Competing charge transfer pathways at the photosystem II-electrode interface.

The integration of the water-oxidation enzyme photosystem II (PSII) into electrodes allows the electrons extracted from water oxidation to be harnessed for enzyme characterization and to drive novel endergonic reactions. However, PSII continues to underperform in integrated photoelectrochemical systems despite extensive optimization efforts. Here we carried out protein-film photoelectrochemistry using spinach and Thermosynechococcus elongatus PSII, and we identified a competing charge transfer pathway at the enzyme-electrode interface that short-circuits the known water-oxidation pathway. This undesirable pathway occurs as a result of photo-induced O2 reduction occurring at the chlorophyll pigments and is promoted by the embedment of PSII in an electron-conducting fullerene matrix, a common strategy for enzyme immobilization. Anaerobicity helps to recover the PSII photoresponse and unmasks the onset potentials relating to the QA/QB charge transfer process. These findings impart a fuller understanding of the charge transfer pathways within PSII and at photosystem-electrode interfaces, which will lead to more rational design of pigment-containing photoelectrodes in general.

Photosystem II (PSII) is a 700 kDa dimeric pigment-protein complex that resides in the membranes of oxygen evolving photosynthetic organisms.1 It is unique in nature in its ability to photocatalyze water-oxidation to liberate H+, O2 and electrons, which is the first step of oxygenic photosynthesis. The study of PSII functionality is important in both fundamental and applied sciences,2 in particular for providing lessons for solar fuel technologies where water-oxidation is a bottleneck.3

Protein-film photoelectrochemistry (PF-PEC) is a direct and sensitive method of characterizing light-induced redox events arising from photoactive proteins on an electrode surface.4,5 Recent advancements in electrode design6,7 has enabled PF-PEC to become a platform in which photosystems could be characterized,4 used in photocatalyst benchmarking,8 and used to drive novel endergonic reactions.6,9 For example, PSII was recently coupled to a hydrogenase photoelectrochemically to drive light-driven water-splitting via an unnatural photobiological pathway.6 The study followed the development of a hierarchically structured inverse opal mesoporous indium tin oxide (ITO) electrode that enabled exceptionally high loadings of PSII and the subsequent quantification of O2 evolution.

The mechanism by which PSII performs water-oxidation photocatalysis is shown in Figure 1A. Light energy is harvested mainly at the chlorophyll (Chl) a pigments in the CP43 and CP47 antenna subunits, and funneled to the reaction center (RC) complex. Due to the energetic disorder, ultrafast charge generation in the RC proceeds via two pathways depending on protein configuration,10,11 and quantum coherence promotes the high efficiency of the initial charge separation events.12 The photogenerated electrons are passed via a pheophytin to an immobile plastoquinone, QA, then to a mobile plastoquinone, QB.13 Upon two electron reductions, the QB diffuses out of its pocket and shuttles the electrons to a cytochrome b6f within the thylakoid membrane. Concurrently, the photogenerated holes at the RC complex move in the opposite direction via a tyrosine to enter the Kok cycle, the catalytic cycle for water-oxidation.

Figure 1

Schematic representation of the energy/charge transfer pathways within (A) PSII in the thylakoid membrane of a cyanobacterium; and (B) PSII adsorbed on a mesoporous indium-tin oxide (ITO) electrode connected to a photoelectrochemical cell.

The sequence of the water-oxidation pathway is given by the numbers in bold. The orange arrows represent energy transfer pathways resulting from light (hυ) absorption. The blue arrows represent electron transfer pathways taking place during oxygenic photosynthesis that give rise to the photoanodic currents detected by PF-PEC. The red arrows represent the competing O2 reduction pathways that give rise to the photocathodic pathways detected by PF-PEC. Electron acceptors such as fullerenes (C60) can enhance both the photoanodic and photocathodic pathways. The solid arrows represent known pathways including: energy transfer at the chlorophyll a (Chl a), charge generation/separation at the reaction center (RC), electron transfer via the pheophytin (Pheo), mediated QB/chemical mimic charge transfer,42,43 direct electron transfer from QA to the electrode,7 hole transfer via the tyrosin (Tyr) and water oxidation at the Mn4Ca cluster. The dashed arrows represent new pathways identified using protein-film photoelectrochemistry under a three-electrode configuration comprising of a working (W), counter (C) and reference (R) electrode. Note that the depiction of the photocathodic pathway at the antenna CP43 subunit is arbitrary.

The slow step in PSII charge transfer is the diffusional step between PSII’s terminal electron acceptor QB and the subsequent electron acceptor.4 However, when a high concentration of electron mediators is present or the PSII is efficiently wired to an electrode to allow direct interfacial electron transfer (ET), turn-over-frequency values of up to 250 mol O2 (mol PSII)–1 s–1 can in principle be achieved, calculated based on the slowest step in the Kok cycle.14 Despite extensive efforts to optimize the ET efficiency between PSII and the electrode, the turn-over-frequency values reported for immobilized PSII on electrodes typically fall below 15 s–1.4,6 The low ET efficiency exhibited by PSII at the electrode interface is currently poorly understood.

