Second and Outer Coordination Sphere Effects in Nitrogenase, Hydrogenase, Formate Dehydrogenase, and CO Dehydrogenase

Abstract

Gases like H₂, N₂, CO₂, and CO are increasingly recognized as critical feedstock in “green” energy conversion and as sources of nitrogen and carbon for the agricultural and chemical sectors. However, the industrial transformation of N₂, CO₂, and CO and the production of H₂ require significant energy input, rendering routines like steam reformation or the Haber process economic and environmental dead ends. Nature, on the other hand, performs similar tasks efficiently at ambient temperature and pressure, exploiting gas-processing metalloenzymes (GPMs) that bind low-valent metal cofactors based on iron, nickel, molybdenum, tungsten, and sulfur. Such systems are studied to understand the biocatalytic principles of gas conversion including N₂ fixation by nitrogenase and H₂ production by hydrogenase as well as CO₂ and CO conversion by formate dehydrogenase, carbon monoxide dehydrogenase, and nitrogenase. In this review, we emphasize the importance of the cofactor/protein interface, discussing how second and outer coordination sphere effects determine, modulate, and optimize the catalytic activity of GPMs. These may comprise ionic interactions in the second coordination sphere that shape the electron density distribution across the cofactor, hydrogen bonding changes, and allosteric effects. In the outer coordination sphere, proton transfer and electron transfer are discussed, alongside the role of hydrophobic substrate channels and protein structural changes. Combining the information gained from structural biology, enzyme kinetics, and various spectroscopic techniques, we aim toward a comprehensive understanding of catalysis beyond the first coordination sphere.

Graphical Abstract

1. INTRODUCTION

Enzymes enhance the probability of a chemical reaction to proceed on biologically relevant time scales. Kinetic rates may differ by a factor of up to 10⁶, from the exclusively diffusionlimited enzymes carbonic anhydrase, catalase, or superoxide dismutase to average performers like RuBisCO that produces only three equivalents of product per second. To catalyze a given reaction, the enzyme must provide a specific binding pocket for reactants, for example, offering electrostatic interactions with polar and charged amino acid side chains, the protein backbone, and water molecules. Additionally, hydrophobic interactions play an important role (e.g., in the binding pockets of ATP synthase the separation of reactants and water drives ATP formation). Many enzymes rely on essential cofactors that are not formed by the protein fold. These may be organic “coenzymes” (covalently bound prosthetic groups and loosely bound “cosubstrates”) or inorganic clusters of metal ions. The later defines the class of metalloenzymes.^1–3

Metalloenzymes catalyze important energy conversion reactions throughout all kingdoms of life. The fast-performing enzymes carbonic anhydrase, catalase, and superoxide dismutase are metalloenzymes as well. Introducing two prominent examples, photosystem II (PSII, oxygenic photosynthesis) and cytochrome c oxidase (CcO, aerobic respiration) are high-valent, gas-processing metalloenzymes (GPMs). At the “oxygen evolving complex” of PSII, a high-potential heterometallic manganese cluster (Mn^III−Mn^V, E° ≈ 1.0 V vs SHE) catalyzes water splitting,^4–6 whereas CcO catalyzes the reverse reaction at a binuclear heme-copper center (Fe^II–Fe^IV, E° ≈ 0.8 V vs SHE).^7–9

In contrast, low-valent GPMs operate under reducing conditions, exploiting soft, electron-rich metal ions in the catalytic activation of N₂, H₂, CO₂, and CO (E° ≈ −0.4 V vs SHE).¹⁰ Such systems carry cofactors reminiscent of abiotic metal clusters like pyrite, mackinawite, and pentlandite. These minerals show residual activity catalyzing CO₂, N₂, and proton reduction,^11–13 but efficient turnover can only be achieved as “molecularly tuned derivates”¹⁴ within the active sites of GPMs like nitrogenase, hydrogenase, or CO dehydrogenase. From an evolutionary perspective, it is debated whether mineral cofactors were a prerequisite for catalysis or optimized the activity of protoenzymes. Russell and co-workers argue that metalloenzymes seem to be older than all-organic enzymes,¹⁵ suggesting atmospheric CO₂ and hydrothermal H₂ and CH₄ as earliest sources of electrons and carbon.¹⁶ Similarly, Martin and co-workers argue that life could have evolved from gases that reacted with the help of transition metals.¹⁷ Embedding metal ions or mineral cofactors, simple peptide “nests” may represent the onset of metalloenzyme evolution^18–20 that produced the GPMs discussed in this review including nitrogenase (Section 2), hydrogenase (Section 3), formate dehydrogenase, and carbon monoxide dehydrogenase (FDH and CODH, Section 4 and Section 5). All of these enzymes rely on iron, either as part of the active site cofactor (hydrogenase and nitrogenase) or in iron–sulfur clusters that primarily serve in long-range electron transfer.²¹ Further metal ions include nickel (CODH and [NiFe]-hydrogenase) and molybdenum (FDH and nitrogenase). In certain isoenzymes, molybdenum is replaced with tungsten (FDH), vanadium, or iron (nitrogenase). Such systems are studied to understand the biocatalytic principles of gas conversion that may inspire the design of biomimetic catalysts for the activation of N₂, CO₂, and CO as well as the production of H₂ as a climate-neutral fuel.

Second and outer coordination sphere effects critically impact the catalytic performance of GPMs. Although the influences are manifold and difficult to predict a priori, comparing the reaction stoichiometries (eqs 1–4) certain similarities become evident:

$$N_2+8H^{+}+8{\text{e}}^{-}\to 2N{H}_3+H_2(\text{nitrogenase})$$ (1)

$$H_2\rightleftharpoons 2H^{+}+2{\text{e}}^{-}(\text{hydrogenase})$$ (2)

$$C{O}_2+H^{+}+2{\text{e}}^{-}\rightleftharpoons HCO{O}^{-}(\text{FDH})$$ (3)

$$CO+H_{2}O\rightleftharpoons C{O}_2+2H^{+}+2{\text{e}}^{-}(\text{CODH})$$ (4)

Each enzyme catalyzes a redox reaction that involves one proton per electron, which hints at proton-coupled electron transfer (PCET)^22–24 as a common principle in GPMs. This demands (i) electron transfer to or from a redox site or redox partner in the outer coordination sphere of the metal ion coupled to (ii) proton transfer via polar amino acid residues or functional water molecules in the second coordination sphere (Figure 1). Proton transfer pathways and (iii) hydrophobic gas channels further modulate the exchange of reactants between active site and bulk solution, sometimes across several nanometers.²⁵ [FeFe]-hydrogenase, shown in Figure 1, illustrates how individual second coordination sphere effects may comprise (iv) hydrophobic interactions and (v) hydrogen-bonding contacts.²⁶ Moreover, the protonation state and polarity of the active site niche influences the (vi) charge distribution across the cofactor and redox centers in channeling distance.²⁷

Figure 1.

Definitions. The figure shows the active site of [FeFe]-hydrogenase (PDB ID 4XDC). For this enzyme, the first coordination sphere (light magenta) is defined by the two iron ions of the catalytic cofactor and its ligands. The second coordination sphere (white) comprises a [4Fe-4S] cluster as well as (iv) hydrophobic interactions and (v) hydrogen-bonding contacts with the protein fold. Outer coordination sphere effects include (i) electron transfer, (ii) proton transfer, (iii) hydrophobic gas channels, and (vi) electric field effects.

Long-distance effects play a crucial role in enzyme catalysis. Within each enzyme family, defined by a common active site structure, homologous enzymes may have very different catalytic properties in terms of turnover rates, catalytic bias (defined here as the ratio of the maximal rates in both directions²⁸), reversibility (the requirement for a thermodynamic driving force to trigger catalysis²⁹), and resistance to “stress” (e.g., sensitivity to O₂ or visible light). The immediate environment of the active sites in each family is very much conserved, but these enzymes differ by their protein sequences, cofactor composition, and quaternary structures. Nature therefore provides a quasi-infinite playground for examining outer coordination sphere effects in catalysis. For example, certain microorganisms produce many homologous enzymes to catalyze the same reaction (isoenzymes), which suggests that distinct catalytic properties are needed for different physiological functions.

In this review, we highlight and compare the role of second and outer coordination sphere effects on the activation of gaseous substrates such N₂, H₂, CO₂, or CO by various low-valent GPMs. We provide a comprehensive overview on protein variants that were found to affect the catalytic properties of nitrogenase, hydrogenase, FDH, and CODH. To understand or mimic GPMs, we emphasize that second and outer coordination sphere effects must not be neglected.

2. NITROGENASE

Nitrogen is an essential element for life, necessary for the synthesis of biomolecules such as deoxyribonucleic acid (DNA), ribonucleic acid (RNA), and amino acids. While nitrogen can be found in many different chemical forms throughout atmospheric and geological sources, the largest reservoir of nitrogen on the surface of the Earth is found in the form of dinitrogen (N₂): a small, gaseous, nonpolar, diatomic molecule with a strong triple bond.³⁰ The strength of this bond is on the order of ~940 kJ/mol, which makes its scission into bioavailable forms, also called “N₂ fixation”, one of the most challenging transformations in Nature.

Two main processes are recognized as natural engines for N₂ fixation within the global nitrogen cycle: lighting strikes and biological N₂ fixation (BNF).^31–33 The high temperature environment in the air surrounding lightning causes N₂ to react with other components in the atmosphere, such as dioxygen (O₂), ozone (O₃), or oxygen-based radical species, forming various nitrogen oxides (NO_x).^31,34,35 The NO_x species are then converted to nitric acid (HNO₃) in the atmosphere, and brought down to the ground through rainfall or surface deposition, where the fixed nitrogen species can be utilized.^30,36 This lightning-based process requires extreme conditions such as high temperature, and is completely unregulated which causes many chemical pathways to converge and ultimately produce the highly oxidized HNO₃. In contrast, biological N₂ fixation occurs in diazotrophic microorganisms that span across multiple phyla between bacteria and archaea.^37–39 These organisms are found in many environments, such as surface soils, sediments, marine and fresh waters, and geothermal sources, in both the presence or absence of oxygen.^40–43 Some organisms can live freely within microbial communities, such as the obligate aerobe Azotobacter vinelandii,⁴⁴ or the thermophilic methanogen Methanococcus thermolithotrophicus,^45,46 while others, such as Rhizobium leguminosarum, form symbiotic relationships with specific plant species.^47,48 In all cases, N₂ is converted into the highly reduced nitrogen species ammonia (NH₃), although more often ammonium (NH₄⁺) is the observed product.³⁰ Despite the wide variation in living conditions, each of the diazotrophs have genes that encode for the common enzyme nitrogenase.

Nitrogenase is a gas-processing metalloenzyme composed of two proteins, the reductase component NifH (also called the “Fe protein”), as well as the catalytic component NifDK (also called the “MoFe protein”) that contains a unique molybdenum-containing iron sulfur cluster called the M-cluster (or “FeMoco”).^49,50 There are also so-called “alternative” nitrogenases that replace the Mo ion in FeMoco for either vanadium (“FeVco” in the VFe protein) or iron (“FeFeco” in the FeFe protein).⁵¹ All nitrogenases are capable of reducing small molecule substrates, such as acetylene (C₂H₂) and cyanide (CN⁻), under ambient temperature and pressure, but the native reaction catalyzed by the enzyme is the reduction of N₂ to NH₃.^51–53 In comparison, humans have developed the Haber-Bosch process that takes N₂ and dihydrogen (H₂) under high temperature and intense pressure conditions to yield NH₃.⁵⁴ Nitrogenase also catalyzes the reduction of protons (H⁺) to H₂, and in all reactions with other substrates (including N₂), some portion of the electron equivalents (or electron flux) is diverted to produce H₂.^50,55 During the native nitrogenase reaction, H₂ is generated as an additional product (eq 1), but the mechanism of its synthesis is distinct from simple proton reduction.^52,56–60

The complexity of nitrogenase has captivated researchers for decades as they seek out answers to the question, how does nitrogenase carry out these difficult chemical transformations? Many scientific approaches have been utilized in its study, each providing complementary information that can add to the larger picture. Since nitrogenase has been extensively reviewed,^{50–52,61–63} this section shall primarily focus on secondary and outer sphere effects that modulate the function of the Mo-dependent enzyme.

2.1. Structural Features of Nitrogenase

The Mo-nitrogenase from A. vinelandii is the most extensively studied nitrogenase protein, and under N₂ fixing conditions up to 10% of all cell proteins are nitrogenase related.^44,64 The MoFe protein is an α₂β₂ heterotetramer of the nif D and nif K gene products that form NifDK (~220 kDa), where two αβ-dimer units pair together.⁶¹ Each αβ-dimer houses two different metalloclusters that are critical for enzyme function, the P- and M-clusters, such that each NifDK tetramer binds a total of four metalloclusters (Figure 2).^61,65,66 The nif H gene product, NifH, serves as the reductase partner of NifDK, and is composed of a homodimeric protein (γ₂) with a molecular weight of ~60 kDa that binds an [4Fe-4S] cluster at the subunit interface.⁶⁷ Additionally, each subunit can bind one molecule of MgATP.^63,67,68 One NifH protein can bind to each αβ-dimer of NifDK, forming an Fe:MoFe protein stoichiometry of 2:1.^69,70

Figure 2.

Crystal structure of the NifH:NifDK complex (PDB ID 1N2C). The relevant metalloclusters are highlighted and the proposed direction of electron flow through the enzyme is depicted as a yellow arrow. The coloring scheme is as follows: NifD (green cartoon), NifK (blue cartoon), NifH (orange and gray cartoon), Fe, orange; S, yellow; C, gray. Reprinted (adapted or reprinted in part) with permission from ref 50. Copyright 2020 American Chemical Society.

The P-cluster mediates electron transfer between the reductase component, NifH, and the M-cluster during catalysis.⁷¹ The cofactor is composed of an [8Fe-7S] cluster positioned at the α/β interface of NifDK, ~ 10 Å below the surface of the protein.^65,72 It appears as the fusion of two [4Fe-4S] clusters that share a μ₆-sulfide vertex, with ligation from six cysteine residues, α-C62, α-C88, α-C154, β-C70, β-C95, β-C153, three of which come from each subunit (Figure 3). Interestingly, the cysteine residues bind terminally to some of the Fe centers, as is typically found with canonical [4Fe-4S] clusters, but two of them bind in bridging modes between the two cubane halves of the P-cluster. Structurally, this fused cubane geometry of the P-cluster is unique and has not been identified elsewhere in biology. The P-cluster has three interconvertible oxidation states that have been identified as relevant for catalysis: the as-isolated P^N state, the one-electronoxidized P¹⁺ state and the two-electron oxidized P^OX (or P²⁺) state.^62,73 The P^N-cluster has been characterized as an all ferrous (Fe²⁺) cluster,^74–77 and subsequent oxidations result in reversible structural changes that physically open the cofactor.^78–80 This structural fluctuation of the P-cluster during redox changes has been proposed as a means of regulating or “gating” electron transfer within NifDK.^62,81–83

Figure 3.

Structure of the P-cluster. (a) Dithionite-reduced state (P^N, PDB ID 3U7Q) and (b) indigo disulfonate-oxidized state (P^OX, PDB ID 3MIN). Atomic coloring: Fe, orange; S, yellow; C, gray; N, blue. Reprinted (adapted or reprinted in part) with permission from ref 84. Copyright 2014 American Chemical Society.

The M-cluster is an asymmetric [Mo-7Fe-9S–C-R-homocitrate] species, with the Mo-capped end ligated by the organic acid R-homocitrate through the 2-hydroxy and carboxyl groups, as well as by the α−442His residue, and the opposite, Fe-capped end ligated by α-C275 residue (Figure 4A).^65,66,85 The cluster is buried within the α subunit of NifDK and is positioned ~19 Å from the P-cluster and over 60 Å from the clusters of the partner αβ-dimer. The structure of the M-cluster is similar to the P-cluster, as both are a fusion of two different cubane species with a common vertex, but the two partial cubane units of the M-cluster, [Mo-3Fe–3S] and [4Fe-3S], are not identical, and the common vertex is an interstitial μ₆-carbide as opposed to a sulfide.^66,86–88 Additionally, there are three “belt” μ₂-sulfido ligands that bridge between the Mo- and Fe-capped halves of the cofactor. These belt sulfide ligands have been shown to exhibit lability in both Mo- and V-nitrogenases, and this apparent flexibility has been implicated to be important during catalysis.^89–95 In V-nitrogenase (or VnfDGK), the V-cluster is a structural analog of the M-cluster (Figure 4B), but with a V ion in place of the Mo ion, and one of the μ₂-sulfide ligands replaced with a carbonate (CO₃²⁻) ligand of unknown origin.^92–95 There are physical metrics that vary between the M- and V-clusters, but they are still remarkably similar.⁵¹

Figure 4.

Catalytic cofactors from the crystal structures of (a) NifDK and (b) VnfDGK. The atoms of the clusters are shown as ball-and-stick models with the amino acid residues, R-homocitrate and carbonate presented as sticks. Atomic coloring: Fe, orange; S, yellow; Mo, teal; V, dark gray; C, light gray; N, blue; O, red (PDB ID 3U7Q, NifDK; 5N6Y, VnfDGK).

The electronic description of the M-cluster has been of interest for decades, as the M-cluster is capable of supporting many redox states,⁵⁶ and so the cluster has been experimentally studied using combinations of Mössbauer, X-ray absorption (XAS), magnetic circular dichroism (MCD), and electron paramagnetic resonance (EPR) spectroscopic techniques.^{62,75,96–98} Many of these techniques probe the sample in an ill-defined state, and as such, there can be great difficulty in deconvoluting the specific contributions of individual clusters. The most commonly observed state of the M-cluster is the resting, dithionite (S₂O₄²⁻) reduced M^N state that is associated with the diagnostic S = 3/2 rhombic signal from EPR spectroscopy (g = 4.3, 3.7 and 2.0).^73,96,99,100 Computations and high-energy resolution fluorescence detected (HERFD) Mo K-edge XAS experiments identified that the Mo center of the M^N-cluster is best described as an S = 1/2 Mo³⁺ center, as opposed to the classical assignment of Mo⁴⁺.⁹⁸ This indicates that the Mo exists in an atypical “non-Hund” configuration, and additional spectroscopy and density functional theory (DFT) calculations support a [3Fe²⁺:4Fe³⁺:Mo³⁺]¹⁻ assignment.^{97,98,101,102} Further, a crystallographic spatially resolved anomalous dispersion (SpReAD) analysis compared electron densities to Fe K-edge XAS experiments to determine the oxidation states of specific Fe atoms of the M-cluster, and this study agreed with the previously described oxidation assignment.¹⁰³ The M^N-cluster can also be oxidized by one electron to the M^OX form, which coincides with a loss of the S = 3/2 EPR signal, and is generally not believed to be relevant for catalysis.^73,104,105

While NifH binds a smaller, more prevalent type of biological [4Fe-4S] cluster, the Fe protein is far from simple. NifH has three recognized physiological functions: (1) Mo and homocitrate insertase in the biosynthesis of the M-cluster, (2) reductase involved in P-cluster biosynthesis on NifDK, (3) obligate reductase for nitrogenase catalysis that couples electron transfer to MgATP hydrolysis.⁶⁸ The [4Fe-4S] cluster binds between the two subunits of NifH and is positioned on a C₂ rotation axis with cysteine coordination by two residues (γ-C97 and γ-C132) from each subunit.⁶⁷ This cluster is located in a solvent-exposed position, which differs from [4Fe-4S] ferredoxins that have clusters buried within the protein.⁶³ Further, MgATP molecules that bind to the Walker motif A protein fold have been shown to affect the position of the [4Fe-4S] cluster despite being bound ~20 Å away from the cluster (Figure 2).^68,106,107 This apparent cluster movement has been studied by various structural and spectroscopic methods and is associated with modulation of the reduction potential of NifH.^63,108–112 The protein can support [4Fe-4S]²⁺, [4Fe-4S]¹⁺, and [4Fe-4S]⁰ oxidation states, which is abnormal compared to other iron–sulfur cluster containing proteins that only support two redox states.¹¹³ In NifH, the [4Fe-4S]¹⁺ state is favored in the presence of dithionite^51,100 and interestingly exists as a mixture of two different spin states, S = 1/2 and S = 3/2, the composition of which changes with additives such as glycerol or urea.^96,114 During catalysis, the [4Fe-4S]^2+/1+ couple is proposed to be operative, transferring one electron to NifDK for every two molecules of MgATP cleaved by NifH, and NifH becomes reduced in vivo through physiological reductants such as ferredoxins or flavodoxins.^{63,83,115–118} The [4Fe-4S]⁰ “all ferrous” or “super-reduced” state has been implicated in reactivity of NifH with carbon substrates like CO and CO₂, yet has also been proposed to play a role in nitrogenase catalysis, but the characterization of this species is fraught with complications that make understanding a physiological role difficult.⁵¹

2.2. Assembly of Nitrogenase Cofactors

Nitrogenase is a complicated system, not only because of the multimeric metalloprotein composition and unique cofactors, but also because of the requirement of many genes for the regulation and biosynthesis of nitrogenase.^50,51,119 For the NifDK protein from the model organism A. vinelandii, a minimum set of nitrogenase proteins that includes NifU, NifS, NifB, NifEN, NifH, NifV, and NifZ are required for in vitro activation of N₂ fixation.^49,120 These proteins are encoded by nif or “nitrogen fixation” genes, and the nif operon is specific to the Mo-nitrogenase system. Organisms that encode for V- or Fe-only nitrogenase (vnf and anf genes, respectively) still require some of the nif operon for proper function.⁵¹ For Mo-nitrogenase, the biochemistry and assembly of the MoFe protein and the cofactors bound therein have been extensively studied over the past two decades, yielding crucial insights into the process.

The biosynthesis of NifDK involves the assembly of both the P- and M-clusters in addition to the polypeptide, but cofactor generation occurs through different routes (Figure 5). While the P-cluster is assembled in situ on the NifDK polypeptide scaffold, the M-cluster is assembled ex situ on a series of different proteins before terminally reaching NifDK.^49,50,120 The P-cluster originates from a pair of [4Fe-4S] clusters, called the P*-cluster, that are synthesized at the α/β subunit boundary of apo NifDK.^84,121–124 The iron–sulfur fragments are assembled and delivered to apo-NifDK through the action of NifU, a scaffold protein that binds iron–sulfur clusters, and NifS, a cysteine desulfurase that provides sulfide from cysteine.^125,126 The P*-cluster is then fused together with concomitant loss of sulfur by the stepwise action of NifH and NifZ in a MgATP- and electron-dependent process. The result is an apo-NifDK protein with two equivalents of P-cluster bound but lacks M-cluster.

Figure 5.

Biosynthetic diagram for the assembly of nitrogenase metalloclusters. The red portion represents the in situ pathway for the synthesis of the P-cluster on NifDK. The blue portion reflects the ex situ synthesis of the M-cluster. Proteins and cluster species are labeled accordingly. Atoms are represented as ball-and-stick models and with atomic coloring as follows: Fe, orange; S, yellow; C, gray; O, red; Mo, teal. Reprinted (adapted or reprinted in part) with permission from ref 51. Copyright 2020 American Chemical Society.

