Edinburgh Research Explorer Hydrogen production through oxygenic photosynthesis using the cyanobacterium Synechocystis sp. PCC 6803 in a bio-photoelectrolysis cell (BPE) system

a Microbial electrolysis cells (MECs) represent an emerging technology that uses heterotrophic microbes to convert organic substrates into fuel products, such as hydrogen gas (H 2 ). The recent development of biophotovoltaic cells (BPVs), which use autotrophic microbes to produce electricity with only light as a substrate, raises the possibility of exploiting similar systems to harness photosynthesis to drive the production of H 2 . In the current study we explore the capacity of the cyanobacterium Synechocystis sp. PCC 6803 to generate electrons by oxygenic photosynthesis and facilitate H 2 production in a two-chamber bio-photoelectrolysis cell (BPE) system using the electron mediator potassium ferricyanide ([Fe(CN) 6 ] 3 (cid:1) ). The performance of a wild-type and mutant strain lacking all three respiratory terminal oxidase activities ( rto ) was compared under low or high salt conditions. The rto mutant showed a decrease in maximum photosynthetic rates under low salt (60% lower P max than wild-type) but signi ﬁ cantly increased rates under high salt, comparable to wild-type levels. Remarkably, rto demonstrated a 3-fold increase in (Fe[CN] 6 ) 3 (cid:1) reduction rates in the light under both low and high salt compared to the wild-type. Yields of H 2 and e ﬃ ciency parameters were similar between wild-type and rto , and highest under high salt conditions, resulting in a maximum rate of H 2 production of 2.23 (cid:3) 0.22 ml H 2 l (cid:1) 1 h (cid:1) 1 (0.68 (cid:3) 0.11 mmol H 2 [mol Chl] (cid:1) 1 s (cid:1) 1 ). H 2 production rates were dependent on the application of a bias-potential, but all voltages used were signi ﬁ cantly less than that required for water electrolysis. These results clearly show that production of H 2 using cyanobacteria is feasible without the need to inhibit photosynthetic O 2 evolution. Optimising the balance between the rates of microbial-facilitated mediator reduction with H 2 production may lead to long-term sustainable H 2 yields. Bioelectrochemical systems have emerged as a promising technology for energy recovery and the production of valuable fuel products such as H 2 gas. In microbial electrolysis cells (MECs), heterotrophic bacteria consume organic compounds to drive the electrochemical production of H 2 . Here we report a two-chamber bio-photoelectrolysis cell (BPE) system for producing H 2 that uses light as a substrate. In the anodic compartment of the BPE the cyanobacterium Synechocystis sp. PCC 6803 was used to generate electrons by oxygenic photosynthesis, with H 2 produced in the cathodic compartment. In addition, we studied thee ﬀ ectsofmutationsabolishingtheterminaloxidasesoftherespiratoryelectrontransportchain,withthestrikingresultthatamutant( rto )showedthree-fold higher rates of reduction of the electron mediator ferricyanide than the wild-type strain. This is one of the  rst examples of O 2 -evolving autotrophs being used to facilitate sustainable H 2 production without the need to inhibit photosynthetic O 2 evolution or establish anaerobic conditions in the culture medium. Further increases in output might be achieved using suitable mutants.