Due to the large dimension of PSII (11 nm × 10 nm × 20 nm per dimer) and the spatial arrangements of the cofactors within the enzyme (Supplementary Results, Supplementary Figure 1), PSII must be interfaced with the electrode via its stromal side to enable efficient electronic communication between the QA/QB and the electrode. Some progress has been made in controlling PSII orientation on electrodes using surface linkers endowed with charged functionalities and Histags.15,16 Additionally, PSII is commonly embedded in electron-accepting matrices or layer-by-layer structures comprising of redox-active polymers or carbon-based sp2 hybridized materials,17,18 which allows multiple layers of PSII to be electronically connected to the electrode regardless of orientation. Thus far, all efforts to more efficiently wire PSII to the electrode have not resulted in the step change in performance desired.

Here, we identified a new light-induced charge transfer pathway at the PSII-electrode interface that competes against the known water-oxidation pathway (Figure 1B). We reveal that the competing pathway arises from antenna Chl a pigments participating in the photo-induced reduction of O2 near the electrode. This pathway could be attenuated/amplified by the O2 concentration, and the presence of an electron-accepting matrix. Anaerobicity enhances the photoresponse of PSII and unmasks the true onset potential of the photoanodic currents, allowing for the QA/QB charge transfer step to be studied using PF-PEC.

Results

This study used PSII core complexes derived from two different sources: the higher plant spinach, and the cyanobacterium Thermosynechococcus elongatus (T. elongatus). Inverse opal mesoporous ITO electrodes (Supplementary Figure 2) were employed for to their ability to accommodate high and stable enzyme loadings.6

Photoelectrochemical water-oxidation with PSII

We first characterized the photoresponse of spinach and T. elongatus PSII by performing stepped chronoamperometry scans under chopped light illumination on the two PSII adsorbed on ITO electrodes (Figure 2). The photoresponses have been normalized to reflect the photocurrent observed per mg Chl a deposited on the electrode. Both PSII gave rise to photoanodic currents at applied potentials above 0.2 V vs SHE, which have previously been assigned to water-oxidation.7,8 Unassigned photocathodic currents can be observed below 0.2 V vs SHE (see next section). Bare (protein-free) ITO electrodes did not give rise to photocurrents under the same conditions (Supplementary Figure 3).

Figure 2

PSII photocathodic currents stem from protein-bound and isolated Chl a pigments.

Stepped chronoamperometry scans of PSII core complexes isolated from T. elongatus (dark blue) and spinach (light blue); and spinach CP43 antenna subunits (orange), spinach RC complexes with the Mn4Ca cluster depleted (red), Chl a molecules (green), and β-carotene (β-car) molecules (black) adsorbed on hierarchically structured ITO electrodes. Potentials steps of 0.1 V were scanned in the anodic direction. All experiments were performed in a MES buffer electrolyte solution (pH 6.5) under aerobic conditions at 25 °C with chopped light illumination (679 nm, 5 mW cm–2). Representative photo-responses are shown and the applied potentials are referenced against the standard hydrogen electrode (SHE); dark current spikes caused by the potential changes were removed for clarity. The photoresponses were normalized to the concentration of Chl a deposited on the electrode (5 μg Chl a cm–2 electrode surface area).

The photoanodic response can be explained by the energy/charge transfer mechanisms depicted in Figure 1B. When the Fermi level of the electrode (controlled by the applied potential) is more positive than the redox potentials of the QA/QB within PSII, PSII in the correct orientation can transfer electrons directly to the electrode when illuminated to result in a rise in anodic current.

Photo-induced Chl a charge transfer in PSII

Previous PF-PEC studies have typically measured photocurrents densities at applied potentials of > 0.4 vs SHE to determine PSII performance.4 As such, the photocathodic currents observed below 0.2 V vs SHE (Figure 2) have not been reported and are unassigned. The origin of the photocathodic current is unlikely to involve the hole-transfer cofactors present in the water-oxidation pathway since the tyrosine and Mn4Ca cluster are both buried 3 nm deep within the protein shell.1 The photocathodic response is more likely to stem from interactions between the electrode and photo-responsive co-factors that reside near the surface of the protein shell such as the many Chl a and β-carotene (β-car) pigments located in the antenna subunits (each antenna subunit has around 15 Chl a pigments, most of which are within 20 Å from the protein shell in the transmembrane region of the enzyme,7,19,20 see Supplementary Figure 1). This is made more likely by the mesoporous substructure of the electrode, which can contact multiple sides of individual enzymes.

To identify the source of the photocathodic currents, we performed stepped chronoamperometry scans with chopped illumination on isolated antenna CP43 and RC complex subunits (Figure 1), as well as on Chl a and β-car molecules. The isolated RC complex used in this study has lost the plastoquinones, non-heme Fe and the Mn4Ca cluster in the isolation process.

We observed clear photocathodic currents for the adsorbed antenna CP43, RC complex, and the isolated Chl a molecules (onset potential ~0.5 V vs SHE, Figure 2), but not for the β-car molecules. The observation that all Chl a-containing components generated photocathodic currents indicates that photo-induced ET can occur from the ITO to free Chl a molecules, as well as to Chl a within the protein shell. The adsorbed RC complex gave rise to larger photocathodic currents than the adsorbed PSII core complexes and the antenna subunit, which may be attributed to the charge transfer-active Chl a pigments within the exposed RC complex forming electronic contact with the electrode surface. It also cannot be excluded that other exposed cofactors, such as the Tyr+ can participate in charge transfer. The antenna CP43 subunit gave rise to a relatively smaller normalized photocathodic response despite having the highest density of embedded Chl a pigments. This may be attributed to less favorable protein-electrode interactions due to the presence of the hydrophobic transmembrane domain.

