{\text{A}}{\text{/Q}}_{\text{A}}^{•-}$ reduction potential compared to the wild type under high light conditions (54). More recently, while investigating the role of PsbS on the efficiency of water use in field crops, a correlation between stomatal conductance and the Q_A redox state was reported (55). As hydrogen peroxide is a signal involved in stomatal opening, and as superoxide rapidly dismutates forming hydrogen peroxide, it is possible that the superoxide generated by ${\text{Q}}_{\text{A}}^{•-}$ reported here contributes to this signaling pathway.

Small carboxylic acids, such as acetate and glycolate, can replace bicarbonate at the nonheme iron binding site in vivo (56, 57) and could thereby protect against photooxidative stress under low-CO₂ conditions (33, 56, 57). The mechanism involves redox tuning of Q_A and the associated binding effects of bicarbonate (33). The observation here (SI Appendix, Fig. S7B) that formate can replace bicarbonate in blocking oxygen access to the iron and thus to reduction by ${\text{Q}}_{\text{A}}^{•-}$ extends this model suggesting a potential role for carboxylic acids in controlling superoxide-initiated signaling.

Materials and Methods

Isolation of the PSII Complex.

Market spinach (Spinacia oleracea) was used for the isolation of PSII-enriched thylakoid membranes (BBY). Samples were prepared following the protocol from ref. 24, obtaining samples with a lower content of Q_B. To obtain samples with a lower Q_B content, preparation was the same, but samples were additionally treated with Triton X-100 at a concentration of 2.5% per milligram chlorophyll and for an incubation period of 50 min.

${\text{Q}}_{\text{A}}^{•-}$ Reoxidation Using Fluorometry.

All fluorescence measurements were recorded using a Fluorometer FL3000 (Photon Systems Instruments) and carried out in the dark. For the kinetic measurements, the samples were prepared in a septum sealed cuvette. Final concentrations were 60 nM PSII and 250 µM hydroxylamine in working buffer (40 mM of 2-(N-morpholino)ethanesulfonic acid [MES] and 15 mM of MgCl₂ pH 6.5). A dark-adaptation period was set to 15 min, during which the samples were mixed and oxygen was removed by flushing with argon. ${\text{Q}}_{\text{A}}^{•-}$ was trapped in 40 to 50% of centers by three saturating flashes spaced by 120 s at room temperature. Fluorescence was monitored from submilliseconds to 100 s after each flash. Oxygen was readded to the system with a gas-tight syringe by adding oxygenated buffer to the degassed samples in different proportions. The ratio of oxygenated to degassed buffer was such that the final concentration of oxygen in the sample ranged from 30 to 230 μM. Final O₂ concentrations, upon each addition, were monitored using the Ocean Optics Neofox Phase oxygen sensor. In the cases where oxygenated buffer was added without bicarbonate, the buffer was bubbled with air passed through Ca(OH)₂ to remove any ${\text{HCO}}_3^{–}$ dissolved in the buffer due to the equilibrium of the atmospheric CO₂ (58). When necessary, 1 mM bicarbonate, 10 μM DCMU, or 100 μM bromoxymil was added to the PSII samples prior to the degassing step. Further experimental details are provided in SI Appendix.

ROS Formation from O₂ Reacting with a Fixed Stable Amount of ${\text{Q}}_{\text{A}}^{•-}$ Using cyt c Reduction.

Reduction of cyt c was used to quantify the concentration of ${\text{O}}_2^{•-}$ generated based on known concentrations of ${\text{Q}}_{\text{A}}^{•-}$ and O₂. A fixed amount of ${\text{Q}}_{\text{A}}^{•-}$ was generated following the protocol previously described for the fluorescence measurements: anaerobic, 250 μM hydroxylamine, three saturating flashes, resulting in the formation of 40 to 50% ${\text{Q}}_{\text{A}}^{•-}$ being stably trapped. The experiment was initiated by the addition of oxygenated buffer corresponding to a specific final concentration of O₂, and after 2 min, the concentration of reduced cyt c was determined. The experiment was performed in the presence and in the absence of 1 mM bicarbonate using 60 nM PSII samples with 250 μM hydroxylamine in 40 mM MES and 15 mM MgCl₂ pH 6.5 with additional 20 μM cyt c and 500 U mL⁻¹ catalase. Catalase was added to remove any H₂O₂ generated by spontaneous dismutation of ${\text{O}}_2^{•-}$, which could oxidize cyt c by diminishing the signal. The samples containing ${\text{HCO}}_3^{–}$ were treated in the same way as for the fluorescence studies. The cyt c absorbance spectrum was measured using a Shimadzu UV-1601PC UV-Visible spectrophotometer, and the absorbance peak at 550 nm was used to determine the concentration of reduced cyt c. SI Appendix provides details on the ROS formation studied under continuous light conditions using cyt c reduction.

${\text{O}}_2^{•-}$ Formation Studied Using EPR.

