Photobiological hydrogen-producing systems

Abstract

Hydrogen photoproduction by micro-organisms combines the photosynthetic properties of oxygenic and non-oxygenic microbes with the activity of H₂-producing enzymes in nature: hydrogenases and nitrogenases. The overall efficiency of the process depends on the separate efficiencies of photosynthesis and enzymatic catalysis. This tutorial review discusses the biochemical pathways for H₂ production in different organisms, barriers to be overcome, and possible suggestions for integrating photobiological H₂ production with fermentative, anaerobic systems for a potentially more efficient process.

---

Jianping Yu, Maria Ghirardi, Pin-Ching Maness and Alexandra Dubini

Jianping Yu, Maria Ghirardi, Pin-Ching Maness and Alexandra Dubini are research scientists at the National Renewable Energy Laboratory (NREL). Dr Ghirardi leads a research group in hydrogen production at NREL’s Chemical and Biosciences Center. She received her PhD degree in Comparative Biochemistry from the University of California, Berkeley, under the supervision of Dr Anastasios Melis. Her doctoral work focused on the composition of the photosynthetic membrane from different organisms under different environmental conditions. She joined NREL in 1995 where she is now a Principal Investigator. Her current research involves photobiological H2 production and algal biodiesel. Her group’s expertise covers metabolic, biochemical and genetic aspects of algal metabolism, and has lately expanded into the area of biomimetic systems for solar energy conversion to hydrogen.

1. Introduction

The term “photobiological H₂ production” covers a wide variety of microbial processes that require light as the energy source, an electron-donating substrate, and a catalyst that combines electrons and protons to generate H₂ gas. Light energy allows photosynthetic organisms to extract electrons from high-redox-potential compounds such as water (in oxygenic photosynthesis) or organic acids (in non-oxygenic photosynthesis), yielding energy (in the form of ATP) and low-redox-potential reductants that can be utilized as substrates for H₂ production by either hydrogenases or nitrogenases.

Figure 1 illustrates the three main types of photobiological H₂-producing processes found in nature: (a) oxygenic photosynthesis coupled to H₂ production by hydrogenases (found in green algae—see section 2.1.1—and, with some variations, in cyanobacteria—see section 2.1.2); (b) oxygenic photosynthesis coupled to H₂ production by nitrogenases (found in a large number of cyanobacteria—see section 2.1.3); and (c) non-oxygenic photosynthesis coupled to nitrogenase-catalyzed H₂ production (present in many purple photosynthetic bacteria—see section 2.2). The ultimate application of each of these systems in the photoproduction of H₂ fuel does directly depend on their light-conversion efficiencies. A comparison between the efficiencies of the three systems described above is discussed in section 3.

Fig. 1 Photobiological H₂-producing processes: (a) oxygenic photosynthesis in green algae linked to H₂ production by the [FeFe]-hydrogenase; (b) oxygenic photosynthesis in cyanobacterial vegetative cells coupled to nitrogenase-catalyzed H₂ production in heterocysts; (c) non-oxygenic photosynthesis in purple non-sulfur bacteria linked to H₂ production by nitrogenase. Green, blue and purple cylinders represent the light-harvesting molecules (chlorophyll, phycobilins and bacteriochlorophylls, respectively) associated with each photosystem. The following electron transport components are also depicted in the figure: the oxygen-evolving complex (OEC), the PSII tyrosine residue Y_Z, the cytochrome b₆/f and the cytochrome b/c₁ complexes, plastocyanin (PC), ferredoxin (Fd), the heterocyst-located ferredoxin (FdH). The hydrogenase is represented by a beige cylinder, and the nitrogenase by a gold one.

Although this review issue focuses on “O₂ makers,” we included non-oxygenic photosynthetic processes (section 2.2) in this chapter. Our rationale is that, although these systems require organic sources of electrons instead of using water, and thus they do not yield O₂, they are essential components of an overall integrated biological system for H₂ production (section 4) that is, in one form or another, under study by research groups worldwide.

2. Photobiological H₂ production

2.1 Oxygenic photosynthesis

Green algae and cyanobacteria are photoautotrophic organisms: they can grow under sunlight and CO₂, without organic sources of carbon. They perform photosynthesis, which converts light energy into chemical energy according to the Z-scheme for linear electron transport (LET, Fig. 1a and b). This process, which takes place in the thylakoid membranes, starts with light absorption by pigment molecules (chlorophylls, phycobilins, and carotenoids) bound to light-harvesting protein complexes associated with two multi-membrane protein complexes, Photosystem I (PSI) and Photosystem II (PSII). The absorbed light energy is then transferred to the reaction center pigments of PSII and PSI where charge separation occurs. Figure 2 shows the electron transfer steps that take place within PSII and PSI. Electron transfer cofactors are arranged in two branches: A and B.¹ In PSII, electron transfer occurs preferentially through branch A,² while both branches are active in PSI.^3a

Fig. 2 Electron transport cofactors and pathways within (a) Photosystem II, (b) Photosystem I and (c) PSII-type bacterial reaction centers in Rhodopseudomonas viridis. The arrows indicate the active electron transfer branch. Chemical structures for PSII and PSI reproduced with permission from Annual Reviews, Palo Alto, CA, USA (see ref. 3a). Chemical structures for bacterial reaction center reproduced with permission from G. Zubay, “Biochemistry”, W.C. Brown Publishers, 1998, p. 380 (see ref. 3b). The active and inactive branches are represented by the letters A and B, respectively. The redox potentials of each component are given in the text.