To better understand the nature of the photocurrents, we obtained the action spectra (photoresponse at different wavelengths) of the PSII core complexes and isolated Chl a pigments on ITO electrodes at wavelengths 720 to 460 nm at applied potentials of 0 and 0.5 V vs SHE (Supplementary Figure 4). At 0.5 V vs SHE, bare ITO exhibited photoanodic response at wavelengths < 540 nm, and adsorbed Chl a molecules exhibited no photoresponse (taking into account the ITO background). Photoanodic responses (maximum at ~680 nm) were observed for both PSII, which corroborates with their respective absorption profile (Supplementary Figure 4A). At 0 V vs SHE, bare ITO exhibited no photoresponse, and adsorbed Chl a exhibited photocathodic responses (maximum at 660-680 nm) consistent with its absorption spectrum. Photocathodic responses (maximum ~680 nm) were observed for both PSII core complexes, which is slightly red-shifted compared to free Chl a and consistent with their absorption profile (Supplementary Figure 4B). This is further evidence that the photocathodic response stems from the PSII core complexes, with large involvement from the Chl a pigments.

The photoresponse of Chl a films on semiconductors has been studied in the past.21,22 Photoexcited states (Chl a*) were shown to transfer electrons to the conduction band of the semiconductor to give rise to photoanodic currents, followed by the regeneration of oxidized Chl a by a chemical electron donor.23,24 However, in the absence of the electron donor and in the presence of O2 under a relatively negative applied potential, Chl a* has been observed to produce a photocathodic current,23 which corroborates with our findings. The authors proposed that the ET from the Chl a* to the O2 is the primary step in the mechanism, followed by hole migration from the electrode to the Chl a+ (Chl a+, Em: ~0.78 V vs SHE)25. The alternative mechanism is that hole migration from the photoexcited Chl a* to the electrode (at applied potential < ~0.78 V vs SHE) is the primary step, followed by the reduction of dissolved O2 (Em (O2/O2•–): ~ –0.33 V vs SHE)26 by the Chl a– (Em: ~ –1.1 V vs SHE)25,27. A recent study has reported that organic dyes adsorbed on ITO electrodes can exhibit analogous photocathodic behavior in the presence of suitable electron acceptors such as I3– under a relatively negative applied potential.28 The authors used transient absorption spectroscopy to show that the primary step is the photo-injection of the hole from the photoexcited dye into the conduction band of the ITO electrode, which supports the second proposed mechanism where Chl a– is generated.

Photoelectrochemical O2 reduction by PSII

In water-splitting systems, the presence of O2 is unavoidable since it is liberated in the process of water-oxidation. Although photoexcited Chl a triplet states (3Chl a) within PSII are known to undergo energy transfer processes with O2 (via 3Chl a-3O2 annihilation)29 to generate 1O2, charge transfer between Chl a and O2 have not been reported. To clarify the role of O2 as an electron acceptor in the observed photocathodic process, we repeated stepped chronoamperometry scans of adsorbed PSII core complexes and Chl a molecules in the presence and absence of atmospheric O2 (Figure 3). In anaerobic conditions, the photocathodic pathway was diminished in all cases, confirming the role of O2 as an electron acceptor in the proposed pathway (Figure 1B). Anaerobicity increased the photoanodic currents arising from the PSII core complexes most significantly at applied potentials of ≤ 0.3 V vs SHE since the onset potential of the competing photocathodic current is around 0.3 V vs SHE.

Figure 3

O2 and the fullerene derivative, C60-DMePyI, are electron acceptors for photoexcited Chl a.

Photoresponse of (A) Chl a, (B) spinach PSII core complexes, and (C) T. elongatus PSII core complexes adsorbed on ITO electrodes in the presence (black traces) and absence (red traces) of a C60-DMePyl matrix. Stepped chronoamperometry was performed under both aerobic (air, solid traces) and anaerobic (Ar purged, dashed traces) conditions in MES electrolyte buffer solution (pH 6.5) at 25°C with chopped red light illumination (679 nm, 5 mW cm–2). Potentials steps of 0.1 V were scanned in the anodic direction and representative photoresponses are shown. (D) Proposed mechanisms of photocathodic current generation by Chl a* in the presence of O2, with enhanced photocathodic currents in the presence of a C60-DMePyl matrix.

To elucidate the rate-limiting step in the photocathodic current generation, we repeated the experiment in electrolyte solutions purged with different concentrations of O2 gas (Supplementary Figure 5). We observed the photocathodic currents increased with the concentration of O2 introduced into the system, which is consistent with the ET from the Chl a to the O2 being the rate-limiting step. This is not surprising as the reduction of O2 by Chl a* is diffusion rate-limited, unlike the charge transfer between the electrode surface and the photoexcited pigment.

