Bicarbonate-induced redox tuning in Photosystem II for regulation and protection

Significance

The E_m of ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ is the reference pegging the thermodynamics of Photosystem II (PSII) to an absolute value. This work resolves a long-standing discrepancy in the literature, removing this ambiguity at the heart of PSII bioenergetics. The discrepancy in the literature reflects the loss of bicarbonate in some titrations. This surprising finding shows that bicarbonate binding controls the E_m of ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$, and thus the redox state of Q_A influences the binding of bicarbonate. This process constitutes a previously unrecognized regulation of PSII: ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ triggers bicarbonate release, slowing electron transfer and tuning the potential of Q_A to protect PSII against photodamage. This work provides an explanation for the existence of bicarbonate in PSII, a mystery for more than half a century.

Abstract

The midpoint potential (E_m) of ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$, the one-electron acceptor quinone of Photosystem II (PSII), provides the thermodynamic reference for calibrating PSII bioenergetics. Uncertainty exists in the literature, with two values differing by ∼80 mV. Here, we have resolved this discrepancy by using spectroelectrochemistry on plant PSII-enriched membranes. Removal of bicarbonate (HCO₃⁻) shifts the E_m from ∼−145 mV to −70 mV. The higher values reported earlier are attributed to the loss of HCO₃⁻ during the titrations (pH 6.5, stirred under argon gassing). These findings mean that HCO₃⁻ binds less strongly when Q_A^−• is present. Light-induced Q_A^−• formation triggered HCO₃⁻ loss as manifest by the slowed electron transfer and the upshift in the E_m of Q_A. HCO₃⁻-depleted PSII also showed diminished light-induced ¹O₂ formation. This finding is consistent with a model in which the increase in the E_m of ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ promotes safe, direct ${\text{P}}^{+\mathrm{•}}{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ charge recombination at the expense of the damaging back-reaction route that involves chlorophyll triplet-mediated ¹O₂ formation [Johnson GN, et al. (1995) Biochim Biophys Acta 1229:202–207]. These findings provide a redox tuning mechanism, in which the interdependence of the redox state of Q_A and the binding by HCO₃⁻ regulates and protects PSII. The potential for a sink (CO₂) to source (PSII) feedback mechanism is discussed.

Photosystem II (PSII), the water/plastoquinone photooxidoreductase, is at the heart of the major energy cycle that powers the biosphere. Chlorophyll-based photochemistry drives charge separation followed by electron transfer reactions that result in the reduction of quinone on one side of the thylakoid membrane (Fig. 1) and the oxidation of water on the other. The photochemistry is intrinsically a one-photon/one-electron process, but the quinone reduction and water oxidation are two- and four-electron processes, respectively. As a result, both processes involve the accumulation of intermediates that are stabilized by the protein and by coupling to protonation reactions (1–3).

Fig. 1.

Structure of PSII in the region of the quinones and the nonheme iron. The structure was made using Pymol based on Protein Data Bank ID code 3WU2 (8). It shows several relevant amino acids and the H-bond network, including several waters (red beads showing the oxygen atoms) near the HCO₃⁻ distal to the Fe²⁺ (rust red sphere). The upper surface of the protein is exposed to the aqueous phase on the stromal side of the membrane.

The electron acceptor side of PSII contains a nonheme ferrous iron (Fe²⁺) that is flanked by the two quinones, Q_A and Q_B (Fig. 1) (reviewed in refs. 3 and 4). The Fe²⁺ is coordinated by four histidine residues, two from D1 and two from D2, and a bicarbonate ion (HCO₃⁻) that provides a bidentate ligand (3–5). The HCO₃⁻ is thought to play a role in the Q_B protonation pathway (reviewed in refs. 3–5). Recent EPR (6) and computation chemistry studies (7) based on the highest-resolution X-ray crystallographic model (8) implicated HCO₃⁻ in the second of two protonation steps that are associated with Q_B reduction. Measurements of the dissociation constant of HCO₃⁻ (K_d = 40–80 μM), compared with estimates of the HCO₃⁻ concentration in the stroma (reviewed in ref. 9), led to the assumption that HCO₃⁻ remains bound under physiological conditions (10). Recently, however, it was reported that the HCO₃⁻ bound to PSII in Chlamydomonas can be displaced by acetate when present in the culture medium (11).

The redox potential of Q_A has been the subject of research for decades (12). The current detailed thermodynamic picture of PSII redox chemistry is based on estimates of energy differences between the electron transfer components, but these estimates rely on measurements of redox potentials to fix absolute values (1–3). The experimental redox potential of ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ (12) provides the key reference value (1–3).

