Biosynthetic Approaches Towards the Design of Artificial Hydrogen Evolution Catalysts

Abstract

Hydrogen is a clean and sustainable form of fuel that can minimize our heavy dependence on fossil fuels as the primary energy source. The need of finding greener ways to generate H₂ gas has ignited interest in the research community to synthesize catalysts that can produce H₂ gas by the reduction of H⁺. The natural H₂ producing enzymes hydrogenases have served as an inspiration to produce catalytic metal centers akin to these native enzymes. In this article we describe recent advances in the design of a unique class of artificial hydrogen evolving catalysts that combine the features of the active site metal(s) surrounded by a polypeptide component. The examples of these biosynthetic catalysts discussed here include i) assemblies of synthetic cofactors with native proteins; ii) peptide-appended synthetic complexes; iii) substitution of native cofactors with non-native macrocyclic cofactors; iv) metal substitution from rubredoxin; and v) a reengineered Cu storage protein into a Ni binding protein. Aspects of key design considerations in the construction of these artificial biocatalysts and insights gained into their chemical reactivity are discussed.

Graphical Abstract

INTRODUCTION

Energy is one of the most pressing challenges of the 21^st century. Fossil fuels such as coal oil and natural gas continue to be the primary source catering to the world’s energy need. However, the fossil fuel reserve is finite. Therefore, this form of energy is not sustainable. Furthermore, burning of fossil fuel emits CO₂ into the atmosphere causing environmental pollution and climate change.^[1] Anthropogenic activities have led to an alarming increase in the atmospheric CO₂ level since the industrial revolution.^[2] Therefore, producing energy from renewable natural sources such as sunlight, air, and water is desirable.^[3] There is an abundant supply of these natural resources that can be employed to generate energy without causing harm to the environment.

Among many different forms of renewable energies, H₂ gas is one of them that can potentially replace our heavy dependence on fossil fuels.^[4] H₂ can be combined in a fuel cell with O₂ to produce heat and electricity to perform work. The byproduct of this reaction- H₂O is innocuous and non-toxic. Thus, if H₂ replaces the fossil fuels, we can obtain energy in a sustainable and environmentally benign way. However, H₂ is not directly available from the atmosphere or from any natural resource. Therefore, in order for H₂ to be considered as an effective, clean, sustainable, and cheap fuel, it must be produced in an environmentally benign manner from other renewable resources.^[5] Potential approaches to produce clean H₂ from natural resources include the electrolysis of water, photochemical water splitting, and from microorganisms that produce H₂.

Hydrogenase enzymes found in many microorganisms produce H₂ naturally and oxidize it as a fuel to drive their cellular processes.^[6] H₂ produced this way is clean, sustainable, and requires a very low driving force for the reaction. Hydrogenases with binuclear active sites reduce H⁺ to H₂ and oxidize H₂ to H⁺ in a reversible manner (2H⁺ + 2e⁻ ↔ H₂). Depending on the organism, H⁺ can be reduced to produce H₂ when there is a need to remove the reducing equivalents or employed as terminal electron acceptors for metabolic purposes, while H₂ is oxidized to produce electrons and protons. The mononuclear enzymes contain Fe as the active site metal and they split H₂ to carryout the reversible reduction of methenyltetrahydromethanopterin (methenyl-H₄MPT⁺) by H⁻ with the subsequent release of a H⁺. The binuclear active sites consist of the [NiFe] and [FeFe] family (Fig. 1). Bridged thiolates and a mixture of CO and CN⁻ are the common ligands in both of the binuclear classes of enzymes. In the [NiFe] hydrogenases two Cys thiolates bridge the Ni and Fe. Two additional terminal Cys ligands exclusively bound to Ni, along with two CN⁻ and one CO bound to Fe, complete the coordination environment. There is another bridging ligand that changes its identity between H⁻ and OH⁻. depending on the catalytic intermediate that the enzyme passes through. In contrast, the [FeFe] hydrogenases are connected to the protein via a single Cys residue bound to the proximal Fe (Fe_p) that bridges the [Fe₄S₄] cluster. The H-cluster also features a bridgehead amide connecting the two bridged S ligands, a bridging CO, a terminal CO and CN⁻ bound to each of the Fe. Although hydrogenases have held a significant promise in producing H₂ as a green source of energy, their widespread use is limited due to their O₂ reactivity, complexity, and low production yield. One approach to circumvent this challenge is to design catalysts inspired by these unique class of enzymes and sustain H₂ production via artificial photosynthesis using solar energy.

Figure 1.

Active sites of [NiFe] (A) and [FeFe] (B) hydrogenases. Fe_p and Fe_d in (B) represent proximal and distal Fe, respectively.

Plants contain photosynthetic reaction centers which can sequester the energy derived from sunlight to oxidize H₂O to O₂ and reduce atmospheric CO₂ into sugars to be used as fuels in a manner that is truly environmentally attractive and sustainable. In order to accomplish artificial photosynthesis, hydrogenase-inspired catalysts have been designed employing earth abundant metals which can reduce H⁺ to H₂ mimicking the reductive half of the natural photosynthesis to produce solar fuels.^[7] The conception of artificial photosynthesis was first put forth by Giacomo Ciamician in the early 20th century.^[8] Ideally, sunlight should be harvested by chemical molecules that can lead to charge separation. The accumulated oxidizing and reducing equivalents should then be transferred to the catalysts for H₂O oxidation to O₂ and H⁺ reduction to H₂ (Fig. 2). While the theoretical efficiency for the conversion of sunlight into chemical energy is ~6.7%, only 0.3% of the solar energy is stored by plants.^[9] Nonetheless, it has been estimated that out of the 120000 TW of available energy from sunlight, if 0.16% of Earth’s land were to be covered with 10% efficient solar cells. ~20TW of energy could be produced - enough to meet the world’s projected energy demand in the mid-21st century.^[10]

Figure 2.

Overview of the reductive half of artificial photosynthesis leading to H₂ production. The reducing equivalents on the catalyst are represented as the reductive pool.

A great deal of research has been directed towards understanding the structure, function and catalytic mechanisms of native hydrogenases. For reviews in this topic see references.^{[6a–d, 11]} A huge number of molecular catalysts as functional analogs of hydrogenases employing earth abundant metals such as Ni, Co and Fe have been synthesized, characterized and employed as electrocatalytic and photocatalytic hydrogen evolution reaction (HER) catalysts as reviewed in references.^[12] A majority of these molecular catalysts are soluble only in organic solvents and require highly acidic conditions for activity. Alternative approaches have been adopted that involve hosting the active metal sites supported with protein or peptide scaffolds that can mimic the protein environment and can be studied in water.^[13] Inspired by the native enzymes, remote H-bonding interactions can be installed in these systems by amino acid substitutions to acidic or basic residues which can facilitate H⁺ transfer to the active site and thus improve activity.^[14] Here we describe specific examples of biomolecular artificial HER catalysts (Fig. 3) inspired by the binuclear classes of hydrogenases which have been constructed by i) self-assembly of photosystem I with well-known molecular catalysts; ii) synthetic catalysts attached to peptides; iii) substitution of native cofactors from proteins with macrocyclic non-native metallic cofactors; iv) substitution of native metal bound to endogenous ligands in an ET protein with non-native metals; and v) repurposing a copper storage protein to a Ni binding protein. A summary of catalytic parameters from photo-induced and electrochemical H₂ production studies of these select biomolecular catalysts are presented in Table 1.

Figure 3.

Various approached towards artificial HER catalysts on protein and peptide scaffolds.

Table 1.

Experimental catalytic parameters for H⁺ rcduction/H₂ oxidation for biosynthetic catalysts discussed here. Entries in blue are from electrochemical data.