In PSII, the pigment P₆₈₀ consists of a chlorophyll dimer bound to conserved histidine residues located in D1 and D2, the two major proteins that comprise the PSII reaction center (Fig. 2a). Two chlorophyll monomers, Chl_A and Chl_B are located at close distances from P₆₈₀ on branches A and B, followed by two pheophytin (Pheo) molecules still further from P₆₈₀. Until recently, it was believed that charge-separation in PSII was initiated at P₆₈₀. However, recent experimental observations have led to the proposal that the initial radical pair actually involves Chl_A⁺ (or Chl_D1) and Pheo_A (−0.75 to −0.35 V^4,5), whereas P₆₈₀ is oxidized immediately afterwards (see ref. 2 for detailed discussion). To prevent charge recombination between the reducing and oxidizing sides of PSII, a rapid charge transfer (<300 ps⁶) occurs between Pheo⁻ and a plastoquinone molecule (Q_A, −0.13 to −0.03 V^4,5) bound to the D2 reaction center protein. Finally, the electron is further transferred to a second plastoquinone molecule (Q_B, +0.03 V) that, when doubly reduced, leaves its binding site at the D1 PSII reaction center protein and diffuses through the thylakoid membrane to dock and reduce intermediate electron acceptors at the cytochrome cyt b₆/f complex. Reduction and oxidation of Q_B is accompanied by the vectorial translocation of two protons from the stroma to the lumen, generating a proton gradient formed across the two sides of the thylakoid membrane. The strong oxidant, P₆₈₀⁺ (+1.12 to +1.27 V; see ref. 4 and 5 for a recent re-evaluation of this value) extracts electrons from a nearby tyrosine residue (Y_Z, +0.97 V) which, in turn, sequentially oxidizes two water molecules bound to the O₂-evolving complex (OEC). This reaction releases oxygen and protons into the lumenal side of the thylakoid membranes, which further adds to the proton gradient.

The pigments associated with the PSI reaction center comprise P₇₀₀, a Chl heterodimer bound to the PsaA and PsaB proteins and two pairs of additional chlorophyll molecules located on both sides of P₇₀₀, Chlec2_A, Chlec3_A (previously known as A₀) on branch A, and Chlec2_B and Chlec3_B on branch B. The overlap between the rings of the chlorophylls ec2 and ec3 is stronger than that between the two chlorophyll molecules that comprise P₇₀₀.^3a Based on this observation, the initial charge separation event in PSI, which had initially been assigned to P₇₀₀⁺/A₀⁻ is also being re-evaluated, and there are currently two suggestions under consideration.⁷ The first suggestion assigns the initial charge separation to either P₇₀₀⁺/(ec2_Aec3_A)⁻ or P₇₀₀⁺/(ec2_Bec3_B)⁻, followed by electron transfer through both branches to the respective phylloquinones, A₁ (−0.8 V). According to the second model, there would be two initial charge separation events between each Chlec2⁺/Chlec3⁻ pair, followed by rapid electron transport from P₇₀₀, generating P₇₀₀⁺, accompanied by reduction of A₁, as above. In this model, two light traps would compete for charge separation. Independent of how charge separation occurs, the negative charge from A₁ is then transferred through a series of highly reducing [4Fe4S] molecules (F_X, F_A, and F_B), and it is eventually used to reduce ferredoxin (Fd, redox potential in the range of −0.45 to −0.40 V, depending on the organism). The oxidant P₇₀₀⁺ (+0.43 V) generated at PSI (Fig. 2b) is not able to oxidize water; it accepts electrons from the cyt b₆/f complex, through a lumen-soluble plastocyanin (PC) molecule (+0.38 V).

Ferredoxin is a [2Fe2S] cluster-containing, soluble protein that serves as a carrier of reductant to a number of different acceptors, including the algal [FeFe]-hydrogenase enzyme (−0.42 V; see section 2.1.1) and the ferredoxin/NAD(P)H oxidoreductase (FNR, −0.32 V). The latter catalyzes the reduction of NADP⁺ (nicotinamide adenosine diphosphate) to NADPH, and is coupled to the fixation of CO₂ into starch, the normal role of photosynthesis under aerobic conditions. In cyanobacteria, H₂ production by the [NiFe]-hydrogenase is not coupled to Fd but to NAD(P)H (see section 2.1.2).

The water-splitting reaction that takes place at the oxidizing side of PSII is unique in nature. It is catalyzed by the OEC, which contains four manganese, one calcium, and one chloride atom bound to PSII protein residues on the lumen side of the thylakoid membranes. The mechanism of water oxidation and the structure of the OEC are still under investigation, and there are currently many proposals in the literature.⁸ The OEC is known to accumulate four successive positive charges upon each PSII charge-separation event, while undergoing the so-called S-state cycle from S₀ to S₄ (with S₁ being the dark-stable state). Based on extensive spectroscopic work (XFAS, XANES, and EPR), most researchers agree that all but the S₂ to S₃ state transition are accompanied by oxidation of manganese, although there is still some controversy on this point. Each S-state cycle (which consume two H₂O molecules) results in the evolution of one O₂ molecule and the release of four protons. The pattern of proton release during the S-state cycle is also still debatable, although the role of Y_Z and D1-histidine 190 as proton-exchange residues is generally accepted. The role of calcium in the chemistry of water oxidation is believed to be in binding a substrate water molecule as a Lewis acid or by interfering with the delivery of protons to reduced, unprotonated Y_Z (Y_Z) at the S₂ state. The chloride requirement for O₂ evolution and its binding location at the OEC is still not clear, although its presence may help maintain the rates of water oxidation (see ref. 8 for more details).

Besides reductant, photosynthetic CO₂ fixation requires ATP, which is produced by the chloroplast ATP synthase. The production of ATP is linked to the dissipation of the proton gradient across the thylakoid membranes, which is established during photosynthetic LET (see above). Besides LET, oxygenic photosynthetic organisms are able to synthesize additional ATP by using a cyclic electron transfer (CET) pathway around PSI. This pathway does not generate net reductant, but it contributes, instead, to the proton gradient across the thylakoid membrane, thus increasing the ratio of ATP/NADPH produced by photosynthesis. In the CET pathway, reductants generated by the excitation of PSI are transferred to Fd, which then shuttles them to the PQ pool (or directly to the cyt b₆/f complex). From there, the reductant is used by PC to re-reduce P₇₀₀⁺, thus completing the cycle. The ATP/NADPH ratio, which changes according to the metabolic requirements of the organism, controls the rates of LET (by means of a signal transduction mechanism that senses the pH of the lumen) or by regulating the relative proportions of LET to CET by a mechanism called “state transitions” (see section 2.1.1).

2.1.1 Green algal [FeFe]-hydrogenases. The hydrogenase enzymes present in oxygenic algae and cyanobacteria belong to different categories, depending on the chemical nature of their catalytic centers (Fig. 3): algal hydrogenases are [FeFe] enzymes, whereas cyanobacterial hydrogenases belong to the class of [NiFe]-hydrogenases (see also section 2.1.2). Some hydrogenases are reversible (mostly [FeFe] ones), and will catalyze either H₂ production or oxidation, depending on the nature of their interacting redox partner.⁹ Most [NiFe]-hydrogenases are primarily H₂-oxidizing enzymes. Surprisingly, no [NiFe]-hydrogenase has yet been found in eukaryotes, while many prokaryotes are known to contain both types of hydrogenases.