Due to the low concentrations of reduced O2 species expected to be produced, we employed electrochemical techniques to detect any products from O2 reduction present on the electrode surface following irradiation. We performed cyclic voltammetry (CV) scans on adsorbed Chl a under aerobic conditions (Supplementary Figure 6A) and observed currents from O2 reduction at potentials below 0 V vs SHE in the absence of light. Following prolonged red-light illumination (> 5 min), we observed a new cathodic wave with an onset of ~0.2 V vs SHE. This was not present in the second or subsequent scans, nor in anaerobic conditions (Supplementary Figure 6B). The cathodic wave with an onset of ~0.2 V vs SHE did not appear when we repeated the experiment under aerobic conditions with catalase present in the electrolyte solution. To determine whether the photogeneration of the cathodic wave was due to H2O2 reduction, we performed a CV scan of a 0.14 mM H2O2 solution using a bare ITO electrode (Supplementary Figure 6C) under anaerobic conditions. We chose this concentration of H2O2 to correspond to half the concentration of dissolved O2 in an aqueous solution since any H2O2 formed would be from the dissolved O2. We observed a cathodic wave with an onset potential of 0.2 V vs SHE, which disappears with the addition of catalase. We therefore assign this wave to H2O2 reduction. This result affirms the activity of the catalase, and supports the photo-reduction of O2 by the Chl a to yield H2O2, most likely via the formation of O2•– followed by disproportionation to H2O2.30

To clarify whether the photo-generated reactive oxygen species (ROS) on the electrode surface serve as electron acceptors and contribute to the photocathodic pathway, we investigated the effects of stirring (to induce changes to the concentration gradient of diffusional species at the electrode surface) on the photocathodic response. We observed no differences in the photocathodic response upon the commencement of stirring during a chronoamperometry scan of adsorbed Chl a at 0.1 V vs SHE (Supplementary Figure 7). The addition of SOD and catalase also had no effect on the photocurrent density, indicating that O2•– and H2O2 do not serve as electron acceptors in the photocathodic pathway.

The embedment of PSII in a fullerene matrix

In addition to O2 acting as an electron acceptor for photoexcited Chl a pigments, the use of electron accepting matrices to improve the PSII-electrode interface may also facilitate undesirable charge escape pathways. To test this hypothesis, we performed PF-PEC on Chl a and PSII co-adsorbed with an excess of the semi-water soluble fullerene C60-(N,N-dimethyl pyrrolidinium) iodide (C60-DMePyI). The fullerene derivative was chosen for the assembly of the conductive matrix due to its high electron mobility, appropriate dimensions (~1 nm) for facile penetration through the porous electrode structure, relative hydrophilicity for interactions with PSII, and appropriate reduction potentials for the QA/QB.31

We performed background stepped chronoamperometry scans of C60-DMePyI (45 nmol) adsorbed on ITO and observed photocathodic response under aerobic conditions, but not under anaerobic conditions (Supplementary Figure 8). This indicates that the fullerene can also participate in photo-induced O2 reduction. Furthermore, the adsorbed C60-DMePyI exhibited photoanodic response both in aerobic and anaerobic conditions, possibly due to the photo-oxidization of buffer components at more positive applied potentials. Regardless, the photoanodic and photocathodic responses were relatively small at the potential window relevant for this study, and will be taken into account in the co-adsorption experiments.

We next performed the same scans on the PSII/Chl a in the presence and absence of the C60-DMePyI matrix under aerobic and anaerobic conditions (Figure 3). In the presence of the C60-DMePyI matrix under aerobic conditions, we observed enhanced photoanodic and photocathodic currents in all cases. This indicates that the non-specific interactions between PSII and the electron conducting matrix has led to enhancements of both ET pathways at the PSII-electrode interface (Figure 1B). The photocathodic enhancements were dampened under Ar purged conditions, indicating that C60-DMePyI enhances the O2 reduction pathways. This may be explained by efficient ET from 1Chl a* or 3Chl a* to the fullerene, which has been previously documented for an analogous solution system.32 The C60– has also been known to transfer electrons to acceptors, including O2.33 If the reduction of O2 by the Chl a is the slow step in the system (Figure 3D) as proposed, then the C60-DMePyI in the photocathodic current enhancement is facilitating charge transfer from the Chl a– to open up the bottleneck. Additionally, the fullerene may act as a mediator between the electrode and the Chl a pigment to further enhance the photocathodic pathway. As such, the electron conductive matrix may be facilitating the photocathodic pathway via two mechanisms.

The presence of the C60-DMePyI matrix enhanced the photoanodic response from spinach and T. elongatus PSII, but not for Chl a (taking into account the background photoanodic response of C60-DMePyI shown in Supplementary Figure 8). This is expected since the fullerene is likely to facilitate the ET from the reduced QA/QB to the ITO electrode. The photoanodic current enhancement was greatest under anaerobic conditions, particularly at applied potentials of 0.3 to 0.5 V vs SHE. Such an enhancement could be attributed to the diminishment of the O2 reduction pathway, which then leads to the electrons being more efficiently mediated from the QA/QB to the electrode, and enhanced PSII water-oxidation.