Catalyst	Metal	TOF (min−1)	TON	pH	η (mV)	FE (%)	Duration (h)	Ref
Co⊂PSI	Co	170a	5.2x103	6.3			1.5	15
PtNP⊂PSI	Pt	350a	8x104	6.2			5	16
Ni⊂PSI	Ni	44a	1.87x103	6.3			3	18
Fld-Ni⊂PSI	Ni	75a	2.8x103	6.3			4	18
Ru-Fld-Ni	Ni	6.8b	620	6.2			6	20
Ru-Fd-Co	Co	0.8b	210	6.3			6	21
Ru-Fld-Co	Co	0.5b	85	6.3			6	22
Thylakoid/PtNP	Pt	35c		6.2			3.5	23
(Spinach)		0.8d
Peptide-[Fe2CO6]	Fe	0.6	84	4.5			2.3	27
Co-GlyGlyHis	Co		275	8	600	91	2.5	28
H-apo cyt c	Fe	2.1	82	4.7			3	29
H-apo minicyt c	Fe	0.5	12	4.7
[FeFe][Ru]-Pep-18	Fe	0.2	9	8.5			2	30
[NiII(PCy2NGly2)2]4+	Ni	1.98x103e		0.7	150			14a
[NiII(PCy2NArg2)2]8+	Ni	1.4x105f		0.5	750			14b
[Ni(Pph2NNNA2)2]2+	Ni	8.6x106e		<3	480			31a, 31b, 31e, 31j
		0.3-2x105f						13d
CoMb	Co	0.5	234	6.5			8	33
		1.5	518	7
		0.9	454	7.5
H64ACoMb	Co		331	6.5			8
H64 A/H97 ACoMb	Co		512	6.5			8
WT Cyt b562	Co	0.3	125	7			8	34
M7A cyt b562	Co	0.7	305	7			8
M7D cyt b562	Co	0.6	275	7			8
M7E cyt b562	Co	0.4	200	7			8
FeFe-NB	Fe	2.3	130	4			6	35
CoMP11-Ac	Co	402	2.5x104	7	852	95	4	36
Ht-CoM61A	Co		2.7x105	7	830	90	6	13e
CoMC6*a	Co		2.3x105	6.5	580	96	3	37
1Bio⊂Sav	Co	0.2	51	11.5			4	38b
2Bio⊂Sav	Co	1.5	166	11.5			4	38b
Fe-Fe⊂Sav	Fe		48	4.5			9	38c
NiRd	Ni	0.5	32	6.5			1	40
	Ni	80		4.5	540	33	2
RuNiRd		0.1	3.5	6.5			2	41
NBP	Ni	1	115	7			2	44
		3.5	210	5	530	93	1

^aCalculated per PSI;

^bcalculated per Ru-photosensitizer;

^cin the presence and

^din the absence of sacrificial electron donor;

^eFor H₂ oxidation;

^ffor H⁺ reduction.

RESULTS AND DISCUSSIONS

I. Assemblies of synthetic cofactors with native proteins.

Utschig and coworkers have developed hybrid bioinorganic systems by combining the light harvesting protein photosystem I (PSI) with self-assembled molecular catalysts for HER. In one such report,^[15] a known H₂ producing catalyst Co(dmgH)₂pyCl (dmgH = dimethylglyoximate, py = pyridine) was mixed with PSI under dark to produce a hybrid system containing 2-4 self-assembled Co catalyst per PSI (Fig. 4A). When irradiated with light in the presence of cytochrome ce (cyt c₆) as a mediator of electron transfer (ET) and ascorbate as the sacrificial electron donor, visual appearance of bubbles were observed from the hybrid Co⊂PSI system due to H₂ production. A turn over frequency (TOF) of 170 min⁻¹ was obtained under near-neutral pH conditions. The reported TOF for Co⊂PSI system was approximately half of a related PtNP⊂PSI hybrid.^[16] The authors hypothesized that the Co complex docked into a hydrophobic patch near the terminal electron acceptor, the [Fe₄S₄] cluster F_B, assuming that the two electrons were delivered to the Co catalyst directly from the reduced F_B⁻ during catalysis (Fig. 4B). It was necessary to add cyt c₆ as a mediator to reduce the photo-oxidized P₇₀₀⁺ and ascorbate as a sacrificial electron donor to sustain H₂ production. After 1.5h, a significant loss of Co (>90%) was observed from Co⊂PSI that could account for the short duration of the activity. The long-lived charge separated state of ~60 ms for P₇₀₀⁺F_B⁻ and a redox potential of −580 mV (vs NHE) for the F_B cluster was expected to provide the required thermodynamic push for H₂ production by Co⊂PSI. However, a 160-300 mV more negative Co(II)/Co(I) redox potential of reported Co(dmgH)₂pyCl complex compared to F_B⁻ argues against the possibility that reduced F_B⁻ is the actual electron donor to the Co complex in Co⊂PSI Several possibilities were put forward to account for this apparent anomaly. For example, one possibility is that the redox potential of the Co complex shifts to anodic potentials when assembled with PSI. A second possibility is that the Co catalyst is accepting electrons from other cofactors with more cathodic redox potentials, such as A₁ or F_x instead of F_B⁻.

Figure 4.

A) Overview of artificial photosynthetic system involving hybrid assemblies of PSI, electron donor cyt c, and molecular metal catalysts. B) Schematic of electron flow upon light-irradiation via the artificial photosynthetic system that leads to H₂ production. C) Zoomed view of the proposed docking site of a molecular Ni catalyst into the hydrophobic groove of flavodoxin. Panels A and B are adapted with permission from J. Am. Chem. Soc. 2013, 135, 13246-13249. Copyright (2013) American Chemical Society.

Subsequently, the DuBois’ Ni-diposphine catalyst^[17] [Ni(P₂^PhN₂^Ph)₂]-(BF₄)₂ (P₂N₂ is 1-aza-3,6-diphosphacycloheptane) was self-assembled with PSI. The Ni⊂PSI hybrid biocatalyst^[18] produced H₂ under similar experimental conditions to the Co⊂PSI system with a TOF of ~44 min⁻¹. Achieving function from the Ni⊂PSI system under neutral pH conditions was a significantly greener pathway of H₂ production compared to the parent Ni catalyst, which required highly acidic conditions for activity. In order to controllably deliver the Ni catalyst in close proximity to the F_B site of PSI, the inherent protein-protein interactions between PSI and its natural electron acceptor flavodoxin was exploited. Since flavodoxin (Fld) docks to PSI and accepts electrons via the flavin mononucleotide (FMN), the authors reconstituted apo Fld with the Ni catalyst. The resultant Fld-Ni metalloprotein was then mixed at 30-fold excess with PSI to allow the docking to occur. Under light illumination, this new Fld-Ni⊂PSI hybrid catalyst was active for H₂O production with ~2-fold higher TOF compared to the Ni⊂PSI system. While the Ni⊂PSI hybrid produced ~1870 mol H₂/mol PSI in 3h, the Fld-Ni⊂PSI complex showed a turnover number of ~2825 over 4h with an initial TOF of ~75 min⁻¹. In comparison, the parent Ni catalyst produced a comparable amount of H₂ but over a longer period of ~150h at pH ~2.2 using Eosin Y and [Ru(bpy)₃]²⁺ as photo sensitizers in 50:50 water: acetonitrile mixture.^[19] These studies demonstrated a successful implementation of the inherent ET efficiency of PSI in creating hybrid biocatalysts that can generate H₂ under environmentally benign reaction conditions. The bulky aromatic nature of the Ni catalyst was proposed to be a significant contributor to enable catalyst incorporation into Fld (Fig. 4C) and subsequent formation of a stable Fld-Ni⊂PSI complex docked into the hydrophobic pocket of PSI, leading to a higher TOF. A Ru-Fld-Ni hybrid was also prepared^[20] by covalent attachment of a Ru photosensitizer [Ru(4-CH₂Br-4’-bpy)(bpy)₂]·2PF₆ to ([Ni-(P₂^PhN₂^Ph)₂](BF₄)₂)-reconstituted apo Fld. This system had a TOF of ~6.8 min⁻¹. In this system the presence of a charge separated Ni(I) intermediate was observed by EPR spectroscopy.