Fig. 3 (a) Model of the algal HYDA2 hydrogenase protein structure based on sequence homology to the crystallized Clostridium pasteurianum CpI [FeFe]-hydrogenase (provided by Dr Paul King, NREL); line drawing of the chemical structure of the [2Fe2S] component of the H-cluster that is bound through a protein cysteine residue to the [4Fe4S] component. The latter is further bound to the protein structure through four cysteine residues. (b) Diagram of the cyanobacterial hydrogenase complex, consisting of a hydrogenase and a diaphorase subcomplex. The hydrogenase subcomplex consists of a large subunit, HoxH (purple) containing the [NiFe] catalytic cluster, and a small subunit HoxY (green) containing a [4Fe4S] cluster (which is not shown in the diagram); 3-D structure of the crystallized hydrogenase sub-complex from Desulfovibrio gigas, with the large subunit shown in purple and the small subunit in green (reproduced with permission from A. Volbeda, M. H. Charon, C. Piras, E. C. Hatchikian, M. Frey, J. C. Fontecilla-Camps, Nature, 1995, 373, 580–587 [ref. 10]); line drawing of the chemical structure of the [NiFe] catalytic cluster which is bound to the protein structure through four cysteine residues.

Algal [FeFe]-hydrogenases are monomeric enzymes with a molecular weight of around 48 kDa (depending on the particular algal species), whereas dimeric, trimeric, and tetrameric hydrogenases are known to be present in certain anaerobic organisms.¹¹ [FeFe]-hydrogenases are characterized by the presence of an H-cluster at their catalytic centers, bound to three conserved protein motifs: L1 (TSCCPxW), L2 (MPCxxKxxE), and L3 (ExMACxxGCxxGGGxP).¹² As shown in Fig. 3a, the active site of [FeFe]-hydrogenases consists of a cubane [4Fe4S] cluster coordinated by three cysteine residues (one from each L motif) and connected to a [2Fe2S] center, whose electrophilic character is stabilized by the presence of CO and CN ligands¹³ and by a dithiolate bridge between the 2S atoms of the active site.^14,15 The chemical nature of the dithiolate bridge had been previously assigned to either dithiomethylamine or propanediol.^14,15 However, a recent report¹⁶ proposes it to be a dithiomethylether ligand, based on higher-resolution crystallography and density function theory (DFT) results. Homology models for the structure of the algal [FeFe]-hydrogenases have been constructed (Fig. 3a), based on X-ray crystallographic structural data for the bacterial [FeFe]-hydrogenase, C. pasteurianum (Cp1) and Desulfovibrio desulfuricans (DdH), as algal [FeFe]-hydrogenases are similar to the C-terminal portion of bacterial [FeFe]-hydrogenases.^14,15 However no crystal structure from the green algal enzymes has yet been obtained.

Hydrogenases catalyze the reversible reduction of protons to H₂:

2H⁺ + 2e⁻⇌ H₂ (1)

The mechanism of H₂ production by [FeFe]-hydrogenases is not completely understood. It is known that electrons are delivered by Fd to a docking site next to the [4Fe4S] cluster, from which they are transferred to the H-cluster, doubly reducing the distal iron atom, Fe₂. The reduced Fe₂ is stabilized by the presence of CO and CN ligands. The proton pathway from the protein surface to the catalytic site is still being deconvoluted, and it must involve a cysteine residue (located next to the H-cluster¹⁵), which probably acts as the final binding site for one of the protons. It is clear, though, that the second proton binds to Fe₂ where it is doubly reduced to a hydride anion. The final reaction involves a process in which the resulting hydride anion recombines with the bound proton and is released as H₂ gas.

[FeFe]-hydrogenases have very high turnover rates of about 6000–9000 s⁻¹. They are, however, particularly sensitive to O₂ and most of them are irreversibly inactivated after exposure to O₂, the by-product of water oxidation. Interestingly, the [FeFe]-hydrogenase from Desulfovibrio vulgaris (Hildenborough), which is purified aerobically, has been reported to become highly sensitive to O₂ after undergoing reductive activation.¹⁷ Similarly, a recent report¹⁸ has demonstrated electrochemically the rapid and reversible inactivation of the D. desulfuricans TCC 7757 hydrogenase at more positive redox potentials (above 0 mV), as well as a decrease in O₂ sensitivity between the anaerobic and active reduced state and the anaerobically oxidized and inactive form of the enzyme. The former becomes irreversibly damaged after exposure to O₂, whereas the latter can be reductively reactivated. The nature of the irreversible inactivation of [FeFe]-hydrogenases remains unclear.

There are two H₂-photoproduction pathways in green algae,¹¹ and a third, light-independent fermentative H₂ pathway coupled to starch degradation, which operates at 1/100th of the rate of the photobiological routes and will not be discussed here.¹⁹ The first H₂-photoproduction pathway is dependent on PSII and PSI activities, whereas the second one is dependent on NADP-PQ oxidoreductase (NPQR) and PSI activities only, as indicated in Fig. 4a. The first mechanism supplies reductants from PSII to PSI through water oxidation, as described above, and the second transfers reductants from the degradation of starch through NPQR to the plastoquinone pool. After illumination, the reductant is then re-energized by PSI and supplies electrons to the hydrogenase through ferredoxin.²⁰

Fig. 4 Hydrogenase-catalyzed H₂-photoproduction pathways in (a) green algae and (b) cyanobacteria.