Introduction
Hydrogen gas (H 2 ) has many merits as a clean energy resource. Although there are several chemical processes by which H 2 can be produced, 1 microbial biotechnologies are particularly attractive as they are renewable and oen relatively cheap to maintain. Under anaerobic conditions, heterotrophic microbes can produce H 2 through fermentation of carbohydrate-rich suspensions, including industrial and municipal waste water. 2,3 Alternatively, photosynthetic microbes are able to facilitate photobiological H 2 production (biophotolysis). [2][3][4][5][6] Using microbes capable of oxygenic photosynthesis, H 2 can be produced using abundant and sustainable natural resources, such as sunlight and water, without the requirement for organic supplements. However, in reports to date with green microalgae or cyanobacteria, substantial restrictions have been imposed on the bioreactor operating conditions in order to generate signicant quantities of H 2 . For example, H 2 production with green algae can occur only under near anaerobic conditions, as the primary catalytic enzyme involved (Fe-Fe hydrogenase) is inhibited by oxygen (O 2 ). Thus far, sustained H 2 production with algae (e.g. Chlamydomonas sp.) has been achieved only in the presence of a major decrease in activity of photosystem IIthe O 2 -evolving complex of the photosynthetic apparatus. This was achieved using sulphur deprivation to decrease the O 2 production rate below the rate of O 2 use through mitochondrial respiration. 5,7,8 To produce H 2 with lamentous cyanobacteria, such as Anabaena sp., nitrogen sources in the culture media must be depleted to allow H 2 -producing heterocysts to form. 4 Unicellular cyanobacteria require the removal of O 2 from the growth medium for sustained H 2 production. 4,9 Extensive work in this eld has been carried out to identify phenotypes/mutants with increased O 2 tolerance and H 2 yields. 6,[9][10][11][12][13] Microbes can also be utilised in bioelectrochemical systems (BESs) to generate an electromotive force, which can then be used for power production (in the form of electricity) or the generation of secondary products at the cathode. Much attention recently has been given to the development of novel BESs for producing a variety of useful secondary products, including H 2 . 3,14 In particular, progress has been made in the eld of microbial electrolysis cells (MECs), which typically use heterotrophic bacteria attached to an anodic electrode to catalyse the cathodic production of H 2 in a single-or two-chamber system. [15][16][17] The production of H 2 in MECs is facilitated by the heterotrophic bacteria and thermodynamically unfavourable. Hence such systems do require additional energy inputs in the form of organic substrate and an applied bias-potential (typically provided by a power supply unit or a potentiostat) to drive the reaction forward. 14 BESs that generate electricity in a light-dependent manner using cyanobacteria and algae (known as biological photovoltaic cells [BPVs]) have also received increased attention recently. [18][19][20][21][22][23] However, to date no studies have reported the use of oxygenic photosynthetic microbes at the anode for the cathodic production of H 2 in a MEC-type system. In MECs the cathode requires a strictly anaerobic environment, as the formation of water from O 2 and protons is energetically more favourable than the production of H 2 from protons alone. 14 The process of H 2 production is currently considered unfeasible in the presence of autotrophic microbes that produce O 2 during the photosynthetic process. Nevertheless, it may be possible to facilitate sustainable H 2 production using autotrophic microbes if the chambers are separated, for example, by an anionic or cationic exchange membrane. 21 A rutheniumbased "articial photosynthesis" photochemical cell was recently described that achieved simultaneous production of O 2 (and H + ) and H 2 by separating the O 2 -producing anodic and H 2 -producing cathodic chambers by a proton permeable membrane. 24 Bora et al. 25 recently highlighted the value of constructing bio-photoelectrochemical cells using biomolecule-modied electrodes separated by a NaonÒ membrane. For both articial and biological systems the challenge then lies in optimising the efficiency of H 2 production, such that external energy inputs (e.g. applied bias-potentials) do not outweigh potential energy gains from the H 2 recovered. 14,26 At a minimum, the microbial electrolysis process should require substantially less energy investment than that required for water electrolysis. 14 Several microbial species, including green algae, diatoms and cyanobacteria, are readily able to photo-catalyse the reduction of soluble extracellular electron mediators, such as potassium ferricyanide ([Fe(CN) 6 ] 3À ) to potassium ferrocyanide ([Fe(CN) 6 ] 4À ). [27][28][29][30] The electrons harvested by (Fe[CN] 6 ) 4À can then be used as a recyclable fuel source at the anode for electricity production (such as in BPV devices 21 ), or production of secondary products at the cathode in what can be referred to as a bio-photoelectrolysis cell (BPE). The model cyanobacterial strain Synechocystis sp. PCC 6803 is an ideal candidate species to investigate BPEs for several reasons: (i) its genome is fully sequenced and the strain is readily transformable 31 34 has been demonstrated to increase H 2 production in a culture at low O 2 concentration. 9 Since terminal oxidases act as alternative electron sinks, we hypothesised that the absence of those terminal oxidases would be likely to lead to a more reduced intracellular environment in the rto mutant, which may then favour the reduction of alternative electron acceptors. Whether specic extracellular or intracellular conditions would substantially affect the rate of exoelectrogenic activity is not well explored, 35,36 but may be of great interest to several BES technologies.