To simplify the catalyst design, a next generation biocatalyst was developed devoid of the multiple cofactors of PSI,^[21] where, a Ru photosensitizer and a Co catalyst, Co(dmgBF₂)₂.2H₂O were covalently attached to two opposite ends of ferredoxin (Fd). The hybrid catalyst Ru-Fd-Co thus prepared (Fig. 5), produced H₂ with a TOF of − 0.8 min⁻¹. reaching a TON of 210 over 6h. Control experiments showed that the presence of the [2Fe-2S] cluster was essential for activity, since no H₂ production was observed either in the absence of Fd or with apo Fd. From EPR and excited state lifetime studies an oxidative quenching mechanism for the Ru photosensitizer was proposed, where a photoexcited [Ru(bpy)₃]²⁺* transfers an electron either directly to the Co(II) catalyst or mediated via Fd to form the oxidized [Ru(bpy)₃]³⁺. A favorable pathway of ET from Ru to the Co catalyst via the [2Fe-2S] cluster was deduced to be the primary mechanism of ET in this catalyst where the Ru(III)-Fd-Co(I) species was produced. In contrast, when apo Fld lacking the internal FMN cofactor was used as a scaffold instead of Fd, this new hybrid Ru-Fld-Co catalyst produced H₂ with a TON of ~85 (TOF 0.5 min⁻¹) via the generation of Ru(I)-Fld-Co(I) following a reductive quenching mechanism^[22] These studies highlight the importance of the protein matrix and the role of internal ET sites in distinguishing the direct vs mediated ET pathways that stabilize the charge separated states critical for H₂ production. Recently, the internal Z-scheme of ET in thylakoid membranes has been employed to obtain photocatalytic H₂ production by in situ self-assembly of PtNPs, Co and Ni catalysts with the membranes.^[23] In these assemblies, photoexcited electrons generated from PSII coupled with H₂O oxidation to O₂ are transferred to the H₂ evolving catalyst bound to the stromal side of PSI via internal redox shuttles of the thylakoid membrane. When ET from PSII to PSI was blocked by an inhibitor 3-(3,4-dichlorophenyl)-1,1-dimethylurea (DCMU), the thylakoid/PtNP preparations produced H₂ with a TOF of ~35 min⁻¹ in the presence of ascorbate as the sacrificial electron donor. However, when DCMU was removed to open up ET from PSII to PSI, the thylakoid/PtNP composites produced H₂ with a TOF of ~0.8 min⁻¹ even in the absence of ascorbate. Control preparations of thylakoid/PtNP containing DCMU but lacking ascorbate did not produce any H₂. These results confirmed that the internal Z scheme pathway of natural PS can be employed to sustain photocatalytic H₂ production removing the need for external sacrificial electron donors.

Figure 5.

Structures of molecular Co catalysts (A, B) and Ru photosensitizer (C). Schematic showing photocatalytic H₂ generation from this hybrid biocatalyst facilitating ET from the Ru to the Co center via the Fe-S cluster. Reproduced with permission from Chem. Sci., 2016, 7, 7068-7078 Published by The Royal Society of Chemistry.

II. Peptide-appended synthetic complexes.

Jones, Dutton, and co-workers have prepared synthetic peptide models inspired by the H-cluster of Fe-Fe hydrogenases.^[24] A Cys-containing peptide (SynHyd1) was designed and reacted with Fe₃(CO)₁₂ to assemble the metallopeptide maquette Fe₂-SynHyd1. Control experiments showed that in the absence of Cys, no attachment of the di-iron site to the peptide was observed. Circular dichroism (CD) showed a stable helical secondary structure, while the UV features confirmed the incorporation of the di-iron site. The Fe₂-SynHyd1 showed three CO bands at 2076,2040 and 2000 cm⁻¹ in FTIR. Using these combined results, the authors demonstrated the synthesis of a first generation hybrid (μ-SR)₂Fe₂(CO)₆ complex attached to a helical peptide scaffold via two Cys residues. A method for preparing peptidomimetic [NiFe] hydrogenase models with defined outer coordination sphere interactions was also developed.^[25] In this case, the AlaCysAspLeuProCysGly peptide was used for binding Ni in an N₂S₂ configuration consisting of N-terminus, an amide nitrogen, and two Cys thiolates that bridge the Ni center with an external organometallic fragment. Various heterometallic fragments (Fe₂(CO)₆, M(CO)₄(piperidine) (M = Mo, W), and Ru(η⁶-arene)²⁺) were successfully incorporated into the Ni-bound peptide. The Fe, Mo and W fragments formed trinuclear peptide complexes whereas the Ru-based fragment formed a binuclear complex. The formation of trinuclear complexes was attributed to the steric and thermodynamic constraints of the peptide. Results from various spectroscopic studies showed that the N₂S₂ coordination environment of the Ni was maintained in all the cases and the Cys thiolates are the reactive sites forming thiol-bridged heterobimetallic complexes with the peptide. Phosphines have been widely employed in models as a surrogate for the CN⁻ ligand of hydrogenases. Using this as a base for their work, Jones and co-workers designed a phosphine-containing peptide to create [FeFe] hydrogenase mimics (Fig. 6).^[26] In the first approach, a Lys-containing peptide (TrpAlaSerLysLeuProSerGly) derived from the N-terminal sequence of nickel superoxide dismutase was modified on-resin to yield a peptide containing a phosphine derivative of Lys. However, this approach faced challenges in terms of low product yield due to the oxygen sensitivity of phosphine. The peptide was reacted with (μ-pdt)Fe₂(CO)₆ [pdt = propane-1,3-dithiol] leading to the formation of 1 (Fig. 6). In another strategy, three phosphine-containing unnatural amino acids boc-Epa-OH, boc-Ipa-OH, and boc-Ppa-OH were used to synthesize peptides for direct incorporation of phosphines. This helped in shortening the side chain and minimizing the number of retainers that facilitated the formation of a more compact metal binding site. The tripeptide ValXpaLeu (Xpa= unnatural amino acid) was reacted to with either (μ-pdt)Fe₂(CO)₆ or (μ-bdt)-Fe₂(CO)₆ [bdt = benzene-1,2-dithiol] giving rise to 4 distinct complexes: [(Val-Epa-Leu)-(μ-pdt)Fe₂(CO)₅] (2), [(Val-Ipa-Leu)-(μ-pdt)Fe₂(CO)₅] (3), [(Val-Ppa-Leu)-(μ-pdt)Fe₂(CO)₅] (4) and [(Val- Ppa-Leu)-(μ-bdt)Fe₂(CO)₅] (5) (Fig. 6). The complexes 2, 3 and 4 showed two irreversible voltammetric waves corresponding to the Fe^IFe^I/Fe^IFe⁰ and Fe^IFe^I/Fe^IFe^II couples. The irreversibility was attributed to complex decomposition. Among the pdt series, complex 4 was the most easily reduced. The change from pdt to bdt showed reductive and oxidative peaks for complex 5, which was attributed to the donating abilities of phosphines and thiolates as suggested by their respective CO stretching frequencies. This demonstrated that a change from bridging thiolate from pdt to bdt changes the electronic properties of the catalyst. The complexes 2–5 also demonstrated their ability to electrocatalyze H⁺ reduction in acetic acid. Sequential addition of acetic acid to the pdt complexes showed increased reductive current for H₂ production. In these complexes, the catalytic wave was not observed at the Fe^IFe^I/Fe^IFe⁰ reduction potential but at more negative potentials. This observation indicated that an ECEC mechanism was likely occurring. The bdt analogue 5 showed a slightly different behavior in the presence of acetic acid where a completely irreversible reduction peak was observed with a small increase in peak current. However, catalysis is observed only at a more negative potential. This suggested that a similar ECEC mechanism is followed. Complex 5 undergoes reduction and protonation to generate 5H, which, upon further reduction produced 5H⁻ that accepts a H⁺ and leads to the reductive elimination of H₂. More electrochemical studies were done in acetonitrile/water mixtures, which showed a shift in reductive waves to a lesser reducing potential with increasing water content. However, the oxidative potential remained unaffected by water, suggesting that the reduction may be coupled to the interaction with solvent. In this study the authors demonstrated the synthesis of phosphine-containing peptides which can be used for one step incorporation of the di-iron site. They also showed the differences in the electronics of pdt vs bdt in tuning catalysis. The catalysts were active for H₂ production in the presence of acetic acid while the presence of water significantly affected the electrochemical properties of the complexes.