For algal hydrogenases, anaerobiosis is a prerequisite condition for H₂ production. In the laboratory, this is achieved either by (a) sealing and purging the culture with inert gas in the dark, (b) providing exogenous reductant, or (c) allowing dark cellular respiration to metabolize the dissolved O₂. It has also been shown that sustained and continuous photobiological H₂ production can occur when cultures are deprived of sulfate.²¹ Sulfate-deprived cultures have attenuated rates of photosynthetic O₂ evolution due to continuous photodamage of the PSII D1 protein, which cannot be replaced quickly. The cells maintain relatively high levels of respiration that exceed the rates of photosynthesis, which leads to anaerobiosis and H₂ production for several days. The maximum rates of H₂ production observed are, however, only about 25% of the corresponding maximum potential of photosynthetic electron transport, suggesting that this process is still limited by different factors.²² These factors include down-regulation of the electron transport rate, caused either by non-dissipation of a proton gradient across the photosynthetic membrane²³ or by state transitions that shift photosynthesis from linear to mostly cyclic electron transport.²⁴ The non-dissipation of the proton gradient under sulfate deprivation is supported by two main observations:²⁵ (a) Rubisco (ribulose bisphosphate carboxylase, the first enzyme in the CO₂ fixation pathway) levels decrease within the first 24 h, thus limiting CO₂ fixation and ATP utilization; (b) proton uncouplers stimulate the rates of H₂ photoproduction. The influence of state transitions on down-regulation of electron transport was confirmed by the recent report of a mutant locked in state I (and thus unable to perform CET) that showed higher rates of H₂ photoproduction than its parental strain.²⁰

In addition, the presence of a large light-harvesting antenna causes under-utilization of absorbed photons, resulting in energy losses as fluorescence or heat. Thus, in a high-density mass culture, cells at the surface over-absorb and waste sunlight, whereas cells deeper in the culture are deprived of light due to shading. Decreasing the chlorophyll antenna size by genetic engineering would therefore increase the light utilization efficiency throughout a dense culture.²⁶ Research efforts have been made to identify and manipulate genes that regulate the Chl antenna size of green algae, but no work has yet demonstrated a positive impact on H₂ production.²²

Another approach to circumventing the O₂-sensitivity problem is to engineer an O₂-tolerant [FeFe]-hydrogenase.²² Molecular dynamics simulations, solvent accessibility maps, and potential mean energy estimates have been used to study potential gas diffusion pathways in the enzyme.²² Based on these data, site-directed and random mutagenesis approaches are being applied to engineer an O₂-resistant catalyst.

Algal H₂ production in general is also limited by the existence of pathways that compete directly with the hydrogenase for photosynthetic reductant from ferredoxin (the physiological electron donor to hydrogenase). These include FNR, FTR (ferredoxin/thioredoxin reductase), nitrite reductase, sulfite reductase, and glutamate synthase. The first pathway occurs during normal, aerobic photosynthesis that generates NADPH required for CO₂ fixation. The second pathway is responsible for the reduction of thioredoxins, which are involved in the regulation of different sets of CO₂ fixation enzymes. The third pathway, through either nitrite or sulfite reductase, is activated only under specific nutrient conditions such as, respectively, ammonium and sulfate depletion. Glutamate synthase reduces glutamine to glutamate, a reaction that is responsible for the synthesis of glutamate under ammonia-limiting conditions. It is clear that a significant amount of reductant could be used by pathways other than the hydrogenase enzyme. To improve H₂ photoproduction from green algae, the competition between hydrogenase and other pathways should be addressed in the future.

2.1.2 Cyanobacterial [NiFe]-hydrogenases. Cyanobacteria perform oxygenic photosynthesis using the same electron transport pathway described for green algae. They display a relatively wide range of morphological diversity, including unicellular, filamentous, and colonial forms. Some filamentous strains form differentiated cells called heterocysts that are specialized in N₂ fixation, in which a nitrogenase functions in the absence of O₂, a potent inhibitor of its activity. In cyanobacteria, as in any diazotrophic bacterium, the reduction of N₂ to NH₃ is accompanied by the formation of molecular H₂. The H₂ produced by the nitrogenase is rapidly consumed by an uptake hydrogenase, an enzyme that has been found in almost all N₂-fixing cyanobacteria. In addition, many strains, both non-N₂-fixing and N₂-fixing, contain a bidirectional hydrogenase that can either produce or consume H₂ according to the cellular redox environment. The bidirectional enzyme probably plays a role in fermentation and/or acts as an electron valve during photosynthesis.²⁷ This section will focus on the latter, which is responsible for catalyzing H₂ photoproduction in the absence of a functional nitrogenase.

The cyanobacterial bidirectional hydrogenase is a heteropentameric complex (Fig. 3b), consisting of a hydrogenase structural unit, HoxYH, and a diaphorase unit (an enzyme capable of catalyzing oxidation of NAD(P)H in the presence of an electron acceptor other than O₂), HoxEFU. The catalytic large subunit (HoxH, about 53 kDa, depending on the organism) contains conserved cysteines that bind Ni and Fe on the active site. The Fe is also coordinated by two CN ligands and one CO ligand. The small subunit (HoxY, 20 kDa) also contains conserved cysteine residues that are involved in coordinating a putative [4Fe4S] cluster. The diaphorase moiety is a flavo-protein composed of a catalytic part (HoxF, 58 kDa and HoxU, 26 kDa) that interacts with the NAD(P)⁺/NAD(P)H couple and a third subunit (HoxE, 19 kDa). Both HoxF and HoxU have typical binding motifs for [2Fe2S] and [4Fe4S] clusters, whereas HoxE may be involved as a bridging subunit in membrane attachment and in electron transport, since its gene sequence also indicates the presence of a [2Fe2S]-binding motif.

The diaphorase subunits were once hypothesized to be shared with the respiratory complex I, based on sequence similarities.²⁹ However, mutants lacking these subunits do not show loss of respiratory activity.³⁰ The molecular weight of the native bidirectional hydrogenase indicates a dimeric complex, Hox(EFUYH)₂.³¹ Under physiological conditions, the bidirectional hydrogenase catalyzes H₂ production using the photo-reduced NAD(P)H (E⁰ = −0.32 V) as the electron donor, mediated via the diaphorase moiety. It has been hypothesized that the selective preference for NAD(P)H is due to the presence of the HoxEFU subunits within the complex, and that the cyanobacterial hydrogenase may be improved for enhanced H₂ production by direct linkage to ferredoxin in mutants devoid of the diaphorase moiety.

The catalytic process is assumed to be the reverse of the well-documented mechanism for H₂ oxidation described for [NiFe]-uptake hydrogenases.²⁸ In that process, H₂ is transported via a hydrophobic channel to the catalytic site, where it is oxidized by a [NiFe] cluster. The Ni atom is probably the site where the H₂ binds as a terminal ligand because the gas channel ends next to it. However, modelling studies suggest that H₂ could bind to the Fe atom instead.²⁸ Hydrogen cleavage is a heterolytic reaction that forms a hydride anion and a proton. The hydride anion gets doubly oxidized by the catalytic center, and a second proton is formed and transferred to a cysteine ligand nearby to complete the reaction. The electrons are transferred from the catalytic center along a redox pathway to a soluble redox partner, and protons are transferred to the protein environment. The electron pathway is formed by a [4Fe4S] cluster proximal to the active site, followed by an intermediate [3Fe4S] cluster and a distal [4Fe4S] cluster exposed to the protein surface. Together, they allow efficient transfer of electrons between the soluble electron acceptor and the active site. Less is known about proton transport pathways, but a glutamate residue close to the initial cysteine ligand seems to be involved in the process.