${\text{O}}_2^{•-}$ was trapped using DIPPMPO and spectra measured with Magnettech Miniscope MS5000. The 50-μL capillary tubes contained 60 μg mL⁻¹ PSII, 10 mM DIPPMPO (28), 500 U mL⁻¹ catalase, and 12 mM bicarbonate. Samples were illuminated for 2, 10, and 20 min using 50 μmol photons m⁻² s⁻¹ of red light (590 nm longpass filter) prior to measuring spectra. PSII in working buffer had ${\text{HCO}}_3^{–}$ removed when necessary by either bubbling with dry N₂ or with air through Ca(OH)₂ (58). Catalase and DIPPMPO were added after ${\text{HCO}}_3^{–}$ removal from the buffer. Spectra were measured at room temperature with 9.2-GHz microwave frequency, 100-kHz modulation frequency, 2-Gauss modulation amplitude, 3,363-Gauss field center, 150-Gauss sweep width, 3-mW microwave power, and 60-s sweep-time (59). Methylene blue was used to artificially generate ¹O₂ to establish that there was no interference between ¹O₂ generated and DIPPMPO.

Molecular Simulations.

All simulations were based on the 1.9-Å resolution crystal structure of PSII from Thermosynechococcus vulcanus (PDB ID: 3WU2) (41), which provided a better resolved Q_A/nonheme iron site compared to the cryoEM structure of S. oleracea (PDB ID:3JCU) (60). The plant and cyanobacterial structures are highly similar around the region studied, and the results obtained are thus likely to apply for both systems (SI Appendix, Fig. S16). The crystal structure of PSII from T. vulcanus (PDB ID: 3WU2) (41) was embedded in a 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine membrane and solvated with TIP3P water molecules and 100 mM NaCl concentration. The total system comprised ca. 425,000 atoms. Parameters for all cofactors were derived from in-house DFT calculations (cf. refs. 61–63), and the remaining system was treated using the CHARMM36 force field (64). Four independent MD simulations, 200 ns each, were carried out with the Q_A site modeled either in the oxidized or anionic semiquinone (${\text{Q}}_{\text{A}}^{•-}$) state, and the nonheme iron center modeled with HCO₃⁻-bound or without the HCO₃⁻. The MD simulations were performed in an NPT ensemble at T = 300 K and P = 1 atmosphere and using a 1 fs integration timestep. Long-range electrostatics was treated using the particle mesh Ewald method. The system was gradually relaxed for 4 ns with harmonic restraints of 1 kcal mol⁻¹ Å⁻¹, followed by 20-ns equilibration without restraints, and the 200-sns production runs. All classical MD simulations were performed using NAMD2 (65), and simulations were analyzed using VMD (66). Tunnels leading to the nonheme Fe site were analyzed using CAVER (67).

DFT Active Site Models and QM/MM Calculations.

DFT models comprising the nonheme Fe²⁺, its four coordinating histidine residues (His215^D1, His272^D1, His214^D2, His268^D2), the Q_A plastoquinone, Glu244^D1, Tyr246^D1, Ser268^D1, Glu242^D2, Tyr244^D2, and Lys264^D2 in addition to three crystal water molecules and the CO₃^2-/HCO₃⁻/O₂ ligand were built based on the crystal structure of PSII (PDB ID: 3WU2) (41). The Fe²⁺ iron was modeled in its high spin state, and anti-ferromagnetically coupled to O₂ using the broken symmetry DFT approach. The quantum chemical models comprised around 160 atoms. Protein residues were cut between C_β and C_α atoms, except for glutamate and lysine residues, which were cut at the C_γ and C_β bond. Terminal carbon atoms were saturated by hydrogen atoms and kept fixed during geometry optimizations at the B3LYP-D3/def2-SVP/def2-TZVP (Fe)/ε = 4 level (68–70). A polarizable dielectric medium using the COSMO (71) was used to model solvation effects. Single point calculations were performed at the B3LYP-D3/def2-TZVP/ε = 4 level as well as by using with B3LYP* (72), CAM-B3LYP (73), CAMh-B3LYP (74), and BHLYP (68) functionals. All calculations were performed using TURBOMOLE v. 7.2/v. 7.5 (75) and the electronic structure was visualized using VMD (66). Hybrid QM/MM calculations were performed based on the MD relaxed crystal structure of PSII (Molecular Simulations). The QM region was defined as in the DFT models but by additionally including the backbone of Ala260^D2 and Phe261^D2 to provide stabilization of the semiquinone at the Q_A site. Link atoms were introduced between C_α and C_β atoms, and the QM/MM systems were relaxed using an adopted basis Newton-Raphson optimizer, followed by QM/MM dynamics at T = 310 K using a 1-fs integration time step. A 12-Å sphere around the QM region was allowed to relax during the QM/MM simulations. The QM/MM calculations were initiated from the ${\text{Q}}_{\text{A}}^{•-}$ state with an HCO₃⁻ or O₂ bound to the nonheme Fe site. All QM/MM calculations were performed using an in-house version of the CHARMM/TURBOMOLE interface (76).

Data Availability

All study data are included in the article and/or SI Appendix.