Figure 6.

Phosphine-containing peptide-Fe-Fe complexes as [Fe-Fe] hydrogenase mimics. Reproduced with permission from Dalton Trans., 2015, 44, 14865–14876 Published by The Royal Society of Chemistry.

A (μ-S–(CH₂)₃–S)[Fe₂(CO)₆] Fe-carbonyl cluster was prepared^[27] by Ghirlanda and coworkers employing a helical peptide containing a dithiol unnatural amino acid (Dt) anchor (Fig. 7). The presence of a nearby Lys at i, i+3 position with respect to the Dt was proposed to stabilize the cluster. The peptide-[Fe₂CO₆] complex was active for photo-induced H₂ production at pH 4.5 with a TON of ~84 over 2.3h (TOF ~ 0.6 min⁻¹). CV experiments showed the presence of a pH-dependent reductive peak at −1.1 V vs NHE. A more negative current was obtained at lower pH values, suggestive of a pH dependence of the Fe redox couples. Bren and coworkers employed the versatile metal binding peptide motif, ATCUN, and demonstrated that in the Co-bound form, a tripeptide is a HER electrocatalyst at near-neutral pH.^[28] The Co-GlyGlyHis tripeptide corresponding to the XxxXxxHis motif of ATCUN was prepared. From voltammograms of Co-GlyGlyHis as a function of pH (4.9-9.4) the maximum current was obtained at pH 6.5. Electrolysis under inert and aerobic conditions at pH 8 produced H₂ with a TON of 275 and a Faradic efficiency (FE) of 91%. A similar FE of 88% was obtained in aerobic conditions. This result demonstrated that Co-GlyGlyHis is not only an active HER catalyst in water but that it is also an oxygen-tolerant metallopeptide catalyst. Control experiments with Ni-GlyGlyHis, Cu-GlyGlyHis, apo GlyGlyHis, and CoCl₂ showed that none of these were catalytically active, confirming the true catalytic nature of Co-GlyGlyHis. This work presented a simple and easily assembled cobalt tripeptide as an oxygen tolerant electrocatalyst for HER from neutral pH.

Figure 7.

A hybrid H₂ generating catalyst employing a synthetic Fe-Fe comples appended to the dithiol amino acid on a peptide scaffold. Reproduced with permission from Chem. Commun., 2012, 48, 9816–9818 Published by The Royal Society of Chemistry.

The Hayashi group has developed H-cluster models using the assemblies of heme proteins and synthetic metal complexes. Inspired by the -CysXxxXxxCys- motif of covalent heme ligation of cyt c, a di-iron carbonyl complex Fe₂(CO)₉ was attached to apo cyt c via Cys 14 and Cys 17 (Fig. 8A).^[29] Apo cyt c was reacted with the Fe₂(CO)₉ to form the H-apo cyt c, a mimic of the H cluster bearing the (μ-S-Cys)₂Fe₂(CO)₆ core. The photocatalytic activity assays performed using [Ru(bpy)₃]²⁺/ASC system led to the production of H₂ with a TON of 80 per H-apo cyt c and a TOF of 2.1 min⁻¹ at pH 4.7. The authors also made an H-apo minicyt c peptide (TyrLysCysAlaGlnCysHis) bearing the -CysXxxXxxCys- motif instead of the full-length protein, which, when reacted with Fe₂(CO)₉, showed a lower TOF of ~0.4 min⁻¹ for H₂ production. This result indicates the importance of protein scaffold surrounding the di-iron center that presumably stabilized the metal cluster leading to a higher catalytic activity. Furthermore, lowering the pH decreased the activity by 16% due to ascorbate inactivation, while increasing the pH showed a loss of activity by 6.8% due to a decrease in proton concentration. Subsequently,^[30] in order to enhance the activity, the authors used -CysXxxXxxCysHis- motif of cyt c to anchor a Ru complex to a model peptide via coordination with His (Fig. 8B). In this work, the Fe₂(CO)₉ complex was reacted with the peptide (Pep 18-TyrIleGlyLysAlaCysGlyAsnCysHisGluAsnPheArgAspLysGluGly) to form the binuclear Fe₂(CO)₆ complex of the peptide. After linking the di-iron carbonyl cluster, the peptide was further reacted with [Ru(bpy)(tpy)(H₂O)](PF₆)2 to be attached to the His residue. Under photochemical conditions, this hybrid system produced H₂ with a TON of 9 and TOF of 0.2 min⁻¹. The peptide Pep-18(His10Ala) lacking the His to coordinate to the Ru produced no H₂ even in the presence of externally added imidazole complex [Ru(bpy)(tpy)(im)]²⁺. This result indicated the importance of direct attachment of the Ru photosensitizer close to the Fe-Fe center which facilitated internal electron transfer to the catalytic site for H₂ production. While internal electron transfer to the Fe-Fe site is important, the peptide TyrLysCysAlaGlnCysHis showed a higher TOF when employing externally added Ru photo sensitizer that lacked a direct attachment in proximity to the active site.

Figure 8.

Incorporation of a synthetic Fe-Fe complex into the -CysXxxXxxCys- motif of peptides inspired by heme ligation to cyt c (A). Schematic of internal ET from the peptide-bound Ru photosensitizer to the Fe-Fe catalyst leading to photocatalytic H₂ generation (B).