Krieger et al. (12) collated 29 reported values for the E_m of ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ showing significant scatter. Much of the variation in the reported E_m values could be explained by the experimental demonstration that there were two E_m values for ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$, differing by ∼150 mV, depending on the nature of the PSII (12–14). The fully active enzyme showed an E_m value that was ∼150 mV lower than that in PSII lacking the Mn₄CaO₅ cluster. This shift was, in fact, due to the binding (or absence thereof) of the Ca²⁺ ion involved in water splitting, but because the Mn cluster provides the Ca binding site, the potential is indirectly determined by the presence of Mn cluster (12–14). Given the high valence state on the Mn cluster even in the lowest redox state of the water splitting cycle (15), it is not surprising that the use of chemical mediators, the role of which is to equilibrate the external redox potential with the cofactors, can lead to the reduction and consequent loss of the Mn cluster. Because loss of the Mn leads to loss of Ca²⁺ binding, then the high potential form of the Q_A/Q_A^−• couple is easily generated during the course of redox titrations, and this technical problem explained some of the scatter in the E_m values reported in the literature (12).

In a parallel study, Johnson et al. (13) showed that the high and low potential forms of Q_A had clear physiological relevance. PSII is assembled as a functional photochemical reaction center lacking the Mn₄CaO₅ cluster and consequently, unable to split water. It then undergoes a light-driven oxidative process by which the cluster is assembled (16), so-called photoactivation. The downshift in the redox potential of ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ occurs during photoactivation (13). A rationale was given for this process, with two competing pathways for the decay ${\text{P}}^{+\mathrm{•}}{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ by charge recombination (13): (i) a direct pathway in which ${\text{P}}^{+\mathrm{•}}{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ recombines in a slow process involving the loss of energy to heat, and (ii) a back-reaction route forming P^+•Ph^−•, which recombines to form the reaction center chlorophyll triplet state, ³P (17), in the majority of the centers. Chlorophyll triplets are relatively long-lived and very reactive with O₂ to generate ¹O₂, a very toxic species. This concept of redox tuning controlling the competition between two main charge recombination routes in PSII (13) has dominated much of the subsequent thinking concerning photoinhibition and redox-based protective mechanisms (18–20).

Recently, Shibamoto et al. (21, 22) used an optically transparent thin-layer electrochemical cell to obtain data that confirmed the 150-mV redox difference with and without the Mn cluster, but they also showed that both values shifted to lower potentials by ∼80 mV compared with the earlier data (12). These titrations were done with a limited set of redox mediators that did not equilibrate with the Mn cluster. Subsequently, some evidence was published indicating that, when titrations were done with the same mediators, the lower potential values were obtained, whereas in the absence of mediators, the higher potential values similar to those by Krieger and coworkers (12–14) were obtained (23). However, low potential values were also recently reported without the use of mediators (24). Since then, the situation has remained unresolved, leaving a doubt about the correct potential and consequently, a serious ambiguity at the heart of PSII bioenergetics.

In this work, we have readdressed the discrepancy in the literature over the E_m of Q_A. We have used spectroelectrochemical titrations, and we have been able to reproduce the two conflicting sets of values. The key component that was not previously taken into account is the HCO₃⁻ ligand to the nonheme iron, which can be lost from PSII during the course of the titration. The data also indicate that this redox effect has physiological significance not only in forward electron transfer but also in controlling the back-reaction pathways and hence, photodamage. Overall, the data indicate the presence of an unexpected Q_A⁻-triggered, HCO₃⁻-mediated redox tuning mechanism, which regulates and protects PSII. This model provides a chemical mechanism rationalizing why HCO₃⁻, an exchangeable ligand to the iron, is beneficial and thus, why it is present in PSII.

Results

Fig. 2A, circles shows redox titrations of plant PSII-enriched membranes at pH 6.5 with a ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ E_m = −144 ± 2 mV. This result is in good agreement with the value reported by Shibamoto et al. (21) for the same biological material and using a comparable spectroelectrochemical method. Fig. 2B, circles shows redox titrations of PSII membranes depleted of the Mn₄CaO₅ cluster. The resulting E_m for the ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ couple is E_m = −32 ± 2 mV. The Mn depletion-induced increase in potential is somewhat smaller than that reported earlier (12–14); however, both values are ∼80 mV lower than those reported originally in potentiometric titrations using the same material (12).

Fig. 2.