PF-PEC characterization of QA and QB

The use of PF-PEC in the characterization of the redox properties of QA/QB has not been reported previously. This may in large part be due to the photocathodic interference reported here, which masks the onset of the photoanodic signals corresponding to the ET events from the plastoquinones to the electrode. Knowledge of the redox properties of the plastoquinones is important since they are the terminal electron acceptors within PSII and serve as the electron gate that ‘opens’ the RC complex for water-oxidation. Appropriate pairing of synthetic electron acceptors to the redox potential of the plastoquinones is essential for overall PSII water-oxidation performance.

We performed stepped chronoamperometry scans on adsorbed T. elongatus and spinach PSII using small potential steps (0.025 V) near the expected redox potentials of the QA and QB, which has been reported to differ by ~80 mV.34,35 We scanned in both the anodic and cathodic directions, and observed a small difference in the onset potential. Scanning in the anodic direction minimizes the presence of O2 near the onset potential (defined here as the potential at which the photocurrent becomes a positive non-zero value) and is likely to yield a more reliable value. Figure 4 shows the normalized photoresponses (100% at 0.6 V vs SHE) scanned in the anodic direction, plotted as a function of the applied potential (raw data are shown in Supplementary Figures 9 and 10). The normalized photoresponses scanned in the cathodic direction is shown in Supplementary Figure 11.

Figure 4

Onset potential determination of (A) spinach PSII and (B) T. elongatus PSII adsorbed on ITO electrodes.

Photoresponse plots from stepped chronoamperometry scans performed in the anodic direction. Experiments were conducted in aerobic (air, grey trace), anaerobic (argon (Ar) purged, black trace) conditions; and in the presence of the QB inhibitor, 3’-(3,4-dichlorophenyl)-1’1’-dimethylurea (DCMU, blue trace) and the QB mimic, 2,6-dichloro-1,4-benzoquinone (DCBQ, red trace). All experiments were conducted in MES electrolyte buffer solution (pH 6.5) at 25°C with light illumination (679 nm, 5 mW cm–2), and the photoresponses were normalized such that the photocurrent density at 0.6 V is equivalent to 100%. The error bars represent the standard error of the mean (n = 3).

In the scans of both PSII, Ar purging produced a significant cathodic shift in the onset potentials due to the elimination of the photocathodic interference. We observed onset potentials of ~75 and 100 mV vs SHE for spinach and T. elongatus PSII, respectively. Since QB is the natural terminal electron acceptor, the onset potentials are likely to correspond to the QB if it is present, with possible flow-over contributions from QA.

The QB can be displaced by high concentrations of an inhibitor such as 3-(3,4-dichlorophenyl)-1,1-dimethylurea (DCMU), in which case the electrons are believed to transfer directly from QA to the electrode,7 allowing QA to be more selectively studied. The presence of DCMU shifted the onset potential of spinach PSII by ~100 mV. Assuming that the QB is displaced by DCMU, we assigned the onset redox potential for QA/QA– in spinach PSII to ~ –25 mV vs SHE. The presence of DCMU did not affect the onset potential of T. elongatus PSII, which may be attributed to QB being lost before the PF-PEC experiment. As such, we estimate the onset potential of the QA in of T. elongatus PSII to be ~100 mV vs SHE.

Based on the onset potentials of the photoanodic currents, the trends in Em values of the plastoquinones can be compared. In the case of spinach PSII, we estimate that the Em (QB/QB–) will be more cathodic than Em (QA/QA–) by 100 mV, which corroborates with the literature.34,35 Similarly, we estimate that the Em (QA/QA–) value of T. elongatus PSII is more positive compared to that of spinach PSII by 100 mV, which also corroborates with literature.36 The difference in Em (QA/QA–) values of PSII from different organisms may stem from variations in the protein environment of the QA pocket. A better knowledge of the structures of eukaryote PSII will help to clarify this issue in the future.

We cannot estimate the Em (QA/QA–) and Em (QB/QB–) values in a straight-forward manner from this study. Both PSII show a linear current-potential response at higher driving forces (> 0.2 V vs SHE, Fig. 4). This is expected for unmediated and randomly oriented enzymes on an electrode surface where a broad distribution in distance (and rates of interfacial ET) between the buried plastoquinones and the electrode exists.37 However, small shoulders are visible at the lower potentials (< 0.2 vs SHE) that may correspond to redox couples from the plastoquinones (Supplementary Figure 12). It may be possible to resolve these shoulders in the future, for example with better control of enzyme orientation on the electrodes or more advanced electrochemical methods.38

Previous methods of estimating Em (Q/QA–) values in PSII include the use of chemical titrations34,39 fluorescence-based spectroelectrochemistry,40 and protein film voltammetry;41 all of which have limitations. Consequently, a consensus of the values has yet to be reached. PF-PEC is a direct means of measuring ET originating from the plastoquinones in open (irradiated) PSII complexes. Based on the onset potentials, the Em (Q/QA–) values derived from PF-PEC would be significantly more positive than those obtained using traditional methods.34,39 It may be that PF-PEC introduces a positive bias in the estimation of the Em due to interfacial resistance. It may also be that the O2 produced by PSII cannot be fully removed by Ar purging alone due to in situ production of O2. To investigate this possibility, we repeated the scans for adsorbed spinach PSII in the presence of the oxygen scavenger glucose-catalase-glucose oxidase system in addition to performing Ar purging (Supplementary Figure 13). We observed a negative shift of ~25 mV in the onset potential, which indicates that further depletion of O2 can help to unmask the true PSII onset potential. However, the presence of residue O2 does not fully explain the relatively positive redox potentials measured by PF-PEC.