We therefore set out to test if a two-chamber BPE system could be used to facilitate H 2 production with Synechocystis strains and (Fe[CN] 6 ) 3À as an electron mediator in either low or high salt grown cultures. Although a bias-potential was required to drive the production of H 2 (1.0-1.4 V), all applied biaspotentials were lower than that required for water electrolysis (ca. 2.2 V). 37 This is, to our knowledge, the rst time that unicellular O 2 -evolving autotrophs have been used to facilitate sustained H 2 production for several hours without the need to inhibit photosynthetic O 2 evolution or establish anaerobic conditions in the culture medium. Furthermore, this is the rst report of an MEC-type system that has successfully harnessed the exoelectrogenic activities of O 2 -evolving autotrophs by spatially and temporally separating O 2 and H 2 evolving processes in a single system.

Photosynthesis and respiration
In the BPE system used here (Fig. 1), photosynthesis and respiration can be considered as the primary sources of electrons for H 2 production. In typical growth medium (i.e. low salt), the maximum photosynthetic O 2 evolution rate (P max ) of the rto mutant was 60% lower than wild-type Synechocystis cultures ( Fig. 2; Table 1). The absence of the three respiratory electron transport chain terminal oxidases in rto resulted in signicantly decreased respiration rates (average 57% lower than wild-type under both high and low salt) (ESI Fig. 1 †).
Growth in high salt medium did not appear to affect photosynthesis or respiration in wild-type cultures (Fig. 2, Table  1). However, photosynthetic rates were signicantly higher than in low salt medium for rto, including a remarkable 85% and 96% increase in P max and apparent quantum efficiency (AQE), respectively. Under high salt conditions, P max and AQE of rto were partially restored to levels similar to wild-type cultures (Fig. 2). For all cultures tested, saturating light levels were reached at approximately 200 mE m À2 s À1 . This intensity was used for all further experiments with the BPE system in the light.

Potassium ferricyanide reduction and regeneration
In the current system the operation of the BPE device was separated into two stages: the "fuelling phase" and "H 2 production phase" (Fig. 1C). The reduction of (Fe[CN] 6 ) 3À during the fuelling phase is closely linked to the theoretically achievable H 2 yields. Therefore identifying strains and environmental conditions that facilitate increased (Fe[CN] 6 ) 3À reduction rates is of great importance. Consequently, we measured (Fe[CN] 6 ) 3À reduction rates under different conditions for the wild-type and rto strains. Each cell culture was initially subjected to dark or light conditions for ca. 12 h following inoculation in the device with (Fe[CN] 6 ) 3À (Fig. 3). As expected, (Fe[CN] 6 ) 3À reduction rates were lower in the dark compared to the light, indicating that photosynthetic reactions facilitate increased reduction rates. Furthermore, (Fe[CN] 6 ) 3À reduction rates were typically higher in high salt compared to low salt conditions for both wild-type and rto cultures.
For cultures at low salt concentrations, rto showed slightly higher reduction rates under light and dark conditions compared to the wild-type. However, at high salt concentrations rto showed a striking 6-fold and 3-fold increase in the specic rate of (Fe[CN] 6 ) 3À reduction compared to the wild-type in high salt grown under dark and light, respectively (Fig. 3). This resulted in signicantly greater (Fe[CN] 6 ) 3À reduction efficiencies (FeCN À QE ) for rto under both low and high salt conditions (Table 1). In all cases, culture densities were measured regularly and did not show any signicant differences before and aer measurements.