Effects of bicarbonate depletion on PSII. A and B show redox titrations of ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ monitoring fluorescence in (A) intact and (B) Mn-depleted PSII membranes in the presence (circles) and absence (triangles) of the HCO₃⁻. The concentration of reduced Q_A was corrected for connectivity (SI Appendix). The data points were fitted to the Nernst equation with n = 1, with error bars indicating the SD of three independent titrations. Black symbols, oxidative titration; white symbols, reductive titration. C shows the influence of HCO₃⁻ depletion on the decay of flash-induced ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ measured as the decay of the variable fluorescence after the third flash given at 1 Hz. Untreated PSII with 1 mM HCO₃⁻ (circles), HCO₃⁻-depleted PSII (black symbols), and HCO₃⁻-depleted PSII with 1 mM HCO₃⁻ added back (triangles). Error bars are the SD of three independent measurements. D shows the relationship between HCO₃⁻ binding/dissociation and the redox state of Q_A in redox and energy terms. SHE, standard hydrogen electrode.

At pH values close to or below 6.35, the pK_a of bicarbonate (HCO₃⁻) with carbonic acid (H₂CO₃), the CO₂ formed from its equilibrium with carbonic acid may be lost from solution, particularly when flushed with N₂ or Ar:

$$\mathrm{C}{\mathrm{O}}_2\mathrm{\uparrow }+{\text{H}}_2\text{O}\mathrm{\leftrightarrow }\underset{\text{carbonic}\hspace{0.17em}\text{acid}}{{\text{H}}_2{\text{CO}}_3}\stackrel{{\text{pK}}_{\text{a}1}6.35}{\leftrightarrow }\underset{\text{bicarbonate}}{{\text{HCO}}_3^{-}}+{\text{H}}^{\mathrm{+}}\stackrel{{\text{pK}}_{\text{a}2}10.5}{\leftrightarrow }\underset{\text{carbonate}}{{\text{CO}}_3^{2-}}+2{\text{H}}^{\mathrm{+}}.$$ [1]

It thus seemed possible that HCO₃⁻ could be lost under the conditions of the redox titration. We, therefore, tested if the removal of HCO₃⁻ could influence the E_m of ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$.

Redox titrations of ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ were performed on HCO₃⁻-depleted, PSII-enriched membranes with (Fig. 2A, triangles) and without (Fig. 2B, triangles) the Mn₄CaO₅ cluster. The E_m values were both upshifted: E_m = −70 ± 2 mV, a shift of +74 mV, for Mn-containing PSII and E_m = +54 ± 2 mV, a shift of +86 mV, for the Mn-depleted PSII. The values obtained in HCO₃⁻-depleted PSII membranes are close to those obtained by Krieger and coworkers (12–14), indicating that HCO₃⁻ had been lost in these studies (Discussion). When HCO₃⁻ was added back to samples that had undergone the HCO₃⁻ depletion procedure, the lower potential E_m values were restored (SI Appendix, Fig. S1).

Fig. 2C shows that the HCO₃⁻ removal procedure used results in the typical HCO₃⁻ depletion-induced slowdown in the rate of ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ oxidation as measured by the fluorescence decay seen after excitation by a flash of light (reviewed in ref. 5). Reversible changes in the kinetics were seen, closely matching those reported earlier (25, 26). When HCO₃⁻ depletion was accompanied by the addition of formate as a replacement ion, more marked changes were seen in the kinetics of ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ oxidation as measured by the fluorescence decay seen after a flash (Fig. 3A). This enhanced inhibition also agrees with the literature (25, 26). In contrast, formate has a significantly smaller effect on the E_m of ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ (Fig. 3B). A positive shift in the E_m still occurred (Fig. 3B), but it was significantly smaller, being about one-half that seen with HCO₃⁻ [E_m(form) ∼ −95 mV].

Fig. 3.

Influence of formate and effect of preillumination on bicarbonate depletion. (A) Decay of flash-induced ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ as described in Fig. 2. HCO₃⁻-depleted PSII with 100 mM sodium formate present (black circles). Untreated PSII with 1 mM HCO₃⁻ (white circles). Error bars: SD of three measurements. (B) Redox titrations of ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ in the presence of formate (squares) or HCO₃⁻ (circles) (from Fig. 2A and other details in SI Appendix). (C) Effect of preillumination on HCO₃⁻ depletion: ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ oxidation (i.e., fluorescence decay as in Fig. 2). Untreated PSII with 1 mM HCO₃⁻ (white circles), degassed PSII (i.e., Ar-bubbled for 15 min; white squares), degassed PSII and then illuminated (4 min, 590 nm, 350 μmol photons m⁻² s⁻¹; black circles), and degassed PSII, illuminated, and then, 1 mM HCO₃⁻ added (triangles). (D) Effect of preillumination on HCO₃⁻ depletion: redox titrations of ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ in intact PSII (triangles). The red triangle shows the level of fluorescence when a sample (in this case, it had been frozen and thawed once) was poised at −65 mV in the dark for 12 min. The sample was then illuminated (4 min, 620 nm, 273 μmol photons m⁻² s⁻¹), dark-adapted (10 min), and poised again at −65 mV (8 min). The arrow shows the light-induced increase in fluorescence. A titration was then performed. The data are the average of two experiments, which gave similar results. A standard “dark titration” is also shown (data from Fig. 2). Black symbols, oxidative titration; white symbols, reductive titration. SHE, standard hydrogen electrode.