[NiFe]-hydrogenases are generally more tolerant to O₂ than [FeFe]-hydrogenases, and their inactivation by O₂ and CO is reversible. Upon exposure to O₂, an oxo- or a hydroxo-group is formed, bridging the Ni and the Fe atoms and rendering the [NiFe]-hydrogenase resistant to further O₂ inactivation. Once returned to an anaerobic environment, the hydrogenase activity is restored upon addition of reductants.¹³ This (a) accounts for the reversibility of the [NiFe]-hydrogenase with respect to O₂ inactivation, (b) explains the lack of H₂ evolution in cyanobacterial cultures during photosynthesis, and (c) provides a rationale for the restoration of hydrogenase activity when cultures are assayed under dark, anaerobic conditions.

Similarly to green algae, the electrons used to produce H₂ in cyanobacteria originate from photosynthetic water splitting (Fig. 4b). Using sensitive mass spectrometry for real-time measurements, light-driven H₂ production was detected in some cyanobacterial strains. However, the reaction was short-lived (lasting less than 30 s in the light) and was followed immediately by H₂ uptake. More sustained H₂ production was reported in the Synechocystis ndhB mutant M55, defective in type I NAD(P)H dehydrogenase complex, where continuous H₂ production lasted for about 5 min. Cournac et al.³³ attributed this phenomenon to a combination of lower O₂ evolution and negligible H₂-uptake activities in the mutant.

Several biochemical barriers are associated with sustained cyanobacterial H₂ production. Some of them are common to both algae and cyanobacteria and have already been discussed in section 2.1.1. Additional issues include (a) the identification of more active strains, (b) H₂ consumption by the bidirectional hydrogenase, possibly via complex I-mediated respiration, and (c) H₂ oxidation mediated by the uptake hydrogenase. These issues, as well as approaches to address them, were discussed extensively in a recent review.³³

2.1.3 Cyanobacterial nitrogenases. As mentioned in section 2.1.2, filamentous cyanobacteria, which are composed of vegetative cells and heterocysts, are able to fix N₂ and photoproduce H₂ through the enzyme nitrogenase, which is present exclusively in heterocysts. The key enzyme is the nitrogenase, which is expressed only under nitrogen-free or nitrogen-limited conditions (i.e., up to 2 mM ammonium chloride). Heterocysts receive organic carbon, thought to be in the form of sucrose, fixed by oxygenic photosynthesis occurring exclusively in the vegetative cells. In return, heterocysts provide neighboring vegetative cells with fixed nitrogen required for cell growth. In heterocysts, metabolism of the organic carbon compounds generates NADPH, which serves as the electron donor for PSI-driven reduction of heterocyst-specific ferredoxin (FdH), required for the nitrogenase reaction.³⁴ FdH may also be reduced by NADPH in the dark. The required ATP can be generated by glycolysis, respiration, and/or PSI-driven photophosphorylation (Fig. 1b). In heterocysts, the O₂-sensitive nitrogenase is protected from inactivation by O₂ by three different mechanisms: (a) the heterocysts are surrounded by an O₂-impermeable, glycolipid envelope layer; (b) O₂-producing PSII is absent from the heterocysts; and (c) the heterocysts have increased rates of respiration, which consume any O₂ that may penetrate the envelope.³⁵ A number of non-heterocystous strains are also able to perform N₂ fixation under anaerobic conditions. In these organisms, N₂ fixation and oxygenic photosynthesis are temporally separated.

The cyanobacterial nitrogenase normally harbors molybdenum at its catalytic center, where the reduction of N₂ and protons occurs. Eqn (2) shows that, during N₂ fixation, the nitrogenase also catalyzes an obligatory side reaction, reducing protons to H₂.

N₂ + 8H⁺ + 8e⁻ + 16ATP → 2NH₃ + H₂ + 16ADP + 16P_i (2)

In the absence of N₂, nitrogenases catalyze solely the reduction of protons to H₂, thus decreasing the ATP requirement from 16 to 4 moles per mole of H₂ produced according to eqn (3).

2H⁺ + 2e⁻ + 4ATP → H₂ + 4ADP + 4P_i (3)

The nitrogenase consists of two components, a MoFe-protein (or dinitrogenase) and an Fe-protein (or dinitrogenase reductase)³² (Fig. 5). The MoFe-protein is a α₂β₂ heterotetramer that harbors a unique iron–sulfur cluster—the FeMo-cofactor (FeMo-co)—in which substrate binding and reduction takes place. A second [8Fe7S] metallocluster, termed the P-cluster, is situated at each αβ-pair interface, likely mediating electron transfer between the Fe-protein and the FeMo-co active site. The Fe-protein is a γ₂ homodimer with a single [4Fe4S] cluster located at the dimer interface near the surface, where either ferredoxin or flavodoxin dock for electron donation. Each γ subunit of the Fe protein also contains a Mg-ATP binding site situated at its dimer interface. Following ATP hydrolysis, the Fe-protein undergoes a conformational change that allows electron donation from its [4Fe4S] cluster to the P-cluster of the MoFe-protein to occur. Two ATPs are hydrolyzed for each electron transferred (see eqn (3)).

Fig. 5 Diagram of the nitrogenase protein complex consisting of an Fe-protein and a MoFe-protein subcomplex. The Fe-protein subcomplex is a dimer of two NifH subunits with an [FeS] cluster (shown as a red diamond) bridging the homodimer. The MoFe-protein contains two NifD subunits and two NifK subunits. The FeMo-cofactors (blue circles) are located in each NifD subunit, and the P-clusters are located at the interface between each NifD and NifK subunit; 3-D structure of the dimeric Fe-protein (top) and one of the two NifD–NifK pairs; line drawing of the chemical structure of the FeMo-cofactor involved in catalysis (reproduced with permission from J. Howard and D. Rees, PNAS, 2006, 103, 17088–17093 [ref 36a]; FeMoCo scheme is from C. M. Kozak and P. Mountford, Angew. Chem., Int. Ed., 2004, 43, 1186–1189 [ref 36b].