When the circuit was closed during the H 2 production phase ( Fig. 1C), (Fe[CN] 6 ) 4À (redox potential of 420 mV at pH 7) was oxidised at a constant rate (Fig. 4A) dependent on the bias-potential applied (discussed below). Oxidation rates during operation were typically greater than reduction rates  6 ] 4À in the "fuelling phase" (C). Following reduction, [Fe(CN) 6 ] 4À was used as a recyclable electron supply at the anode in the "H 2 production phase". Facilitated by a bias-potential provided from an external power supply (D), current harvested from the anode and protons diffusing through the CEM separating the chambers were used to drive H 2 production at the cathode. under open circuit conditions (Fig. 4B), indicating that the rate of (Fe[CN] 6 ) 4À production was not high enough to facilitate a continual substrate supply for H 2 production. Following complete re-oxidation, the device was disconnected and the cultures were given time to re-reduce (Fe[CN] 6 ) 3À either in dark or light conditions. No signicant differences in culture reduction rates were observed between successive reduction cycles.

H 2 production rates and efficiencies
The production of H 2 in the cathodic chamber of the BPE system required an applied bias-potential and (Fe[CN] 6 ) 4À as the anodic substrate (ESI Fig. 2 †). No H 2 was produced when (Fe[CN] 6 ) 4À/3À was not present in the anodic chamber with Synechocystis cells, regardless of the bias-potentials typically applied (1.0-1.4 V) (Fig. 5). Under those conditions, appreciable H 2 production was only seen when 4.0 V of bias-potential was applied. At such high voltages this observation was likely to be attributable to water electrolysis in the cathodic chamber. In the absence of added (Fe[CN] 6 ) 4 , the only source for H 2 production between 1.0 and 1.4 V is therefore that generated by biological reduction of (Fe[CN] 6 ) 3À .
Following reduction of (Fe[CN] 6 ) 3À to (Fe[CN] 6 ) 4À in the presence of cells in the anodic chamber, the device was connected and a bias-potential applied. During these experiments, all H 2 production proceeded under dark conditions. The rate of H 2 production increased with increasing bias-potential (Fig. 5). Rates were typically higher under high compared to low salt conditions, with a maximum rate of 2.23 AE 0.22 ml H 2 l À1 h À1 (specic rate of 0.68 AE 0.11 mmol H 2 [mol Chl] À1 s À1 ) in high salt achieved at an applied bias-potential of 1.4 V (Fig. 5). Based  6 ] 3À (1 mM). Respiration rates were measured in the dark period following each light measurement. The maximum photosynthetic O 2 evolution rate (P max ) and apparent PSII quantum efficiency (AQE) for wild-type (black circles, squares) and rto (white circles, squares) are indicated in Table 1. Table 1 Photosynthetic performances and substrate turnover efficiencies of cultures used to drive H 2 production. The maximum photosynthetic O 2 evolution rate (P max ), dark respiration rate (R d ), maximum respiration rate in the light (R max ), apparent PSII quantum efficiency (AQE), actual quantum efficiency at P max (QE max ) and the efficiency of [Fe(CN) 6 ] 3À reduction at P max (FeCN À QE ) were derived from measurements in Fig. 2 and 4B. The values are the mean AE SE (N ¼ 4) and are followed by letters indicating significant difference as in Fig. 3 Wild-type Rto     conversion efficiency of (Fe[CN] 6 ) 4À into usable charge at the anode. Overall hydrogen recovery (r H 2 ) increased with increasing applied bias-potentials, reaching a maximum r H 2 at 1.4 V of ca. 20%. It is likely that at lower voltages, slower reaction rates and system inefficiencies allow for competition of electrons at the cathode from more thermodynamically favourable acceptors (e.g. small amounts of O 2 leakage) and decrease possible H 2 yields. 17 No differences in efficiencies were observed for BPEs without cells in the anode (ESI Fig. 2 †), indicating that the presence of cells did not inuence the H 2 production phase.