In contrast, the assembly of the M-cluster starts with the construction of [4Fe-4S] clusters by NifU and NifS that are subsequently delivered to NifB, a radical S-adenosyl-l-methionine (SAM)-dependent enzyme.^127,128 NifB binds three [4Fe-4S] clusters; one is specifically for catalyzing SAM reactivity, while the other two are the substrate clusters (K1 and K2) known collectively as the K-cluster.^129–131 Under reducing conditions, two equivalents of SAM are cleaved to first transfer a methyl group to the K2 cluster on NifB, and the second equivalent of SAM initiates a hydrogen atom abstraction of the transferred methyl group.^87,88,131 In a series of uncharacterized subsequent steps, carbide (C₄⁻) is generated upon removal of the remaining hydrogen atoms. Concurrently, the two K-cluster cubane units are fused yielding a [8Fe-8S–C] cluster, called the L*-cluster, that contains an interstitial carbide bound to the six central Fe atoms of the cluster.^132,133 A final “9th sulfur” is then incorporated into the L*-cluster through the reduction of a sulfite (SO₃²⁻) ion to produce the [8Fe-9S–C] L-cluster. The L-cluster on NifB is transferred to another scaffold protein, NifEN, that facilitates the conversion of the L-cluster into the M-cluster through the loss of an Fe atom and the incorporation of Mo and R-homocitrate.^134–137 In vitro experiments demonstrate that M-cluster-loaded NifEN can then interact directly with apo NifDK (containing only P-clusters) to transfer the M-cluster and produce active NifDK, although additional chaperone proteins have been suggested to assist with this process in vivo.^49,50

2.3. Nitrogen Reduction Mechanism

N₂ reduction requires a high degree of regulation to successfully carry out the reaction, and one of the key factors to this process is the shuttling of electrons. Electrons are transferred from NifH to NifDK through direct complexation of both proteins and the hydrolysis of two MgATP molecules for each electron delivered.⁶³ This creates a complex system that has been the subject of study for decades. Extensive investigation of the mechanism of N₂ reduction in the 1970s and 1980s led to a kinetic framework known as the Lowe-Thorneley model or cycle.^51,52,63 This model can be viewed from two different perspectives; one is an Fe protein-centric view known as the “Fe protein cycle” and the other, more common view is that of the MoFe protein, or the “MoFe protein cycle”.

The Fe protein cycle begins with the NifH protein being in the reduced [4Fe-4S]¹⁺ state (NifH^R), binding two equivalents of MgATP.^63,83 The reduction to NifH^R has been facilitated in vitro by dithionite, but other reducing agents such as Eu(II) chelates have been used as well. The nucleotide-bound Fe protein then interacts with NifDK, forming a 2:1 complex that has been crystallographically characterized.^69,70,138 Subsequently, there are a series of steps that involve conformational changes, electron transfer, ATP hydrolysis, and phosphate release, resulting in oxidized NifH (NifH^OX, [4Fe-4S]²⁺) with two equivalents of MgADP bound in complex with NifDK.^63,83 NifH^OX and NifDK undergo dissociation, releasing the reductase before NifH^OX is reduced back to the NifH^R state, which becomes capable of exchanging MgADP for MgATP, thereby restarting the cycle. This cycle is presumably active during the reduction of most substrates by NifDK, though in the absence of NifDK, NifH has also been shown to carry out C1 substrate reduction through alternative mechanisms.^51,68

The Lowe-Thorneley model for N₂ reduction by NifDK, or the MoFe protein cycle, describes the catalytic intermediates in the reaction, each designated by an E_n notation where n is the number of proton and electron equivalents that have accumulated on one αβ-dimer of NifDK (Figure 6).⁵⁶ However, the value of n does not indicate where proton and electron equivalents are located, only that they have been delivered to the system, which has sparked considerable research efforts in the field.⁵² The first state, E₀, reflects the resting, as-isolated state of NifDK associated with the well-characterized S = 3/2 EPR signal of the M-cluster.⁶² As each proton and electron pair is accumulated through the early (n = 2–4) steps of the cycle, it is possible for a “relaxation” to occur, releasing H₂ with concomitant transition to an n − 2 state.^52,60 N₂ has been proposed to bind to the M-cluster in the E₃ and E₄ states, but reduction of N₂ only occurs in the E₄ (or E₄(4H)) state.^52,56,139 This makes the E₄ state one of the most studied of the cycle, because either E₄ can relax backward to lower E_n with release of H₂, or the cycle can proceed forward with the cleavage of the N₂ bond. Intriguingly, when N₂ binds to E₄(4H), one equivalent of H₂ is produced with concomitant formation of the E₄(2N2H) state through a proposed mechanism of reversible reductive elimination or oxidative addition.^52,60,140 The E₄(4H) to E₄(2N2H) reaction represents the first lowering of the N₂ bond order, though several possibilities have been put forward as to the identity of the resultant E₄(2N2H) species.^51,52

Figure 6.

Modified Lowe-Thorneley model for N₂ reduction by Mo-dependent nitrogenase. The E_n labeling indicates n number of proton and electron pairs that have been loaded on one of the αβ-dimers of NifDK. Reprinted (adapted or reprinted in part) with permission from ref 51. Copyright 2020 American Chemical Society.

The subsequent steps of the MoFe protein cycle E₅–E₈ promote the further reduction of the N₂ bond. Initially, there were two competing proposals for this latter half of the cycle that differed in the location where proton and electron equivalents are delivered to the reduced nitrogen species: a “distal” and an “alternating” pathway. The distal pathway posits that protons are delivered to the unbound N atom of N₂ first, forming an M = N–NH₂ hydrazido species that undergoes further proton/electron addition to yield one equivalent of NH₃ and a terminal M ≡ N metal nitrido species by the E₅ state.^141–146 The E₅ species is then sequentially protonated to release the second equivalent of NH₃ after the E₈ state. The alternating pathway, based on synthetic Fe-catalyzed versus Mo-catalyzed N₂ reduction,^147,148 proposes that protons are distributed between both N atoms of N₂, starting with diazene (HN=NH), proceeding through hydrazine (H₂N–NH₂) and releasing NH₃ after E₆ and E₈. While there is still uncertainty in distinguishing between these two mechanistic possibilities, mounting evidence supports the alternating pathway for biological N₂ reduction,^51,52 including crystallographic evidence of a putative N₂-bound species.⁹¹ However, evidence from the V-dependent nitrogenase potentially suggests a distal pathway may be operative in that system, but kinetic analysis and calculations propose that the Mo-, V-, and Fe-only nitrogenases all share a common mechanism.^{51,93,149,150} Clearly, further investigation will be necessary.

2.4. Nitrogenase Active Site Pocket

To understand the outer sphere environment surrounding the M-cluster, it is necessary to take a critical look at the residues that make up the active site pocket. However, there is no “best” way to discuss the location of amino acid residues with relation to the catalytic cofactor of nitrogenase. High resolution crystallography has provided three-dimensional representations of nitrogenase and the cofactors with precision, but it can be difficult to gain a sense orientation of the cofactor with respect to its protein environment. The 3-fold rotation axis that runs from the Mo-capped end to the Fe-capped end of the cluster does not inherently provide a landmark around the M-cluster by which relative position can be compared other than the Mo and Fe subcubanes. The crystallographic labeling of the cofactor atoms provides some basis of orientation, but the labeling convention is also not particularly intuitive; for instance, the three belt sulfur positions are given labels of S2B, S3A, and S5A, whereas other SXA and SXB labeled sulfur atoms are bound to the Fe-capped (or Fe1) and Mo-capped subcubane structures, respectively (Figure 7). In this discussion, a helpful means to navigate the M-cluster site is to establish an origin at the central carbide atom, followed by an “x-axis” that incorporates the C₃ symmetry axis running between Fe1 and Mo. The Fe1 end of the M-cluster is pointed toward the surface of the α-subunit adjacent to the bulk solvent (designated as the “negative” x direction), whereas the Mo end of the cluster is directed toward the αβ-subunit interface (designated as the “positive” x direction). The S2B sulfur site has been established as a labile position on the catalytic cofactor,^89,90 and is adjacent to several residues critical for reactivity, so a “z-axis” would bisect S2B as well as the central carbide atom of the cluster, with the direction of the S2B atom being positive. There is no convenient atom to position a “y-axis” relative to the carbide, but the S3A side of the xz plane can be the positive direction of y and the S5A side can be negative. Within this framework of (x, y, z) coordinates, α-H442 would be in the (+, +, −) direction relative to carbide, α-C275 would have a (−, 0, 0) position, and the nearest P-cluster would have a bearing of (+, −, +). While also an imperfect system due to constraints, such as α-R359 (Figure 8) and the R-homocitrate ligand (HC) running through positive and negative values of y, the coordinates should generally provide a spatial navigation within the active site.

Figure 7.

A relative coordinate system is (blue arrows) overlaid with the crystal structure of M-cluster. The coordinate system assumes CX as the origin (0, 0, 0), the x-axis passes through Fe1, CX and Mo1 with positive values toward Mo1, the y-axis passes through CX into the page with positive values pointing out of the page, and the z-axis passes through S2B and CX with positive values toward S2B. The cluster is presented as a ball-and-stick model with atomic coloring: Fe, orange; S, yellow; C, gray; O, red; N, blue.

Figure 8.

Display of amino acid residues targeted for mutagenesis within the active site M-cluster of NifDK from A. vinelandii. (a) Side view and (b) view along the Mo-C-Fe1 axis of the M-cluster site are shown (PDB ID 3U7Q). The cluster is presented as a ball-and-stick model while the homocitrate ligand (HC) and amino acid side chains are shown as sticks. Atomic coloring: Fe, orange; S, yellow; C, gray; O, red; N, blue.

The catalytic M-cluster of nitrogenase is one of the largest, most complicated biological cofactors known so far but it is only bound to the protein scaffold through two amino acid residues, α-C275 and α-H442.^{61,65,66,85,151} While R-homocitrate does bind to the Mo-end of the cluster, the lack of additional ligation to the M-cluster indicates that second coordination sphere effects play a crucial role in catalysis. Many amino acids have side chains with nitrogen-containing functional groups such as histidine, arginine and glutamine that are capable of hydrogen-bonding interactions. Nonpolar or otherwise bulky residues such as valine, phenylalanine and tryptophan have been found to play roles in the steric environment for substrate access and binding, as well as appropriate positioning of the M-cluster within the active site. All of these residues are connected to each other in a tightly regulated way, and disruptions in this network tend to be detrimental to the function of the enzyme. Figure 8 shows the positions of some of the important residues discussed within this review relative to the M-cluster.

In 1991, before crystal structures of NifDK were solved, a series of point mutations of conserved histidine residues were reported, hoping to identify those interacting with the M-cluster. Making use of electron spin echo envelop modulation (ESEEM) spectroscopy,¹⁵² microwave pulses are used in a specific sequence within an EPR experiment to probe nuclei (such as ¹⁴N or¹H) that are coupled to a paramagnetic species.⁶² Specifically, the intense S = 3/2 EPR signal (Figure 9) associated with the M-cluster was probed to search for ¹⁴N atoms from histidine residues that may be hyperfine-coupled with or covalently bound to the cofactor.¹⁵²

Figure 9.

EPR characterization of the M-cluster. The figure depicts wild-type (black), α-H195Q (red), and α-H195N (blue) NifDK proteins from A. vinelandii. The g ≈ 2 region is not shown for α-H195N due to contamination. Adapted with permission from ref 153. Copyright 1995 American Chemical Society. Data digitalized via WebPlotDigitizer v4.5.

Residues α-H83 (+, −, +), α-H195 (−, 0, +), α-H196 (−, 0, +), α-H274 (−, −, −), and β-H90 (+, +, +) were targeted for single point mutations. ESEEM studies on whole cells were carried out and signals assigned to N atoms from histidines were observed in all variants, only the variant α-H195 NifDK proteins showed no ESEEM signal, and it was concluded that α-H195 must bind to the M-cluster. However, once the crystal structure of NifDK became available in 1992, it was clear that α-H195 was too far from the M-cluster to bind directly, but instead was likely involved with a hydrogen-bonding interaction to the S2B sulfur of the cofactor. In 1995, Dean, Hoffman, and co-workers studied the EPR and ESEEM signature of partially purified NifDK variants, namely α-H195N and α-H195Q.¹⁵³ Consistent with the previous observations, it was found that the ESEEM signal was dramatically attenuated in the α-H195N NifDK variant but α-H195Q NifDK showed nearly identical signals in both EPR and ESEEM experiments to those of the wild-type enzyme. If the hydrogen bond of α-H195 was responsible for the ESEEM signals, then both variants should display similar spectra, but they did not. This suggested that a nitrogen-containing residue that is sensitive to the position of the α-H195 residue was responsible for the signal. Subsequently, ESEEM studies were expanded to include purified variant proteins targeting residues α-R96 (+, −, 0), α-H195 (−, 0, +), α-R359 (0, +, −), α-F381 (0, +, +), and α-H442 (+, +, −).¹⁵⁴ The spectra of α-H195N NifDK showed that the ESEEM signal (labeled as “N1”) was indeed eliminated as previously reported,^152,153 but a weaker signal “N2” that was buried underneath was also observed.¹⁵⁴ Mutation at α-R96 only showed the N1 signal, eliminating this residue as the source of N1, and substitution of α-H442 resulted in a lack of observable EPR signal, consistent with loss of the M-cluster entirely. However, mutation of α-R359 or α-F381 resulted in a loss of the N1 signal while the N2 signal remained intact. Together, these experiments were used to assign the N1 signal to α-R359 and the N2 signal to the peptide NH groups of one or both of the residues α-G356 (+, +, 0) and α-G357 (−, +, 0).¹⁵⁴ However, the catalytic activity of the variant NifDK proteins could not be correlated to the ESEEM signals. Despite this, the study demonstrates that changes in the second coordination sphere on one side of the M-cluster can perturb the interactions occurring on the other side of the cluster, as seen by the loss of the N1 signal at α-G359 (0, +, −) when α-H195 (−, 0, +) is modified. In fact, changing the steric environment as seen with α-F381L NifDK variants (0, +, +) was enough to disrupt hydrogen-bonding interactions around the cluster even though α-F381 is not directly involved in such an interaction, reinforcing that the active site residues are highly interconnected.

Another salient observation about the active site of nitrogenase comes from the crystal structure of an M-cluster depleted NifDK, called Δnif B NifDK.¹⁵⁵ In 2002, Burgess, Rees, and co-workers managed crystallized the NifDK protein without the catalytic cofactor, which resulted in changes to the tertiary structure as compared to wild-type enzyme. Rearrangements of the amino acids left a “funnel” on the surface of the protein that was positively charged, with a loop of residues from α−353 to α−364 positioned at the entrance to the funnel. This loop also contained residues that had positively charged, nitrogen-containing side chains, and the authors proposed that the electrostatic charge of the funnel would help direct the negatively charged M-cluster into the appropriate position so it could be locked into the active site through movement of the funnel loop. Interestingly, a similar concept was discussed for the insertion of the active site cofactor of [FeFe]-hydrogenase (Section 3.2.2). Subsequent point mutation studies further explored this proposal by targeting specific residues that may assist in guiding the M-cluster into NifDK.^156–158 Thus, the side chain selection within the nitrogenase active site has the demonstrated capability to support a robust network of hydrogen bonds, while at the same time functioning in a biosynthetic capacity to orient the M-cluster optimally within the protein for catalysis.

2.5. Assessing the Nitrogenase Second Coordination Sphere

Before the publication of the crystal structures in the 1990s,^{65,67,69,80,85,159} the study of nitrogenase was based on genetic, spectroscopic, and reactivity experiments that were used to gain insight into the structure of the protein as well as the bound metallocofactors. At that time, direct information regarding the properties of the individual P- and M-clusters was scarce, and their three-dimensional relationship. Even after the appearance of high-quality crystallographic data, the study of structure–function relationships in nitrogenase was difficult due, in no small part, to the multiple, unique iron–sulfur clusters that are bound to the enzyme. Some of the early efforts to characterize and understand the nitrogenase cofactors involved chemical extraction into organic solvents.^160,161 The P-cluster is integral to the nitrogenase structure, so the acid quenching required for extraction destroys the P-cluster and disrupts the protein fold such that the M-cluster can be removed. This allowed for new avenues of study and analysis, but to date, there have been no successful reports of recrystallized M-cluster outside of the nitrogenase protein.¹⁶²

To better assess the nitrogenase protein structure, amino acid sequencing was carried out through genetic analysis and protein digestion in the 1970s and 1980s to compare nitrogenases from different species.^163–168 Conserved residues were identified and targeted for single-point mutations as a way to connect information about the cofactors to the protein structure and function. Initially, conserved cysteine residues were investigated for their widespread use in biology as ligands for iron–sulfur clusters, as it was understood that nitrogenase likely contained multiple [4Fe-4S] clusters. These cysteine residues were changed into serine or alanine, and the resulting bacterial strains (of A. vinelandii or Klebsiella pneumoniae) were studied.^169–172 Growth rates in the absence of a fixed nitrogen source were used to establish the N₂ fixation phenotype, and whole cell and/or crude extract activity assays were carried out in an acetylene atmosphere to compare diazotrophic growth rates with nitrogenase activity.^169,172 In addition, EPR spectra of whole cell variants or crude extracts of these cells would be collected to search for the intense S = 3/2 signal associated with the M-cluster as a means to cross reference the observed reactivity.¹⁷² Generally, when the cysteine residues (that would later be found to bind to the P- and M-clusters) were mutated to alanine, diazotrophic growth and acetylene-based nitrogenase activity would be eliminated. Serine subsitutions of cysteine residues bound to cofactors would also eliminate activity, but in one case, Newton and co-workers found that substitution of C183 (+, −, +) on the NifD subunit (or α-C183) from A. vinelandii to serine did not arrest activity but resulted in diminished diazotrophic growth rates.¹⁶⁹ Conserved residues such as α-Q191 (+, +, +) were mutated to change the polarity or charge; for instance, changing the polar, amide-bearing amino acid side chain of glutamine to the carboxyl group of glutamate (α-Q191E).¹⁶⁹ In this case, the activity was eliminated, despite the fact that glutamine residues are not typically found as ligands to iron–sulfur clusters in biology.^113,173 These findings were unanticipated, but provided indications of the active site of nitrogenase in the absence of structural data, and marked the beginning of targeted efforts to understand the secondary sphere effects within nitrogenase. In the years that followed, many mutagenic studies were reported with a focus on determining effects to nitrogenase activity as well as the number and location of substrate binding sites on the M-cluster. The residues that have been targeted for mutagenic studies related to the M-cluster site in NifDK from A. vinelandii have been summarized in Table 1 and Table 2.

Table 1.

List of Single Point Mutations Reported for Azotobacter vinelandii NifDK with Proximity to the M-Cluster Site^a,b

NifDK residue	Arg (R)	His (H)	Lys (K)	Asp (D)	Glu (E)	Ser (S)	Thr (T)	Asn (N)	Gln (Q)	Cys (C)	Gly (G)	Pro (p)	Ala (A)	Ile (i)	Leu (L)	Phe (F)	Tyr (Y)	Trp (W)	ref
α-G69						x													174, 175
α-V70											x		x	x					176, 177, 178, 179
α-H83				x				x		x						x	x		152
α-R96		x	x						x				x		x				154, 179, 180, 181
α-Q151					x														169
α-C154						x													169
α-D161					x														169
α-C183						x													169
α-S190											x								174
α-Q191			x		x							x	x						139, 169, 174, 178, 181, 182, 183, 184, 185
α-S192	x			x	x		x	x	x						x			x	174, 186
α-H195							x	x	x		x				x		x		139, 152, 154, 179, 182, 183, 184, 187, 188, 189
α-H196	x			x	x									x	x				152
α-H274									x				x						152, 156, 158
α-C275						x													169
α-R277		x	x				x			x	x	x			x	x			181, 186, 190
α-G356c																			186
α-R359			x						x										154, 174, 181
α-H362				x									x						157
α-F381														x	x				154, 174
α-H442								x		x									154
α-W444											x		x			x	x		158
α-H451													x						156, 158

^aNifD = α, NifK = β.

^bNifDK residues listed have been mutated to the marked amino acids.

^cNo specific mutations are listed.

Table 2.

List of A. vinelandii NifDK Residues near the M-Cluster That Have Been Investigated through Mutagenesis

NifDK residue	location near M-clustera	coordinateb	proposed role(s)
α-G69	outer sphere near S2B	(+, −, +)	Hydrophobic interactions with substrate
α-V70	2nd sphere near S2B	(0, −, +)	Hydrophobic interactions with substrate
α-R96	2nd sphere near S5A	(+,−, 0)	Steric/Hydrogen-bonding interactions with M-cluster, substrates
α-S190	outer sphere near S2B	(0, +, +)	Steric or hydrogen-bonding interactions
α-Q191	2nd sphere near S2B, HC	(+, +,+)	Hydrogen-bonding interactions with R-homocitrate ligand ofM-cluster
α-S192	outer sphere near S2B	(−, +, +)	Hydrogen-bonding interactions with conserved H2O molecules
α-H195	2nd sphere near S2B	(−, 0, +)	Hydrogen-bonding and steric interactions with M-cluster, substrates
α-H196	outer sphere near S2B, Fe1	(−, 0, +)	Unknown
α-H274	outer sphere near Fe1, S4A	(−,−,−)	Assists with M-cluster incorporation, adjacent to α-C275
α-C275	cofactor bound	(−, 0, 0)	Ligates Fe-capped end of M-cluster
α-R277	2nd sphere near S1A	(−, +, +)	Hydrogen-bonding interactions with conserved H2O molecules
α-G356	2nd sphere near S3A	(+, +, 0)	Hydrogen-bonding interaction with M-cluster
α-R359	2nd sphere near S5A	(0, +, −)	Hydrogen-bonding interaction with M-cluster
α-H362	outer sphere near Fe1	(−, +, −)	Assists with M-cluster incorporation
α-F381	2nd sphere between S2B, S3A	(0, +, +)	Hydrophobic interactions with M-cluster
α-H442	cofactor bound	(+, +, −)	Ligates Mo-capped end of M-cluster
α-W444	outer sphere near S4B	(+,−, −)	Hydrophobic interactions with M-cluster
α-H451	outer sphere near S4A	(−,−,−)	Assists with M-cluster incorporation

^aLocations are based on proximity to the M-cluster atoms from the crystal structure of the resting state NifDK (PDB ID 3U7Q).³⁸

^bThe general location of the amino acid side chain in the coordinate system defined in Figure 7.

One complicating factor when discussing the results of site-directed mutagenic studies is that often the modification substantially reduces or eliminates nitrogenase activity, but it is not always clear why this occurs. In studies of the MoFe protein from K. pneumoniae, Buck and co-workers tested the activity of cell extracts from several cysteine to alanine variants and found almost no acetylene reduction, implying their mutations have made nitrogenase nonfunctional.¹⁷² They employed a parallel experiment to assess the stability of the variant proteins where wild-type apo NifDK (Δnif B NifDK) was added to their cell extracts and would show activity only if free M-clusters were present in the extracts to enable reconstitution of the Δnif B NifDK protein. Low acetylene reduction activity was observed for many variants following such a treatment, but the α-C275A NifDK protein that targeted the M-cluster-binding residue showed an increased activity by an order of magnitude.¹⁷² This indicated that some variant proteins are unable to retain the M-cluster, a conclusion that was validated by adding additional M-cluster to cell extracts of the NifDK variants alone with no observed increase in activity. Indeed, extracts were commonly used because of the difficulty associated with purification, and this problem becomes exacerbated when the stability of the protein is affected. Several reports from Newton and Dean acknowledge that certain proteins could only be partially purified because they would not survive heat treatment steps.^{139,153,183,185,191} In some instances where purification was successful, the Mo content of the NifDK variant would be lowered in the wild-type enzyme (<50%), implying that cofactor loading was impaired resulting in variable mixtures of apo and holo NifDK proteins.^139,183,191 These studies still provide valuable insight, but it is important to recognize that it can be difficult to make definitive assessment of outer sphere effects when activities can be attenuated from simple loss of the M-cluster or degradation of the protein.