2.2 Non-oxygenic photosynthesis

Non-oxygenic or anoxygenic photosynthetic bacteria have attracted interest because of their ability to conduct photosynthesis in the absence of air and without producing O₂, using organic acids as electron donors rather than water. In this tutorial review, we will focus on the purple non-sulfur (PNS) photosynthetic bacteria in the family of Rhodospirillaceae due to their versatility in carrying out alternative modes of metabolism including photosynthesis, respiration using organic acids, N₂ fixation into NH₃ and coupled photoproduction of H₂, and dark fermentation using sugars.

Rhodospirillaceae contain a PSII-type photosynthetic reaction center.³⁷ However, in contrast to PSII, these PNS photosynthetic bacteria perform only CET (Fig. 1c) and thus generate only ATP. Reductant for CO₂ fixation and nitrogenase activity is provided by the photoassimilation of organic acids (discussed in more detail below). Light is captured by an antenna containing several bacteriochlorophyll and carotenoid pigments that absorb visible light from 400 to 600 nm and near-infrared light from 800 to about 1010 nm. In Rhodopseudomonas viridis, light excites the reaction center bacteriochlorophyll dimer (P₈₆₅) (Fig. 2c). Charge separation generates oxidized P₈₆₅ (+0.45 V) and reduced bacteriopheophytin (Bpheo, −0.6 V), the primary electron acceptor. As in PSII, the reaction centers of PNS bacteria contain two branches of cofactors, of which only one is active in electron transport (Fig. 2c). The reduction of Bpheo is followed by the sequential reduction of two quinones, a menaquinone, Q_A (−0.2 V) and a ubiquinone, UQ_B (+0.08 V). Doubly reduced UQ_B diffuses out of the bacterial photosystem and transfers electrons to the cyt b/c₁ complex. This reaction is accompanied by protonation and deprotonation of UQ_B, generating a proton gradient across the photosynthetic membrane. CET is completed by electron transfer from the cyt b/c₁ complex through soluble cytochrome c₂ (+0.38 V) to oxidized P₈₆₅. ATP is synthesized via a membrane ATP synthase driven by the ΔpH, and subsequently used for H₂ production by nitrogenase.

The reductants required for H₂ production by nitrogenase are supplied by the photoassimilation of organic acids (Fig. 1c). The most common organic acids utilized by PNS bacteria include formate, acetate, lactate, butyrate, etc., which are the typical waste by-products of the H₂ fermentators being proposed for the integrated system (see section 4). The initial steps in the oxidation of organic acids supply electrons to the soluble ubiquinones, from where they are “pushed” uphill by the ΔpH created by photosynthetic CET. This reverse-electron transport reduces NAD⁺ or ferredoxin; the latter drives nitrogenase-catalyzed H₂ production.

2.2.1 Bacterial nitrogenases. Almost all PNS photosynthetic bacteria can carry out the nitrogenase-catalyzed reactions described in section 2.1.3. Since O₂ is not generated during non-oxygenic photosynthesis, O₂ inhibition and H₂/O₂ gas separation during H₂ photoproduction are not a concern. Based on these attributes, bacterial, nitrogenase-catalyzed H₂ production has attracted interest in recent years.³³

Five important factors currently limit the utilization of the non-oxygenic, nitrogenase-mediated photofermentation process.³⁸ They are: (a) the concomitant occurrence of H₂ uptake (which can be genetically inactivated); (b) the low photofermentation efficiency (PE) of H₂ production (the efficiency by which light energy can be converted into H₂ gas energy); (c) the low turnover rate of nitrogenases (6.4 s⁻¹); (d) the relatively low actual carbon conversion yield (mol of H₂ produced per mol carbon substrate consumed as a percentage of its theoretical maximum); and (e) the availability of sources of organic acids.

Eqn (4) shows the calculation of the theoretical maximum yield of H₂ obtained from the utilization of different organic acids. The coefficients x, y and z reflect the different C, H, and O composition of the most common organic acids utilized by PNS bacteria: formate, acetate, lactate, malate, butyrate, etc.

C_xH_yO_z + (2x−z)H₂O → (y/2 + 2x− 2)H₂ + xCO₂ (4)

Carbon conversion yields as high as 70–84% have been reported, depending on the organic acid composition of the substrate. PNS photofermentation is currently of interest not so much as a stand-alone process, but rather as part of an integrated system (section 4).

3. Light-conversion efficiencies

In this review, we use the term “light conversion efficiency” to denote the fraction of the energy content of the incident solar spectrum that is converted into chemical energy by the organism. This calculation takes into consideration E₀, the portion of the solar spectrum that is absorbed by different organisms (45% of the energy of the solar spectrum for green algae and cyanobacteria and 71% for photosynthetic bacteria), times E₁, the efficiency with which the organism converts the absorbed light to reductants capable of reducing protons to H₂ (reduced ferredoxin). For an oxygenic organism, requiring 4 photons per H₂ produced through a hydrogenase-catalyzed reaction, the latter corresponds to a 31% conversion efficiency: approximately 52 kcal for each average PAR (photosynthetically active radiation) absorbed photon, 550 nm, and 64 kcal for H₂. See the end of this section for additional energy requirements for nitrogenase-catalyzed reactions.

Based on the above calculations, the light-conversion efficiency of oxygenic photobiological systems can be, potentially, very high, up to about 13% of sunlight (Table 1), resulting from the high quantum yield of the primary charge-separation reactions.³⁹ However, the rates of photosynthesis saturate at fairly low light intensities. This phenomenon is due to the presence of large, pigment-containing, light-harvesting antennae associated with the photosystems, such that the rate of photosynthesis is limited by light absorption at low light intensities but saturates at an intensity of about 1/5th to 1/10th of sunlight. From this point on, photosynthesis becomes limited by biochemical factors such as the rate of electron transport between Photosystem II and Photosystem I, and by intracellular mechanisms that sense the ratio between reductant generated and ATP produced and that down- or up-regulate photosystem activity accordingly. As a consequence, the light-conversion efficiency of photosynthetic processes declines at intensities above the light saturation point, resulting in very inefficient sunlight utilization of about 2% or lower. Furthermore, the overall light-conversion efficiency to H₂ depends on the nature of the catalyst that is coupled to photosynthetic electron transport. Whereas hydrogenases do not require extra chemical or energetic input besides protons and electrons, nitrogenases require four ATPs for each H₂ molecule produced. This leads to decreased light-conversion efficiency for nitrogenase-based versus hydrogenase-based systems. However, hydrogenase-catalyzed H₂ production is a reversible reaction, with a ΔG′° of 3.78 kJ mol⁻¹ and an apparent equilibrium constant of 0.218, with only 30 mV between reductant and oxidant. Nitrogenase-catalyzed H₂ production, on the other hand, has a ΔG′° of −463 kJ mol⁻¹ and an apparent equilibrium constant of 1.4 × 10⁸¹, implying its spontaneous and irreversible character.⁴⁰