The effect of outer coordination sphere proton shuttles on the rates and overpotentials of both H₂ oxidation and H₂ production activity of DuBois’ Ni catalyst [Ni(P₂^RN₂^R’)₂]²⁺ has been systematically studied.^{[13d, 14, 31]} The parent Ni catalyst has served as an excellent biomimetic model system for the bridgehead amine group of the H-clustcr.^{[12f, 17, 32]} Detailed mechanistic studies have revealed how the protonation state of the bridgehead amine and the isomeric forms of the catalyst dictate the nature of intermediates formed in the catalytic cycle influencing turnovers. Combined, these studies have led to an understanding of the critical parameters in molecular catalyst design for reversible H⁺ reduction and H₂ oxidation. Shaw and coworkers introduced ionizable functional groups, single amino acids substitutions that can deliver protons during catalysis (Fig. 9), and small peptide fragments at the outer coordination sphere to mimic the protein environment of native enzymes. Upon substitution of the benzyl (Bn) groups from the parent H₂ oxidizing catalyst^[17] [Ni(P^Cy₂N^Bn₂)₂]²⁺ with (3-pyridazyl)methyl, the resultant [Ni(P^Cy₂N^{3-pyridazyl)methyl}₂)₂]²⁺ catalyst improved the overpotential for the reaction by 300 mV by facilitating H⁺ movement from the active site to the N-atoms of pyridazyl groups via the pendant amines of the ligand scaffold.^[31d] The mechanism of H₂ oxidation in the parent catalyst is thought to proceed by oxidation of H₂ by the Ni(II) form of the catalyst, which protonates the pendant amines that make H-bonds to the resultant Ni(0) form. The strength of the NH---Ni H-bond is the highest in the endo-endo conformation of the catalyst, which makes the subsequent oxidation process more difficult, thus increasing the overpotential. The addition of 3-pyridazyl groups at the outer sphere promoted fast proton movement that weakened the strength of the NH---Ni H-bond in the endo-endo isomer and promoted its conversion to more readily oxidizable isomers, such as the endo-exo form, thus lowering the overpotential for H₂ oxidation. Amino acids were next introduced at the outer sphere to make a closer connection to the enzymes. Initially, the Ala and Phe esters were introduced on the para position of Bn groups of [Ni(P^Cy₂N^Bn₂)₂]²⁺, which was active for H₂ oxidation,^[31i] although the rate was slower than the parent catalyst.^[17] The lower rate was attributed to a higher pK_a of the pendant amines due to an electron withdrawing effect of the esters and steric hindrance to the active site caused by the benzyl groups. To address the limitations with tins system, Gly was introduced in the ligand scaffold with the -CH₂-COOH group of Gly hanging out from the pendant amines.^[14a] The presence of -COOH groups of Gly at the outer coordination sphere facilitated fast H⁺ movement resulting in a 150 mV overpotential for H₂ oxidation at pH 0.1 reaching the fastest TOF of 33 s⁻¹ (1980 min⁻¹) at pH 0.7. The Gly derivative was also active for H₂ production with a rate of ~2400 s⁻¹ (1.4x10⁵ min⁻¹) at pH 0.5 with a corresponding overpotential of 750 mV. When covalently attached to an electrode, the [Ni(P^Cy₂N^Gly₂)₂]²⁺ derivative demonstrated better catalytic performance compared to native [NiFe] hydrogenase at low pH, while at pH 7 the enzyme was a better catalyst.^[31f] Along the same lines, an Arg derivative was prepared instead of -COOH, to add a third proton relay at the outer coordination sphere.^[14b] The Arg derivative [Ni(P^Cy₂N^Arg₂)₂]⁸⁺ oxidized H₂ with 6-fold higher rate at 1 atm H₂. compared to the Gly derivative and operated at 180 mV. At higher H₂ pressures and lower pH (<3), while the rate of oxidation was very high- 144 000 s⁻¹ (8.6x10⁶ min⁻¹). the overpotential also increased to 480 mV. The higher rate in this Arg derivative was attributed to a faster H₂ addition to the catalyst, as opposed to an enhanced proton transfer from the active site to the outer sphere Arg via the pendant amines. Intramolecular interactions between the Arg side chains that lead to an ideal distance of the outer sphere N atoms from the active site was also thought to be crucial to achieve a faster rate. At pH >3 the deprotonated -COO₋ groups of Arg disrupted the proton transfer pathway leading to higher overpotentials. The observations that even though faster rates of H₂ oxidation were observed at lower pH and high H₂ pressures, the associated high overpotentials were a result of either altering the proton transfer or electron transfer process from equilibrium conditions. The beneficial role of -COOH groups of Arg was further proved by preparing a methyl ester derivative of Arg, which demonstrated slower TOFs in both H₂ oxidation and H₂ production compared to the Arg derivative.^[31c] The fact that the outer sphere proton relays enhance the rates for reversible H⁺/H₂ interconversion is evidenced by the engineered functionalities on the ligand scaffold of these classes of catalysts. The high H₂ oxidizing activity of the Arg derivative was exploited as an active anodic catalyst in fuel cells.^[31h] For this purpose, the catalyst was assembled on carbon nanotubes (CNTs) via electrostatic interactions of the Arg residues and the -COO⁻ groups on modified CNTs. The catalyst/CNT composites were then employed as anodic H₂ oxidation catalyst in constructing enzymatic fuel cell using bilirubin oxidase the cathode or in proton exchange membrane fuel cell (PEMFC) using Pt/C as the cathode. The assembled fuel cells reached power densities of ~2 mW/cm² and 14 mW/cm² for the enzymatic and PEMFC, respectively. This study opens up the possibility of direct applicability of biomimetic catalysts in fuel cells to generate green energy from renewable resources.

Figure 9.

Schematic representation of reversible Ni-P catalysts with strategically positioned inner and outer coordination spheres where the introduction of functional groups capable of H⁺ delivery enhances turnovers.

To more closely mimic the outer sphere of enzymes, peptide fragments were introduced to the parent catalyst. Initially, single and dual peptide fragments were appended.^[31a] These substitutions led to an increase in the TOF for H₂ production up to 1000 s⁻¹ (6x10⁴ min⁻¹) at an overpotential of ~300 mV. When H₂O was used as the H⁺ source a rate enhancement of ~ 2-7 times was achieved. A series of dipeptides were then prepared by attaching 3-(4-aminophenyl)propionic acid (NNA) to the parent ligand framework followed by coupling amino acids and esters to NNA.^{|31b, 31j]} The identity of the amino acid was chosen to vary the side chain polarity, aromaticity, and size of the side chain to probe how these changes affect the H₂ production activity. The Glu, Ala, Ser, Phe, Tyr, Lys and Asp derivatives thus prepared, all were active electrocatalysts. Depending on the side chain polarity and pK_a, rate enhancements of an order of magnitude were found among these dipeptide derivatives (Fig. 10) with TOFs ~1000 s⁻¹ (6x10⁴ min⁻¹). The rates were not dependent upon the size and polarity of side chains; however, amide bonds increased the rates. Esterification of the -COOH and -NH₂ groups of the dipeptides led to a decrease in the rate by upto 70%, demonstrating the importance of H⁺-donating functionality at the outer sphere in facilitating catalysis. Subsequently, a structured peptide-based outer sphere was introduced, rather than flexible dipeptides, by attaching a β hairpin peptide on two sides of the parent complex that contained only the NNA.^[31e] The peptide-Ni catalyst was also an active catalyst with rates that are ~2-fold compared to the parent Ni complex. Overall, these approaches aimed towards mimicking the protein environment around the active site of hydrogenases represent preparation of peptide-inorganic hybrids as active HER catalysts where the outer coordination sphere influences the turnovers by not only providing H⁺ delivery to the active site but also by securing the metal site and providing suitable local dielectric environments amenable to catalysis.

Figure 10.

TOFs for H₂ production by Ni catalysts with NNA-dipeptides on the outer sphere. Reprinted with permission from Acc. Chem. Res. 2014, 47, 2621–2630. Copyright (2014) American Chemical Society.

III. Substitution of native cofactors with non-native macrocyclic cofactors.

A Co-protoporphyrin IX (CoP)-substituted myoglobin (Mb) protein^[33] has been demonstrated to be an active H₂ generation catalyst both electrochemically and photochemically by the Ghirlanda group. Optimal catalytic current observed from cyclic voltammetry was at pH 7.5, while at lower pH values the activity decreased due to protonation of the coordinating proximal His 93, which led to CoP loss from the protein. This possibility was verified by rinse tests, where catalytic current was sustained at lower pH due to electrodeposited free CoP dissociated from the protein, while at higher pH no catalytic current was observed after rinse test. No significant loss of current was observed under aerobic conditions, although the possibility of O₂ reduction contributing to the current has not be eliminated. The CoP-Mb was photocatalytically active, producing H₂ with a TON of 518 at pH 7 (TOF ~0.5 min⁻¹). This TON is ~3-fold higher compared to what is observed for free CoP, demonstrating a beneficial role of the protein scaffold in modulating the activity. Similar to electrochemistry experiments, pH dependence was also observed for photo-induced catalysis with activity decreasing when the pH deviated from neutral to more acidic or more basic conditions. Proximal and distal His mutants were prepared in the form of His64Ala and His97Ala, and the corresponding double mutant His64AlaHis97Ala, to probe how the protonation states of these residues modulates activity. A higher activity was observed for His64Ala and the double mutant compared to that of the WT protein. The His97Ala mutant showed activity that is similar to free CoP. The higher activity for the distal His mutant His64Ala was rationalized by the fact that more H⁺ are available near the Co site to produce the Co(II)-hydride intermediate during catalysis, since there is no demand to protonate the His64 in this mutant. However, given the pK_a of this His is ~5, at the experimental condition of pH 6.5 this His is expected to be deprotonated, making minimal competition of H⁺ for Co(0). The lower activity for His97Ala was attributed to increased CoP mobility since this mutation obliterated the H-bond with one of the propionate groups. A CoP-substituted ET protein cyt b₅₆₂ was also employed as a H⁺ reduction catalyst^[34] producing H₂ with TON of 120 over 8h (TOF ~0.26 min⁻¹) under photocatalytic conditions. Variants prepared by replacing the axial Met, Met7Ala, Met7Asp, and Met7Glu were shown to produce H₂ with higher TONs of 310, 270, and 195, respectively, at neutral pH. The higher activity in these variants was attributed to a combination of lower sterics and more structural flexibility of CoP that led to a rapid production of the rate-limiting Co(II)-hydride intermediate. With a gradual increase in the O₂ concentration, the TON for photocatalytic H₂ production decreased. In air, the TON decreased by ~3.75-fold compared to strictly anaerobic conditions, possibly due to competitions with O₂ and CO₂ reduction. Further expanding the choice of scaffolds within heme proteins, Hayashi and coworkers employed a rigid β-barrel heme protein nitrobindin to site-specifically incorporate a synthetic di-iron complex (μ-S)₂Fe₂(CO)₆ into the apo protein cavity (Fig. 11).^[35] The complex was introduced into the protein by covalent attaclunent to Cys96 via a maleimide moiety appended to the di-iron complex. The hybrid biosynthetic catalyst was active for H₂ production under photocatalytic conditions with a TON of 130 over 6h and an initial TOF of 2.3 min⁻¹. The naked di-iron catalyst under similar conditions but in the absence of the protein matrix produced H₂ with a higher TON and TOF, which was attributed to the fact that upon protein encapsulation access of Ru(bpy)₃²⁺ to the Fe-Fe site is hindered.