However, there are investigations targeting the assembly of NifDK that use the retention or loss of the M-cluster as means to probe the function of certain amino acids.^156–158 Ribbe and co-workers have carried out site-directed mutagenesis on several outer coordination sphere residues ≥6 Å from the M-cluster (Figure 10): α-W444 (+, −, −), α-H274 (−, −, −), α-H362 (−, +, −), and α-H451 (−, −, −). A crystal structure of apo NifDK from a Δnif B background revealed that α-H442 in the absence of the M-cluster shifts ~5 Å within the active site.^155,158 The α-H442 residue is rearranged and placed nearby two other histidine residues, α-H274 and α-H451, forming a “triad” while another residue, α-W444, is found to take up the same space as α-H442 does in the holo enzyme (Figure 10). The apparent position swapping of α-H442 and α-W444 was intriguing, so mutagenesis was carried out to replace α-W444. In the α-W444Y and α-W444F NifDK variants, the M-cluster loading was ~90% of the wild-type enzyme, whereas α-W444A and α-W444G contained only 14% and 3% of wild-type M-cluster, respectively.¹⁵⁸ As the side chain of the α−444 residue was changed to smaller, less sterically bulky groups, the Mo concentration decreased. The reactivity assays of the variant proteins with N₂, C₂H₂, and protons, as well as the intensity of the EPR signals all roughly decreased by the same magnitude as the Mo concentration. Paired with heat treatment studies, it was suggested that the variant proteins themselves were stable, leading to the conclusion that α-W444 must be involved in a steric interaction that keeps the M-cluster secure within the active site pocket.¹⁵⁸ It was also observed that in the α-W444A and α-W444G variants, the EPR signal of the P-cluster appeared to distort relative to the larger substituent-containing variants, implying that α-W444 may also have longer range effects in NifDK than immediately around the M-cluster site.

Figure 10.

Crystal structure of Δnif B NifDK (in gray; PDB ID 1L5H), overlaid with the M-cluster and selected residues from the crystal structure of wild-type AvNifDK (in pink; PDB ID 3U7Q). The two structures were aligned, and the M-cluster and specific residues from the wild-type protein were superimposed on top of the Δnif B NifDK structure. The residues α-H274, α-H442, and α-H451 represent the identified “histidine triad”. The cluster atoms are shown as ball-and-stick models, and the amino acid residues are presented as sticks. Atomic coloring: Fe, orange; S, yellow; C, gray (in DnifB NifDK) or pink (in wild-type NfDK); O, red; N, blue.

The other two residues of the histidine triad, α-H274 and α-H451 (Figure 10), were also targets for point mutations.¹⁵⁶ A. vinelandii cells separately expressing NifDK variants α-H274A, α-H451A and a double mutant α-H274A/α-H451A grew more slowly under N₂-fixing conditions, indicating issues with nitrogenase activity, with the double mutation having the most deleterious effect on the growth rate. Assessment of the standard substrate reduction assays showed decreased activity of 67%, 56%, and 43% for the α-H274A, α-H451A, and α-H274A/α-H451A NifDK variants, respectively. The decreased activity also correlated well to the Mo concentration, and to the intensity of EPR signals for each of the variants. The α-H451A NifDK protein in the dithionite-reduced state had a very similar EPR spectrum for the S = 3/2 M-cluster as wild-type NifDK, but the introduction of α-H274A as either the single or double mutant yielded a spectrum with large distortions in g-values as well as line shapes. The rationalization for this observation was that α-H274 is adjacent to α-C275, which binds the M-cluster and may have more of a direct impact on the M-cluster environment than the more distant α-H451.¹⁵⁶ Additionally, none of the studied mutations affected the EPR signals of the P-cluster, indicating that these residues have less of a long-range steric effect than α-W444. Introduction of isolated M-cluster into solutions of M-cluster deficient α-H274A, α-H451A, and α-H274A/α-H451A Δnif B NifDK variants did not allow for full activation of the protein to wild-type levels of reactivity. Thus, it was concluded that both α-H274A and α-H451A are required during the biosynthesis of NifDK through sequential binding events that draw the M-cluster into the protein.¹⁵⁶ These residues may also play roles in the proton shuttling within the active site, as seen from shifts from N₂ reduction toward proton reduction in the variants, but one factor that was not explored is what steric effect the size of the side chain played in going from histidine to alanine. Alanine substitutions were certainly deleterious, but it would be interesting to see how the protein would retain the M-cluster if uncharged residues larger in size than alanine were studied.

Along similar lines, a final outer coordination sphere residue, α-H362 (−, +, −), had been studied with respect to M-cluster insertion into the NifDK protein (Figure 10).¹⁵⁷ In the crystal structure of the apo NifDK protein, α-H362 is surface exposed and is at the entrance to the insertion funnel of positively charged residues that have been proposed to guide a negatively charged M-cluster into the protein.^155,157 After M-cluster incorporation, the residue moves with the protein secondary structure to pin the cofactor in place. Single point mutants were generated and two variant NifDK proteins were studied, α-H362D and α-H362A.¹⁵⁷ Much like the previous mutations described, the α−362 variants grew more slowly under N₂-fixing conditions, expressed a decreased reactivity profile that correlated with Mo content, and had perturbations to the line shape and intensity of EPR signals, relative to wild-type enzyme. The decreased activities of the α-H362D and α-H362A NifDK variants are ~80% and ~40%, respectively. While modification of this residue did not cause a cessation of activity like that observed for α-W444,¹⁵⁸ substitution of α-H362 for aspartate induces a ~ 20% reduction in activity while alanine substitution results in a more drastic reduction of ~60%. This suggests that either α-H362 is involved in directing the M-cluster into the NifDK active site (supported by the decreased Mo concentration), or the residue is structurally important for adequately positioning the M-cluster within the active site (supported by the observed perturbations of the M- and P-cluster signals). The best answer is likely a combination of both, but further structural characterization of these NifDK variants may shed light on the M-cluster insertion process.

2.5.1. Reduction of Nitrogen, Acetylene, and Protons.

As mentioned earlier (eq 1), the reduction of N₂ by nitrogenase is a challenging transformation that requires the delivery of 8 electrons, 8 protons, and 16 equiv of MgATP to produce NH₃, H₂, and MgADP.^51,52,100 Not only is NH₃ produced, but H₂ is generated as an integral part of the N₂ reduction mechanism.^57,58,192 As discussed in Section 2.2, electron and proton equivalents are sequentially delivered to the catalytic cofactor by way of NifH and the P-cluster, leading to distinct E_n steps along the Lowe-Thorneley model reaction pathway.^56,60 At the E₄(4H) level, N₂ has been proposed to bind through a reductive elimination/oxidative addition mechanism, forming a reduced nitrogen species E₄(2N2H) and liberating one equivalent of H₂ (Figure 6). It is important to emphasize that while the bond order of N₂ is lowered when binding to E₄, the conversion between E₄(4H) and E₄(2N2H) technically requires no additional energy input, and the process is reversible in the presence of H₂. In fact, this reversibility is what allows H₂ to act as an inhibitor of N₂ reduction. This phenomenon has also been exploited to investigate N₂ binding. If isotopically labeled D₂ (²H₂) is mixed with N₂, the wild-type NifDK can produce HD gas, and, if separately, T₂ (³H₂) is mixed with N₂, NifDK does not significantly accumulate mixed label TOH.^56–60 This means that the reversible H₂ production through N₂ binding is not a simple proton exchange process with solvent, which supports the notion that hydrides (H⁻) are generated in the Lowe-Thorneley model.^52,60 Moreover, the relative rates between the forward and reverse reactions have been proposed to favor N₂ reduction over N₂ release,^52,149,150 but these rates change when moving to the alternative nitrogenases.

Protons are also reduced by nitrogenase in a hydrogenase-like process (hydrogenases are discussed in Section 3) that competes with other substrate reduction reactions, but also occurs in the absence of substrates such as N₂, C₂H₂, or CO.⁵⁵ Additionally, there is no known additive that can increase the proton reduction of the wild-type enzyme, and it is unclear what the site(s) of proton reduction is. This can complicate analysis of nitrogenase catalysis, as often enough N₂ or acetylene reduction is reported without proton reduction under the same conditions. Consequently, it becomes difficult to rigorously compare systems, particularly from single-point mutagenic studies. For instance, if N₂ reduction decreases between two NifDK variants, the proton reduction under those same conditions could either increase or decrease, and each scenario would lead to a different conclusion. There is also difficulty in the accurate measurement of NH₃/NH₄⁺ produced from catalysis compared to gaseous H₂, particularly at low concentrations. For this reason, acetylene is often used as an N₂ surrogate due to the analogous triple bond and gaseous nature of both the substrate and product states. Acetylene was first discovered as a competent substrate for nitrogenase in 1966 by Michael J. Dilworth,¹⁹³ and since then it has become crucial to the study of the nitrogenase mechanism despite the nonphysiological role the substrate plays.^{51,52,100,194} Wild-type Mo-nitrogenase reacts with C₂H₂ such that >90% of electron equivalents are funneled to produce exclusively ethylene (C₂H₄), substantially attenuating the H⁺ reduction pathway.^100,195 The alternative substrate is thought to bind to the nitrogenase E₁ or E₂ states, although, C₂H₄ is not released from the protein until three proton/electron equivalents are delivered to the M-cluster so that the E₁ state can be reformed.¹⁹⁵ This is in contrast to the proposed N₂ binding states of E₃ or E₄, the latter of which is proposed to be the only productive binding state that results in N₂ reduction (Figure 6). Further, the analysis of acetylene reduction in the presence of the nitrogenase inhibitor carbon monoxide (CO) has led to the proposal that there are multiple substrate binding sites on the M-cluster, and that N₂, C₂H₂ and CO occupy different positions.¹⁹⁵ Finding where the substrate binding sites are located as well as the properties of each has been a driving force behind the analysis of single-point mutations of NifDK. There have been many amino acid residues that have been targeted for single-point mutagenesis, and herein, only examples pertinent to the outer sphere effects on catalysis will be discussed. Table 3 summarizes some of the available assay data for NifDK variant proteins.

Table 3.

Reactivity of Select Purified NifDK Variants^a

NifDK protein type	NH3b (N2)	H2b (N2)	C2H4c (C2H2)	C2H6c (C2H2)	H2c (C2H2)	H2d (Ar)	NH3e (N2/CO)	H2e (N2/CO)	C2H4f (C2H2/CO)	C2H6f (C2H2/CO)	H2f (C2H2/CO)	H2g (Ar/CO)	ref
Av wild-type	860	615	1706	0	276	2494	0	2727	1134	0	959	2841	185
α-V70A	800					1800							176
α-V70I	170	1940	130	0	2000	2340							177
α-R96K	373		804			972							154
α-Q191K	0	514	40	7.2	528	540	0	182	14	2.3	340	254	185
α-H195Q	-	-	1350	0	1115	2860							183
α-H195N	-	-	535	135	355	1470							183
α-R277C						1400						2400h	186
α-R359K	709		1523			1587							154
α-F381L	816		1656			1844							154
Av Wild-type	989	555	2096			2211							158
α-W444Y	1036	507	2066			2069							158
α-W444F	906	505	1906			1795							158
α-W444A	212	101	364			410							158
α-W444G	0	6	8			11							158
Av ΔnifV						840							181
Kp Wild-type	725	471	1127	-	-	1527							196
Kp ΔnifV	318	842	1044	-	-	1242							196
Kp ΔnifV	-	812	-		202	1369	-	574	-	-	585	718	197

^aSpecific activity is reported with units of nmol product × min⁻¹ × (mg protein)⁻¹. Gas in parentheses reflects the substrate conditions used for the assay.

^bAssays are reported in an atmosphere of 100% N₂.

^cAssays reported in an atmosphere of 10% C₂H₂, 90% Ar.

^dAssays reported in an atmosphere of 100% Ar.

^eAssays are reported in an atmosphere of 3% CO, 97% N₂.

^fAssays are reported in an atmosphere of 3% CO, 10% C₂H₂, 87% Ar.

^gAssays are reported in an atmosphere of 3% CO, 97% Ar.

^hAssays are reported in an atmosphere of 10% CO, 90% Ar.

In the 1990s, NifDK variants at the α-Q191 (+, +, +) and α-H195 (−, 0, +) residues were reported based on sequence conservation among NifDK analogs, and between NifDK and the biosynthetic protein NifEN.^{182,185,187,188} The cells of variant α-Q191K and α-H195N NifDK expressing strains had a Nif⁻ phenotype where the cells did not appear to grow without the addition of a nitrogen source (such as ammonium chloride or urea).¹⁸² This indicated that these NifDK variants were unable to reduce N₂ under standard conditions. Crude extracts of cells expressing α-Q191K and α-H195N NifDK variants were found to yield no N₂ fixation products, however, reaction with C₂H₂ produced C₂H₄ with <10% of the activity of the wild-type enzyme. Furthermore, ethane (C₂H₆) was found as a product from the variant protein strains, with 12.5% and 35% of the electron equivalents diverting toward ethane production for the α-Q191K and α-H195N NifDK proteins, respectively.¹⁸² This was an interesting result at the time because it demonstrated that the M-cluster site was likely nearby the targeted residues prior to the publication of crystallographic data.

The α-Q191K NifDK protein was purified in 1992 and the reactivity of the enzyme was found to be ~30% of the activity of the wild-type NifDK. It produced no N₂ reduction products, and diverted 90% of electron equivalents to the production of H₂ in the presence of C₂H_2, compared to wild-type enzyme that uses 90% of electron flux to generate C₂H₄.¹⁸⁵ Subsequent characterization has shown that the Michaelis constant (K_m) for C₂H₂ to produce C₂H₄ is 35 kPa, 2 orders of magnitude higher than that for the wild-type enzyme (K_m = 0.5 kPa), indicating that acetylene is not a great substrate for α-Q191K NifDK.¹⁸³ It has been further demonstrated that α-Q191K NifDK does not react with C₂H₄ to produce C₂H₆, in contrast to wild-type NifDK that will produce only a small amount (<1% of electron flux) of C₂H₆ from C₂H₄ when C₂H₂ is absent. This and isotope labeling experiments led to the assertion that there must be a longer-lived, common ethylenic intermediate that can be differentiated to release ethylene, or ethane.¹⁸³ The lack of N₂ reduction was also explored through the introduction of N₂ gas into C₂H₂ and proton reduction assays.¹³⁹ Under standard conditions, wild-type NifDK will divert electron flux from proton reduction to N₂ reduction if N₂ is introduced into the reaction vessel, indicating that N₂ binds to or otherwise interacts with the M-cluster. HD will also be observed as a product in the presence of both N₂ and D₂. In contrast, α-Q191K NifDK does not appear to even bind N₂ because the catalytic performance is unaffected when N₂ is added into the typical acetylene or proton reduction assays.¹³⁹ When mixtures of N₂ and D₂ are added to reactions of α-Q191K NifDK, the product HD is absent, consistent with a lack of N₂ binding as well as an inability to access the E₄ state. A rationalization provided for these observations is that the α-Q191K variant is unable to even reach the E₃ level of the Lowe-Thorneley cycle (Figure 6).¹³⁹ From crystallography, it is known that α-Q191 is in proximity to the R-homocitrate ligand, and both the protein residue and the ligand are involved in hydrogen-bonding interactions,^65,85,151 potentially causing deleterious modulation of the redox properties of the M-cluster. Interestingly, when R-homocitrate is exchanged for citrate in a Δnif V variant of NifDK from K. pneumoniae (see below), N₂ reduction decreases to 40% of the wild-type activity, diverting more electron flux toward the production of H₂ that exceeds that for the wild-type NifDK; yet, the Δnif V NifDK variant is still capable of binding N₂.¹⁹⁶ Clearly, α-Q191 plays a key role in regulating the energetics in the active site, but the mechanism by which this is accomplished remains elusive.

In parallel, α-H195 (−, 0, +) variants have also been purified and studied.^{139,183,187,188} The initial report of α-H195N showed that the enzyme was relatively unstable, did not have a high reactivity toward C₂H₂ and proton reduction compared to wild-type, and was unable to form N₂ reduction products.^182,185 EPR spectra of whole cells indicated that while α-H195N NifDK showed a low intensity S = 3/2 M-cluster signal, the α-H195Q variant had an EPR signal of comparable intensity to wild-type enzyme with some perturbations of the g-values, suggesting that the electronic properties of the cluster had not substantially changed.¹⁸⁷ The α-H195Q variant showed a modest decrease in activity toward C₂H₂ compared to wild-type enzyme, but there was no observable products of N₂ reduction, even under acid-quenched conditions that would yield N₂H₄ in wild-type NifDK.^198,199 However, in the presence of N₂, the α-H195Q protein proton reduction was reversibly inhibited by slowing the reaction rate, but did not divert electron flux toward other observable products.¹⁸⁷ This indicated that N₂ would bind to the M-cluster of the variant, but would not become reduced. To further explore this notion, Fisher and co-workers increased the pressure of N₂ used in reduction assays of α-H195Q NifDK and saw that under hyperbaric conditions, NH₃/NH₄⁺ was observed at a rate <2% of the wild-type enzyme.¹⁸⁸ Additionally, in the presence of a mixture of N₂ and D₂, α-H195Q NifDK produced measurable HD gas, indicating that the protein variant could access the same E₄ state as the wild-type enzyme, but could just not reduce N₂ well. A crystal structure of the α-H195E NifDK protein was later published, showing that in agreement with EPR spectra, the N–H–S2B hydrogen bond stayed intact in the α-H195Q variant like that in the wild-type protein.¹⁸⁴ Thus, the hydrogen bond does not sufficiently control N₂ reduction capabilities, and additional factors, such as proton donors and redox potentials, should be considered. The α-H195N variant was also tested for the ability to bind N₂, and it was found that N₂ would inhibit proton reduction much like in the α-H195Q variant, but no HD evolution was measured. This suggests that the α-H195N variant could bind N₂ (that is, achieve the E₃ state) but was unable to reduce N₂ (that is, unable to access the E₄ state).¹³⁹ In this case, it is unlikely that any hydrogen-bonding interaction between α-N195 and the M-cluster exists, but the protein is still able to bind, but not reduce, N₂. In comparison, α-Q191K NifDK likely maintains the hydrogen bond between α-H195 and S2B of the cofactor, but the protein variant is unable to bind N₂ at all.

Acetylene reduction was also investigated with the variant α-H195 proteins, and it was found that N₂ would inhibit the production of both C₂H₄ and C₂H₆, but this inhibition could be reversed with H₂.^139,183,188 In the presence of 10% acetylene, the α-H195N and α-H195Q proteins used 69% and 55% of electron equivalents to generate reduced products, but the former protein yields both C₂H₄ and C₂H₆ while the latter yields C₂H₄ exclusively.¹⁸³ The α-H195Q protein behaves similarly to wild-type enzyme, with a K_m for C₂H₂ of 0.5 kPa for both enzymes, and neither producing C₂H₆ from C₂H₂; however, the α-H195Q NifDK variant is less efficient than the wild-type enzyme in C₂H₂ reduction. In comparison, the α-H195N variant has an affinity for C₂H₂ (K_m = 1.0 kPa) similar to that of the wild-type enzyme but produces ethane which accounts for 23% of the total electron flux. Ethylene was also tested as a substrate for the variant proteins in comparison to the wild-type enzyme, with a K_m value of 120 kPa measured for both wild-type and α-H195Q proteins (though the latter uses less than 1% of total electron flux toward ethane production) and a K_m value of 48 kPa measured for the α-H195N variant (alongside a 10-fold increase in activity). These experiments revealed that the high affinity for acetylene in the wild-type enzyme and α-H195Q variants likely prevents ethane production, because once ethylene is produced, it is rapidly replaced by a new molecule of acetylene.¹⁸³ The slightly lowered affinity for the substrate in α-H195N NifDK may explain why ethane can be observed as a product, but it is possible that the steric environment also contributes, as asparagine has a shorter side chain than glutamine. These acetylene reduction experiments, in comparison to N₂ reduction, illustrate that more rigorous conditions are required for the cleavage of the triple bond of N₂ compared to alkyne or alkane substrates.

To better identify and understand substrate binding sites on the M-cluster, additional mutations of nitrogenase were studied.^{174–178,180} In wild-type NifDK, several observations have been made with respect to the competition between N₂ and acetylene binding to the catalytic cofactor; specifically, it has been shown that two binding sites with different affinities for acetylene exist, and that acetylene is a noncompetitive inhibitor of N₂ reduction while N₂ is a competitive inhibitor of acetylene reduction.^{193,200–203} Additionally, C₂H₂ has also been found to be a more potent inhibitor of N₂ reduction under low electron flux conditions, where the rate of electron delivery is decreased, compared to those under high flux conditions.²⁰⁴ In 2000, a study was reported that used an Azotobacter vinelandii strain expressing a NifDK variant with a β-Y98H (+, −, +) mutation, which was shown to have a lower electron flux, as a progenitor for a genetic study.¹⁷⁴ The β-Y98H variant would not grow well under diazotrophic conditions in the presence of 2.5% C₂H₂, because N₂ reduction by the low flux NifDK protein would be strongly inhibited by C₂H₂. The goal was to screen for a mutation that would remove the acetylene sensitivity while maintaining similar levels of N₂ reduction activity. After analysis, it was found that in addition to the β-Y98H mutation, an α-G69S (+, −, +) mutation would allow for cells to grow under N₂ fixing condition in the presence of C₂H₂. A separate protein was then generated that only possessed the α-G69S mutation. The Michaelis–Menten parameters of the α-G69S NifDK protein for C₂H₂, N₂, and proton reduction were similar to those of the wild-type enzyme, indicating that this mutation did not affect nitrogenase activity. In contrast, the affinity for C₂H₂ decreased for α-G69S NifDK (K_m = 14.2 kPa) compared to wild-type enzyme (K_m = ~0.71 kPa), and the variant protein now showed a competitive inhibition of N₂ reduction in the presence of C₂H₂, instead of a noncompetitive inhibition.¹⁷⁴ The rationale for these observations was that the mutation of α-G69 removed the high-affinity C₂H₂ binding site through steric changes nearby the M-cluster. However, the authors proposed that α-V70 was the actual residue that blocked the substrate binding site, due to the closer proximity to the cofactor. They also identified that α-R96 is positioned nearby α-V70 and is likely to be involved in determining access to the binding site. A separate study concluded that using the α-G69S NifDK protein, two substrate binding sites could be better distinguished on the α-V70 (0, −, +) face of the cluster.¹⁷⁵ One site is denoted as a “high-affinity site” for acetylene binding that may also accommodate CO but not N₂, N₂O, and N₃⁻. The other site is designated the “low-affinity” acetylene-binding site that also binds CO, N₂, N₂O, and N₃⁻.

Substitutions of the α-R96 (+, −, 0) and α-V70 (0, −, +) positions in NifDK were then explored for their effects on the access and binding of substrates and inhibitors.^176,180 In 2001, multiple mutations of α-R96 were reported, and it was shown that the EPR spectra of all of the NifDK variants were essentially the same as the spectrum of the wild-type enzyme.¹⁸⁰ This result was used to conclude that the substitution had little to no effect on the electron properties of the M-cluster, and as such, the authors selected the α-R96L NifDK variant as a representative example to further study. When high concentrations of C₂H₂ were added to the protein, a new EPR signal appeared in a 70% acetylene atmosphere with g-values of 4.5, 3.5, and <2.0, constituting 40% of the total spin (Figure 11). The addition of N₂ or CO did not affect the appearance of this signal, which is totally absent in the wild-type enzyme. It was found that addition of CN⁻ to α-R96L NifDK would yield a new paramagnetic species with a g-value at 4.5, that also was absent from the wild-type system. Pulsed electron nuclear double resonance (ENDOR) experiments with ¹³C-enriched acetylene and cyanide showed that CN⁻ definitively binds to the M-cluster, while it was inconclusive if C₂H₂ did as well, but in both cases the new EPR signals were derived from the M- as opposed to the P-cluster.¹⁸⁰ These results led to the conclusion that α-R96 must play a steric “gatekeeping” role during catalysis. The change in size of the α-R96 residue presumably opened the active site allowing the small molecules to bind on or near the M-cluster. However, this finding should be taken with some scrutiny: the concentrations required to affect a change in the EPR spectrum are far beyond what is required under usual catalytic conditions, and other molecules, such as N₂, CO, and N₃⁻ do not cause perturbations in the EPR spectrum of α-R96L NifDK. Interestingly, the α-R96 variants do not product ethane, in contrast to α-Q191 and α-H195 variants.²⁰⁵ The α-R96 variants do all have different affinities and rates for the reduction of C₂H₂, some similar to wild-type enzyme while α-R96L NifDK has a 3-fold higher affinity, so it is unclear to what extent physiologically relevant conclusions can be made.