A shift of +74 mV in the E_m on dissociation of HCO₃⁻ (−74 mV on HCO₃^– binding) indicates that the binding of the HCO₃⁻ should become weaker when Q_A⁻ is present as shown in Fig. 2D based on the relationship in the following equation:

$$E_m^b=\hspace{0.17em}{E}_m^d+\hspace{0.17em}\frac{RT}{nF}\text{ln}\hspace{0.17em}\left(\frac{K_d^{ox}}{K_d^{red}}\right),$$ [2]

where E_m^b and E_m^d are the reduction potentials when the HCO₃⁻ is either bound or dissociated, respectively; K_d^ox is the HCO₃⁻ dissociation constant when Q_A is oxidized; and K_d^red is the HCO₃⁻ dissociation constant when Q_A is reduced.

This effect is confirmed in Fig. 3C, which shows the influence of a period of illumination on electron transfer from ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ to Q_B (or ${\text{Q}}_{\mathrm{B}}^{-\mathrm{•}}$) as monitored by fluorescence decay after a flash. The characteristic slowdown in the rate of forward electron transfer from ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ associated with HCO₃⁻ removal is seen after the illumination of a sample in a low-HCO₃⁻ medium. This slowdown of ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ oxidation is reversed when HCO₃⁻ is added back to the medium.

Fig. 3D shows a redox titration experiment aimed at testing the prediction that ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ formation could result in HCO₃⁻ release. A sample without added HCO₃⁻ was poised at −58 mV, close to the E_m of the HCO₃⁻-depleted, Mn₄CaO₅-containing PSII. The red triangle in Fig. 3D shows the fluorescence level after 15 min at this potential. As expected, the fluorescence remained low, because no ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ was present at this potential in an HCO₃⁻-containing sample. The poising potential was then switched off, and the sample was allowed to relax in the dark for 10 min before being illuminated by 4 min of blue light to generate ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$. After the illumination, the dark sample was again poised at −58 mV. The fluorescence level had increased to 50% of the maximum value (as shown by the arrow in Fig. 3D) [i.e., the level expected if HCO₃⁻ had been released (Fig. 2A)]. An abbreviated redox titration was then performed on the same sample, and a reversible redox transition was seen corresponding to the value obtained earlier for the HCO₃⁻-depleted PSII. In a control experiment, exactly the same procedure was carried out with a dark period given instead of the illumination and when repoised at −58 mV; the fluorescence level was the same as that recorded before the dark incubation. The experiment in Fig. 3D is an additional indication that illumination favors release of the HCO₃⁻. The positive shift in the redox potential of ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ is predicted to increase the energy gap between ${\text{P}}^{+\mathrm{•}}{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ and ${\text{P}}^{+\mathrm{•}}{\text{Ph}}^{-\mathrm{•}}$ and thus, decrease back-reactions via the P^+•Ph^−• and consequently, the formation of the reaction center triplet state and ¹O₂ (13) as shown in Fig. 4A. Fig. 4B shows the effect of HCO₃⁻ depletion on light-induced ¹O₂ generation monitored using EPR to detect an ¹O₂-specific spin trap. The HCO₃⁻-depleted sample shows decreased levels of ¹O₂ formation (<60%) compared with the controls (SI Appendix, Table S1), verifying the expectations of the model (13).

Fig. 4.

Charge recombination pathways and singlet O₂ generation in PSII. (A) Energy scheme illustrating the outcomes of charge recombination pathways in PSII and the influence of bicarbonate release. The scheme shows the competition between two dominant routes (13): (1) the triplet route is favored when HCO₃⁻ is present and the energy gap between the ${\text{P}}^{+\mathrm{•}}{\text{PheQ}}_{\text{A}}^{-\mathrm{•}}$ and ${\text{P}}^{+\mathrm{•}}{\text{Phe}}^{-\mathrm{•}}$ is small, and (2) the direct route is favored when HCO₃⁻ is not bound when the energy gap is bigger by ∼74 mV as measured here. The question mark at the P^+•Phe^−• energy indicates that this has not been measured for the HCO₃⁻-depleted PSII. The changes in standard free energy are not to scale. This scheme is simplified: for example, (i) it omits details of the first radical pair formation, and (ii) it does not specify which pigments are considered to be P, P*,³P, and P⁺ (see 2 and 3). (B) Generation of ¹O₂ in HCO₃⁻-depleted PSII (row 1), after addition of 1 mM NaHCO₃ (row 2), and in control samples (rows 3 and 4) measured by EPR using the spin probe TEMPD. Typical spectra are shown (additional data are in SI Appendix).