Figure 11.

Incorporation of a non-native Fe-Fe complex into the cavity of apo heme protein NB as an artificial H₂ evolving catalyst.

Bren and co-workers have employed engineered hemoproteins as catalysts for HER (Fig. 12).^[13e] In one example,^[36] a cobalt-containing microperoxidase-11 (MP11) was reported as an H₂ evolving catalyst. MP11 was produced by pepsin digestion of cyt c that produced a covalently attached heme to an 11-residue peptide fragment. The peptide was acetylated to protect the free amines followed by iron extraction using reductive demetallation and subsequent introduction of Co into the porphyrin ring by heating the peptide with Co(II)-acetate, which produced the bis-acetylated Co(III) microperoxidase-11 (CoMP11-Ac). Under neutral conditions, the H⁺ reduction current increased linearly with an increase in catalyst concentration. After 10 min of CPE using 1 μM catalyst under anaerobic conditions at − 1.5V vs Ag/AgCl at pH 7, 20 μmol of H₂ was produced corresponding to a TOF of ~6.7 and a Faradic efficiency (FE) of ~ 95%. However, the catalyst operated at a high overpotential of 852 mV and after 10 min of CPE porphyrin degradation was observed. Control experiments with the Zn derivative and demetallated Ac-MP11 confirmed that Co porphyrin was the catalytically active species. Oxygen tolerance is an important feature of H₂ evolution catalysts. CoMP11-Ac, upon electrolysis in the presence of O₂ showed a FE similar to that in presence of nitrogen. Upon prolonged electrolysis (~4 h) the catalyst showed a TON of ~ 2.5×10⁴ and 1.9×10⁴ under anaerobic and aerobic conditions, respectively. Combined, these results demonstrated that CoMP11-Ac is an efficient oxygen tolerant biomolecular catalyst for H₂ evolution at neutral pH. The degradation of CoMP11-Ac was attributed to the high flexibility of the scaffold. To alleviate this problem, a thermostable cytochrome c₅₅₂ was chosen as a scaffold. The new electrocatalyst, Co-substituted M61A c₅₅₂ (Ht-CoM61A) sustained activity for 24h and demonstrated a high TON of >10⁵ for H₂ production after 6h of electrolysis with a high FE of ~90% at an overpotential of ~830 mV.^[13e] Subsequently, to develop robust and stable electrocatalysts, a cobalt-mimochrome VI*a (CoMC6*a) synthetic miniprotein was designed as an electrocatalyst for HER.^[37] The CoMC6*a consists of a covalently bound cobalt deuteroporphyrin with a distal 10-mer peptide and a proximal tetradecapeptide. CoMC6*a is an active catalyst under both inert and aerobic conditions with FE of 96% and 97%, respectively, indicating the robustness of the metallopeptide. The activity results of CoMC6*a were compared with the CoP bound to a single proximal peptide chain (CoMP11-Ac) to discern the effects of peptide and its secondary structure in CoMC6*a that can influence the activity. The CoMC6*a showed enhanced α-helical secondary structure compared to CoMP11-Ac, which was primarily present in a random coil conformation. Electrocatalytic activity of both the systems was also investigated in the presence of TFE, which showed a ~90 mV decrease in overpotential for CoMC6*a compared to CoMP11-Ac. Tins difference in activity was attributed to the TFE-induced folding of CoMC6*a. CoMC6*a was also shown to be more stable under electrolysis conditions and produced H₂ with a ~9-fold high TON of 2.3x10⁵ compared to CoMP11-Ac.

Figure 12.

Models of cobalt-substituted M61A Cyt c₅₅₂ (A) prepared from PDB ID 1AYG and CoMC6* (B). Chemical structure of CoMP11-Ac (C). Reproduced with permission from Chem. Sci., 2018, 9, 8582–8589 Published by The Royal Society of Chemistry.

Biotin-streptavidin technology has been employed to construct artificial hydrogenases featuring Co and Fe-Fe cofactors.^[38] Biotinylated macrocyclic Co complexes employing pentapyridyl^[38a] and 1,4-di(picolyl)-1,4,7-triazacyclononane (^HPy₂tacn)^[38b] ligands with varying linker lengths have been synthesized followed by investigating the H₂ production activity of protein-encapsulated cofactors (Fig. 13). Upon incorporation of the biotinylated pentapyridyl Co complex into streptavidin (Sav), a significant decrease in the activity was observed for photocatalytic H₂ production. Introduction of both acidic and basic residues at the second coordination sphere of Sav was able to restore some of the initial activity observed with the naked catalyst. For the ^HPy₂tacn ligand, incorporation of these biotinylated Co complexes into Sav resulted in an anodic shift of the pH-dependent catalytic waves ranging from 20-100 mV (overpotential ~ 700 mV) compared to the naked Co catalysts, suggesting the beneficial effect of the protein matrix in tuning the redox potentials of Co. While the peak potentials at half of maximum current showed approximately 2-electron 1-proton process vs pH for 1^Bio, 2^Bio and 1^Bio⊂Sav, the 2^Bio⊂Sav was intriguingly almost unchanged with pH (~9 mV/pH). CPE also suggested some influence of the protein environment on the electrochemical behavior of these hybrid bioinorganic catalysts. Both the Sav-incorporated catalysts showed higher TON (1^Bio = 7; 1^Bio⊂Sav = 51; 2^Bio = 46; 2^Bio⊂Sav = 166) and TOF (1^Bio ~ 0.03 min⁻¹; 1^Bio⊂Sav ~ 0.17; 2^Bio −0.26; 2^Bio⊂Sav ~ 1.46) for photo-induced H₂ production compared to the naked catalysts. A bell-shape pH vs TON curve was observed in these cases with optimum activity being observed in pH values ~10.5-11.5. The origin of higher activity in 2^Bio⊂Sav was investigated by MD simulations. The 2^Bio⊂Sav hybrid having a longer and more flexible linker between the biotin and ligand scaffolds, sampled two different conformations with the Co center ending up in close proximity to polar residues and internal H₂O molecules of Sav, which presumably influenced catalysis by making H-bonding interactions. A biotinylated Fe-Fe complex featuring CO and bridging thiols was incorporated into Sav where encapsulation of the Fe-Fe complex into the protein enhanced its activity.^[38c] Under photocatalytic conditions, while naked biotinylated Fe-Fe complex produced H₂ with a TON of ~6.4 after 3h, the Fe-Fe⊂Sav had a TON of ~48 over 11h. The parent Fe-Fe complex was active for 100 min while the protein-encapsulated complex produced H₂ for 9h due to the longer lifetime of the catalyst upon its incorporation into the protein scaffold.