Figure 11.

EPR spectra of α-R96L NifDK under nonturnover conditions with acetylene. The α-R96L NifDK protein was prepared under 1 atm of Ar (top spectrum, black) or under 1 atm of acetylene (bottom spectrum, red). Reprinted (adapted or reprinted in part) with permission from ref 180. Copyright 2001 American Chemical Society. Data digitalized via WebPlotDigitizer v4.5.

In 2002, a follow up to the mutagenesis study of α-G69 NifDK was reported with the substitution of α-V70 (0, −, +) for alanine.¹⁷⁶ The α-V70A NifDK protein was able to grow under diazotrophic conditions, indicating that N₂ reduction was unaffected by the substitution. The active site access of the variant compared to wild-type enzyme was assessed through the introduction of the water-soluble acetylene derivate propargyl alcohol (HC≡C–CH₂–OH). When added to the growth media, the wild-type enzyme was unaffected while the α-V70A protein was unable to grow, presumably due to a larger substrate pocket that allows the larger alcohol in. Propargyl alcohol was a potent inhibitor of α-V70A NifDK with respect to proton and N₂ reduction, but direct reduction of propargyl alcohol resulted in propene (H₂C = CH–CH₃; 4-electron product) as a minor product, as well as allyl alcohol (HC = C–CH₂–OH; 2-electron product) as the major product. Propyne (HC≡C–CH₃) was also found to serve as a substrate for α-V70A NifDK, producing the 2-electron product propene, in contrast to the wild-type enzyme where propyne is a poor substrate.^55,59 Additionally, α-V70A NifDK was unable to produce ethane from the reduction of acetylene.¹⁷⁶ Dean, Seefeldt, and co-workers then published a study of the NifDK protein, using the smaller (α-V70A) and larger (α-V70I) side chain variants to study the effect the substitutions at α-V70 had on N₂ reduction.¹⁷⁷ The ability of the α-V70I variant to reduce N₂ was decreased by ~70% compared to wild-type enzyme while electron flux was diverted toward proton reduction, with a concomitant increase in the K_m values for N₂ reduction from K_m = 0.1 atm to >1.5 atm. Acetylene reduction was analogously affected with a substantial decrease in C₂H₄ production, and an increase to both proton reduction and the K_m value compared to wild-type NifDK. These results were rationalized by stating that the longer side chain of isoleucine must impinge on the “catalytic face” of the M-cluster where N₂ and C₂H₂ bind, including Fe2, Fe3, Fe6, and Fe7 from the crystallographic labeling scheme (Figure 7). Substitution of glycine at position α−70 resulted in a nitrogenase protein (α-V70G) that is unable to react with N₂ at all, but maintains activity with acetylene and larger carbon-based substrates.¹⁷⁸ The lack of N₂ reduction with the smallest side chain variant was attributed to the inability to access higher redox states of the M-cluster, but this hypothesis was not further tested. A double mutation consisting of α-V70A and α-Q191A was also generated, and the enlarged substrate pocket allowed for longer chain alkynes, such as 2-butyne (H₃C–C≡C–CH₃), to be reduced near the proposed reactive site at Fe6 on the M-cluster.¹⁷⁸ Undoubtedly, the second coordination sphere environment is important for determining substrate access through steric effects, but with the drastic changes that sometimes occur “globally” in variant proteins, it is logical to be cautious in extending insights gained from the variants to the native system. The α-V70I and α-V70A NifDK proteins have been used to trap and study a variety of intermediate states with a battery of spectroscopic and computational techniques, the like of which have been described at length elsewhere^52,60,206 and will not be further discussed in this review.

Recently, a crystal structure was published by Ribbe, Hu, and co-workers that provides some structural insight into substrate binding on wild-type NifDK.⁹¹ Previous attempts at trapping intermediates of nitrogenase have employed single-point mutations in combination with rapid freeze-quench techniques that can capture transient species.^52,60,206 The strategy used in the recent crystallographic study involved isolation of NifDK in an anaerobic environment that did not contain the exogenous reductant dithionite (S₂O₄²).⁹¹ Previous work has shown that when dithionite and other sulfur sources are stringently excluded from in vitro maturation assays, an M-cluster precursor called the L*-cluster ([8Fe-8S–C]) can be isolated on NifB that lacks a μ₂-sulfide ligand that it otherwise binds in the presence of dithionite.¹³³ This sulfur-deficient cluster cannot be matured into the active M-cluster unless a source of SO₃²⁻ is added, which is usually provided by the decomposition of dithionite. Drawing from this inspiration, it was reported that when NifDK was anaerobically purified under a N₂ atmosphere in the absence of dithionite (denoted NifDK*), the resultant protein was initially unreactive toward substrate reduction.⁹¹ However, reactivity could be restored upon the addition of dithionite. When cells expressing NifDK* were grown in the presence of¹⁵N₂, the subsequent protein was reported to have¹⁵N₂ bound, as acid quenching samples of NifDK* yielded labeled N₂ by gas chromatography mass spectrometry (GC-MS) analysis. The labeled N₂ ligand was not observed when dithionite was included as part of the purification protocol.

Isolated NifDK* was then crystallized, and a structure was solved that contained M-clusters with additional electron density between central Fe ions of the cluster (Figure 12).⁹¹ Not only that, the two M-cluster sites within the heterotetrametric NifDK each had different sets of electron density; at ‘Site 1’, this density appeared in the usual place of S2B (0, 0, +), and at ‘Site 2’, density was observed in the location of S3A (0, +, −) and S5A (0, −, −). On the basis of anomalous sulfur density, it was concluded that Site 1 lacked a sulfur atom at the S2B position, and Site 2 lacked sulfur atoms at both S3A and S5A. During refinement, best fits of the data resulted in N₂ molecules displacing the respective sulfur atoms at the two different M-cluster sites. At Site 1 (Figure 12A), the N₂ unit appears in a pseudo μ_1,2-bridging mode with N6A sitting 1.8 Å from Fe2 and N6B being 2.3 Å from Fe6. N6A is also within 2.9 Å of α-H195 (−, 0, +), indicative of a potential hydrogen-bonding interaction. A stabilizing interaction between the putative N₂ unit and α-H195 would be consistent with observations from the mutagenesis studies described in this review.^{139,182–185,187} At Site 2 (Figure 12B), the N₂ units are bound in asymmetric μ_1,1-bridging modes where the proximal N atom is within 1.8 Å of one Fe ion, and 2.1 Å from the other Fe (with the Fe pairs Fe4/Fe5 for the “S3A position” and Fe7/Fe3 for the “S5A position”, respectively).⁹¹ The distal N atom at the S3A position is 2.9–3.4 Å from the backbone amide groups of α-G356 (+, +, 0) and α-G357, and at the S5A position, the distal N is within 3.2 Å of the side chain of α-R96 (+, −, 0) and a water molecule, also suggesting potential hydrogen-bonding interactions. These positions have not been previously proposed as substrate binding sites, which is an interesting finding, although in crystal structures of V-nitrogenase, an unusual carbonate (CO₃²⁻) ligand is found to displace the equivalent S3A atom.^92–95 Additionally, in both Site 1 and 2 of sulfur-deficient NifDK*, the R-homocitrate ligand appears to be bound to the Mo center through one O atom exclusively; at Site 1 the O atom from the −OH group is 2.7 Å from Mo and the −COO⁻ group is closer, while at Site 2 the opposite is observed.⁹¹ This could potentially indicate protonation, which may elongate the refined Mo–O distances, consistent with the proposed role R-homocitrate may play in proton transfer, but additional evidence is required to validate the protonation states. Strikingly, when NifDK* is put under turnover conditions in the presence of dithionite and then recrystallized (PDB ID 6VXT), the cofactor returns to the normal resting state of NifDK, indicating that the displacement of sulfur is not deleterious but rather reversible.⁹¹ The resolution of the crystallographic data for NifDK* does not allow for the absolute determination of the N₂ unit as a reduced species or otherwise;^207–209 however, the assignment of the N₂ unit is supported by available structural data, and it would further validate the importance of residues such as α-H195 and α-R96, as well as the “catalytic face” of the M-cluster, while putting forward sulfur displacement as a critical mechanistic necessity for substrate reduction by Mo-nitrogenase. A recent study by Hu, Ribbe and coworkers⁷²³ confirmed binding of N₂ to NifDK* in a catalytically competent conformation, showing formation of C₂H₃D in the C₂H₂ reduction assay (an equivalent to HD formation in the absence of C₂H₂) of NifDK* where only D₂ was supplied instead of a D₂/N₂ mixture, appearance of new S = 1/2 features in the EPR spectrum of NifDK* concomitant with a decreased intensity of the M-cluster-associated S = 3/2 signal, and release of¹⁵NH₃/NH₄⁺ upon turnover of the¹⁵N₂-bound NifDK*. In addition, this report revealed that product release only occurred via displacement by sulfite-derived belt-sulfide, and that catalysis involved all belt sulfur locations that were shown to be labeled with selenite-derived selenides in a XAS pulse chase experiment. These observations provide support to a previously proposed mechanism involving a stepwise reduction of N₂ at position S3A, S2B, and S5A that occurs asynchronously in the two αβ-dimers of NifDK.⁹¹ Further crystallographic, biochemical, spectroscopic, and computational work will be required to explore and validate these unique findings.

Figure 12.

Crystal structure of NifDK*. The M-cluster and key interacting residues at Site 1 (a) and Site 2 (b) are depicted (PDB ID 6UGO). The cluster and N molecules are presented as ball-and-stick models while the residues are shown as sticks. The M-clusters are superimposed with the F₀ − F_c omit map of the N₂ ligands contoured at 10σ (mesh). Atomic coloring: Fe, orange; S, yellow; C, gray; O, red; N, blue. Reprinted (adapted or reprinted in part) with permission from ref 91. Copyright 2020 American Association for the Advancement of Science.

2.5.2. Inhibition by and Reduction of Carbon Monoxide.

Carbon monoxide has long been characterized as an inhibitor of nitrogenase reactivity in the Mo-dependent systems that affects the reduction of all substrates apart from protons.^{52,100,200,210} Mo-nitrogenase has multiple CO-related states that have been spectroscopically characterized under different conditions and concentrations of CO.^52,62 Under low electron flux conditions (1:5 Fe:MoFe protein) and in the presence of low concentration of CO (<1% in argon), the typical S = 3/2 EPR signal associated with the M-cluster converts to an S = 1/2 signal with g values of 2.09, 1.97, and 1.93 (designated the “lo-CO state”), whereas under high concentrations of CO (>50% in argon) a new S = 1/2 signal arises (designated the “hi-CO state”) with g = 2.17 and 2.06 (Figure 13).184,202,203,211–213 Under higher electron flux conditions (1:1 Fe:MoFe protein) and high concentrations of CO, a third state known as “hi(5)-CO” was observed in wild-type NifDK, with g values of its EPR signature at 5.78, 5.15, and 2.7 (Figure 13).^181,184,211 A low intensity EPR signal was observed for VnfDGK in the presence of 1 atm of CO with g values at 2.09, 1.99, and 1.91, and the intensity of this signal could be increased with more potent reducing agents.^214,215 By comparison to NifDK, the EPR signal in VnfDGK was assigned to the lo-CO form of the enzyme. In the presence of strong Europium-based reducing agents and 2.6 atm of CO, a new set of EPR signals were observed that could be described as the superposition of the lo-CO state, and a newly identified species with g-values at 2.13, 2.01, and 1.97.^215,216 Although not identical, this new signal was assumed to be analogous to the hi-CO state of NifDK. No equivalent of the hi(5)-CO state has been reported for the VFe protein.

Figure 13.

EPR characterization of A. vinelandii NifDK during enzymatic turnover in the presence of CO. (a) α-H195Q NifDK in the presence of 1.0 atm CO (top, black) and 0.001 atm CO (bottom, red) showing the hi-CO and lo-CO signals, respectively. The small inflections are associated with the S = 3/2 signal of the resting state of the M-cluster. (b) EPR spectra of α-H195Q NifDK (top, black) and wild-type NifDK (bottom, blue) during turnover with 1 atm of CO, with the three lowest inflections (1)–(3) of the hi(5)-CO signal. The CO-related spectra of α-H195Q and wild-type NifDK are nearly identical. Reprinted (adapted or reprinted in part) with permission from ref 184. Copyright 2001 American Chemical Society. Data digitalized via WebPlotDigitizer v4.5.

As the crystallographic characterization of CO-bound states was not reported until 2014,⁸⁹ many of the earlier studies employed competitive substrate reduction experiments, where separate N₂, C₂H₂, and proton reduction assays were compared in the presence and absence of CO.^51,52 The behavior of nitrogenase in the presence of multiple substrates could then be analyzed and extrapolated to provide insight into CO-bound states of the M-cluster. Additionally, a series of spectroscopic and computational studies revealed EPR signals associated with CO binding to Mo-nitrogenase. As described above, an S = 1/2 spin species trapped under low concentrations of CO was designated “lo-CO”, and at higher CO concentrations, another S = 1/2 signal would appear that was assigned “hi-CO”.^{202,211,217–221} When the initial characterization of V-dependent nitrogenase was undertaken, it was found that CO inhibited reactivity in a similar manner to Mo-nitrogenase, so it was reasonable to anticipate analogous EPR signals in V-nitrogenase. Hales and co-workers constructed a hybrid nitrogenase that used the polypeptide of V-nitrogenase from A. vinelandii and the M-cluster from the A. vindelandii Mo-nitrogenase (hereafter labeled M-VFe protein).²²² It was found that the hybrid M-VFe protein had an activity profile more like the native V-nitrogenase. However, in the presence of CO, no EPR signals analogous to the lo-CO or hi-CO signals of Mo-nitrogenase could be observed, although the M-VFe protein carried the same cofactor as the native Mo-nitrogenase. This result was puzzling, but was a strong indication that second and outer coordination sphere effects played a pivotal role in determining the properties of nitrogenase.^51,222 Later, it was shown by Ribbe and co-workers that Av V-nitrogenase (and to a lesser degree, wild-type Av Mo-nitrogenase) could catalyze the coupling of gaseous CO into longer chain hydrocarbon products following Fischer–Tropsch-like chemistry, but with an observed decrease of ~75% in the specific activity for proton reduction by the V-nitrogenase.^223–225 This finding further expanded the known capabilities of the Mo- and V-nitrogenases, but still lacked a comprehensive structural understanding of CO activity.

With the advent of crystallographically characterized, CO-bound intermediates, insight into structural features of CO binding to nitrogenase could be obtained.^89,94,95,226 In 2014, Spatzal, Rees, and co-workers published the first structure of CO binding to the M-cluster of NifDK (designated NifDK-CO), which corresponded to an inhibited state of the enzyme (Figure 14A).⁸⁹ The CO molecule is ligated between Fe2 and Fe6 of the M-cluster in a μ₂-binding mode, apparently displacing the μ₂-S2B sulfur atom that is usually found in resting state structures of NifDK.^61,66,151 The CO ligand is bound symmetrically between the two Fe ions, with Fe–CO distances of 1.86 Å to both metal centers, and the oxygen atom of the CO moiety within close proximity to the side chains of residues α-H195 (2.8 Å) and α-V70 (3.4 Å). The interaction between α-H195 and CO is within the range of a hydrogen bond, which would affect the stability of the CO moiety, while nonbonding interactions from α-V70 may provide stability. When crystals of NifDK-CO were dissolved into buffer in the absence of exogenous CO, the S2B atom and acetylene reduction activity were quantitatively recovered.⁸⁹ EPR spectra of NifDK-CO in the solution state, obtained under analogous conditions to crystallography, demonstrated the appearance of an S = 1/2 signal previously assigned to the lo-CO species, connecting observation derived from spectroscopy and structural biology.²¹⁴

Figure 14.

Crystal structure of NifDK-CO (a) and NifDK-(2CO) (b) from A. vinelandii. The cluster atoms are shown as ball-and-stick models, the CO, R-homoctirate (HC), and amino acid residues are presented as sticks. Black dotted lines represent a bonding interaction, blue dotted lines represent potential hydrogen bonding interactions. Atomic coloring: Fe, orange; S, yellow; C, gray; O, red; N, blue. PDB IDs: NifDK-CO, 4TKV; NifDH-(2CO), 7JRF.

Recently, Spatzal, Rees, and co-workers reported another CO-bound crystal structure of NifDK (Figure 14B), but with two CO ligands (designated NifDK-(2CO)) as opposed to one.²²⁶ This species was isolated when crystals of NifDK-CO were pressurized with CO at 5.4 atm. The first CO ligand (“μCO”) is found in a bridging position between Fe2 and Fe6 of the M-cluster in a μ₂-binding mode analogous to that in NifDK-CO, with Fe–CO distances of 1.93 and 1.92 Å, respectively. The second CO ligand (“tCO”) is bound in a terminal mode to Fe6, with an Fe–CO distance of 2.03 Å. Binding of tCO causes the Fe6–C distance to slightly elongate from 2.01 Å in NifDK-CO to 2.06 Å in NifDK-(2CO).^89,226 The oxygen atom of tCO may interact with the amide N of the α-Q191 residue (~3.3 Å) and a carboxylic acid group of R-homocitrate (~2.7 Å), while μCO maintains analogous interactions with the side chains of α-V70 and α-H195, in a similar manner as in NifDK-CO. Compared to μCO that is modeled with 100% occupancy, tCO is modeled with only 50% occupancy, likely because the binding of this ligand is relatively weak. Parallel EPR spectra of the solution state and crystal slurry of NifDK-(2CO) were subsequently used to assign μCO and tCO to the lo- and hi-CO states, respectively.²²⁶

Analogous crystal structures were solved by Einsle and co-workers of the V-dependent nitrogenase with one (VnfDGK-CO) and two (VnfDGK-(2CO) CO molecules bound (Figure 15).^94,95 In the 1.0-Å resolution structure of VnfDGK-CO, the CO molecule is bound in a μ₂-bridging mode between Fe2 and Fe6, much like that found in the lo-CO state of NifDK-CO, but with a slight asymmetry (Fe2–CO = 2.03 Å, Fe6–CO = 1.94 Å).⁹⁴ The μCO ligand has similar interactions with local amino acid side chains, being within 2.9 Å of α-H180 and 3.6 Å of α-V57 (analogous to α-H195 and α-V70 in NifDK), respectively. Unlike NifDK-CO, when VnfDGK-CO crystals were used for activity assays in the absence of exogenous CO, the typical S2B ligand was not found between Fe2 and Fe6. Instead, a monatomic ligand interacting with the nearby α-Q176 residue was observed and assigned as a putative μ₂–OH species. The authors proposed that this oxygen-based ligand might result from reduction of a previously bound CO ligand.⁹⁴ When crystals of VnfDGK-CO were pressurized with 1.5 atm of CO for 1 min, a 1.05 Å crystal structure of the VnfDGK-(2CO) was obtained (Figure 15B).⁹⁵ The μCO ligand in VnfDGK-(2CO) is positioned in the same manner as that in VnfDGK-CO, and the terminal CO ligand is bound to Fe6 with an occupancy of 50%, analogous to that observed in the structure of NifDK-(2CO); however, the tCO in VnfDGK-(2CO) has an Fe6–tCO distance of 1.89 Å, compared to 2.03 Å in NifDK.^95,226 The oxygen atom of tCO in VnfDGK-(2CO) is also in close proximity to the R-homocitrate ligand (~2.9 Å) and the α−176Gln residue (~3.1 Å). Despite the high degree of similarity in active site for the two structures, the tCO ligand appears to be more strongly bound in VnfDGK-(2CO) versus NifDK-(2CO), which may align with the difference in reactivity observed for both systems.⁵¹ However, there are strong indications that the lo-CO and not the hi-CO state of V-nitrogenase is catalytically competent for CO reduction.^215,216 A parallel set of EPR measurements using solution and crystal slurry samples of VnfDGK-(2CO) was not reported by Einsle and co-workers,⁹⁵ so the crystallography could not be directly connected to the available spectroscopic characterization of the lo- and hi-CO states. Instead, in situ Fourier-transform infrared (FTIR) difference spectroscopy was conducted using a film of VnfDGK in the presence of N₂ and CO isotopes.⁹⁵ Supporting the crystallographic assignment, isotopically sensitive features at 1931 and 1888 cm^–1 were assigned to t¹²CO and t¹³CO, respectively. Because of strong background signals at lower frequencies, a vibration for the μ¹²CO could not be assigned unambiguously, but a putative signal was identified at 1720 cm^–1. More work will be needed in the future to fully understand the structure–function relationships gleaned from these structural studies.

Figure 15.

Crystal structure of VnfDGK-CO (a) and VnfDGK-(2CO) (b) from A. vinelandii. The cluster atoms are shown as ball-and-stick models, the CO, R-homoctirate (HC), carbonate, and amino acid residues are presented as sticks. Black dotted lines represent a bonding interaction, blue dotted lines represent potential hydrogen bonding interactions. Atomic coloring: Fe, orange; S, yellow; C, gray; O, red; N, blue. PDB IDs: VnfDGK-CO, 7ADR; VnfGDH-(2CO), 7AIZ.

On the basis of the crystallographic information, what can be said is that α-Q191, α-H195, and α-V70 in NifDK (and equivalent analogs in VnfDGK) are residues that can interact with the catalytic cofactor when CO molecules are bound. The α-H195 residue is in hydrogen-bonding distance to the μCO ligand and has been proposed to be generally involved in proton delivery during substrate turnover.^227–229 The α-Q191 residue is within an appropriate distance from the tCO ligand to potentially interact with this ligand; additionally it helps stabilize the R-homocitrate ligand that binds to the Mo or V ion of the catalytic cofactor.^95,226 As stated previously, the hi-CO state does not appear to be competent for CO catalysis,^215,216 thus binding to tCO may not be relevant, although the equivalent residue in VnfDGK, α-Q176, has been shown to stabilize a putative N₂ reduction intermediate through a hydrogen-bonding interaction, which further reinforces the importance of this residue.⁹³ These details regarding CO-bound species provide a measure of perspective that was previously inaccessible and allows for a retrospective look on experiments conducted to assess the structural basis of CO inhibition within nitrogenase.

2.5.2.1. Influence of Point Mutations on the Reaction with CO.

The research groups of Dean and Newton have generated a multitude of single-point mutants of NifDK that have been used to study substrate interactions with the enzyme, including CO. Many of these studies yield overlapping information regarding the properties of NifDK variants, including insights gleaned from experiments with crude extracts of A. vinelandii cells and purified variant proteins, so it can be difficult to comprehensively describe the results. Instead of addressing every possible variant reported, the effects of mutations to the α-H195 and α-Q191 residues on CO reactivity will be the primary focus of our discussion below, as well as relevant comparisons to other variants, as necessary.