Discussion

E_m of ${\mathbf{Q}}_{\mathbf{A}}/{\mathbf{Q}}_{\mathbf{A}}^{-\mathbf{•}}$: Explaining the Discrepancy in the Literature.

Here, we show that removal of bicarbonate (HCO₃⁻) from its site on the nonheme iron (Fe²⁺) in PSII shifts the E_m of Q_A/Q_A^−• by ∼+74 mV. The E_m value obtained for PSII in the presence of the HCO₃⁻ (−144 mV) agrees with those in the work by Shibamoto et al. (21), which used a comparable spectroelectrochemical method (21, 22). Furthermore, the values obtained here without HCO₃⁻ correspond quite well to those reported in the potentiometric studies of Krieger and coworkers (12–14), strongly suggesting that the PSII used in those studies was HCO₃⁻-depleted. Comparing the two methods, we explain the loss of HCO₃⁻ in the latter work (12–14, 23) as follows: (i) the lack of mediators meant longer equilibration times; (ii) in the potentiometric method, the sample was stirred under a stream of Ar or N₂, whereas the sample in the thin-layer cell is not stirred, and there is little or no sample surface exposed to inert gas; and (iii) frozen and thawed samples were used, and these samples seem to lose the HCO₃⁻ more easily (SI Appendix).

Relevant to this argument is a report in which Allakhverdiev et al. (24) verified the values reported by Shibamoto et al. (22) but used the method by Krieger et al. (12). This finding, however, does not contradict the present explanation, because the work in ref. 24 was done at pH 7.0, well above the pK_a 6.35 of H₂CO₃/HCO₃⁻, and this pH would significantly inhibit the loss of HCO₃⁻. A comparison of the most relevant E_m values in the literature with these results is given in SI Appendix, Table S2.

These findings resolve the long-standing discrepancy and reveal another factor influencing the E_m of ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$, HCO₃⁻ binding. Overall, four valid values for the E_m values for ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ have now been defined depending on the state of the PSII: with and without Ca²⁺ (and the rest of the Mn₄CaO₅ cluster) and with and without the HCO₃⁻ ligand to the nonheme iron.

Table 1 shows the measured E_m values and those corrected for (i) “connectivity” of the PSII centers (27) (details are in SI Appendix) and (ii) both connectivity and the measurement temperature (15 °C) calculated based on the assumption that, for a small temperature range, the average dE/dT was 1 mV/K (28). The functional value at 25 °C of −145 mV can be used for anchoring bioenergetics schemes of PSII. The other three corrected values are also relevant, because they represent conditions that are likely to be encountered by PSII in vivo (e.g., during photoassembly of the Mn₄CaO₅ cluster, before repair when photoinhibited, and/or when ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ is long-lived).

Table 1.

Experimental and corrected Q_A/Q_A⁻ reduction potentials

Sample	Measured Em	Corrected Em
		Em (con)*	Em (con, 25 °C)†
PSII + HCO3	−138 ± 2	−134 ± 2*	−144 ± 2†
PSII − HCO3	−61 ± 2	−60 ± 2*	−70 ± 2†
−Mn PSII + HCO3	−23 ± 2	−22 ± 2*	−32 ± 2†
−Mn PSII − HCO3	+65 ± 2	+64 ± 2*	+54 ± 2†

^*E_m corrected for the nonlinear behavior of chlorophyll fluorescence (connectivity) based on the excitation energy transfer in PSII dimer complexes (details are in SI Appendix).

^†E_m corrected for the connectivity and the temperature difference between the experimental temperature (15 °C) and the standard conditions (25 °C) with an average dE/dT of −1 mV/K.

Relevance to Donor-Side Effects of Bicarbonate.

A role for HCO₃⁻ on the electron donor side of PSII has been suggested (5). However, its absence in the high-resolution crystal structures (8) confirmed the results of other studies (29, 30) that HCO₃⁻ was not bound to the electron donor side of PSII. A secondary nonbound donor-side role has not been ruled out, however (5). A role of HCO₃⁻ in photoassembly of the cluster has been suggested previously, and remains possible (5, 16). A role of HCO₃⁻, enhancing proton removal from the site of water splitting, was recently reported (10), and it was suggested that this could represent a regulatory link between CO₂ and water splitting (10). The present work also shows a mechanism in which the CO₂ level can control PSII activity (see below). Future research will need to distinguish between these two mechanisms.