Table 1 Characteristics of the three major biological H₂-photoproduction processes

	Green algae and cyanobacteria (hydrogenase-based)	Cyanobacteria (nitrogenase-based)	Purple bacteria (non-oxygenic, nitrogenase-based)
Light absorption spectrum	400 to 700 nm	400 to 700 nm	400 to 600 nm and 800 to 1010 nm
Photons/H2 generated	4	15	15
Estimated maximum light conversion efficiency (EMLCE)	10–13%	6%	6%
Electron donor	Water	Water	Organic acids

---

The theoretical light-conversion efficiency over the utilizable portion of the solar spectrum for most PNS bacteria (400 to 600 nm and 800 to 1010 nm) is calculated to be maximally 9.5% (see first paragraph for explanation). This calculation is based on the approximately 15 photons that are needed to (a) raise the energy of the electrons to a level at which they can reduce ferredoxin, the electron donor of nitrogenase, and (b) meet the nitrogenase’s requirement of four ATP per H₂ produced. Because only 71% of the incident solar light is photosynthetically active, the maximum light-conversion efficiency of non-oxygenic H₂ production is closer to 6% (see Table 1).

4. Integrated systems

Many anaerobic organisms are able to produce H₂ through dark fermentation. Fermentation has several attributes that make it an attractive technology: (a) it requires simple reactor design and dark operation; (b) fermentative microbes are readily available in sewage sludge, garden soils, and anaerobic compost; (c) diverse waste materials can be used as substrates; and (d) high rates of H₂ production can be achieved, unsurpassed by other biological processes, with reported values ranging from 184 to 2710 mL H₂ L⁻¹ h⁻¹.^41–43 However, fermentative H₂ production faces a major barrier, which is the relatively low theoretical molar yield of 4 H₂/glucose. So, at present, none of the four biological processes for H₂ production, either photobiological or fermentative, is ready for implementation as a “stand-alone” system. This is due to a combination of lower than expected performance under natural conditions, high cost of substrates and processes, and slow progress toward solving basic biochemical issues such as the O₂ sensitivity of hydrogenases. Although we scientists are optimistic that these issues will eventually be solved, we have also started to address alternative methods to circumvent the current limitations, by integrating two or more of the biological processes into a single system.

4.1 Coupling of photosynthetic bacterial H₂ production to fermentation

The nature of the substrate required for non-oxygenic photosynthetic H₂ production (a mixture of organic acids) is similar to that of the effluent of the H₂-producing fermentor. Indeed, Claassen’s group in the Netherlands has been exploring the coupling between the two processes for quite a few years.⁴⁴ The approach utilized by the Dutch group and many others consists of selecting and anaerobically growing fermentative consortia from sewage or waste water, enriched in sugars, to allow for maximum H₂ accumulation. This step is followed by transfer of the effluent from the fermentor (enriched in organic acids such as acetate, propionate, butyrate, etc.) to the photobioreactor containing the photosynthetic bacteria. In the light, the photosynthetic bacteria will further convert the organic acids to H₂ and CO₂.

Although this coupled system is better than any of the two processes (photobacterial and fermentative H₂ production) separately, it still has a number of disadvantages, as described in Table 1: (a) photosynthetic bacteria cannot utilize all of the solar spectrum; (b) the second process is nitrogenase-based and therefore has a low light-conversion efficiency; and (c) the substrate for photosynthetic H₂ production is not water. To address these issues, a second integrated system has been proposed⁴⁵ that includes oxygenic photosynthetic microalgae.

4.2 Three-component integrated biological system for H₂ production

This proposed system takes advantage of the fact that the absorption spectra of green algae (or cyanobacteria) and photosynthetic purple bacteria are complementary (Table 1): the chlorophyll and phycobilin molecules in green algae and cyanobacteria absorb maximally between 400 and 700 nm (covering about 45% of the energy in the solar spectrum), whereas the bacteriochlorophylls in photosynthetic bacteria have absorption maxima at 400 to 600 nm and 800 to 1010 nm. The near-infrared absorption maxima of photosynthetic bacteria adds another 25% of the energy of the solar spectrum to that of green algae/cyanobacteria, thus potentially increasing the light-conversion efficiency if the two organisms are co-cultivated under H₂-producing conditions⁴⁵ or cultivated in separate, stacked reactors (Kosourov et al., unpublished results).

As shown in Fig. 6, the role of the photobioreactors is to produce H₂, using both oxygenic and non-oxygenic photosynthesis, and to accumulate cell biomass. The biomass is then used as substrate for dark fermentation by microbes that are able to convert it into H₂ and CO₂, while waste organic acids from the latter are fed back into the photobioreactors to stimulate biomass accumulation. Although the integrated system is still in development, many of the individual steps have been tested successfully.

Fig. 6 Three-component integrated biological system for H₂ production. See text for more details.

Acknowledgements

This work was supported by the US Department of Energy Office of Science (Basic Energy Science and BER) and the Hydrogen, Fuel Cells and Infrastructure Technologies Program under Contract No. DEAC36-99GO10337 with the National Renewable Energy Laboratory. We acknowledge Dr Paul King for the homology model of HYDA2 shown in Fig. 3a, and Dr Peter Wolk for discussions on carbon metabolism in heterocysts.