Figure 13.

Schematic showing streptavidin-encapsulated Co complexes as artificial H₂ evolving biocatalysts.

IV. Metal substitution from an ET protein.

Moura and co-workers isolated rubredoxins from three different strains of sulfate-reducing bacteria and probed their H₂ evolution and H/D exchange rates upon substituting the native metal Fe with Ni.^[39] The D. desulfuricans (ATCC 27774) NiRd was the most active among the three, producing H₂ at a rate of ~10 nmol/min/mg of protein showing an H/D exchange rate of 29 mnol/min/mg of protein. From the ratios of H₂ to H₂₊H(D) of 0.45-0.6 found in the NiRd proteins, it was deduced that the activity of NiRd had a closer resemblance to the NiFeSe hydrogenases than the NiFe hydrogenases. The NiRd protein was also inactivated by CO requiring only ~0.5 μM CO for 50% inhibition of hydrogenase activity compared to native enzymes that require ~20-30 μM CO for the same level of inhibition. This report is one of the earliest examples of an artificial biomolecular hydrogenase. Shafaat and co-workers have further developed the NiRd as a model system to probe into the mechanism and fine-tuning the activity.^[40] The activity of NiRd was probed by protein-film electrochemistry (PFE) and photocatalytic conditions. In contrast to FeRd which was non-catalytic, the NiRd displayed catalytic currents due to H⁺ reduction in a pH-dependent manner. With a decrease in pH from 5 to 3 the onset potentials shifted cathodically with a concomitant increase in catalytic currents. The initial reduction potentials from pH 3 to 5 showed a linear relationship with a slope of ~67 mV/pH, indicating a proton-coupled electron transfer (PCET) mechanism. NiRd operated at an overpotential of ~540 mV. When electrolyzed for 2h at ~0.9V vs NHE, NiRd produced ~3 μmoles of H₂ with a FE of ~33%. Under photocatalytic conditions, NiRd produced H₂ with a TOF of ~0.5 min⁻¹ and a TON of ~32 after 1h, while at lower catalyst concentrations a TON of ~100 was obtained over 8 h. About 90% of the sample could be recovered, indicating the stability of NiRd under photocatalytic conditions. Solution-based assays in the dark with titanium citrate as the reducing agent showed a TON ~300 over a period of 8h with almost no catalyst degradation.

Subsequently, photocatalytic H₂ evolution was studied (Fig. 14) by covalent attachment of a free Cys (Cys31) near the metal binding site of NiRd using a ruthenium-based chromophore [Ru^II(2,2’-bipyridine)₂(5,6-epoxy-5,6-dihydro-[1,10]-phenanthroline)]²⁺.^[41] The hybrid RuNiRd protein produced ~110 mnol H₂ within 45 min corresponding to a TON of 3.5 and TOF of ~ 0.1 min⁻¹. The optical spectra obtained before and after photocatalysis indicated degradation of the ruthenium complex which was attributed to the observed loss of activity after 45 min. The influence of the chromophore’s location on the protein surface relative to active site was also studied by preparing three different variants Glu17Cys/Cys31Ala, His18Ala/Cys31Ala/Ser38Cys, and Cys31Ala/Ala45Cys, each of which contained a Cys for Ru attachment. Results from these variants indicated a strong distance dependence for activity. All three mutants exhibited a lower activity compared to the WT RuNiRd, which was attributed to a weaker coupling between the chromophore and the active site. This was also supported by the fact that the enhancement factor (calculated as the ratio of H₂ produced by Ru-attached protein to H₂ produced by the protein with an equivalent amount of [Ru(bpy)₃]²⁺ in solution) was ~8 for RuNiRd and only ~2 for the other variants. Steady state and time-resolved emission spectroscopy were used to look into the photophysical properties of these hybrids. A significant component of the transient emission traces exhibited a short lifetime in all the constructs, revealing a direct electron transfer between ruthenium and nickel center. These results also suggested that intramolecular ET might play a key role in the activity of covalently bound RuNiRd and the mutants.

Figure 14.

Schematic of internal ET from a covalently appended RuPS to the Ni site of Ni-Rd. Reproduced with permission from ChemSusChem 2017, 10, 4424 – 4429 Published by The Royal Society of Chemistry.

The mechanism of H₂ evolution in NiRd was probed by protein film electrochemistry (PFE), spectroscopy, and computational studies.^[42] The non-catalytic Fe^III/II redox couple was used as an internal standard with NiRd to obtain the total electroactive protein coverage and quantitative information from the PFE studies. In NiRd catalysis, pH and solvent isotope-based effects indicated that the intramolecular H⁺ transfer is the rate limiting step, which occurs via a thiol inversion to generate the Ni^III–hydride species. Temperature-dependent studies revealed an exponential change in activity and a negative entropy for the activation process indicating an ordered transition state and supported the intramolecular proton transfer step to be the rate limiting step. In the presence of O₂, while the peak corresponding to O₂ reduction was observed at −500 mV this didn’t seem to affect the activity of NiRd toward H⁺ reduction in the experimental timescale. Secondary sphere mutations with His in two different sites revealed interesting tunability of the NiRd system. The Val8His mutation exhibited 3-fold enhanced activity with slightly higher overpotential and kinetics similar to that of NiRd. The Val34His mutant on the other hand, showed similar activity like NiRd but different kinetics with respect to pH dependence. The authors also investigated the pH-induced structural changes by employing optical absorption and resonance Raman spectroscopy. These results indicated that NiRd mutants undergo structural changes at low pH. DFT calculations and MD simulations complemented the experimental results leading to a proposed model for the reaction process (Fig. 15). MD simulations on NiRd revealed that only two Cys sites located on the surface of the protein are more likely to interact with solvent molecules and changes at the active site do not affect the global structure of the protein significantly. MD studies on the mutants of NiRd indicated significant structural changes based on the different degree of protonation of His residues. MD simulations of the Val34His mutant suggested that the protonation of Cys9 could be occurring from water or buffer which could account for the pH-dependence on activity. Results from ValHis8 on the other hand, led to the hypothesis that His8 may help in the protonation of Cys35 leading to enhanced catalysis.

Figure 15.

Proposed mechanism of Ni-Rd catalyzed H₂ generation. Reprinted with permission from J. Am. Chem. Soc. 2018, 140, 10250–10262. Copyright (2018) American Chemical Society.

Based on studies from a series of secondary sphere NiRd mutants^[43] it was found that the mutations modulated the activity, H-bonding network and active site solvent accessibility. The mutant library was made by mutagenesis of three residues in the metal binding loops, Val8, Val34 and Val37. When the TOF vs overpotentials were compared for the various mutants introduced at these sites, it was observed that the site of mutation dictates the correlation between TOF and overpotential as opposed to the nature of the specific side chain in the mutant library (Fig. 16). For example, the Val8 mutants exhibit ~10-fold higher activity and overpotentials within ~20 mV of WT. The Val34 variants exhibited a similar activity like WT with ~25 mV decrease in overpotentials, while the Val37 mutants showed a 3-fold decrease in activity and ~62 mV increase in overpotential for the Val37Asn variant. Similar to NiRd, a majority of the mutants showed a pH dependence of the onset potentials between pH 3-5, indicating a PCET, while a few mutants showed a shallower dependence on pH. Some mutants showed a decrease in activity with pH, which could be due to changes in the secondary sphere. While the spectral features of NiRd remained unaffected in the pH range of 4-8, mutants with side chain pK_a of 4-8 showed changes in the spectral features, indicating sensitivity of the spectra to the side chain protonation state. MD simulations showed that the secondary sphere mutants impacted the H-bonding network and the solvent accessibility. Perturbations in the H-bond by shortening the length between the amide backbone and the active site thiolates coordinated to the Fe or Ni center led to increased potentials, because hydrogen bond provides a means to delocalize and stabilize negative charge in the reduced state. The highest increase in activity found for mutations at Val8 also suggested that solvent accessibility plays a significant role in catalysis. This result was consistent with MD simulations which indicated that a majority of the Val8 mutations opened up the active site, exposing the Cys35 residue to solvent, which in turn correlates with an increase in TOF. Out of all the mutants studied, Val8Asp and Val8Asn were particularly interesting due to ~10-fold and a ~2-fold increase in activity, respectively. While the MD simulations of these mutants suggested a similar solvent accessibility, the presence of an ionizable side chain in Val8Asp suggested the likely involvement of this residue in forming a H⁺ shuttle that promoted turnovers. These studies demonstrated the importance of protein environment in tuning the activity for HER.