The “donor-side” studies of PSII are often based on the idea that the acceptor-side HCO₃⁻ has a single fixed binding constant for HCO₃⁻ bound to the Fe²⁺, keeping it permanently bound to PSII under physiological conditions (10). These results change that premise, and thus it may be worthwhile reassessing some phenomena attributed to the donor side. (Additional discussion on donor-side effects on the E_m values of Q_A is provided in SI Appendix.)

Redox Effect of HCO₃⁻: Significance on Charge Recombination.

An HCO₃⁻-dependent shift in the E_m of ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ has not been reported previously as far as we are aware (3–6). An effect of HCO₃⁻ removal on the E_m of ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ was looked for but was not observed (31). There are two potential explanations for this: (i) formate was added to the sample to aid HCO₃⁻ depletion, and we show here that, when formate is present, the redox shift is much less marked; and (ii) the titrations reported (31) were not reversible, presumably because of the loss of the Mn cluster as shown subsequently (12–14).

Since the introduction of the concept of competing charge recombination routes to explain the influence of switching between high and low potential forms of Q_A in PSII (13), any changes in the energy gap between P^+•Ph^−• and ${\text{P}}^{+\mathrm{•}}{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ have been interpreted within this model as favoring or disfavoring back-reactions via the ³P-mediated ¹O₂ formation (18–20, 32). The HCO₃⁻ effect reported here can be interpreted in the same way. Loss of the HCO₃⁻ shifts the E_m by +74 mV, thus increasing the energy gap between the P^+•Ph^−• and ${\text{P}}^{+\mathrm{•}}{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ and making the back-reaction to ${\text{P}}^{+\mathrm{•}}{\text{Ph}}^{-\mathrm{•}}$ less likely, and consequently, less ³P-mediated ¹O₂ is expected to be formed (Fig. 4A). Here, we show that this prediction holds true (Fig. 4B). Thus, the loss of HCO₃⁻ is predicted to protect PSII from this kind of photodamaging reaction.

In the extensive literature on photoinhibition of PSII, there are some reports investigating the removal of HCO₃⁻, and they show both positive and negative correlations with photoinhibition, depending on the conditions (33–35). Given the wide range of conditions used and the multiple mechanisms possible for damaging PSII, the contradictory observations are not too surprising. The mechanistic role of HCO₃⁻ and its correlation with diminished ¹O₂ formation shown in this work will act as a starting point for future research aimed at assessing the physiological significance of this protective mechanism.

Effect of the Redox State of Q_A on the Binding of HCO₃⁻.

The +74-mV shift in the E_m of ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ implies that the presence of ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ should diminish the binding affinity of HCO₃⁻. The binding of HCO₃⁻ should be weaker by a factor of 10 for every 59-mV shift in the potential, according to Eq. 2. In this case, then the +74-mV shift implies a weakening of the binding constant by a factor of ∼17.9 (Fig. 2D). This weak binding would be predicted to result in loss of the bicarbonate when Q_A^−• is reduced, depending on the ambient concentration of HCO₃⁻. Under the conditions of our experiment, illumination of the sample to maintain Q_A^−• reduced led to HCO₃⁻ loss as shown by the typical slowing of forward electron transfer from ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ (Fig. 2C) and characteristic shift in the E_m of ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ (Fig. 2D). We duly interpret this as confirming that formation of Q_A^−• results in HCO₃⁻ release from the nonheme iron.

The influence of illumination on HCO₃⁻ binding has been investigated previously (36–39), but it has not been mentioned in recent work (3–5). In general, the observations made in this work that Q_A⁻ formation results in HCO₃⁻ release are consistent with the diverse observations in the literature (36–39). However, in the previous work, there was no suggestion made linking the redox state of Q_A and the binding of the HCO₃⁻.

The nature of the interaction between the HCO₃⁻ and ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ could be (i) at least partially electrostatic, because the two anions are within 10 $\AA$ of each other, and (ii) electronic, transmitted through the Fe²⁺ to the D2-His214 ligand, weakening its H bond to Q_A (Fig. 1). It seems reasonable to expect that both mechanisms would lower the redox potential of ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$.

These findings are relevant to an earlier suggestion based on X-ray absorption studies that the bidentate carboxylic acid binding to the nonheme iron becomes transiently monodentate on ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ formation (40). That said, it also seems possible that the presence of an uncompensated anion on Q_A could make binding of the HCO₃⁻ anion less strong without the formation of a discrete monodentate intermediate. This point is worth investigating further to clarify the specific molecular mechanism.