References

1. F. Rappaport and B. A. Diner, Coord. Chem. Rev., 2008, 252, 259.
2. Y. Li, A. van der Est, M. G. Lucas, V. M. Ramesh, F. Gu, A. Petrenko, S. Lin, A. N. Webber, F. Rappaport and K. Redding, Proc. Natl. Acad. Sci. U. S. A., 2006, 103, 2144.
3. (a) N. Nelson and C. F. Yocum, Annu. Rev. Plant Biol., 2006, 57, 521; (b) G. Zubay, Biochemistry, W.C. Brown Publishers, 1998, p. 380.
4. F. Rappaport, M. Guergova-Kuras, P. J. Nixon, B. A. Diner and J. Lavergne, Biochemistry, 2002, 41, 8518.
5. M. Grabolle and H. Dau, Biochim. Biophys. Acta, 2005, 1708, 209.
6. A. Cuni, L. Xiong, R. Sayre, F. Rappaport and J. Lavergne, Phys. Chem. Chem. Phys., 2004, 6, 4825.
7. M. G. Müller, J. Niklas, W. Lubitz and A. R. Holzwarth, Biophys. J., 2003, 85, 3899.
8. J. P. McEvoy and G. W. Brudvig, Chem. Rev., 2006, 106, 4455.
9. P. M. Vignais and B. Billoud, Chem. Rev., 2007, 107, 4208.
10. A. Volbeda, M. H. Charon, C. Piras, E. C. Hatchikian, M. Frey and J. C. Fontecilla-Camps, Nature, 1995, 373, 580.
11. T. Happe, A. Hemschemeier, M. Winkler and A. Kaminski, Trends Plant Sci., 2002, 7, 246.
12. J. Meyer, Cell. Mol. Life Sci., 2007, 64, 1063.
13. M. Frey, ChemBioChem, 2002, 3, 153.
14. Y. Nicolet, C. Piras, P. Legrand, C. E. Hatchikian and J. C. Fontecilla-Camps, Struct. Fold Des., 1999, 7, 13.
15. J. W. Peters, W. N. Lanzilotta, B. J. Lemon and L. C. Seefeldt, Science, 1998, 282, 1853.
16. A. S. Pandey, T. V. Harris, L. J. Giles, J. W. Peters and R. K. Szilagyi, J. Am. Chem. Soc., 2008, 130, 4533.
17. A. J. Pierik, W. R. Hagen, J. S. Redeker, R. B. Wolbert, M. Boersma, M. F. Verhagen, H. J. Grande, C. Veeger, P. H. Mutsaers and R. H. Sands, Eur. J. Biochem., 1992, 209, 63.
18. K. A. Vincent, A. Parkin, O. Lenz, S. P. Albracht, J. C. Fontecilla-Camps, R. Cammack, B. Friedrich and F. A. Armstrong, J. Am. Chem. Soc., 2005, 127, 18179.
19. R. P. Gfeller and M. Gibbs, Plant Physiol., 1984, 75, 212.
20. F. Mus, L. Cournac, V. Cardettini, A. Caruana and G. Peltier, Biochim. Biophys. Acta, 2005, 1708, 32.
21. A. Melis, L. Zhang, M. Forestier, M. L. Ghirardi and M. Seibert, Plant Physiol., 2000, 122, 127.
22. M. L. Ghirardi, M. C. Posewitz, P. C. Maness, A. Dubini, J. Yu and M. Seibert, Annu. Rev. Plant Biol., 2007, 58, 71.
23. O. Kruse, J. Rupprecht, K. P. Bader, S. Thomas-Hall, P. M. Schenk, G. Finazzi and B. Hankamer, J. Biol. Chem., 2005, 280, 34170.
24. C. de Vitry, Y. X. Ouyang, G. Finazzi, F. A. Wollman and T. Kallas, J. Biol. Chem., 2004, 279, 44621.
25. L. Zhang, T. Happe and A. Melis, Planta, 2002, 214, 552.
26. A. Melis, J. Neidhardt and J. R. Benemann, J. Appl. Phycol., 1999, 10, 515.
27. J. Appel, S. Phunpruch, K. Steinmuller and R. Schulz, Arch. Microbiol., 2000, 173, 333.
28. S. Niu and M. B. Hall, Inorg. Chem., 2001, 40, 6201.
29. J. Appel and R. Schulz, Biochim. Biophys. Acta, 1996, 1298, 141.
30. G. Boison, H. Bothe, A. Hansel and P. Lindblad, FEMS Microbiol. Lett., 1999, 174, 159.
31. O. Schmitz, G. Boison, H. Salzmann, H. Bothe, K. Schütz, S. H. Wang and T. Happe, Biochim. Biophys. Acta, 2002, 1554, 66.
32. L. Cournac, G. Peltier and P. M. Vignais, J. Bacteriol., 2004, 186, 1737.
33. P. Tamagnini, E. Leitao, P. Oliverira, D. Ferreira, F. Pinto, D. J. Harris, T. Heidoen and P. Lindblad, FEMS Microbiol. Rev., 2007, 31, 692.
34. P. Wolk and A. Ernst and J. Elhai, in The Molecular Biology of Cyanobacteria, ed. D. A. Bryant, Kluwer Academic Publishers, Dordrecht, The Netherlands, 1995, p. 769.
35. Q. Fan, G. Huang, S. Lechno-Yossef, C. P. Wolk, T. Kaneko and S. Tabata, Mol. Microbiol., 2005, 58, 227.
36. (a) J. Howard and D. Rees, Proc. Natl. Acad. Sci. U. S. A., 2006, 103, 17088; (b) C. M. Kozak and P. Mountford, Angew. Chem., Int. Ed., 2004, 43, 1186.
37. A. W. Roszak, T. D. Howard, J. Southall, A. T. Gardiner, C. J. Saw, N. W. Isaacs and R. J. Cogdell, Science, 2003, 302, 1969.
38. I. Akkerman, M. Janssen, J. Rocha and R. H. Wijffels, Int. J. Hydrogen Energy, 2002, 27, 1195.
39. M. L. Ghirardi, Indian J. Biochem. Biophys., 2006, 43, 201.
40. R. A. Alberty, J. Biol. Chem., 1994, 269, 7099.
41. R. Datar, J. Huang, P. C. Maness, A. Mohagheghi, S. Czernik and E. Chornet, Int. J. Hydrogen Energy, 2007, 32, 932.
42. N. M. Demain and J. H. Wu, Microbiol. Mol. Biol. Rev., 2005, 69, 124.
43. Z. Kadar, T. DeVrije, M. A. Budde, Z. Szengyel, K. Reczey and P. A. Claassen, Appl. Biochem. Biotechnol., 2003, 105–108, 577.
44. P. A. M. Claassen, J. B. van Lier, A. M. Lopez-Contrearas, E. W. J. van Niel, L. Sigtsma and A. J. M. Stams, Appl. Microbiol. Biotechnol., 1999, 52, 741.
45. A. Melis and M. R. Melnicki, Int. J. Hydrogen Energy, 2006, 31, 1563.

---

Footnote

† Part of the renewable energy theme issue.

---