Discussion
This is, to our knowledge, the rst example where oxygenic photosynthesis in cyanobacteria has been harnessed in an MEC-type system to produce a source of reducing equivalents for the production of H 2 . In the present study the production of H 2 takes place without the need to inhibit the O 2 -evolving photosystem II complex. Separation of "O 2 evolution" and "H 2 production" was achieved using a two-chamber BPE system that provided a spatial and temporal partitioning of photosynthetic O 2 evolution in the anodic chamber and subsequent H 2 synthesis in the cathodic chamber (Fig. 1). Charge was initially captured from Synechocystis cells using a recyclable extracellular electron mediator ([Fe(CN) 6 ] 4À/3À ) that, with the addition of an applied bias-potential, could then be utilised as a substrate for H 2 production. Although previous work has described similar systems for electricity production 18,19,21,38 the results of the current study suggest that the production of secondary products and even desalination as demonstrated with BES systems employing heterotrophic microbes 39-41 may also be feasible with photosynthetic microbes. The two-stage cycle method outlined in this paper (Fig. 1C) relied on an initial substrate accrual period (the fuelling phase) to facilitate the complete reduction of (Fe[CN] 6 ) 3À to (Fe[CN] 6 ) 4À in the anodic chamber, followed by a H 2 production phase (Fig. 5). Although it is likely that (Fe[CN] 6 ) 3À is still reduced in the anodic chamber during the production of H 2 , the time taken for complete oxidation of (Fe[CN] 6 ) 4À depended on the biaspotential applied, with a maximum rate of 2.23 AE 0.22 ml H 2 l À1 h À1 (specic rate of 0.68 AE 0.11 mmol H 2 [mol Chl] À1 s À1 ) for ca. 4 hours. The resulting yields compared favourably with existing photobiological H 2 production studies. The specic H 2 production rates achieved here are comparable but more sustained than a recent study with transgenic Chlamydomonas reinhardtii. 12 Using an inducible chloroplast gene expression system, Surzycki et al. 12 showed that the activity of photosystem II could be transiently inhibited, resulting in a specic H 2 production rate of 1 mmol H 2 [mol Chl] À1 s À1 that was sustained for a maximum of 1.5 hours. The overall rates we observed were also similar to those reported in other published photobiological H 2 production studies using sulphur-deprived Chlamydomonas, which range from 0.58 to 2.35 ml H 2 l À1 h À1 . 7,8 It is difficult to compare our specic rates directly with those studies, as reactor cultures that undergo cycles of sulphur deprivation typically show large variations in chlorophyll content depending on the particular point of the cycle. 7,8,42 The energy inputs required for achieving similar rates of H 2 production are substantially smaller for BPEs in comparison with water electrolysis and photobiological techniques. BPE systems primarily need an energy input in the form of a biaspotential. This energy input is required to overcome the potential gap between the anodic reaction (oxidation of electron mediator) and the cathodic reaction (production of H 2 ) (ESI Fig. 3 †). Using (Fe[CN] 6 ) 4À/3À as the electron mediator, this gap cannot be smaller than 0.84 V (at standard atmospheric pressure and pH 7). To allow for additional internal systemic constraints (e.g. electrode over-potentials), 14 the bias-potentials used in the current study were between 1.0 and 1.4 V. These values are 55-36% smaller than the potential required to generate H 2 via water electrolysis (ca. 2.2 V). 37 The performance of the BPE system used in the current study still compares unfavourably to the typical H 2 recoveries and efficiencies attained in MEC systems. [15][16][17] Typical r H 2 for MECs range from 70-95%, depending on reactor design and organic substrate used, which are achieved at signicantly lower biaspotentials (0.2-1.0 V) than used here (Table 2). Nevertheless, BPE systems do appear comparable in terms of C E , with values of greater than 95% for all applied potentials tested. One of the principal advantages of BPEs such as that described here is that the microbes, and consequently the system, are driven by light. No organic substrate is required. This decreases the inherent complexity of the system compared to MFCs/MECs, where feed stocks must be carefully monitored, as differences in substrate can cause large uctuations in performance. 