As described earlier, the site-directed mutagenesis in NifDK of α-H195 (−, 0, +) and α-Q191 (+, +, +) to α-H195N and α-Q191K were reported for A. vinelandii.¹⁸² The crude extracts of variant cells were individually assayed in an atmosphere of 10% acetylene (C₂H₂), demonstrating that the variant NifDK proteins were much less active than wild-type enzyme for the generation of C₂H₄, but they also produced some amount of ethane (C₂H₆), whereas wild-type NifDK does not. This raised interest in these variants, as crystallographic evidence would not be available for several years, and the pattern of reactivity was similar to that found for V-nitrogenases.⁵¹ When C₂H₂ reduction assays were carried out, the α-H195N variant was less sensitive to inhibition by CO, showing 50% activity in the presence of 0.5% CO, compared to wild-type NifDK that showed the same decrease in activity already at 0.02% CO. It is important to note that these effects were observed with only the crude extracts, as the protein could not be purified at that time. EPR spectra of whole cells indicated that the S = 3/2 M-cluster signal was still present in the variant, albeit with axial perturbations, and at a low intensity (~6% compared to wild-type enzyme).^182,185 Several years later, five additional variants of α-H195 were reported.¹⁸⁷ Similar to the previous reports of α-H195N,^182,185 the crude extracts of the tyrosine, threonine, and glycine variants were not particularly active, but the glutamine and leucine variants expressed slightly higher activity, with α-H195Q NifDK maintaining ~70% the activity of crude extracts of wild-type enzyme.¹⁸⁷ Whole cell EPR of all variants had similar line shapes to wild-type enzyme, but the glycine and leucine variants had much lower intensities of the S = 3/2 signal (≤10%), whereas α-H195Q NifDK had comparable intensity to native enzyme, with slight perturbations of the g-values. The proton reduction activities in the presence and absence of CO were assayed, and it was demonstrated that the asparagine, glutamine, and leucine variants were relatively insensitive to CO, but α-H195T NifDK was sensitive to the presence of CO, with ~50% reduction of activity. The α-H195Q NifDK protein was further purified for more accurate comparisons to wild-type enzyme. It was found that CO was a noncompetitive inhibitor of C₂H₂ and N₂ reduction for both proteins, but CO had a more potent inhibition of α-H195Q NifDK compared to wild-type enzyme. The authors suggested that α-H195 could be involved in steric interactions, as the much smaller glutamine residue would be unable to fill the same space as a histidine side chain, leaving room within the active site. Additionally, it was noted that the glutamine residue could still provide a hydrogen-bonding interaction to the S2B sulfur and, although the differential electronic effect of the histidine- and glutamine-derived interactions could modulate the redox potential of the M-cluster, it was not sufficient to completely change the activity profile. We now know that the S2B site is the location of the μCO ligand in the lo-CO state,⁸⁹ and by comparing a 2.5 Å crystal structure of α-H195Q NifDK¹⁸⁴ to the structure of the wild-type NifDK, we know that glutamine and histidine residues at position 195 have a similar distance to the S2B sulfur (Figure 16). This could rationalize why some of the other α-H195 variants do not generally exhibit any sensitivity to CO, that is, the responsible residues are simply too far from the μCO site to strongly interact with CO. The crystal structure also shows that the substitution of histidine with glutamine greatly destabilizes the local hydrogen-bonding network surrounding the M-cluster, complicating the interpretation. Moreover, different substitutions at α-H195 result in protein variants with CO-insensitive proton reduction and CO-inhibited acetylene reduction,¹⁸⁷ leading to the proposal of multiple CO binding site at the M-cluster. In pursuit of this idea, Newton, Fisher, and co-workers used the α-H195Q and α-H195N NifDK proteins in reactivity experiments with HCN, CN⁻, and N₃⁻ in combination with CO.^183,188,191 For the variants, CO was found to remove the inhibitory effects of CN⁻ and N₃⁻, implying that these classic nitrogenase inhibitors all bind to the same sites, albeit with different affinities. In this context, it should be noted that other GPMs like formate dehydrogenase (Section 4) and CO dehydrogenase (Section 5), CN⁻ and N₃⁻ inhibit activity, which highlights a functional connection between the low-valent metalloenzymes discussed herein. With respect to the binding site(s) of CO in nitrogenase, recent structural analysis shows that a second CO binds terminally to Fe6 of the M-cluster, and α-H195 is located near Fe2 and is generally unperturbed going from the NifDK-CO structure to the NifDK-(2CO) structure.^89,226 It may be possible for CO molecules (or other inhibitors) to bind to the Fe2 site on the M-cluster in α-H195 variants due to the change in the relative steric protection provided by the imidazole ring, underlining the importance of second coordination sphere effects. Additional structural information with other inhibitory molecules would be important for better understanding these complex effects.

Figure 16.

Crystal structure of α-H195Q NifDK (in gray; PDB ID 1FP4), overlaid with the crystal structure of wild-type NifDK (in cyan; PDB ID 3U7Q) from A. vinelandii. The cluster atoms are shown as ball-and-stick models, the R-homocitrate (HC), and amino acid residues are presented as sticks. Atomic coloring: Fe, orange; S, yellow; C, gray; O, red; N, blue.

The α-Q191 (+, +, +) residue of NifDK has been a target of mutagenic studies for as long as α-H195.^182,185 Crude extracts of α-Q191E and α-Q191K NifDK were assayed with C₂H₂ and in the presence of 0.02% CO, activities were inhibited by 50% and 75%, respectively.¹⁸⁵ The α-Q191K NifDK was then further purified, but the α-Q191E variant was not stable enough to sustain such a treatment. The purified α-Q191K NifDK protein exhibited an EPR spectrum that had similar line shape to that of the wild-type NifDK, but with perturbations to the g-values, suggesting that the M-cluster was bound to the protein similarly in both proteins. The introduction of 3% CO into substrate reduction assays showed that electron flux for reactions with α-Q191K NifDK decreased by about 70%, 40%, and 50% for N₂, C₂H₂, and proton reduction, respectively, compared to in the absence of CO.¹⁸⁵ This is an unusual finding considering that wild-type NifDK generally redistributes electron equivalents from the reduction of N₂ and C₂H₂ toward proton reduction in the presence of CO.^51,185 Much like for the α-H195 variants, the purified α-Q191K NifDK was also assayed for proton reduction in the presence of both CN⁻ and CO.¹⁹¹ The α-Q191K variant showed a mild inhibition of proton reduction in the presence of CN⁻, and subsequent introduction of a 10% CO atmosphere further decreased the proton reduction activity. In contrast, the wild-type and α-H195 variant proteins show inhibition of total electron flux by CN⁻ but this activity is restored in the presence of CO. The result agrees with the previous observation of CO-derived inhibition of the electron flux for α-Q191K NifDK,¹⁸⁵ and additionally supports that α-Q191 likely can interact with both substrates and inhibitors.

It is now possible to glean additional insight from the crystal structures of NifDK-CO and NifDK-(2CO). Figure 14 shows how the α-Q191 residue of NifDK in both structures is bent away from the M-cluster, but the residue is in proximity to several molecules, 3.6 Å from an H₂O molecule, 3.3 Å from the tCO ligand and 2.9 Å from a terminal carboxylate of the R-homocitrate ligand.^89,226 Similar distances are also found for the analogous α-Q176 residue in the CO-bound structures of the V-nitrogenase (Figure 15).^94,95 In addition, it has been noted that the α-Q176 residue can swing in to interact with the V-cluster at the S2B position, so the residue is less static.⁹³ The closest and likely strongest interaction is between α-Q191 and R-homocitrate, so when glutamine is mutated to lysine, the side chain is extended by an additional carbon atom, which may introduce steric misalignment of the Lys residue in the active site. Lysine may clash with the tCO binding site, preventing substrate molecules from properly accessing the M-cluster. This would be consistent with an observation made by Hales, Newton, and co-workers that α-Q191K NifDK does not generate a hi-CO EPR signal and, by inference, does not bind tCO; instead, it only shows the lo-CO signal even under 100% CO atmosphere.¹⁸¹

2.5.2.2. Role of the Homocitrate Ligand.

While a physical blockage by lysine may be the source of the aberrant properties of α-Q191K NifDK, another important factor is that a disruption of the α-Q191–homocitrate interaction may destabilize the hydrogen-bonding network within the active site M-cluster, which is thought to be critical for proton delivery during catalysis.^{102,181,229–232} When the nif V gene is deleted in a nitrogenase-expressing organism, the resulting Δnif V NifDK protein has an M-cluster with citrate in place of R-homocitrate.^196,197 In the presence of 3% CO, crude extracts of Δnif V NifDK from K. pneumoniae (Kp) demonstrated a ~50% decrease of proton reduction, similar to what was observed for α-Q191K NifDK.^185,196,233 Further investigation showed that CO would decrease the electron flow through Kp Δnif V NifDK, and uncouple MgATP hydrolysis from electron transfer with a corresponding decrease to the observed proton reduction.²³³ This analysis by Dixon and co-workers also indicated that CO was unlikely to be reduced by the variant Mo-nitrogenase from K. pneumoniae. Analogous Δnif V NifDK variants were also generated in A. vinelandii (Av), showing a range of activities in the presence of CO.^181,234,235 Newton, Hales, and co-workers showed that proton reduction by Av Δnif V NifDK was inhibited by ~40% in the presence of CO, a similar magnitude to that of Kp Δnif V NfiDK, and that Av Δnif V NifDK was capable of adopting the lo-CO, hi-CO, and hi(5)-CO states.^181,234 However, CO reduction by a Δnif V NifDK protein was not reported until 2021 by Ribbe, Hu, and coworkers.²³⁵ This protein, compared to wild-type NifDK, expressed lower activities for all substrates, including CO. It was reported that the Mo content of the in vivo isolated Av Δnif V NifDK was only ~30% of that in the wild-type protein, which may in part explain the observed low reactivity of the protein. To improve cofactor incorporation, an in vitro reconstitution of Δnif V NifDK was carried out, which increased the M-cluster content and rendered the in vitro reconstituted variant higher in activity but similar in reactivity pattern as compared to its in vivo isolated counterpart. In the presence of CO, Δnif V NifDK diverts more electron flux from proton reduction toward hydrocarbon formation while still maintaining lower specific activity than the wild-type enzyme. These observations were not made previously for the A. vinelandii variant because it was not reported that Av NifDK could reduce CO to hydrocarbons until 2011 when scaled up reaction conditions were used to detect low-yielding products for V-nitrogenase.²²³ An analogous Δnif V variant was also generated for V-nitrogenase, showing that citrate also binds to the V-cluster.²³⁶ However, unlike the scenario in the Mo-dependent system, the Δnif V VnfDGK protein produces a decreased amount of longer chain hydrocarbon products compared to wild-type VnfDGK, favoring a slight increase in proton reduction.

Initially, it was unclear what role the nif V gene product played in nitrogenase biosynthesis, but in 1990, Burris and co-workers reported that the organic ligand bound to the M-cluster was identified as citrate, which was subsequently confirmed in 2002 by Lawson and co-workers, who reported a 1.9 Å resolution crystal structure of the Kp Δnif V NifDK.^197,237 The structure of the Δnif V protein is highly similar to wild-type NifDK, but while R-homocitrate can form hydrogen-bonding interactions with the backbone amide of α-I423 (~2.8 Å), the citrate ligand in Δnif V NifDK is further away from α-I423 (~3.7 Å) and requires H₂O molecules to facilitate an interaction.²³⁷ This would potentially leave the citrate ligand loosely bound to the protein scaffold, disrupting the hydrogen-bonding network necessary for proton delivery to the M-cluster. An additional ramification would be that the loss of second coordination sphere interactions may also modulate the redox properties of the catalytic cofactor, similar to what was proposed for certain α-H195 NifDK variants (vide infra). This also means that it might be rather difficult to deconvolute the effects of α-Q191 substitutions from those of Δnif V knockout proteins, because the Gln residue is involved in stabilizing interactions with the citrate ligand. However, there are distinct differences. The α-Q191K NifDK protein only supports the lo-CO EPR signal, while Av Δnif V NifDK supports all three CO-based EPR signals. In fact, Newton and co-workers expanded investigations of CO on the reactivity of NifDK by introducing mutations to residues α-R96 (+, −, 0), α-R277 (−, +, +), and α-R359 (0, +, −).^181,186 In this study, CO-bound states were generated for a series of NifDK variants and EPR spectra were collected along with data from proton reduction assays to correlate the spectroscopic signals to catalytic activity. Wild-type NifDK, as well as α-R96Q NifDK, and Av Δnif V NifDK all showed the lo-CO, hi-CO, and hi(5)-CO states. However, the variants α-R96K, α-R277C, and α-R359K adopted only the lo-CO and hi(5)-CO states, but not the hi-CO state. Interestingly, α-R277H only showed the hi(5)-CO signal. Inhibition of proton reduction by CO was also equally dynamic, with instances of no inhibition in α-R96Q but ~50% inhibition with the α-R96K substitution, among others.¹⁸¹ Based on these observations, the authors concluded that the CO-related EPR signals and proton reduction activity could not be correlated, noting that the single-point mutations did not introduce additional CO-binding sites.¹⁸¹ The α-R277C NifDK variant was slightly different, in that it demonstrated a 70% increase in proton reduction in the presence of 10% CO compared to the wild-type enzyme.¹⁸⁶ With higher electron flux conditions, the proton reduction activity was enhanced, but dropping tchhe flux resulted in a lack of inhibition of proton reduction, although this reaction occurred at slower rates, similar to that observed in the case of the wild-type NifDK. It was proposed that CO binding to α-R277C NifDK affected catalytic proton transfer and additionally prevented the M-cluster from accessing more reduced states (i.e., E₃ and E₄ from the Lowe-Thorneley cycle). Instead, the variant may divert all the electron equivalents into the E₂ state which subsequently generates H₂.¹⁸⁶ It is unclear why CO binding would have such a different effect on the reactivity of the α-R277C variant as compared to the others, and more work is needed to gain further understanding of CO reactivity.

2.5.2.3. Reduction of Carbon Monoxide.

Attempts to stimulate CO reduction in Mo-nitrogenase have been carried out using combinations of point mutations to α-V70 (0, −, +), α-R96 (+, −, 0), α-H195 (−, 0, +), and α-Q191 (+, +, +).¹⁷⁹ Modifications of α-V70 in Av NifDK have been shown to increase the substrate scope of the enzyme by changing the steric environment around the M-cluster.^52,194 The α-V70G and α-V70A NifDK variants, were reported to react with CO, producing hydrocarbons ethylene (C₂H₄), propene (C₃H₆), ethane (C₂H₆), and propane (C₃H₈) in a ratio of 7:4:2:1.¹⁷⁹ These rates are much lower than those reported for the V-nitrogenase²²³ and not directly compared with those for the wild-type Mo-nitrogenase under the same experimental conditions. Double mutations were then incorporated to assess the effect of α-R96, α-H195, and α-Q191 on the α-V70-mediated CO reactivity.¹⁷⁹ The α-V70A/α-R96H and α-V70A/α-Q191A NifDK variants show a different product pattern than the single substitutions, with similar values for all C₂ and C₃ products, as well as the inclusion of trace amounts of methane (CH₄). The α-V70A/α-Q191A NifDK, though, produces twice as much C₃H₆ than the α-V70A/α-R96H variant. Additionally, α-V70A/α-H195Q was shown to drastically decrease product formation compared to the single α-V70A NifDK variant. It is difficult to make definitive statements about the exact effect of point mutations in this case because of the interconnected nature of the second coordination sphere effects. What can be said is that the α-V70 residue seems important for regulating the size of substrates and products, consistent with previous observations.⁵² The size reduction of the α−70 residue side chain presumably enlarged the active site to accommodate multiple CO molecules, and by extension, additional modification of α-R96 and α-Q191 further expanded the active site to favor the generation of longer chain hydrocarbons. However, the loss of activity in the α-V70A/α-H195Q NifDK protein shows that the size of the cavity is not the only factor governing reactivity toward CO. The α-Q191 and α-H195 residues are known to be involved in proton transfer,^{102,181,229–232} but the present data may reflect a more direct role of α-H195 in protonation of reduced, CO-coupled species whereas α-Q191 may play more of a structural role. It may also be possible that substitution of α-Q191 could affect the rate of protonation and allow for intermediate species to accumulate, increasing the hydrocarbon length. Rigorous studies involving additional mutations that vary the size and proton donor ability of the side chains may be necessary to shed further light on the “steric vs protonation” arguments for the reaction of CO with nitrogenase. It is also possible that yet unidentified factors, such as electron transfer rate or allosteric changes, may play a role.

2.5.3. Reduction of Carbon Dioxide.

The triatomic carbon dioxide (CO₂) is a highly oxidized small molecule that was found to react with Mo-nitrogenase in 1995.²³⁸ CO was initially found as the product of the reaction, as the concentration of CO would increase only in the presence of NifH and MgATP, indicating it was a nitrogenase-dependent phenomenon. However, the exact substrate form, CO₂, CO₃²⁻, or HCO₃⁻, was unclear because of the equilibrium between CO₂ and bicarbonate (HCO₃⁻) in solution. It was later reported by Ribbe and coworkers that NifDK was capable of generating CH₄ and C₂H₄ from from CO₂, although the rate of hydrocarbon formation was an order of magnitude slower than CO formation.²³⁹ Interestingly, V-nitrogenase was found to reduce CO₂ to CO in H₂O buffers, but in deuterated buffers, deuterated hydrocarbons (CD₄, C₂D₄, C₂D₆) were observed while NifDK was unable to generate deuterated products under these same conditions. For both Mo- and V-nitrogenase, the specific activity of CO evolution from CO₂ increased in the deuterated buffer system resulting in an inverse kinetic isotope effect (KIE).²³⁹ This could indicate that the formation of CO is carried out through similar mechanisms in both nitrogenases, but the change in hydrocarbon product profile between the two nitrogenases is reflective of something more complex that requires further investigation.

Subsequently, Mo-nitrogenase variants α-V70A and α-H195Q were shown to convert CO₂ into methane (CH₄), confirmed by use of¹³C-labeled bicarbonate.²⁴⁰ Under the reported conditions, C₂ and C₃ hydrocarbon products were not observed. As the NifH:NifDK ratio in the reaction was increased, the rate of the competing proton reduction reaction decreased, diverting electron flux to the formation of CH₄. Even with a high electron flux (50:1), the primary product was H₂, which makes sense considering that proton reduction is a two-electron process, while the conversion of CO₂ to CH₄ requires eight electrons. To separate the reaction of the α-V70A and α-H195Q NifDK variants between CO and CO₂, deoxyhemoglo-bin was added to the activity assays so that any released CO molecule would be rapidly removed from the analysis.²⁴⁰ The resultant reactions demonstrated a ~25% decrease in the production of CH₄, indicating that two different processes were being observed: one where CO₂ remains bound to nitrogenase throughout the reduction reaction and the other where CO dissociates and presumably rebinds to the M-cluster before reduction to CH₄. CO₂ reduction was further explored when the NifDK variant was incubated in an atmosphere of 45% CO and Ar, including 1–3% acetylene. Propylene (C₃H₆) was observed as the major product with propane (C₃H₈) as a minor component, and isotopically labeled¹³C atom from bicarbonate was found to incorporate into propylene. This indicates that one molecule each of CO₂ and acetylene can couple to form the product. Adjustments of the electron flux would favor CH₄ formation at a higher nitrogenase ratio and C₃H₆ formation at lower ratios. Additionally, introduction of ethylene to the CO₂ reaction did not yield propylene, suggesting that the mechanism requires that both molecules must bind and be activated by the cluster before they can be coupled.

It is interesting that in the double mutant α-V70A/α-H195Q NifDK, the reactivity with CO is inhibited compared to α-V70A NifDK, but the same α-H195Q mutation enhances the reactivity of the variant NifDK with CO₂. Electron transfer rates seem to be important factors for determining the result of CO₂ reduction. A slower rate would favor reduced intermediate species that are long-lived, encouraging C–C bond formation, whereas faster electron delivery appears to more quickly generate CH₄, discouraging carbon coupling.²⁴⁰ However, electron transfer in nitrogenase must also be understood in the context of proton transfer, because the two events are closely linked to each other.⁶³ In the α-V70A NifDK variant, CO molecules are readily coupled to form longer chain hydrocarbons, but this activity becomes disrupted when the α-H195 residue is exchanged for one less capable of proton transfer. The timing of proton-coupled electron transfer to the substrate is affected for CO reduction but happens to function well for the eight-electron reduction of CO₂. What should be emphasized is that a new activity was generated by controlling the second coordination sphere via the steric environment to house larger substrates in the active site niche and by tuning protonation of one of the oxygen atoms of CO₂ to generate CO and H₂O. Recently, CO₂ has also been converted to CH₄ by the wild-type Fe-only nitrogenase from Rhodopseudomonas palustris,²⁴¹ but without additional mutational or structural information about the Fe-only nitrogenase it is difficult to gain further insight into the factors that control CO₂ reduction. The structure and catalytic mechanism of other low-valent GPMs active in CO₂ reduction like formate dehydrogenase and CO dehydrogenase are discussed in Section 4 and Section 5, respectively.

3. HYDROGENASE

Hydrogen turnover (eq 2) is a important reaction in many microorganisms that thrive under reducing, anoxic conditions, and also plays a role in certain aerobes.^242–244 Trace amounts of H₂ provide electrons to power the anabolism of numerous archaea and bacteria in the soil, aqueous environments, or host tissue (H₂ oxidation or “H₂ uptake”).^245–247 Under fermentative or “microaerobic” conditions, proton reduction and H₂ release have been shown to be a key in the redox regulation of autotrophs like photosynthetic bacteria and algae.^248–250 Additionally, H₂ contributes to the cellular redox equilibrium as a side product of N₂ fixation by nitrogenase, as discussed in Section 2. The GPMs responsible for proton reduction and H₂ oxidation are referred to as hydrogenases.²⁵¹

Three classes of phylogenetically unrelated hydrogenases have been defined.²⁵² In archaea, [Fe]-hydrogenase is involved in the conversion of CO₂ to CH₄ (methanogenesis), catalyzing H₂ splitting and hydride transfer at a cofactor with a monometallic iron center.^253–255 Not much is known about the influence of second or outer coordination sphere effects, thus we will focus on the remaining two classes in this article. Unlike [Fe]-hydrogenase, [NiFe]- and [FeFe]-hydrogenase are iron–sulfur enzymes and carry a bimetallic active site cofactor (Figure 17).

Figure 17.

Schematic representation of hydrogenase cofactors. (a) [NiFe]-hydrogenase active site cofactor and (b) [FeFe]-hydrogenase active site cofactor. The [4Fe-4S] cluster is coordinated by three further cysteines (not shown). Note the presence of an azadithiolate ligand (ADT, S₂(CH₂)₂NH) at the diiron site. ‘X’ marks the catalytic binding site in both cofactors.

[NiFe]-hydrogenases have been found in archaea and bacteria, and are typically involved in heterotrophic H₂ uptake and the “recycling” of H₂ during methanogenesis or N₂ fixation (Section 2).^255–257 A classification into Groups 1–4 has been proposed,²⁴⁶ in which the water-soluble “standard” [NiFe]-hydrogenases are distinguished from membrane-bound, multimeric, as well as O₂-tolerant and bidirectional [NiFe]-hydrogenases.^257–259 Although highly diverse, all [NiFe]-hydrogenases share the same active site cofactor. It is composed of a nickel and an iron ion, covalently attached to the enzyme by either four cysteine residues, or three cysteines and one selenocysteine (Figure 17A). The iron ion is coordinated by two CN⁻ and one CO ligand, the latter being a common feature of hydrogenases.²⁶⁰

[FeFe]-hydrogenases have been found in bacteria, lower eukaryotes, and green algae, representing the phylogenetically most recent class.^261–263 The physiological diversity among [NiFe]-hydrogenases is less pronounced than in [FeFe]-hydrogenases:²⁴⁶ often, [FeFe]-hydrogenases are involved in redox regulation (H₂ release upon proton reduction) but in complex with other metalloproteins, additional roles in CO₂ reduction (Section 4) or bifurcation have been suggested.^264–266 “Standard” [FeFe]-hydrogenases belong to group A and have been investigated intensively whereas knowledge about group B–C [FeFe]-hydrogenase is only beginning to emerge.^267,268 Figure 17B shows the active site cofactor including the covalently attached [4Fe-4S] cluster. The metal ions of the diiron site are modified with a bridging carbonyl ligand (μCO) and two terminal CN⁻ and a CO ligand. Here, we shall elaborate on the active site architecture, catalytic mechanism, and second/outer coordination sphere effects in [NiFe]-hydrogenase (Section 3.1) and [FeFe]-hydrogenase (Section 3.2).