When considering electrostatic and electronic effects, both formate and HCO₃⁻ have the same charge, and yet, formate has a much smaller effect. A possible explanation for this apparent anomaly is that HCO₃⁻ differs from formate and most other relevant carboxylic acids in that it can undergo an additional ionization, forming the CO₃²⁻ anion with a pK_a 10.3. Cox et al. (41) proposed that the presence of CO₃²⁻ rather than HCO₃⁻ as a ligand to the Fe²⁺ could explain the specific EPR signal arising from the semiquinone complex in PSII (6, 41). The lower pK_a needed to make the model feasible was attributed to an association between the HCO₃⁻ and a lysine, D2-K264 (41). Subsequently, with the availability of the high-resolution crystal structure (8), it was pointed out (7) that this lysine was, in fact, involved in an ion pair with a nearby glutamate, D1-E244 (Fig. 1). Nevertheless, we consider that the pK_a on HCO₃⁻ could still be significantly shifted by other factors: (i) binding to the Fe²⁺, (ii) the presence of H bonds to HCO₃⁻ from the symmetrical tyrosine residues D2Y268 and D1Y246, and (iii) the distal O of the HCO₃⁻ being part of an extended H-bond network (Fig. 1), which could allow dynamic deprotonation and protonation reactions. We, thus, consider that CO₃²⁻ remains a valid candidate as the native ligand to the Fe²⁺.

The light-induced loss of the HCO₃⁻, shown here, occurred in a low-CO₂ environment. Our experiments were done at pH 6.5, which was chosen for the stability of the Mn cluster. The pK_a of HCO₃⁻/carbonic acid is at pH 6.35, and therefore, at pH 6.5, the concentration of HCO₃⁻ present in the buffer was low (17.5 µM) and relatively easily lost by flushing with Ar or N₂. The question thus arises whether HCO₃⁻ loss will occur under physiological conditions where the pH is higher. The dissociation constant for HCO₃⁻ has been estimated to be 40–80 µM, whereas the concentration of HCO₃⁻ in the stroma at ∼pH 8 is estimated to be 220 µM (9, 42). Thus, it has been generally assumed that the HCO₃⁻ should be constantly bound to the site (10). A weakening of the HCO₃⁻ binding, with the estimated K_d increasing by ∼18 times (Fig. 2D), on formation of ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$, as reported here, should allow dissociation of the bicarbonate even under physiological conditions. This conclusion needs to be verified experimentally. However, because it is known that, under physiological conditions, acetate is able to replace HCO₃⁻ in vivo (11), it seems very likely that the ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$-dependent HCO₃⁻ exchange will occur under some physiological conditions. It is also worth examining the possibility that other natural ligands (e.g., glycolate, malate, and glycerate) in the stroma could replace HCO₃⁻ under some conditions.

Conclusions and Working Model.

This study resolves the long-standing discrepancy concerning the E_m of ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ (described in the Introduction and refs. 12 and 21) and thus, provides a firm basis for understanding the bioenergetics of PSII (Fig. 4A). It also shows that HCO₃⁻ binding to the nonheme iron has a significant influence on the redox potential of Q_A. Correspondingly, the redox state of Q_A influences HCO₃⁻ binding, with the ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ state favoring its dissociation. This work also provides a mechanistic explanation for the presence of HCO₃⁻ in PSII, and it rationalizes the long-studied bicarbonate effect in terms of an unexpected redox-based protective mechanism.

The principle aspects of this protective mechanism are described in what follows. Normal photosynthesis can lower the ambient concentration of CO₂ to the point when it becomes limiting, resulting in the overreduction of the electron transfer components and leading to the accumulation of ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ (43). From this work, the presence of ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ would be expected to result in weaker binding of HCO₃⁻ to the nonheme iron, and thus, HCO₃⁻ would be released. This HCO₃⁻ loss would not only slow down electron transfer to the plastoquinone pool (9, 42) (and thus, water splitting and O₂ formation), but importantly, it would also change the potential of ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ to a more positive value (Fig. 4A). This higher potential increases the energy gap between P^+•Q_A^−• and ${\text{P}}^{+\mathrm{•}}{\text{Ph}}^{-\mathrm{•}}$, favoring the harmless, direct recombination of ${\text{P}}^{+\mathrm{•}}{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ over the back-reaction via P^+•Ph^−• (13)_. This direct route is beneficial, because the back-reaction route forming P^+•Ph^−• leads to formation of the chlorophyll triplet, ³P, which reacts with O₂ to form the highly toxic reactive oxygen species ¹O₂ (Fig. 4). When the CO₂ level returns to normal, which would be favored by the inhibition of PSII activity, linear electron transfer and CO₂ fixation would resume, reoxidizing the plastoquinol pool and ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$. HCO₃⁻ would then rebind to the nonheme iron, returning the potential of the ${\text{Q}}_{\mathrm{A}}/{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ couple to its low potential form, restoring normal proton-coupled electron transfer between the quinones, and returning electron transfer out of PSII to normal rates.