17 Moreover, the requirement for feed stocks will incur several additional costs. For example, waste water feed stocks must typically be demineralised to avoid deposits on the electrodes and corrosion. 2 One of our main aims was to compare the performance of wild-type cultures to the mutant strain rto. Photosynthetic O 2 evolution of rto returned to near wild-type levels when grown in high salt (Fig. 2). The mutant also showed increased (Fe[CN] 6 ) 3À reduction rates compared to the wild-type under high salt (Fig. 3), suggesting that cyanobacteria can be genetically and environmentally optimised to increase extracellular electron transfer rates and thus energy yields in BPE-type systems. The reason(s) for the remarkable increase in photosynthesis and (Fe[CN] 6 ) 3À reduction rates in rto mutants under high salt conditions remains to be fully investigated. Increased media salinities typically result in an increase in conductivity, which can lead to a rise in cellular exoelectrogenic activity. 14 High salt conditions may also have facilitated an alternative respiratory route in the rto mutant (i.e. an alternative electron sink), resulting in the increase in FeCN À QE observed (Table 1). It is not yet clear whether the observed increase in photosynthetic activity for rto (as indicated by O 2 evolution) was the direct cause of the increase in cellular exoelectrogenic activity as high salt also led to increased rates of (Fe[CN] 6 ) 3À reduction in the dark. Salt stress response is well studied in Synechocystis and leads to a multitude of changes in both intracellular and plasma membrane protein levels. 43,44 Drawing on that work, a comparison of the changes in plasma membrane protein abundance between wild-type and rto in response to high salt may lead to a clearer understanding of proteins involved in extracellular electron transfer.
Our results indicate that there is substantial scope for optimising the balance between mediator turnover rates with H 2 production, which may lead to long-term sustainable H 2 yields using BPE systems. In terms of optimising the current setup, increases in Synechocystis cell concentrations, (Fe[CN] 6 ) 3À and light have previously been shown to improve performance in BPV systems . Recent evidence does suggest that other cyanobacterial species, such as Synechococcus sp. WH 5701, may have a higher capacity for extracellular electron transport compared to Synechocystis. 22,30 It is possible that alternative electron mediator compounds with a lower electrode potential could further reduce the bias-potential required, thus improving energy efficiencies and potential H 2 yields. 45 For example, 2,6dichlorophenol-indophenol (DCPIP) has a potential of 290 mV at pH 7, which is 130 mV less than (Fe[CN] 6 ) 4À/3À . 46 Even in the current study, the highest (Fe[CN] 6 ) 3À reduction rates seen for rto in high salt (Fig. 4B) were similar to the average (Fe[CN] 6 ) 4À oxidation rate observed at applied potentials of 1.0 V or 1.2 V (Fig. 4A). Future work will focus on screening different species and mutants under varying growth conditions, exploring the potential of alternative redox mediators and improvements in reactor design. The latter should include further characterisation of the multitude of electrochemical-related bottlenecks that may affect performance as seen in other BES systems, such as electrode material, spacing, orientation and the area and type of exchange membrane used. 14,47,48

Conclusions
The current work advances on previous photobiological H 2 production methods by providing a novel and robust BES-based solution that overcomes many of the problems associated with traditional single-chamber reactor approaches. By separating the process of biological O 2 evolution and H 2 production, O 2 -evolving autotrophs (such as cyanobacteria and green algae) can be used to facilitate H 2 production without the need to inhibit photosynthetic O 2 evolution directly. This approach negates the requirement for anoxygenic photosynthesis and thus cultures do not require regular cycles of oxygenic regeneration. As H 2 is produced in a separate chamber the product is relatively purer and more concentrated than in single-chamber designs. Furthermore, the process is not directly light-dependent, so H 2 can also be generated in the dark. BPE technology shows promise as a technique with exciting potential practical applications and appears to exhibit many unique and attractive attributes for renewable energy production.