3.1. [NiFe]-Hydrogenase

3.1.1. Structural Features of [NiFe]-Hydrogenase.

[NiFe]-hydrogenases are divided into Groups 1–4, according to the classification by Morales, Greening, and co-workers.²⁴⁶ In these groups, [NiFe]-hydrogenases are further distinguished into “H₂ uptake” (Groups 1 and 2), “bidirectional” (Group 3), and “H₂ evolving” (Group 4) clades. Each group is divided into subgroups, for example, the O₂-sensitive prototypical or standard-type hydrogenases in Group 1b and the O₂-tolerant membrane-bound hydrogenases in Group 1d. The complete list is available in the “hydrogenase database”, HydDB.²⁶⁹ [NiFe]-hydrogenases are multisubunit proteins, which contain at least the catalytic subunit and an iron–sulfur cluster subunit. For example, standard [NiFe]-hydrogenases (Group 1b) are composed of two subunits, large and small, with molecular weights of approximately 60 kDa and 30 kDa, respectively (Figure 18A).²⁷⁰ The large subunit binds the catalytic cofactor and the small subunit embeds the accessory iron–sulfur cluster. The first crystal structure of a standard [NiFe]-hydrogenase was determined for the enzyme from Desulfovibrio gigas (Dg) in 1995 at 2.85 Å resolution,²⁷¹ followed by the structures of [NiFe]-hydrogenases from Desulfovibrio vulgaris Miyazaki F (DvMF)^272–276 and other species.^277–290 The Ni–Fe active site cofactor is located inside the large subunit, where the Ni and Fe ions are bridged with two cysteine thiolates, and two other cysteine residues are bound to the Ni ion in a terminal fashion.^273,291,292 In addition to two CN⁻ ligands and one CO ligand coordinated to the Fe ion, a bridging hydride ligand (μH^–) may exist between the metal ions (Figure 18B).²⁷³ The Ni site changes its oxidation state (i.e., Ni³⁺, Ni²⁺, and Ni⁺) among various intermediate states, some of which (Ni³⁺ and Ni⁺) have been characterized by EPR spectroscopy. On the other hand, the Fe site maintains the low oxidation, low spin state (Fe²⁺, S = 0).^{251,293–298} In the O₂-tolerant NAD⁺-reducing [NiFe]-hydrogenase (Group 3d) from Hydrogenophilus thermoluteolus TH-1, three cysteine thiolates bridge the Ni and Fe ions while one cysteine residue terminally ligates to the Ni ion in the oxidized state.²⁸⁴ In the case of [NiFeSe]-hydrogenase, a terminal cysteine is replaced by a selenocysteine.^299–303 In standard [NiFe]-hydrogenases, three iron–sulfur clusters are located in the small subunit and mediate the electron transfer between the Ni–Fe active site and physiological redox partners.^304–307 In the crystal structure of the O₂-tolerant membrane-bound [NiFe]-hydrogenase (Group 1d, see Figure 18C), a proximal [3Fe-4S] cluster harbored by six cysteine residues has been identified.^286,288,289

Figure 18.

Crystal structures of [NiFe]-hydrogenase. (a) Crystal structure of the standard [NiFe]-hydrogenase from D. vulgaris Miyazaki F (PDB ID 1WUJ). (b) Stick representation of the Ni–Fe active site with two CN⁻, one CO, and a bridging ligand X that may be μOH⁻, μOOH⁻, or μH⁻ (PDB ID 1WUJ shows the Ni-B state). The Ni–Fe active site cofactor is located inside the large subunit, whereas three iron–sulfur clusters are located almost linearly (each ~12 Å apart) in the small subunit. (c) Crystal structure of O₂-tolerant membrane-bound [NiFe]-hydrogenase from E. coli (PDB ID 4GD3). Atomic coloring: Ni, green; Fe, orange; S, yellow; C, gray; O, red; N, blue.

The substrate (H₂) or inhibitors (O₂ and CO) are transferred via hydrophobic channels from the molecular surface to the Ni–Fe active site.^308–311 After splitting H₂ into electrons and protons (eq 2), proton transfer pathways provide an exit route to the molecular surface where protons are release into the solvent. The crystal structures of [NiFe]-hydrogenases revealed that highly conserved amino acid residues like arginine, serine, glutamate, and histidine are located in the second coordination sphere of the Ni–Fe active site (Figure 19). The CO ligand on the Fe ion interacts with the hydrophobic side chains of valine and leucine (V500 and L482 in DvMF). On the other hand, the two CN⁻ ligands accept hydrogen bonds from arginine R479 and serine S502, respectively, in DvMF.²⁷³ One of the bridging cysteine thiolates is hydrogen-bonded to histidine H88.³¹² These interactions have been confirmed in the electron density map of the hydrogen atoms in the subatomic resolution structure of DvMF [NiFe]-hydrogenase.²⁷³

Figure 19.

Second and outer coordination spheres of the Ni–Fe active site. Crystal structure of the standard [NiFe]-hydrogenase from D. vulgaris Miyazaki F in the Ni-B state (PDB ID 1WUJ). Glutamate E34, histidine H88, arginine R479, and serine S502 are important residues located in the second coordination sphere of the Ni–Fe active site. The dotted lines indicate the hydrogen bond network. Atomic coloring: Ni, green; Fe, orange; S, yellow; C, gray; O, red; N, blue.

3.1.2. Ni–Fe Cofactor and Catalytic Cycle.

Standard [NiFe]-hydrogenases form two inactive oxidized states, Ni-A and Ni-B (Ni³⁺, S = 1/2).^259,313,314 In the presence of H₂ or other reductants, activation of Ni-A requires long times, whereas the Ni-B is readily activated.^251,296 A hydroxide is present at the bridging position in the Ni-B state (μOH^–, Figure 20C);²⁷⁵ however, the identification of the bridging ligand in the Ni-A state remains controversial.^312,315,316 The crystal structure of the Ni-A state in DvMF was interpreted to represent a peroxide ligand (μOOH^–, Figure 20A), while a μOH⁻ ligand was found in Allochromatium vinosum (Av, Figure 20B).^275,277,278 The later is in agreement with DFT calculations and single-crystal ENDOR spectroscopy that indicated a μOH⁻ ligand in both Ni-A and Ni-B.³¹⁵ It was proposed that the differences in the EPR spectra of the two states are caused by a ligand rotation of 7–10° involving C546 and C549 in DvMF. Accordingly, the bridging μOH⁻ ligand of the active site could be same in Ni-A and Ni-B, and the differences of the activation rates between these states may result from the slight distortion in the coordination of the cysteine residues. Further investigation is needed to clarify the difference of these states.

Figure 20.

Crystal structures of the Ni–Fe active site in various oxidation states. (a) The Ni-A state of DvMF hydrogenase (PDB ID 1WUI) and (b) of Av hydrogenase (PDB ID 3MYR). (c) Ni-B state (PDB ID 1WUJ). (d) CO-inhibited Ni-SCO state (PDB ID 1UBH). (e) The Ni-SI_a state (PDB ID 6QGR). (f) Ni-R state (PDB ID 4U9H). A bridging ligand exists in the Ni-A (μOOH⁻ and μOH⁻), Ni-B (μOH⁻), and Ni-R states (μH⁻), while the bridging position is vacant in the Ni-SI_a state. Atomic coloring: Ni, green; Fe, orange; S, yellow; C, gray; O, red; N, blue.

One-electron reduction of the Ni-A and Ni-B states leads to the EPR-silent inactive states, Ni-SU and Ni-SI_r (Ni²⁺), respectively (Figure 21).³¹⁷ The midpoint redox potential for the interconversion between the Ni-A and Ni-SU states was shown to decrease by ~60 mV per pH unit in various [NiFe]-hydrogenases, indicating that the one-electron reduction of Ni-A is coupled to the uptake of one proton.^{296,308–311,318–321} While the crystal structure of the Ni-SU state is not yet available, DFT calculations proposed coordination of a water molecule to the Fe ion and modification of the Ni coordination structure.³²²

Figure 21.

Proposed catalytic cycle of [NiFe]-hydrogenase. The catalytic cycle comprises four states (Ni-SI_a, Ni-R, Ni-C, and Ni-L) that interconvert in direction of H₂ oxidation or proton reduction. The EPR-active states are shown in red, diamagnetic states are shown in blue. Among the inactive states, a μOH⁻ ligand is tentatively assigned in the Ni-SU state.

A glutamate residue in the second coordination sphere of the Ni–Fe cofactor plays an important role for the inactivation process. This highly conserved glutamate residue (E34 in DvMF) is located next to the Cys residue which is terminally bound to the Ni ion (C546 in DvMF, see Figure 19). When the catalytically active [NiFe]-hydrogenase is oxidized under anaerobic conditions, the enzyme converts from the Ni-SI_a state to the Ni-B state via the Ni-SI_r state (Figure 21). The first order rate constant of the anaerobic inactivation process in the E25Q (equivalent E34 in DvMF) variant of [NiFe]-hydrogenase from Desulfovibrio fructosovorans (Df) was 5.6 times smaller than that of the wild-type enzyme, although the activation rate constant was hardly affected by the replacement.^323,324 Thus, the glutamate residue is important for the incorporation of the bridging OH⁻ ligand to the Ni–Fe cofactor. Under anaerobic conditions, formation of the Ni-B state with a bridging OH⁻ ligand from the Ni-SI_a state requires not only the transfer of a water molecule into the active site but also the extraction of a proton from a water molecule and oxidation of the Ni ion (Ni²⁺ to Ni³⁺). The carboxylic group of the glutamate residue may deprotonate the bound water molecule at the active site, producing the bridging OH⁻ ligand of the Ni–Fe cofactor.³²³ Interestingly, the Ni-B state was not produced when the E28Q variant (E34 in DvMF and E25 in Df) of the [NiFe]-hydrogenase 1 from Escherichia coli (EcHyd-1) was oxidized on an electrode in the presence of H₂, indicating that the interaction of H₂ with the Ni–Fe cofactor is maintained in the E28Q variant and that the early stages of H₂ activation outpace electrochemical oxidation.³²⁵ It may become much easier to start another cycle by activating H₂ and generating the Ni-R state in the E28Q variant, since protons may not easily leave from the Ni–Fe cofactor in the variant and thus it becomes difficult to produce OH⁻ from water.

During the catalytic cycle, the Ni–Fe cofactor changes the oxidation states in the following order: from Ni-SI_a (Ni²⁺), Ni-R (Ni²⁺), and Ni-C (Ni³⁺) to the Ni-L (Ni⁺) state (Figure 21).^{251,293–297,308} These states convert among each other by addition or release of H₂, protons, and/or electrons. In the first step of the cycle, the Ni-SI_a state (with the bridging ligand position vacant) reacts with H₂ (Figure 20E). H₂ is cleaved heterolytically to a proton (H⁺) and a hydride (H^–), which initiates the transition of the Ni-SI_a state to the fully reduced Ni-R state. The high-resolution crystal structure of DvMF revealed that the Ni-R state possesses a μH⁻ ligand and a protonated cysteine (C546-SH) at the Ni–Fe cofactor (Figure 20F).²⁷³ The μH⁻ ligand and the protonated cysteine residue in the Ni-R state was also elucidated by nuclear resonance vibrational spectroscopy (NRVS), which is particularly suited to monitor ⁵⁷Fe ligand vibrations, in combination with DFT calculations.^326–328 The hydride ligand was located closer to the Ni ion in the Ni-R state of the enzyme, whereas a short Fe–H⁻ bond and a long Ni–H⁻ bond were found in nearly all synthetic Ni–Fe models.^329–332 The protonated cysteine residue is considered as a part of the proton transfer pathways (see below).

The H₂ binding site at the first step of the cycle is still unclear, and either of the two metal ions may be involved in the formation of a (side-on) H₂ σ-bond complex. Although experimental data on H₂ binding to the Ni–Fe cofactor are not yet available due to the transient nature of intermediate, theoretical studies suggest the Ni ion as the initial site of H₂ binding.^333–337 In addition, complementary studies showed that the competitive inhibitor CO binds to the Ni ion at the Ni–Fe active site (Figure 20D).^274,338 The coordination geometry of the Ni ion is key to the thermodynamically favorable interaction of H₂ with the Ni–Fe cofactor. Theoretical calculations predicted that a peculiar seesaw-shaped geometry in the Ni-SI_a state with trans S–Ni–S angles near 120° and 180° is necessary for favorable binding of H₂ to the Ni site.³³⁵ In line with this proposal, the crystal structure of the Ni–Fe cofactor of F₄₂₀-reducing [NiFe]-hydrogenase from Methanosarcina barkeri (Mb) exhibited no ligand at the bridging position with unusual trans S–Ni–S angles of 107° and 171° in the Ni-SI_a state (Figure 20E).³³⁹ Similar angles have also been suggested by resonance Raman studies of the O₂-tolerant, membrane-bound [NiFe]-hydrogenase from Cupriavidus necator, ReMBH (noted by its former name Ralstonia eutropha).³⁴⁰

One-electron oxidation of the Ni-R state leads to the paramagnetic Ni-C state (Ni³⁺, S = 1/2) with a μH⁻ ligand and deprotonated C546-S⁻ at the Ni–Fe active site.^341,342 The Ni-C state is then converted to the paramagnetic Ni-L state (Ni⁺), with a vacant bridging position and a (re-)protonated cysteine C546-SH (see below).^{296,297,342–345} In the last step of the catalytic cycle, one-electron oxidation coupled with removal of the proton from C546-SH results in conversion of the Ni-L state to the Ni-SIa state.^{295,346–351}

3.1.3. Proton Transfer and Proton-Coupled Electron Transfer.

The proton acceptor during the catalytic H₂ oxidation cycle has been proposed by theoretical³⁵² and Raman spectroscopic studies.^340,353 Apparently, the proton is transferred from the Ni–Fe cofactor to the Ni-coordinating terminal cysteine thiolate (C546 in DvMF), which is the first step in the proton transfer pathway from the Ni–Fe cofactor to the outer coordination sphere (Figure 19).^{273,297,346,354–356} The H2 oxidation activity of the E25Q variant (E34 in DvMF) in Df [NiFe]-hydrogenase decreased to less than 0.1% compared to the wild-type enzyme.³⁵⁷ These results indicated that the glutamate residue located close to the C546 ligand plays an important role for the proton transfer during the catalytic reaction, which was supported by theoretical studies.^358,359 Time-resolved infrared spectroscopy addressing the E17Q variant of O₂-tolerant Pyrococcus furiosus [NiFe]-hydrogenase (Pf SH-1) showed that glutamic acid E17 (E34 in DvMF) is a proton relay for the interconversion between the Ni-C and Ni-SI_a states.³⁵¹ The replacement E17Q did not interfere with the μH⁻ photolysis of the Ni-C state but it disrupted PCET from the Ni-L state to the Ni-SI_a state, preventing formation of the Ni-SI_a state.

Alternatively, the highly conserved arginine (R479 in DvMF and R509 in EcHyd-1) has been proposed as a proton acceptor, forming a frustrated Lewis pair for the oxidation of H₂ in EcHyd-1 from studies using variants R509K, D574N, D118A, P508A, and D118N/D574N (R479, D544, D123, P478 and D123/D544 in DvMF, see Figure 19).^360,361 The turnover rate of the R509K variant decreased by a factor of 100 compared to that of the wild-type enzyme. The H₂ oxidation efficiency of the D118A variant, in which aspartate D118 forms a salt bridge to arginine R509, also decreased compared to native EcHyd-1. In the case of the soluble [NiFe]-hydrogenase 1 from Pf SH-1, the R355K (R479 in DvMF) variant altered the ligand binding environment at the Ni–Fe cofactor and destabilized the Ni-C state, resulting in a Ni-C/Ni-L tautomeric equilibrium.³⁶²

The SH stretching frequency at 2505 cm⁻¹ has been detected by highly sensitive Fourier-transform infrared (FTIR) difference spectra utilizing the photoconversion of the Ni-C state to the Ni-L and Ni-SI_a states in DvMF [NiFe]-hydrogenase. The data showed that C546 is protonated in the Ni-L state (Cys-SH) and deprotonated in the Ni-C and Ni-SI_a states (Cys-S⁻).^297,346 Glutamic acid E34-COOH was found to be double hydrogen-bonded in the Ni-L state and single hydrogen-bonded in the Ni-C and Ni-SI_a states (Figure 22) according to the COOH stretching frequency between 1700–1730 cm^–1. Additionally, a stretching mode of a “dangling” water molecule was observed in the Ni-L and Ni-C states. These results suggest that the Ni–Fe cofactor and its surrounding amino acids function as an optimized proton transfer system in standard [NiFe]-hydrogenases, and the direction of the proton transfer is regulated by the rearrangement of the hydrogen bond network around C546, E34, and a dangling water molecule (Figure 22).

Figure 22.

Proposed proton transfer mechanism for [NiFe]-hydrogenase. The hydrogen bond network between C546 (DvMF nomenclature, the respective sulfur atom is bold), E34, T18, A548, and a dangling water molecule is rearranged during the catalytic reaction. An involvement of R479 (gray) is possible. Reprinted (adapted or reprinted in part) with permission from ref 346. Copyright 2019 Wiley.

The Ni-L and Ni-R states of DvMF [NiFe]-hydrogenase are likely to construct a similar hydrogen bond network between C546-SH, E34, T18, a backbone contact including A548-NH, and a dangling water molecule (Figure 22).³⁴⁶ Several Ni-L states have been identified by light irradiation of the Ni-C state under anaerobic conditions at T < 100 K.^{342,363–366} From temperature-dependent FTIR studies of DvMF, the ΔH and ΔS values for the equilibrium between the protonated/deprotonated forms (i.e., Cys-SH and Cys-S⁻) of two Ni-L states were obtained as 6.4 ± 0.8 kJ mol⁻¹ and 25.5 ± 10.3 J mol⁻¹ K⁻¹, respectively.³⁴⁸ The small ΔH and ΔS values indicate efficient proton transfer at the cysteine residue C546 of the Ni–Fe cofactor between the two Ni-L states.

The Ni-SI_a and Ni-C states may form a similar hydrogen bond network between C546-S^–, E34, A548-NH, and a dangling water molecule (Figure 22). Theoretical studies have indicated involvement of a threonine residue in the proton transfer of the transmembrane proton pump bacteriorhodopsin;³⁶⁷ however, T18 seems to be nonessential for proton transfer in Df [NiFe]-hydrogenase and rather stabilizes the local protein structure. This observation was suggested by structural and spectroscopic studies of variants in which threonine was replaced with serine, valine, glutamine, glycine, and asparagine.³⁶⁸

3.1.4. Accessory Iron–Sulfur Clusters.

In standard [NiFe]-hydrogenases, three accessory iron–sulfur clusters (denoted proximal [4Fe-4S]^2+/+, medial [3Fe-4S]^+/0, and distal [4Fe-4S]^2+/+) are located almost linearly (each ~12 Å apart) in the small subunit (Figure 23),²⁷³ and mediate electron transfer between the Ni–Fe active site and physiological redox partner, such as cytochrome c₃.^304–307 The proximal [4Fe-4S]^2+/+, medial [3Fe-4S]^+/0, and distal [4Fe-4S]^2+/+ clusters of Dg [NiFe]-hydrogenase exhibit redox potentials of −315, −80, and −445 mV versus SHE, respectively.³⁶⁹

Figure 23.

Location of the accessory iron–sulfur clusters. The large subunit with the Ni–Fe cofactor is shown in white cartoon, the small subunit with the proximal [4Fe-4S], the medial [3Fe-4S], and the distal [4Fe-4S] cluster is shown in green (DvMF [NiFe]-hydrogenase, PDB ID 1H2R). The three accessory iron–sulfur clusters are located almost linearly. Atomic coloring: Ni, dark green; Fe, orange; S, yellow; C, green; O, red; N, blue.

The accessory iron–sulfur clusters are not only important for electron transfer but may also play a role with respect to hydrogen turnover in the presence of O₂: some [NiFe]-hydrogenases are “O₂-sensitive” (i.e., inhibited by O₂) while the “O₂-tolerant” [NiFe]-hydrogenases maintain activity under aerobic conditions. The latter have been shown to quickly reactivate from the Ni-B state in a process that appears to be very dependent on the nature of the proximal cluster.^370–372 The reactivity toward O₂ clearly distinguishes the bimetallic hydrogenases; as discussed in Section 3.2.2, [FeFe]-hydrogenases are irreversibly destroyed by O₂.

3.1.4.1. Proximal Cluster.

When the proximal [4Fe-4S]^2+/+ cluster is reduced in standard [NiFe]-hydrogenases, the [4Fe-4S]⁺ cluster interacts magnetically with the Ni^3+/+ center, which causes splitting of the Ni-C/Ni-L EPR signals below 10 K.³⁷³ Spectroscopic studies revealed that the conversion of Ni-C to Ni-SI_a via the Ni-L state is controlled by the redox state of the proximal [4Fe-4S]^2+/+ cluster in DvMF [NiFe]-hydrogenase.³⁴⁷ The transition of the Ni-L state to the Ni-SI_a state may occur when the proximal [4Fe-4S]^2+/+ cluster is oxidized but not when it is reduced. These results suggest that the catalytic cycle is controlled by the redox state of the proximal [4Fe-4S]^2+/+ cluster, which may act as a gate for the catalytic electron transfer.

Most of the standard [NiFe]-hydrogenases possess a glutamate near the proximal [4Fe-4S] cluster (Figure 24A), while some [NiFe]-hydrogenases possess an aspartate instead. The crystal structure of membrane-bound Av [NiFe]-hydrogenase in the Ni-A state showed that there are two different forms for the proximal [4Fe-4S] cluster; one is a standard cubane, while in the other form the proximal [4Fe-4S] cluster is distorted with one of the Fe ions bound to aspartate D75, which is located nearby the cluster (Figure 24B).²⁷⁸ It was proposed that the negatively charged side chain of aspartate stabilizes the distorted form of the [4Fe-4S] cluster. A similar coordination was found in the O₂-tolerant [NiFe]-hydrogenase from Citrobacter sp. S-77.²⁸⁵ Additionally, the water molecules near the proximal cluster relocated upon reduction, emphasizing the importance of the water network in this hydrogenase.²⁸⁵

Figure 24.

Structural changes of the proximal iron–sulfur cluster in various [NiFe]-hydrogenases. (a) Oxidized form (PDB ID 1UBH) in DvMF and (b) oxidized-distorted form (PDB ID 3MYR) in Av. (c) Reduced and oxidized forms of the [4Fe-3S] cluster in HmMBH (PDB ID 3AYX and 5Y34). (d) Reduced, partially reduced, and oxidized forms of the [4Fe-3S] cluster in ReMBH (PDB ID 3RGW, 4IUD, and 4IUB). The proximal [4Fe-4S] cluster is distorted with one of the Fe ions bound to D75 in one of the oxidized forms (b). Atomic coloring: Ni, green; Fe, orange; S, yellow; C, gray; O, red; N, blue.