This simplified model does not take into account the numerous processes that regulate electron transfer, CO₂ fixation, and photorespiration, but the basic idea is straightforward, intuitive, and consistent with the results here and in the literature (43). Also, the model as presented does not specify or require that CO₂, the true terminal electron acceptor, through its equilibrium with HCO₃⁻ regulates the function of PSII. Indeed, with the dissociation constants reported in the literature (40–80 µM), the HCO₃⁻ bound to PSII would be released when Q_A⁻ is present even under the normal (i.e., nonlimiting) concentrations of CO₂ in the stroma, where HCO₃⁻ is estimated to be 220 µM (9, 42). However, we have two doubts about this point. First, we doubt the validity of the 40–80 µM dissociation constant: our samples were maintained at pH 6.5 for long periods, the expected HCO₃⁻ concentration in our samples at equilibrium with atmospheric CO₂ should be 17 µM, and what we see is just a fractional loss of HCO₃⁻. It thus seems likely that the K_d is lower than the current estimates. Second, the stroma can reach pH values higher than pH 8.0, leading to significant increases in HCO₃⁻ concentration. Given these two factors, it remains possible that the accumulation of Q_A^−• may result in HCO₃⁻ release only when the HCO₃⁻ concentration is low, potentially coinciding with CO₂-limiting conditions. If so, the mechanism could then represent a sink (CO₂) to source (PSII) regulation mechanism. To test this possibility, it is worth investigating the following: (i) the conditions in which this regulation mechanism actually occurs in vivo; (ii) the detailed chemical parameters that determine the mechanism, including a more accurate determination of the HCO₃⁻ K_d in a range of conditions; and (iii) the rate of binding and dissociation of HCO₃⁻ to determine how dynamic the mechanism can be. (SI Appendix has an example of published data that may be reinterpreted in light of the mechanism.)

The protective/regulation model also provides the explanation for the presence of HCO₃⁻ in PSII. Nonoxygenic Type II reaction centers lack HCO₃⁻ as a ligand to the nonheme iron. Instead, a glutamate provides a nonexchangeable but otherwise structurally similar bidendate ligand. It has been pointed out that these nonoxygenic Type II reaction centers do not undergo back-reactions from ${\text{P}}^{+\mathrm{•}}{\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ to ${\text{P}}^{+\mathrm{•}}{\text{BPh}}^{-\mathrm{•}}$ (where BPh is bacteriopheophytin), because they have a much bigger energy gap between these states than is present in functional PSII; therefore, this back-reaction does not need to be regulated in the same way as it does in PSII (18). It also seems reasonable to suppose that the nonoxygenic Type II reaction centers, with their cyclic electron transfer system, single reaction center, and low-O₂ environment, would need much less in terms of protective regulation compared with the situation encountered by PSII in oxygenic photosynthesis. A protective role for HCO₃⁻ aimed at diminishing ¹O₂ formation also implies that this type of regulation evolved in an O₂-containing environment, perhaps during the phase of PSII evolution when water could be split but still only inefficiently (44).

Materials and Methods

PSII-enriched membranes were prepared from spinach (45) with modifications shown in ref. 46. Mn depletion was done using NH₂OH (47). O₂ evolution was assayed with a Clark-type oxygen electrode (Oxylab; Hansatech) at 25 °C with 0.5 mM dichlorobenzoquinone/1 mM K₃Fe(CN)₆ using saturating red light (590-nm cutoff filter; 5,000 μE m⁻² s⁻¹). The activity in the various preparations was 300–500 μmol O₂ mg chlorophyll⁻¹ h⁻¹. HCO₃⁻ was depleted (48), and the PSII was stored on ice in the dark under Ar until use. If not indicated otherwise, the samples used for the spectroelectrochemical titrations were prepared and used on the same day. Formate treatment of PSII membranes was as described in ref. 48. Spectroelectrochemical redox titrations were performed as reported earlier (21, 22) with minor differences (details are in SI Appendix). The redox state of ${\text{Q}}_{\mathrm{A}}^{-\mathrm{•}}$ was monitored by measuring chlorophyll a fluorescence and corrected for connectivity (additional details are in SI Appendix). Singlet oxygen was trapped using the water soluble spin probe 2,2,6,6-tetramethyl-4-piperidone hydrochloride (TEMPD) (11) and measured with EPR (SI Appendix). Samples (10 µg chlorophyll mL⁻¹) were illuminated for 2 min with 500 µmol quanta m⁻² s⁻¹ red light (RG 630).