Regardless, PF-PEC provides valuable insights into the terminal electron acceptors of the water-oxidation pathway in PSII. For example, soluble redox mediators such as 2,6-dichlorobenzoquinone (DCBQ) are commonly used as a QB mimic or to mediate ET from any remaining QB to the electrode (Figure 1B).42,43 In the presence of DCBQ, we observed onset potentials for both PSII to be ~250 mV vs SHE (Figure 4). We can derive the Em (DCBQ/DCBQ–) value of ~310 mV vs SHE from the limiting photocurrents (Supplementary Figure 12), which is consistent with the value of 320 mV vs SHE derived from CV scans performed in the same buffer conditions (Supplementary Figure 14). We learned that although DCBQ can mediate electrons from PSII to the electrode, the Em value of DCBQ is mismatched with the terminal electron acceptor in PSII by at least 200 mV.

Discussion

This study has provided a fuller picture of the ET pathways at the PSII-electrode interface, and showed how the overall PSII photoresponse is the net result of different competing pathways. Figure 5 is a summary of energy/charge transfer pathways occurring within PSII and at the PSII-electrode interface, and includes the new mechanism of H2O2 photogeneration by PSII uncovered in this study.

Figure 5

Summary of the photoanodic (blue) and photocathodic (red) charge transfer pathways within PSII and at the PSII-electrode interface.

The applied potential (Eapp) range used in this study is shown. Energy levels of materials and cofactors are based on numbers reported within references.4,26,47 The PD1/RC is also commonly referred to as the P680. Grey bars represent uncertain energy levels. The dashed arrows represent pathways and energy levels identified by this study.

We propose that the H2O2 photogeneration pathway identified may also exist in nature. The photo-generation of ROS by PSII, in particular H2O2, has been known to occur both in vitro and in vivo;44,45 but the source/mechanism of H2O2 generation has been contentious.45,46 Until now, antenna Chl a have only been associated with energy transfer, but the reduction of O2 to form O2•– by Chl a/Chl a* (in the antenna and RC complex) has the greatest thermodynamic driving force than any other potential sources. Chl a* may be reduced by a nearby reductant such as a β-car or components within the thylakoid membrane when an electrode is not available.

Knowledge of the competing pathway will impact on photobioelectrodes incorporating photosystems, reaction centers and whole cells with respect to energy conversion efficiency and device stability. Conductive matrices or high surface area electrodes are often employed to improve loading or interfacial wiring within photoelectrodes. Additionally, electrode materials and electron mediators are often chosen based on their low reactivity for O2 reduction to limit charge recombination and improve the open circuit voltage of biophotovoltaic cells. It is now apparent that ET to Chl a should also be minimized to limit side charge leakage pathways and H2O2 generation, which is detrimental to device stability.

Lastly, the advancement of PF-PEC as a chemical biology tool will heavily depend on improvements in the enzyme-electrode interface. In the future, advanced spectroscopic techniques, such as ultrafast laser spectroscopy, can be coupled to PF-PEC to provide a more holistic approach to characterize the kinetics and energetics of photo-active enzymes.

Online Methods

Source data

Experimental data that support the findings of this study are available at the University of Cambridge data repository (http://dx.doi.org/10.17863/CAM.747) or from the corresponding author.

Materials

All chemicals, including 2,6-dichloro-1,4-benzoquinone (DCBQ, 98%, Sigma Aldrich), 2-(N-morpholino)ethanesulfonic acid (MES, >98%, Alfa Aesar), CaCl2 (>99%, Sigma Aldrich), MgCl2 (98%, Fisher Scientific), KCl (99%, Alfa Aesar), glycerol (>99%, VWR Chemicals), KOH (>99%, Sigma Aldrich), betaine (98%, Fisher Scientific), NH4OH (30%) solution (Fisher Scientific), H2O2 (30%) solution (Fisher Scientific), 3’-(3,4-dichlorophenyl)-1’1’-dimethylurea (DCMU, 98%, Alfa Aesar), spinach chlorophyll a (Sigma Aldrich), C60-(N,N-dimethyl pyrrolidinium) idodide (>99% over all isomers, Solaris Chem. Inc.), bovine liver catalase (Sigma Aldrich), bovine erythrocyte superoxide dismutase (Sigma Aldrich), glucose oxidase from Aspergillus niger (Sigma Aldrich), polystyrene beads (Polysciences, Inc.), ITO nanoparticles (Sigma Aldrich) and fluoride-dope tin oxide (FTO) coated glass (Sigma Aldrich) were purchased from commercial suppliers and used without further purification unless otherwise noted. Reagents for the analytical part of the work were of the highest available purity. T. elongatus PSII was kindly provided by Prof William A. Rutherford, from the Imperial College London.