Contrary to the standard [NiFe]-hydrogenases, some of the O₂-tolerant, membrane-bound [NiFe]-hydrogenases, such as EcHyd-1, ReMBH, Hydrogenovibrio marinus (HmMBH), and Aquifex aeolicus (AeMBH), possess a unique proximal [4Fe-3S]^5+/4+/3+ cluster ligated by six cysteine residues (Figure 24CD).^{286,288–290} This cluster can be oxidized twice within a very small potential range.³⁷⁴ Its electronic structure has been characterized and debated.^375–378 The importance of the coordination by six cysteines to the proximal iron–sulfur cluster has been shown by site-directed mutagenesis studies in ReMBH and EcHyd-1.^379–382 Replacement of the two supernumerary cysteine residues C19 and C120 in ReMBH with glycine residues altered the electronic structure of the proximal cluster, and the double variant was unable to sustain activity under prolonged O₂ exposure.³⁷⁹ Similar mutagenesis studies in EcHyd-1 showed that O₂ tolerance depends on C19 and not on the two-electron oxidation at the proximal cluster, which is detected with the O₂-sensitive C19G variant but not with the O₂-tolerant C120G variant.³⁸⁰ For another example, the O₂-tolerant NAD⁺-reducing soluble [NiFe]-hydrogenase from C. necator, ReSH (noted by its former name R. eutropha), can produce trace amounts of superoxide in H₂-driven NAD⁺ reduction with O₂.³⁸¹ The C41S variant of ReSH displayed up to 10% of wild-type activity, suggesting that the coordinating role of C41 at the proximal [4Fe-4S] cluster might be partly substituted by the nearby C39 residue, which is present only in O₂-tolerant pyridine nucleotide-dependent [NiFe]-hydrogenases, whereas the C39G, C39A, and C39S variants increased the O₂ sensitivity compared to wild-type ReSH.³⁸²

EPR and ⁵⁷Fe Mössbauer spectroscopic studies have shown that the proximal [4Fe-3S]³⁺ cluster donates two electrons to the Ni–Fe active site, resulting in the “superoxidized” state [4Fe-3S]⁵⁺.³⁷⁴ A charged glutamate near the cluster (COO⁻, e.g. in HmMBH and EcHyd-1) or a deprotonated water molecule (OH⁻, e.g. in ReMBH) stabilizes the [4Fe-3S]⁵⁺ cluster.^{286,288,289,307,383} The deprotonated amide nitrogen of the backbone formed by cysteine C26 (in HmMBH) may bind to one of the Fe ions of the cluster, also stabilizing the highly oxidized state. In AeMBH, the proximal [4Fe-3S]^5+/4+/3+ and medial [3Fe-4S]^+/0 clusters show midpoint potentials that are higher than potential necessary value for the conversion of the Ni-SI_a to Ni-L/Ni-C states in standard [NiFe]-hydrogenase.^345,374,384 Additionally, a kinetic argument favoring the appearance of the Ni-L state in EcHyd-1 is that the proximal [4Fe-3S]³⁺ cluster is fully reduced at the potentials that are required for both the Ni-C and Ni-L states, impeding the elementary electron transfer step that converts the Ni-L to Ni-SI_a state.

The conserved glutamic acid residue (E34 in DvMF) plays an important role for the mechanism of O₂ protection in the O₂-tolerant NAD⁺-reducing [NiFe]-hydrogenase from H. thermoluteolus TH-1, which is revealed by the crystal structures (Figure 25).²⁸⁴ The Ni–Fe cofactor of HtTH-1 in the H₂-reduced state exhibits a similar coordination structure as those of standard [NiFe]-hydrogenases, whereas in the air-oxidized state, an octahedral Ni geometry was found; three bridging thiolate ligands between the Ni and Fe ions, one terminally bound cysteine thiolate, and an unprecedented bidentate ligation of the glutamate side chain (E32 in Ht and E34 in DvMF) (Figure 25). The fully occupied octahedral Ni geometry protects the Ni–Fe cofactor from direct attack by O₂. Higuchi, Shomura, and coworkers suggested that the conformational change of the Ni–Fe active site was triggered by the reduction of the proximal [4Fe-4S] cluster (Y1 in Figure 25). The coordination structure of the amino acid residues located in between the Ni–Fe cofactor and the proximal [4Fe-4S] cluster also changed upon reduction. In the reduced state, E32 does not bind to the Ni ion and forms a hydrogen bond network including S464 and E56, notably unrelated to the catalytic proton transfer pathway as discussed above (Figure 22). The side chain of arginine R58 flips between E56 and the proximal [4Fe-4S] cluster (Figure 25), which may affect the efficiency of catalytic electron transfer.

Figure 25.

Proposed proton transfer in the O₂-tolerant [NiFe]-hydrogenase HtTH-1. The oxidized state is shown in cyan (PDB ID 5XF9) and the H₂-reduced state is shown in white (PDB ID 5XFA). Y1 refers to the proximal iron–sulfur cluster. The side chain of E32 binds to the Ni ion in the oxidized state, while it does not bind to the Ni ion and forms a hydrogen bond network including S464 and E56 in the reduced state. The metal centers in the reduced state are omitted for clarity. Atomic coloring: Ni, green; Fe, orange; S, yellow; C, gray; O, red; N, blue.

3.1.4.2. Medial and Distal Clusters.

In standard [NiFe]-hydrogenases, the medial [3Fe-4S]^+/0 cluster possesses a rather high midpoint potential, which may play a role in effectively “trapping” electrons from the Ni–Fe cofactor, facilitating the reaction of the Ni-SI_r state with H₂ via the acid–base equilibrium between the Ni-SI_r and Ni-SI_a states.^373,385,386 Substitution of a proLine, a potential ligand to the [3Fe-4S] cluster, to cysteine (P238C in Df [NiFe]-hydrogenase, equivalent P242 in DvMF) triggered the conversion of the medial [3Fe-4S] cluster to a [4Fe-4S] cluster decreasing the midpoint potential from +65 mV to −250 mV versus SHE. This potential decrease caused a 30% reduction in H₂ oxidation activity and a 2-fold increase in H₂ evolution activity without significantly altering the spectroscopic properties of the Ni–Fe active site and the proximal and distal [4Fe-4S] clusters.³⁸⁷ The distal [4Fe-4S] cluster is located very close to the solvent and interacts with electron acceptors.²⁵¹ Three cysteines and one histidine coordinate this cluster (Figure 23). The H184C and H184G variants (distal [4Fe-4S] cluster ligand variants) of Df [NiFe]-hydrogenase (H188 in DvMF) showed only 1.5% and 3% oxidative activity compare to the wild-type enzyme, respectively.³⁸⁸ The activity of the H184G variants decreases or increases upon addition of mercaptoethanol or imidazole to the assay, respectively: these ligands modulate electron transfer rates upon binding to a free coordination site on the distal cluster.^388,389 The affinity of the variants H184G, P238C/H184G, and P238C/H184C for the electron acceptors methyl viologen and cytochrome c₃ was similar or greater than that of the wild-type enzyme, suggesting that H184 does not govern the partner recognition.³⁸⁹ DFT calculations demonstrated that substitution of the histidine ligand to a cysteine in Df [NiFe]-hydrogenase does not change the reorganization energy of the distal [4Fe-4S] cluster.³⁹⁰ The calculated rate of electron transfer was, however, reduced by three orders of magnitude, resulting from a change in electronic donor–acceptor coupling including histidine H184 and phenylalanine F193 (F197 in DvMF). These results indicated that the protein environment is tuned for efficient electron transfer. A systematic survey of the substitution of amino acid residues related to the medial and distal iron–sulfur clusters has been reported for the [NiFe]-hydrogenase from Alteromonas macleodii (which belongs to Group 1e but shows modest O₂ tolerance).^391–393 For example, the double substitution of P285C at the medial [3Fe-4S] cluster (P242 in DvMF) and H230C at the distal [4Fe-4S] cluster (H188 in DvMF) increased the H₂ evolution activity three- to four-fold compared to the wild-type enzyme.³⁹³ Electrochemistry measurements showed that in the R193L variant of O₂-tolerant EcHyd-1, in which R193 is located near the histidine ligand (Figure 26, L194 in DvMF), the midpoint potentials of the medial [3Fe-4S]^1+/0 and distal [4Fe-4S]^2+/+ clusters are more negative than those of the wild-type enzyme.³⁹⁴ The R193L variant also enhanced bias toward H₂ evolution and slightly diminished the O₂ tolerance, whereas the catalytic activity of the K189N and Y191E variants (equivalent Q190 and P192 in DvMF) did not change compared to that of wild-type EcHyd-1.³⁹⁴

Figure 26.

Amino acids surrounding the distal [4Fe-4S] cluster. The [NiFe]-hydrogenase EcHyd-1 is shown (PDB ID 3UQY). Arginine R193 in the second sphere affects the electron transfer rate. Atomic coloring: Fe, orange; S, yellow; C, green; O, red; N, blue.

The electron transfer route between the Ni–Fe cofactor and the protein surface is complicated, and theoretical calculations proposed that multiple routes coexist between the metal centers during catalysis.³⁹⁵ Thus, enzymatic activity is tuned by not only in the second coordination sphere of the Ni–Fe cofactor but also via the electron transfer trajectory in the outer coordination sphere. Furthermore, in the case of O₂-tolerant [NiFe]-hydrogenases, most variants at the proximal iron–sulfur cluster became O₂ sensitive.³⁷⁹ In addition to controlling electron transfer, the accessory iron sulfur clusters play an important role in the O₂ tolerance.

3.1.5. Gas Channels in [NiFe]-Hydrogenase.

The Ni–Fe cofactor is deeply buried in the center of the large subunit (Figure 27).²⁵¹ Both substrate and inhibitory gas molecules, such as H₂, O₂, and CO, can access the Ni–Fe active site from the solvent region through hydrophobic gas channels predicte from the crystal structure of the protein.^{309,354,396–398} Cavity calculations of these cavities using a 1 Å radius probe showed that there are four gas access points at the protein surface, combining into one channel that leads to the Ni–Fe active site.^396,397 In vicinity of the Ni–Fe cofactor, the end of the channel becomes narrower due to a “bottleneck” composed of two hydrophobic amino acid residues, typically valine and leucine (Figure 27).^397,398 Noble gases like Xe and Kr bind to the hydrophobic area of the protein and are well visible in electron density maps; thus, protein crystal structures pressurized with Xe or Kr provide useful information for hydrophobic gas channels.^{354,396,399,400} The crystal structures of the Xe-bound standard [NiFe]-hydrogenases from Df and DvMF have been determined at 6.0 and 1.8 Å resolution, respectively, where the Xe atoms were observed in the hydrophobic cavities of these enzymes.^354,396 Controversially, molecular dynamics (MD) simulations suggested that gas molecules like H₂ and O₂ could reach the Ni–Fe active site via several pathways.^398,401

Figure 27.

Gas transfer channels. (a) Standard Df [NiFe]-hydrogenase. Reprinted (adapted or reprinted in part) with permission from ref 309. Copyright 2008 National Academy of Sciences. (b) O₂ pathway in O₂-tolerant ReMBH. Two entrances for gas transfer channels have been revealed, which are combined and extend to the Ni–Fe active site. Reprinted (Adapted or Reprinted in part) with permission from ref 400. Copyright 2018 National Academy of Sciences.

The O₂ transfer pathways of O₂-tolerant ReMBH has also been investigated by X-ray crystallography together with computational studies and has been compared with those of O₂-sensitive [NiFe]-hydrogenases.^399,400 The channels of [NiFe]-hydrogenases span distances of 28–76 Å between the protein surface and the Ni–Fe active site, wherein O₂-sensitive [NiFe]-hydrogenases have a wider bottleneck and a more complex channel network than those of their O₂-tolerant counterparts. Several O₂ binding sites have been revealed within the hydrophobic channels that have two entrances and extend to the Ni–Fe active site (Figure 27).

Various site-directed mutagenesis studies have tested the suggestion that the presence of bulky residues at the end of the gas channel in the so-called O₂-tolerant, regulatory hydrogenases (RH) may prevent O₂ access to the active site (e.g., isoleucine and phenylalanine instead of valine and leucine).³⁹⁷ The H₂ oxidation activity of the I62V and F110L variants of the regulatory [NiFe]-hydrogenase from C. necator, ReRH (noted by its former name R. eutropha), decreased under aerobic conditions 25-fold and 5.5-fold, respectively, compared to the O₂-tolerant wild-type enzyme.⁴⁰² These results suggest that Re RH exhibits O₂ and CO tolerance owing to interruption of the gas channel at the bottleneck to the Ni–Fe active site with the bulky amino acid residues isoleucine I62 and phenylalanine F110. Furthermore, the double variant I62V/F110L was completely inactive, indicating that the shape of the gas channels may play a key role in O₂ tolerance.⁴⁰² The I65V/F113L double variant of the regulatory hydrogenase from Rhodobacter capsulatus is also more active and more O₂-sensitive than the native enzyme.⁴⁰³ The hypothesis is that narrower gas channels and fewer gas entrances limit the mass transport to the catalytic active site, presumably contributing to the net O₂ tolerance of these enzymes. However, site-directed mutagenesis studied aimed at narrowing the end of the gas channel of the O₂-sensitive Df [NiFe]-hydrogenase showed that steric hindrance slows down intramolecular diffusion and O₂ inhibition, but without making the variant O₂-tolerant. These gas diffusion processes have been studied intensively using many residue replacements at the bottleneck position, namely V74 and L122 in Df [NiFe]-hydrogenase.^{309,310,404–406} The rates of intramolecular gas diffusion were measured by using the bimolecular rate constant of inhibition by CO as a proxy of the rate of ligand access to the active site, by analyzing the results of isotope exchange assays, and by measuring the Michaelis constant for H₂.^309–311 The results show that the variations have the same relative effects on the transport of all ligands, CO, O₂, and H₂, and that not only the size but also the charge of the residues in the bottleneck have a very strong effect (up to nearly four orders of magnitude) on the kinetics of ligand access. Slowing down H₂ egress has the direct consequence of selectively slowing H₂ production and increasing the catalytic bias of the enzyme in the direction of H₂ oxidation.³¹¹ Some of the observed effects of the variations on the rates of diffusion could be explained by the results of MD simulations, according to which two distinct gates determine the kinetics, namely V74/R476 and V74/L122.^401,407 Although site-directed mutagenesis slows down the intramolecular diffusion rates, the resulting effect on the kinetics of inhibition by O₂ is 10-fold or less, because the transport toward the active site limits the reaction with O₂ only when it is very slow (e.g., in the V74 M and V74Q variants).^406,408 However, some V74 substitutions (in particular V74C and V74H) have a very beneficial effect on O₂ inhibition that is not related to the kinetics of diffusion: for reasons that have not been clarified, these variations strongly increase the rate of reactivation after O₂ inhibition. The rate of reactivation of Df [NiFe]-hydrogenase V74H is close to that observed with the O₂-tolerant enzyme from A. aeolicus,^404,409 and mutation to cysteine further increases the O₂ tolerance of EcHyd-1.⁴¹⁰

According to the type of [NiFe]-hydrogenase, the gas channels differ in size, shape, and branching, with respect to the physiological requirement of each [NiFe]-hydrogenase. The crystal structure of the Mb F₄₂₀-reducing [NiFe]-hydrogenase in the Xe-bound form revealed an additional gas channel (Figure 28).³³⁹ Interestingly, the [NiFeSe]-hydrogenase from Desulfovibrio vulgaris Hildenborough (DvH) has a similar hydrophobic channel,^339,411 although this channel is not observable in the crystal structures of standard [NiFe]-hydrogenases.^{354,396,399,400} In the case of DvH [NiFeSe]-hydrogenase site-directed mutagenesis experiments suggest that O₂ reaches the active site through a hydrophilic water channel.⁴¹²

Figure 28.

Gas tunnels of Mb F₄₂₀-reducing [NiFe]-hydrogenase. Seven Xe atoms (red spheres) were detected within an additional gas tunnel from the surface to the Ni–Fe cofactor (PDB ID 6QII). Reprinted (adapted or reprinted in part) with permission from ref 339. Copyright 2019 Wiley.

3.2. [FeFe]-Hydrogenase

3.2.1. Structural Features of [FeFe]-Hydrogenase.

Standard [FeFe]-hydrogenases are in most cases monomeric enzymes with a molecular weight ranging between 50–80 kDa. The [FeFe]-hydrogenase from Clostridium pasteurianum, CpI, whose X-ray crystal structure has been published in 1998, shows a mushroom-like shape with four iron–sulfur clusters binding to the “stipe” (Figure 29A).^413–415 Because of the sequence similarity with ferredoxin, this fold is referred to as “F-domain”. The [FeFe]-hydrogenase from C. acetobutylicum, CaI, albeit not crystallized, is likely of similar architecture.⁴¹⁶ In both enzymes, the iron–sulfur clusters form a conductive chain that serves in electron exchange between soluble redox partners and the catalytic active site. This chain consists of two buried [4Fe-4S] clusters, one surface-exposed [2Fe-2S] cluster, and one surface-exposed [4Fe-4S] cluster, the later which is bound by a three cysteine and one histidine ligand.⁷²⁴ From a comparison of CaI variants, it was concluded that the [2Fe-2S] cluster is the main electron transfer conduit, at least when the enzyme is wired to an electrode.⁴¹⁷ On the other hand, docking simulations with CpI suggest that the physiological electron donor, ferredoxin, interacts predominantly with the surface-exposed [4Fe-4S] cluster. According to EPR titrations this cluster has the lowest potential.⁴¹⁸ The authors investigated a catalytically inactive cofactor variant to facilitate steady-state conditions, notably below the H⁺/H₂ Nernstian potential, where CpI spontaneously reoxidizes as a result of proton reduction, which prevents equilibrium conditions during the redox titration at low potential.^418,419

Figure 29.

Crystal structures of various [FeFe]-hydrogenases. (a) C. pasteurianum CpI (PDB ID 4XDC) and (b) D. desulfuricans DdH (PDB ID 1HFE). The characteristic peptide belt of DdH is shown in green. Panel (c) depicts a homology model of the [FeFe]-hydrogenase from C. reinhardtii CrHydA1. Here, the characteristic “loop” that is specific for this enzyme is shown in yellow. Accessory iron–sulfur clusters and the catalytic H-cluster are highlighted as spheres. Note the successive miniaturization of the F-domain from (a) to (c). Atomic coloring: Fe, orange; S, yellow; C, gray; O, red; N, blue.

The [FeFe]-hydrogenase from Desulfovibrio desulfuricans, DdH, whose X-ray crystal structure was published in 1999, features a smaller F-domain with only two [4Fe-4S] clusters and is surrounded by a peptide “belt” (Figure 29B).^420,421 The [FeFe]-hydrogenase from Megasphaera elsdenii, MeHydA, lacks this subunit but displays an equally small F-domain.⁴²² Voltammetry experiments suggested that intramolecular electron transfer in MeHydA is rate limiting in H₂ oxidation, which primarily determines the catalytic bias of this enzyme.⁴²³ The electron transfer chain of the recently crystallized hydrogenase from Clostridium beijeirincki, CbA5H, consists of only two [4Fe-4S] clusters and an additional N-terminal cluster located more than 20 Å away from any other cluster.⁴²⁴ Because of significant disorder, these clusters are not visible in PDB entry 6TTL. A crystal structure of the functional [FeFe]-hydrogenase from the green alga Chlamydomonas reinhardtii, CrHydA1, has yet to be obtained; however, based on several lines of evidence,^425–430 it can be concluded that CrHydA1 completely lacks the F-domain and any additional metal centers. The homology model shown in Figure 29C highlights an insertion or “loop” that is typical for [FeFe]-hydrogenases of the Chlorophyta-type.²⁶

The above-mentioned hydrogenases have been crystallized and thoroughly studied using biophysical methods; however, they are not representative of the diversity of the [FeFe]-hydrogenases, some of which are much larger and associated in multienzyme complexes.²⁶⁶ As an example, the hydrogen-dependent carbon dioxide reductase from Acetobacterium woodii has four subunits and 11 predicted iron–sulfur clusters, with an apparent mass around 169 kDa^431,432 while the respiratory formate hydrogen lyase complex (FHL) of Thermococcus onnurineus consists of 18 subunits, with a molecular weight of approximately 600 kDa.⁴³³ The coupling of CO₂ and H₂ turnover is discussed in Section 4. The small number of [FeFe]-hydrogenases that have been crystallized so far is certainly an obstacle to learning about how diverse structures and outer sphere effects modulate activity; this is actually true for all GPMs discussed in this review.

Unlike the F-domain, the catalytic “H-domain” shows a high degree of similarity in all [FeFe]-hydrogenases.²⁴⁶ Four cysteine residues bind a [4Fe-4S] cluster that is part of the active site cofactor and facilitates electron exchange with the accessory iron–sulfur clusters (or soluble redox partners, in case of CrHydA1). These residues are part of at least three protein loops 1–3 (Figure 30).²⁶ A strictly conserved cysteine residue in loop 3 additionally coordinates the diiron site, which is equipped with CO and CN⁻ ligands and an azadithiolate group (ADT, compare Figure 17B).^434–436 Together, the [4Fe-4S] cluster and the diiron site form the active site “H-cluster”. Figure 30 depicts the H-cluster, including a number of functionally relevant amino acid residues and water molecules in the second and outer coordination sphere.

Figure 30.

Active site cavity and catalytic cofactor of [FeFe]-hydrogenase CpI (PDB ID 4XDC). The H-cluster is composed of an [4Fe-4S] cluster and the diiron site. Loops 1–3 provide four cysteine residues to bind the [4Fe-4S] cluster and the diiron site. In loop 1 (green), C299 is part of the proton transfer pathway (further including water cluster W1, E279, and other groups). In loop 2 (blue), M352 may interact with the μCO ligand while K358 was suggested to form a hydrogen bond with the CN⁻ ligand at the distal iron ion, Fe_d. In loop 3 (magenta), M497 may form a hydrogen bond with the ADT group. Cysteine C499 was proposed to be the 1st proton acceptor of the proton transfer pathway. Serine S232 provides a hydrogen bond to the CN⁻ ligand of the proximal iron ion, Fe_p, presumably involving a backbone contact with P231. Atomic coloring: Fe, orange; S, yellow; C, gray; O, red; N, blue.

The standard or group A [FeFe]-hydrogenases CpI, CaI, DdH, and CrHydA1 have been studied intensively and most of what is known about the influence of second and outer coordination sphere is based on the biochemical and biophysical analysis of these enzymes. The diversity of [FeFe]-hydrogenases has been emphasized;^437–439 advances in understanding the second and outer coordination sphere effects in the lesser studied or more complex [FeFe]-hydrogenase groups is currently an emerging area of study. For example, the sensory or bifurcating [FeFe]-hydrogenases from Thermotoga maritime (TmHydS, TmHydABC) and Thermoanaerobacter mathranii (TamHydS) have been characterized.^440–443 A cryo-EM structure of TmHydABC has recently been published as a preprint.⁴⁴⁴ Table 4 shows a comparison of important amino acid residues in the H-domain, highlighting the level of similarity among group A [FeFe]-hydrogenases and the differences to group C and D [FeFe]-hydrogenases. In addition to the above-mentioned enzymes, the table also includes CpII and CpIII, MeHydA1, and CrHydA2.⁴²⁴ This comparison is not comprehensive and restricted to experimentally characterized [FeFe]-hydrogenases. As such, Table 4 allows correlating amino acid composition and catalytic function and will serve as a guideline for the analysis of second and outer coordination sphere effects. Additionally, Table 5 provides a compilation of all amino acid and H-cluster variants that have been published in the last 20 years. We will discuss the influence of natural and artificial variations of the H-domain in Section 3.2.2. Next to the biochemical and electrochemical analysis of turnover activity, the spectroscopic signatures of the active site cofactor have been a major source of information. We define the various H-cluster states hereafter.