Figure 16.

Correlation between TOF and overpotential of WT, Val8, Val34 and Val37 mutant library of NiRd showing dependence of catalytic parameters on the site of mutation. Reprinted with permission from ACS Catal. 2019, 9, 8928-8942. Copyright (2019) American Chemical Society.

V. Reengineering a Cu storage protein into a Ni binding protein.

We have reported the redesign of a naturally existing copper storage protein (Csp1) into a mononuclear nickel binding protein (NBP) featuring a NiS₄ active site as a HER catalyst.^[44] Csp1 is a four-helix bundle that binds 13 equivalents of Cu(I) to 13 Cys residues along the protein core.^[45] NBP was designed by replacing 9 out of the 13 Cys residues with non-coordinating hydrophobic residues. A majority of these were Ala since it is similar in size to Cys and it is likely to maintain a similar helix-helix distance at the core to facilitate metal binding to the target Cys residues. Bulky hydrophobic residues were introduced at exposed ends of the bundle to avoid fraying. We chose a central Cu bound to Cys 26, Cys 62 and Cys 87 as the potential location to introduce the desired Ni site where the 4th Cys was provided by the nearby Cys 113. To avoid competition, 6 surface His were mutated to noncoordinating but polar Asn and Gln, totaling 15 mutations into Csp1 to produce the NiS₄ active site (Fig. 17 A–E). Combined CD, UV-vis, X-ray absorption, NMR and computational studies showed that NBP is α-helical, thermostable, and bound Ni(II) in a distorted square planar geometry to four Cys with Ni-S bond distances of 2.2-2.4 Å (Fig. 17F), comparable to the Ni-S distances of [NiFe] hydrogenases (2.2-2.6 Å). NiNBP produced H₂ under photochemical conditions with a TON of −115 over 2h, corresponding to a TOF of ~1 min⁻¹ (Fig. 18A). The H₂ production followed a reductive quenching mechanism where the photoexcited *Ru(bpy)₃²⁺ is quenched by ASC to produce Ru(bpy)₃⁺, which reduces NiNBP into the catalytically active Ni(I) state for H₂ production. PFE studies on NiNBP demonstrated that NiNBP is also active for electrochemical H₊ reduction showing pH-dependent catalytic waves in pH in the range of 4-6 operating with an overpotential of ~560-510 mV (Fig. 18 B) depending on the pH. While these overpotentials are higher than the native [NiFe] hydrogenases these values are lower or similar to many biomolecular and molecular catalysts^{[12m, 13e, 27, 30, 40]} absence of Ni(II)/Ni(I) redox peak in the voltammogram suggested that the ET, H⁺ transfer and chemical reactions occur within the catalytic wave itself. The Pourbaix diagram suggested that the NiNBP followed a PCET mechanism. At pH 5 NiNBP produced H₂ with a TON of ~210 at a FE of 93 ± 5 % after 1h of electrolysis at −1 V (vs NHE) (Fig. 18 C–D). Rinse tests and suitable control experiments showed the homogeneous nature of NiNBP. Some changes in the UV-vis (Fig. 18D inset) of the samples after electrolysis were observed, which we attributed to slight changes in helix orientations (Fig. 18E) that can alter the electronic properties of the Ni site. Mass spectra of NiNBP after electrolysis did not show any fragmentation (Fig. 18F), suggesting that the protein is stabile under these conditions. In summary, the NiNBP system highlights an example where rational protein engineering can lead to metalloproteins with new functions that are not native to the starting protein constructs.

Figure 17.

Redesign strategy (A-B) of Csp1 into NBP. C) 1ns NAMD structure of NBP, D) the thiol pocket, E) overlay of NBP (green), apo Csp1 (purple: PDB 5FJD) and Cu-Csp1 (cyan: PDB 5FJE), F) is the QM/MM model of NiNBP. Reprinted with permission from ACS Catal. 2019, 9, 5847-5859. Copyright (2019) American Chemical Society.

Figure 18.

Summary of catalytic assays by NiNBP. A) Photocatalytic H₂ production by NiNBP, B) pH-dependent voltammograms (inset shows the Pourbaix diagram), C) charge passed during 1h electrolysis at −1V vs NHE at pH 5, D) the corresponding TON (inset shows the UV-vis), (E) CD and (F) MALDI of samples before (red) and after (green) electrolysis. In A, C, D: NiNBP (blue); apo NBP (red); NiNBP in the absence of photosensitizer (green); NiSO4 (black); NiNBP after rinse test (orange); buffer alone (gray). Reprinted with permission from ACS Catal. 2019, 9, 5847-5859. Copyright (2019) American Chemical Society.

CONCLUSIONS AND FUTURE PERSPECTIVES

Important progress has been made during the last two decades in constructing hybrid biosynthetic catalysts containing active sites surrounded by a polypeptide matrix. The choice of redox-active metal center, the nature of amino acid residues near the metal site, and the presence of ET sites within these hybrid constructs are shown to influence the activity and mechanisms. From a protein design perspective, however, significant territories need to be covered to advance the field of artificial hydrogenases. An “ideal” example of an artificial hydrogenase should not only mimic the active site metals and ligands of native enzymes, but also the non-covalent interactions around the active site. These include both amino acid and water-mediated H-bonding networks, and the internal ET sites. This design aspect is important to achieve reversible reactivity of native hydrogenases while still maintaining very low overpotentials. Computational protein design using versatile platforms such as Rosetta^[46] may be a viable handle to create artificial hydrogenases in a suitable foreign protein keeping in mind the critical design considerations such as the inner dielectric environment, positioning the active site in an optimum location within the protein pocket, construction of ET sites, and the H-bonding pathway originating from the active site to the protein’s surface. The computational approach should also provide insight into the requirement for the minimal but a large enough protein pocket to stabilize the active site and maintain the reversible functionality. In addition, the design of artificial hydrogenases that can function in an oxic environment at elevated temperatures is highly desired if these are to be employed in enzymatic fuel cells. Parametric protein design may be useful in this regard.^[47] Another important area that need to be explored is a systematic incorporation of versatile unnatural amino acids at the first coordination sphere and to evaluate the effect of such ligand alterations on the electronic and catalytic activity of such biosynthetic hydrogenases. The application of genetic engineering by directed evolution methods^[48] should be explored to create artificial hydrogenases that are optimized for stability and function. In addition, methods to suitably attach the designed proteins on the electrode surfaces in orientations that can lead to productive catalysis should be developed by careful mutagenesis and by decorating the protein surface with suitable redox-active moieties. Mechanistic insights gained from this design process will lead to a complete understanding of the native enzymes paving way to the creation of biosynthetic proteins with stability and functionality paralleling or even better than the native enzymes.

SYNOPSIS.

Green Biosynthetic Hydrogen:

Approaches to engineer artificial hydrogen evolution catalysts are discussed in this review. These bioinorganic hybrids are based on the redesign of various natural proteins and the assemblies of molecular catalysts with polypeptide scaffolds. Important design considerations in the first coordination sphere and remote interactions around the active site in tuning the physical catalytic, and mechanistic properties of these artificial biohybrids are also covered.