Preparation of enzymes

Thermosynechococcus elongatus

BP-1 PSII core particles were isolated from a CP47 His-tagged mutant by Ni2+-affinity chromatography based on previously reported methods,48 with minor alterations.6 The resultant PSII solution was determined to be ca. 2.4 mg Chl a mL–1 in concentration, with activity of ca. 3500 μmol O2 mg (Chl a)–1 h–1.

Spinach PS II core particles were isolated as previously described with some modifications.49 PSII membranes were diluted in a BTS400 buffer solution [20 mM BisTris (pH 6.5), MgCl2 (20 mM), CaCl2 (5 mM), MgSO4 (10 mM) and sucrose (0.4 M)]. A 10% (w/v) β-dodecyl maltoside (β-DM) solution was made using the BTS400 buffer solution, which was added to the PSII membrane solution to give a final concentration of 1.25% (w/v) β-DM and 1.75 mg Chl a ml–1. The suspension was gently stirred whilst incubated at room temperature for 20 min. Any non-solubilized material was removed by centrifugation for 20 min at 40 000 x g. The supernatant was transferred to a Q-Sepharose column (Pharmacia LKB, 1 g dry weight (mg Chl)–1) that has been equilibrated with BTS400 with 0.03% β-DM to separate LHC II from the PS II core. The column was washed with BTS400 plus 0.03% β-DM until the elute became almost colorless and the ratio of absorption amplitude at 675/650 nm was larger than 3. The PSII core particles were eluted from the column by adding MgSO4 (75 mM) to the BTS400 with 0.03% β-DM buffer. The PSII core dimers were purified from the monomers by gel filtration chromatography (Supplementary Figure 15), using a Superdex 200 HR 10/30 column (GE Healthcare) with BisTris (20 mM pH 6,5), MgCl2 (10 mM), NaCl (20 mM), 1.5% Taurine and 0.03% β-DM as mobile phase, with a flow rate of 25 ml h–1.50 A Shimadzu LC-10AT VP diode array detector was used for absorption amplitude detection.

The antenna CP43 was isolated as described previously with some modifications.51 The first step of the isolation and purification of the CP43 involved the preparation of PSII core complexes from spinach as described above. Briefly, these preparations were Tris-washed, solubilized with β-DM, and subjected to ion-exchange chromatography. The CP43 was found in the fraction that did not bind to the column. This fraction was mixed with 50 mM Tris-HCl (pH10) and 0.03% β-DM until the final pH was 7.8. This fraction was loaded into a Q-Sepharose column equilibrated with 50 mM Tris-HCl (pH 7.8) and 0.03% β-DM. After the elute (which contained mainly free Chl) became colourless, CP43 was eluted with the same buffer with 50 mM MgSO4. The purity of CP43 was verified by fast protein liquid gel filtration chromatography (Supplementary Figure 16), using a Superdex 200 HR 10/30 column (GE Healthcare) with 20 mM BisTris (pH 6,5), 10mM MgCl2, 20mM NaCl, 1.5% Taurine and 0.03% β-DM as mobile phase and a flow rate of 25 ml h–1.50

The isolated RC complex used in this study is also known as the D1-D2-cyt b559 particle and was isolated according to a previous literature procedure,12 and the purity was also verified using the same method as for the CP43 (Supplementary Figure 17). It contains four Chl a, two pheophytin molecules, two peripheral Chl a molecules, two β-car and a cytochrome b559. The plastoquinones, non-heme Fe and the Mn4Ca cluster were all lost in the isolation process.

Stock solutions of T. elongatus PSII (2.4 mg Chl a mL–1, 77 μM), spinach PSII (0.35 mg Chl a mL–1, 11 μM), antenna CP43 (1.0 mg Chl a mL–1) and RC complex (0.15 mg Chl a mL–1) were stored in small aliquots in a liquid N2 dewar, and a new aliquot was thawed in each day of experiments. One in three dilution of the stock T. elongatus PSII solution was made immediately before adsorption on the electrodes to give a final concentration of 25.6 μM. Stock solutions of Chl a (100 μg ml–1) and β-car (400 μg ml–1) were prepared in ethanol in the dark and used immediately.

Preparation of inverse opal mesoporous ITO electrodes

The hierarchically structured ITO electrodes were prepared from ITO nanoparticles (< 50 nm diameter) and polystyrene beads (750 nm diameter, 2.54% w/v suspension in water) according to a previously published procedure.6 In this study, the electrodes were prepared with an ITO film thickness of 12 μm, and contain macroporous inverse opal chambers and channels with diameters of 750 nm and 100 nm, respectively, to facilitate the penetration of the PSII and other macromolecules into the electrode structure (Supplementary Figure 2). The rough mesoporous sub-structure offers further enhancement in surface area and thus anchoring sites for the enzymes. Before use in the experiments, the ITO electrodes were cleaned by incubation in a solution mixture containing H2O2 (30%), H2O, NH4OH (30%) in a 1:5:1 v/v ratio, at 70 °C for 15 min. The electrodes were then rinsed with water, air dried for 1 hour at 180 °C, and allowed to cool to room temperature.

The surface morphology of the electrodes was analyzed by scanning electron microscopy (SEM; Philips XL30). Feature dimensions have been measured by built-in software. The elemental composition was verified using Powder X-ray diffraction (XRD) analysis was carried out using an X’Pert PRO X-ray diffractometer (PANalytical B.V.).