Cultures and growth
A wild-type strain of Synechocystis sp. PCC 6803 (referred to subsequently as Synechocystis) was from a laboratory stock. 24 A "triple-knockout" Synechocystis mutant from which all the three respiratory terminal oxidases (cytochrome bd oxidase [cyd], cytochrome c oxidase [cox] and alternative respiratory terminal oxidase [ARTO]) were absent was generated as described by Lea-Smith et al. 49 Cultures were grown and then analysed in BG11 medium 50 (low salt) or BG11 medium containing NaCl (0.25 M) (high salt). All cultures were supplemented with 5 mM NaHCO 3 and maintained at 22 AE 2 C under low light (ca. 5 W m À2 ) in a 24 h light cycle (12 h light/dark) under sterile conditions. Media conductivities were calculated using a Jenway conductivity meter 4310 (Jenway, Cambridge, UK) set at a reference temperature of 25 C.
Strains were periodically streaked out and grown on agar plates containing agar (0.5-1.0%) and BG11, which were then used to inoculate fresh liquid cultures. Culture growth and density were monitored by spectrophotometric determination of chlorophyll content. Chlorophyll was extracted in 99.8% (v/v) methanol (Sigma-Aldrich, Gillingham, UK) as described previously. 51
During operation, the anodic chamber was inoculated with Synechocystis (10 nmol Chl ml À1 ) containing (Fe[CN] 6  To initiate H 2 production, the cathodic chamber was sparged with nitrogen gas for 20 minutes and the light was turned off. A xed bias-potential (1.0-1.4 V) was applied to the reactor circuit using a power box. A resistor (100 U) was connected in series with the power supply, and the voltage across the resistor was measured using an ADC-20 High Resolution Data Logger (HRDL) and PicoLog Data logging soware version R5.21.5 (Pico Technology, St Neots, UK) to calculate the current.

Potassium ferricyanide measurements
At regular intervals, samples (1 ml) were taken from the anodic chamber, cells removed by rapid centrifugation, and the concentration of (Fe[CN] 6 ) 3À in the supernatant measured spectrophotometrically at 420 nm (E 420 ¼ 1.02 mM À1 cm À1 ) as described by Davey et al. (2003).

Hydrogen and oxygen production
Photosynthetic O 2 evolution rates were determined on concentrated cultures ($10 nmol Chl ml À1 ) at 22 AE 2 C with a Dissolved Oxygen Meter (Extech Instruments Corporation, MA, USA) in media containing (Fe[CN] 6 ) 3À (1 mM). Following dark equilibration (5-10 min), O 2 exchange rates were recorded for 5 min at increasing light intensities (0-280 mE m À2 s À1 ) followed immediately by 5 min in darkness to calculate the average photosynthetic O 2 evolution and respiration rate, respectively. The respiration rate following illumination at each light intensity was subtracted to estimate the true rate of photosynthetic O 2 evolution (Fig. 2). Light response curves were analysed as described in Akhkha et al. 52 to determine the rate of maximum photosynthetic O 2 evolution (P max ) and apparent PSII quantum efficiency (AQE). Predicted values for actual quantum efficiency at P max (QE max ) and the efficiency of (Fe[CN] 6 ) 3À reduction at P max (FeCN À QE ) were derived for each culture as follows: where P is the photosynthetic O 2 evolution rate at P max (mmol O 2 m À2 s À1 ), I is the corresponding light intensity used and F r is the rate of (Fe[CN] 6 ) 3À reduction. In the cathode, O 2 evolution and uptake were measured using a Clark-type O 2 electrode. H 2 concentrations were measured using a Laboratory H 2 Microsensor (AMT Analysenmesstechnik GmbH, Germany). Coulombic efficiency (C E ) was calculated as the ratio of electrons recovered from (Fe[CN] 6 ) 4À that were available for H 2 production at the cathode: where E is the charge (mmol electrons per h) and F o is the rate of (Fe[CN] 6 ) 4À oxidation. Cathodic H 2 recovery (r cat ) and overall H 2 recovery (r H 2 ) were calculated as described in Logan and Call. 17 Statistical analysis Results were subjected to analysis of variance (ANOVA) or Student's t-tests to determine the signicance of the difference between responses to treatments. When ANOVA was performed. Tukey's honestly signicant difference (HSD) post-hoc tests were conducted to determine the differences between the individual treatments (SPSS Ver. 11.5; SPSS Inc., Chicago, IL, USA).