Adsorption of bio-films onto electrodes

The Chl a pigment content within each biological sample used in the study was quantified using the absorption amplitude at λ665 (ε = 79.95 (mg Chl a)–1 mL cm–1)52 minus the background at λ710nm. Unless otherwise stated, T. elongatus PSII (1 μL, 25.6 μM), spinach PSII (2 μL, 11 μM), antenna CP43 solution (2 μL, 1.0 mg Chl a mL–1), RC complex solution (2 μL, 0.15 mg Chl a mL–1), Chl a (2 μL,100 μg ml–1) and β-car (2 μL, 400 μg ml–1) were drop-cast onto cleaned ITO electrodes with a geometrical surface area of 0.25 cm2 and left to stand for 15 min in a closed Petri dish in the dark before being used in photoelectrochemical experiments.

To estimate the degree of protein-film retention on the electrodes following experiments, the protein-film modified electrodes were incubated for 10 min in the MES electrolyte buffer solution in the dark (to avoid confounding photodegradation effects). The ITO film from each electrode was then scraped into MeOH (250 μL) and centrifuged for 5 min at 10 000 rpm. The absorbance Δλ(665-710nm) of the supernatant was measured to determine the amount of Chl a content that was adsorbed on the electrode surface. The adsorbed Chl a concentration was then expressed as a percentage of the original amount that was drop-cast on the electrode. Electrodes modified with films of T. elongatus, antenna CP43 and the RC complex exhibited on average 85% retention (n = 3), whereas those adsorbed with spinach PSII and Chl a molecules exhibited on average 65% and 40% retention (n = 3), respectively. Similarly, the retention of β-car molecules was determined based on the absorption amplitude at λ450nm; adsorbed β-car molecules exhibited on average 45% retention (n = 3).

In experiments where C60 was co-adsorbed with PSII or Chl a, a suspension of C60 in MES electrolyte buffer solution (50 mg mL–1, 45 mM) was first prepared and sonicated for 30 min. This suspension was combined with the PSII (25.6 μM) or Chl a (0.45 mM) in a ratio of 1:1 v/v. A volume of 2 μL of the PSII/C60 mixture and 2 × 2 μL of the Chl a/C60 mixture was drop-cast onto cleaned ITO electrodes. The electrodes were left to stand for 15 min in a closed Petri dish in the dark before being used in photoelectrochemical experiments.

Electrochemical experiments

All electrochemical experiments were performed on Ivium Compactstat or PalmSens Emstat potentiostats, and the temperature was maintained at 25 °C by the water jacket around the photoelectrochemical cell and a thermostat. Electrochemistry was conducted using an inverse opal mesoporous ITO or a glassy carbon electrode as the working electrode, a platinum counter electrode, and an Ag/AgCl reference electrode unless stated otherwise. The MES buffer electrolyte solution employed in the experiments contained CaCl2 (20 mM), MgCl2 (15 mM), KCl (50 mM) and MES (40 mM) and glycerol (5%), and adjusted to pH 6.5. Stationary working electrodes were employed and the electrolyte solution was not stirred unless otherwise stated. To achieve anaerobic conditions, a stream of Ar was bubbled through the electrolyte solution for 15 min before introducing the working electrode inside the cell. The set-up was purged for an additional 15 min before the start of the experiment, the headspace was purged during the electrochemical experiment. Anaerobicity was also achieved by the addition of glucose (5 mM), glucose oxidase (0.1 mg mL–2) and catalase (0.05 mg mL–2) in a separate control experiment.53 Where O2/N2 purge mixtures were required, a mass flow controller (Brooks Instrument, GF series) was used to control the ratio of O2:N2 gases.

The light source (Kodak S-AV 2000) employed in the photoelectrochemical experiments was a halogen lamp collimated by a plano-convex lens and filtered by a bandpass filter (679 nm, full width at half maximum: 10.3 nm). The light intensity was adjusted to 5 mW cm–2. The final volume of the electrolyte solution used was 4 mL. All redox potentials are quoted against the standard hydrogen electrode (SHE), with the potentials being calculated using the following correction factor: ESHE = EAg/AgCl + 0.209 V.

In chronoamperometry scans, the experimental regimes were very similar in each case. Each potential was held for at least 60 s before being stepped up by 0.1 V, starting at 0 V vs SHE. This regime allowed the electrode to equilibrate for 20 s in the dark at each new potential before being irradiated with red-light (679 nm, 5 mW cm–2) for 30 s, followed by light off for 10 s. Variations to the potential step sizes and the number of steps were introduced according to the needs of the experiment. Where DCMU or DCBQ were needed, stock DCMU and DCBQ solutions (100 mM, DMSO) were prepared, and 40 μL was added to the electrolyte solution to give a final concentration of 1 μM. Dark current spikes caused by stepping the potential were removed in all figures for clarity. All experiments were replicated three times, and the most representative scans are shown.

Action spectra measurement

A 300 W Xenon lamp coupled to an MSH300 monochromator (LOT Quantum design) was used in the action spectra measurements. The light intensity measured as a function of wavelength (full width half maximum: 20 nm) with a photodetector (SEL033/F/QNDS1/W) and power meter (ILT1400).