Table 4.

Conservation of Amino Acid Residues in the H-Domain of Experimentally Characterized [FeFe]-Hydrogenases^a

	CpI	CpII	CpIII	CaI	DdH	MeHydA	CbA5H	CrHydA1	CrHydA2	TmHydA	TmHydS	TamHydS
HB	S232	S101	A162	S231	A109	A113	A294	A94	A97	A227	S79	S81
PT	E279	E148	E202	E278	E156	E151	E341	E141	E144	E274	E118	F119
PT	E282	E151	E205	E281	E159	E154	E344	E144	E147	E277	E121	E122
PT	R286	R155	R209	R285	R163	R158	R348	R148	R151	R281	G125	K126
PT	C299	C169	C222	C298	C178	C171	C367	C169	C172	C294	A131	A137
CC	C300	C170	C223	C299	C179	C172	C368	C179	C173	C295	C132	C138
PT	S319	S189	S242	S318	S189	S192	S387	S189	S192	S314	A151	L157
HP/HB	M353	T223	G268	M352	M232	T227	M421	M223	M226	M348	G177	S191
CC	C355	C225	C270	C354	C234	C229	C423	C225	C228	C350	C179	C193
HB	S357	A227	A272	D356	A236	A231	A425	R227	R230	A352	A181	A195
HB	K358	K228	K273	K357	K237	K232	K426	K228	K231	K353	K182	K196
HB	M497	M367	M381	M497	M376	M381	M565	M415	M426	M480	S267	L291
CC	C499	C369	C383	C498	C378	C391	C567	C417	C428	C482	C269	C293
CC	C503	C382	C387	C502	C382	C395	C571	C421	C432	C486	C273	C297
PDB IDs	6N59	-	-	-	1FEH	-	6TTL	3LX4	-	-	-	-

^aRepresentative PDB accession codes for CpI, DdH, CbA5H, and the apoprotein of CrHydA1 are given in the last row. Legend: HB, hydrogen bonding; HP, hydrophobic; PT, proton transfer; CC, cluster coordination.

Table 5.

List of [FeFe]-Hydrogenase Amino Acid and H-Cluster Variants^a

	variant	PDB ID	target	ref
CpI	G412H, G414A/H, G418A/H, G421H, G422H, K358N, R449A/H	N/A	Maturation	445
	E279A	5LA3	Proton transfer, formation of Hhyd	446
	C299A	6GLY	Proton transfer	447, 448, 449
	C299D	6GLZ
	C299S	N/A
	E279A	N/A
	E279D	6YF4
	E279Q	6GM0
	E282A	6GM1
	E282D	6GM2
	E282Q	6GM8
	R286A/L	6GM3
	S319A	6GM4
	C299S, M353L, K358N, M497L	N/A	Hydrogen bonding	450
	S31P, E47G, R86H, C153Y, N160D, D186G, T188A, I197V, K208E, K252R, I253V, A280V, N289D	N/A	Hydrogen turnover, O2 sensitivity	451
	L192G, G194C, N189C, T356C/T/V, S357C/G/P/T, M397C, A498C	N/A	Hydrogen turnover, O2 sensitivity	452
	M70L, M211L, M243L, M277L, M387L, M481L, M485L, M551L	N/A	O2 sensitivity	453
	C48A, C100A, C298D	N/A	Proton transfer, electron transfer	417, 454, 455
CrHydA1	C169A/S	N/A	Proton transfer, formation of Hhyd	456, 457, 458, 459
	C169A/D/S	N/A	Proton transfer	448
	E141A	6GM5
	E141D	N/A
	E141Q	6GM6
	E144A	6GM7
	E144D/Q	N/A
	R148A	N/A
	S189A	N/A
	C417Hb	6GL6	[4Fe-4S]	460
	C170A/D/S, C225A/D/S, C417A/D/S, C421A/D/Sb	N/A	[4Fe-4S]	461
	T226K/V	N/A	[4Fe-4S]	462
	R171D/Wc	N/A	Electron transfer	463
	A92S, A94S, S193C, E231D	N/A	Hydrogen bonding to CN−	464
	C169S, M223L, K228N, M415L	N/A	Hydrogen bonding	450
	C169D, F290Y, V296F	N/A	O2 diffusion	465
	R96Q K179Q K262Q R349Q R353Q R379Q K396Q K397E, K433Q	N/A	Interaction with ferredoxin	466
Other	L364F, C367D/A, P386L, A561F (in CbA5H)	N/A	O2 inhibition, formation of Hinact	424
	C48A, C100A, C298D (in CaI)	N/A	Proton transfer and electron transfer	417, 454, 455
	V229T (in CrHydA2)	N/A	[4Fe]H	467 c
Cofactor Variants	PDT, ODT, SDT, and other headgroups	N/A	Maturation, hydrogen turnover	418, 468, 469, 470, 471, 472, 473, 474
	Other modifications	N/A	Hydrogen turnover	475, 476, 477, 478, 479
	ODT	5BYQ	Hydrogen turnover, hydrogen bonding	479, 480
	PDT	5BYR
	SDT	5BYS
	EDT	6H63

^aThe majority of amino acid variants were produced for CpI and CrHydA1, with most H-cluster variants being produced for CrHydA1. Other enzymes include CaI, CbA5H, and CrHydA2.

^bIn refs 460 and 461, the numbering is different due to the deletion of 55 C-terminal amino acids.

^cIn ref 463, arginine R171 resembles R227 following the nomenclature used for CrHydA1 in this work.

3.2.2. H-Cluster, Catalytic Cycle, and Artificial Maturation.

3.2.2.1. Oxidized States.

In the absence of H₂ or other reductants, the H-cluster in standard [FeFe]-hydrogenases adopts the “active-ready”, oxidized state, H_ox. Moderate spin exchange coupling between [4Fe-4S] cluster and diiron site gives rise to the characteristic, rhombic EPR signal (g = 2.105, 2.044, 2.001, see Figure 31), indicating that the [4Fe-4S] cluster is oxidized (+2 state, a pair of Fe(II)–Fe(III)) while the diiron site adopts a mixed-valence +3 state (Fe(II)–Fe(I)).^481–483 The distribution of charges among Fe_p and Fe_d is debated.²⁶⁷ [FeFe]-hydrogenase in the H_ox-state was analyzed by EPR spectroscopy within cells, isolated in solution, and in crystallized form.⁴⁸⁴ X-ray diffraction structures of such crystals indicated a Fe–Fe bridging CO ligand,^413,420 which reflects in a low-frequency IR absorbance band at 1802 cm⁻¹, accompanied by two terminal CO and CN⁻ bands at higher frequencies (Table 6).^485–487 A small upshift in frequency relative to H_ox has been explained by a protonation of a cysteine residue that coordinates the [4Fe-4S] cluster (see below);⁴⁸⁸ the EPR signal shifts accordingly (g = 2.106, 2.048, 2.001, see Figure 31).²⁶⁷ This state is referred to as H_oxH. In the presence of CO, no H₂ oxidation and proton reduction activity is observed⁴⁸⁹ as the oxidized H-cluster reacts with exogenous CO to form the CO-inhibited state, H_ox-CO.^414,415 The rhombic EPR signal of H_ox converts into an axial signature (g = 2.05, 2.01, see Figure 31), indicative of pronounced spin exchange coupling.^490–492 While the crystal structure of H_ox shows an open, apical binding site at Fe_d, the crystal structure of H_ox-CO hints at a labile, inhibiting ligand in apical position. Spin polarization favors an apical CO ligand,⁴⁹² which contrasts from IR spectroscopy that has suggested ligand rotation and the stabilization of an apical CN⁻ ligand.^479,493,494 The IR spectrum shows four CO ligands instead of three (Table 6), but due to significant vibrational energy transfer^495–497 the assignment of normal modes is less intuitive than in H_ox.⁴⁹³ Under acidic conditions, H_oxH-CO is formed.⁴⁸⁸ The EPR spectrum of this state has not yet been reported.

Figure 31.

EPR spectra of the oxidized H-cluster. Simulations of EPR spectra for the oxidized states H_ox, H_oxH, and H_ox-CO as observed in CrHydA1. The g-values are indicated. Reprinted (adapted or reprinted in part) with permission from ref 267. Copyright 2020 American Chemical Society.

Table 6.

Vibrational and Electronic Properties of the Main H-Cluster States^a

	vCN−/cm−1		vCO/cm−1			[4Fe-4S]	[FeFe]	μ/t
H ox	2088	2070	1964	1940	1802	+2	II/I	CO/−
H ox H	2092	2074	1970	1946	1812	+2	II/I	CO/−
H ox -CO	2091	2081	1968	1962c	1808	+2	II/I	CO/CN−
H ox H–CO	2094	2086	1972	1966c	1816	+2	II/I	CO/CN−
H red′	2084	2066	1962	1933	1792	+ 1	II/I	CO/−
H red′ H	2086	2068	1966	1938	1800	+1	II/I	CO/−
H red′ -CO	2086	2076	1967	1951c	1793	+1	II/I	CO/CN−
H red	2070	2033	1961	1915	1891	+2	I/I	H/COb
H hyd	2088	2076	1980	1960	1860	+1	II/II	CO/H−
H sred	2068	2026	1953	1918	1882	+1	I/I	H/COb
H inact	2106	2087	2007	1983	1848	+2	II/II	CO/?
H trans	2100	2075	1983	1977	1836	+1	II/II	CO/?

^aAll data stem from DdH (H_inact and H_trans) or CrHydA1 (other H-cluster states).

^bUnder cryogenic conditions, H_redH+ and H_sredH+ are formed that bind a bridging carbonyl ligand and are likely to feature an open coordination site (μ/t = CO/−).

^cIn the CO-inhibited states, the coupled pCO/dCO stretching vibration gives rise to an additional band at 2012 cm⁻¹ (H_ox-CO), 2006 cm⁻¹ (H_oxH–CO), or 2002 cm⁻¹ (H_red′-CO).

3.2.2.2. One-Electron Reduced States.

The H-cluster may adopt two one-electron reduced states, both of which are largely diamagnetic. Reduction of the [4Fe-4S] cluster from +2 to +1 gives rise to the characteristic IR signature of H_red′ (sometimes referred to as just “H_red”), which looks similar to H_ox but shifted to lower frequencies (Figure 32A).^470,498,499 In H_ox and H_red′ the H-cluster binds a μCO ligand and maintains the “rotated’ structure” with an open, apical biding site at Fe_d (Figure 33A).⁵⁰⁰ As such, the H-cluster readily interacts with exogenous CO, forming the reduced, CO-inhibited state H_red′-CO.⁴⁹⁴ As in H_oxH, protonation of a cysteine has been proposed, stabilizing the excess electron of the [4Fe-4S] cluster in H_red′.^488,501 This may also explain the small upshift of frequencies under acidic conditions that has been assigned to the H_red′H state (Figure 32A).

Figure 32.

Influence of the [4Fe-4S] cluster on the IR signature of the H-cluster. (a) ‘H₂ − N₂’ difference spectra of CrHydA1 cofactor variant PDT at pH 5 (‘H_red′H − H_oxH’, blue trace) and pH 8 (‘H_red′ – H_ox’, red trace). Negative and positive bands are assigned to H-cluster states with an oxidized and reduced [4Fe-4S] cluster, respectively. The small frequency shifts between the oxidized and reduced states are assigned to a protonation at or near the [4Fe-4S] cluster. Data taken from ref 488. (b) Spectro-electrochemical difference spectrum of CrHydA1 that shows the accumulation of H_sred (magenta labels, −650 mV vs SHE) over H_red (green labels, −450 mV vs SHE) at pH 9. Negative and positive bands are assigned to H-cluster states with an oxidized and reduced [4Fe-4S] cluster, respectively. At ambient temperature, the bridging carbonyl ligand (μCO) converts into a terminal ligand, which results in three tCO bands.

Figure 33.

Changes in geometry upon reduction. (a) In the paramagnetic II/I state of the diiron site (H_ox and H_red′), the μCO ligand stabilizes the “rotated structure” that is characterized by a squarepyramidal/inverted square-pyramidal arragenment and an open coordination site at Fe_d (X). In H_hyd, the diiron site is “superoxidized” (II/II) and X is a terminal hydride. (b) Upon reduction (I/I), the diiron site adopts the symmetrical square-pyramidal/square-pyramidal arrangement. No μCO ligand is observed in the respective redox states Hred and Hsred.

Reduction of the diiron site from +3 to +2 ((Fe(I)–Fe(I)) gives rise to the characteristic IR signature of H_red. Reduction and protonation of the diiron site under ambient conditions triggers the release of the μCO ligand into a terminal position⁴²¹ and the formation of a Fe–Fe bridging hydride ligand, μH.⁵⁰² Accordingly, H_red shows three terminal CO ligands and the apical position is occupied by a CO ligand, most likely (Figure 32A).⁴⁷⁹ In variance to H_ox and H_red′, the square-pyramidal geometry of the diiron site is symmetrical (Figure 33B). Under cryogenic conditions, such changes in geometry are precluded and result in an IR signature that clearly shows a μCO ligand.^503–505 The latter is likely to be protonated at the ADT ligand and should be referred to as ‘H_redH⁺’. Whether H_red or H_redH⁺ are catalytically relevant cofactor intermediates or representative of an H₂-inhbited H-cluster geometry has been debated.^506–509

3.2.2.3. Two-Electron Reduced States.

Reduction of the [4Fe-4S] cluster starting from H_red results in the formation of H_sred.⁵¹⁰ This redox state was described for CrHydA1, where it accumulates under H₂, presumably due to the lack of accessory iron–sulfur clusters that distribute reducing equivalents between H-domain and F-domain.⁵¹¹ The H-cluster is diamagnetic in this “super-reduced” state; however, the assignment of the rhombic EPR spectrum with g = 2.08, 1.94, 1.87 has recently been questioned by Zebger, Horch, and coworkers who assigned an axial signal to H_sred (g = 2.15, 1.86).⁵⁰⁴ The IR difference between H_red and H_sred is similar to H_ox and H_red′, that is, a small shift to lower energies (Figure 32B).⁵¹² Just like H_red, the rotation of the μCO ligand into terminal position is precluded under cryogenic conditions (‘H_sredH⁺’).^503–505

A catalytic intermediate with a terminal hydride ligand was first suggested by Krasna and Rittenberg in 1954, based on H₂ and D₂ exchange experiments with the [FeFe]-hydrogenase from Proteus vulgaris.⁵¹³ 60 years later, the FTIR, Mössbauer, and EPR signatures (g = 2.07, 1.93, 1.88) of the so-called hydride state, H_hyd, were identified in protein and cofactor variants (Table 5).^456–458 In wild-type enzyme, accumulation of H_hyd was achieved upon H₂ oxidation starting from H_oxH, which confirmed the hydride state as a “natural”, catalytically relevant H-cluster intermediate (Table 6).⁴⁴⁶ Subsequent NRVS and NMR studies further established our knowledge about H_hyd.^459,474,514 The diiron site resides in a formally “superoxidized” state (Fe(II)–Fe(II)), binding the terminal hydride in apical position at the distal iron ion. This position trans to μCO has been exploited to demonstrate the hydride character of the ligand in FTIR H/D exchange experiments.^446,457 Similar to H_red′, the [4Fe-4S] cluster is reduced (+1) and likely to be protonated.⁵⁰⁶ The geometry of H_hyd resembles H_ox and H_red′ (Figure 33A).

3.2.2.4. Catalytic Mechanism of Hydrogen Turnover.

The ongoing debate about the catalytic mechanism of [FeFe]-hydrogenase^506,509 is inseparable from questions regarding the influence of second and outer coordination sphere effects that are discussed in Section 3.2.3 and Section 3.2.4. In the following, the two main models will be introduced. The key difference between the models is the assumption of a protonated and reduced diiron site in the “5-step model” or a protonation event at the [4Fe-4S] cluster in the “3-step model”.

In the absence of H₂ or other reducing components, all standard [FeFe]-hydrogenases adopt the H_ox state,^481–483 which is widely accepted as the starting and end point of a catalytic cycle that accounts for hydrogen turnover (eq 2). Figure 34 shows a simplified proposal for the succession of redox intermediates in the direction of H₂ evolution. Under the assumption that the H-cluster maintains the μCO geometry upon reduction of the diiron site,^503–505 H_redH⁺ and H_sredH⁺ may be catalytic intermediates. The 5-step model starts with reduction of the [4Fe-4S] cluster and the formation of H_red′ over H_ox (transition A in Figure 34). It is conceivable that the ADT ligand of the diiron site is protonated via the catalytic proton transfer pathway (Section 3.2.5), which may induce intramolecular electron transfer from the [4Fe-4S] cluster to the diiron site, resulting in the formation of H_redH⁺ over H_red′ (transition B). From here, H_sredH⁺ is formed upon reduction of the [4Fe-4S] cluster (transition C).⁵¹⁰ Although there is no experimental evidence for a protonation of the ADT ligand in any H-cluster state,⁵⁰⁶ the formation of a terminal hydride at Fe_d upon intramolecular proton transfer and oxidation of the diiron site seems a reasonable assumption (H_hyd over H_sredH⁺, transition D).⁴⁹⁸ In the final step of the 5-step model (transition E), protonation of the terminal hydride would lead to H₂ evolution and the recovery of the oxidized resting state, H_ox.⁵¹⁴

Figure 34.

Proposed catalytic cycle. The 5-step model (A–E) includes H_redH⁺ and H_sredH⁺, presuming the H-cluster retains the μCO ligand upon reduction of the diiron site. The 3-step model (A*–B*–E) suggests a “short-cut” from H_red′ to H_hyd as both species share the same geometry. The geometry of the reduced diiron site (I/I) under ambient conditions is debated. Reprinted (adapted or reprinted in part) with permission from ref 267. Copyright 2020 American Chemical Society.

On the basis of the observation that the H-cluster undergoes geometry changes upon reduction of the diiron site at ambient temperatures (Figure 32), H_red and H_sred have been excluded as catalytic intermediates.^505–508 In the alternative “3-step model”, H_red′ is formed over H_ox upon proton-coupled electron transfer to the [4Fe-4S] cluster (transition A*).⁵⁰¹ Here, the protons may reach the H-cluster via the so-called regulatory proton transfer pathway (Section 3.2.4) while in the 5-step model, transition A is a simple reduction process. H_red′ may convert directly into H_hyd in a second step of proton-coupled electron transfer (transition B* in Figure 34) before protonation of the terminal hydride ligand leads to H₂ evolution, analogous to the 5-step model (transition E). It is debated whether H_ox or the protonated resting state H_oxH is recovered upon H₂ evolution.⁴⁸⁸

Other H-cluster species are also described in the literature. Electrochemistry experiments have shown that various inactive states are formed under reducing conditions; whether these forms correspond to any spectroscopically identified species is unclear, although it can be said that not all species that have been isolated and characterized are necessarily catalytic intermediates.⁵¹⁵ Inactive states also accumulate under certain oxidizing conditions. H_inact (or ‘H_ox^air’) and H_trans represent unready, “superoxidized” H-cluster states that have been observed in some [FeFe]-hydrogenases.^419,516,517 These states differ in the redox state of the [4Fe-4S] cluster, and the IR difference between H_inact and H_trans is similar to H_ox and H_red′ (Table 6). H_trans is paramagnetic and gives a substoichiometric rhombic EPR spectrum with g = 2.06, 1.96, 1.89.⁵¹⁶ Remarkably, [FeFe]-hydrogenases in the H_inact state do not react with CO or O₂.⁵¹⁸

The reductive activation of unready DdH was long believed to be irreversible but it has recently been demonstrated that H_ox can be converted into H_trans and H_inact in the presence of exogenous sulfide, which binds to Fe_d.^519–521 The formation of H_inact in the absence of sulfide was considered to cause the reversible high potential inactivation of the enzyme as detected in electrochemistry experiments.⁵²² Later, it became clear that sulfide-independent oxidative inactivation results from the inhibition of the enzyme by traces of halide ions like chloride or bromide.⁵²³ Recently, the formation of H_inact was observed in the [FeFe]-hydrogenase CbA5H, notably in the absence of either sulfide or halides.⁵²⁴ The structure of the air-oxidized H-cluster suggests that Fe_d is coordinated by the sulfur atom of the conserved cysteine residue C367 (Figure 35). The formation of H_inact in CbA5H and the resulting resistance to O₂ are lost when this cysteine is replaced with aspartate (Table 5).⁴²⁴ This mechanism explains why the same H_inact IR signature is observed in air-oxidized CbA5H,⁵²⁴ in DdH natively produced in the sulfate-reducing bacterium D. desulfuricans,⁴⁸⁷ and in certain standard [FeFe]-hydrogenases exposed to sulfide.⁵²⁰ However, it is unclear why CpI is reportedly insensitive toward exogenous sulfide, unlike other standard [FeFe]-hydrogenases like CrHydA1 and DdH.⁵²¹ Moreover, direct coordination of Fe_d by the cysteine side chain was observed exclusively in CbA5H and not in standard [FeFe]-hydrogenases. It has been shown that C367 in CbA5H is located on a loop whose unique flexibility depends on residues up to 13 Å away from the H-cluster (Figure 35).⁴²⁴ The investigation of how the conversion between active and inactive states depends on the nature of these residues provided a clear demonstration of long-range outer coordination sphere effects on active site reactivity.

Figure 35.

Comparison of CpI (white) and CbA5H (green) in the H_ox and H_inact state. The distance of the sulfur atom of C299/C367 shrinks from ~6 Å to ~3 Å, making a direct coordination of the distal iron ion (Fe_d) in the H_inact state of CbA5H likely (dashed line). The annotated amino acid residues have been identified to cause the flexibility of the unique loop in CbA5H. Drawn after PDB IDs 4XDC and 6TTL.

Overall, the mechanism of aerobic inactivation is not well understood. While O₂ inhibition of DdH and CbA5H is reminiscent of the reaction of [NiFe]-hydrogenase with O₂ (Section 3.1.4), most [FeFe]-hydrogenases are irreversibly destroyed under aerobic conditions. It is believed that O₂ binds to the H-cluster like a ligand^525–528 and becomes reduced to superoxide, H₂O₂, or other reactive oxygen species (ROS) that subsequently damage the H-cluster.^529–533 Outer coordination sphere effects are likely to play a role, for example, considering O₂ diffusion through the enzyme⁴⁶⁵ and the influence of electron transfer⁴²³ or proton transfer⁵³² on ROS formation. The comparatively sluggish reaction of sensory [FeFe]-hydrogenases with O₂ had been explained with inefficient proton transfer.⁴⁴³ Additionally, second coordination sphere effects may influence the formation of the H_ox-O₂ state where the superoxide ligand is stabilized by hydrogen-bonding interactions with the ADT headgroup, not unlike H_ox-CO and H_hyd.⁴⁷⁹ As the hydride state is a key intermediate in the catalytic cycle (Figure 34) truly O₂-tolerant [FeFe]-hydrogenases may not be feasible.

3.2.2.5. Artificial Maturation of the H-Cluster.

In vivo, the iron–sulfur clusters of [FeFe]-hydrogenase are inserted by the regular iron–sulfur machinery, namely the isc system.⁵³⁴ This includes the [4Fe-4S] cluster of the H-domain that serves are “nucleation core” for the synthesis of the H-cluster.⁵³⁵ This task is performed by three auxiliary enzymes, HydEFG.⁵³⁶ A precursor of the diiron site is assembled on HydF^537,538,539 by the help of radical S-adenosylmethionine (rSAM) enzyme HydG.^540–544 Then, HydF delivers the active site precursor to the hydrogenase apoprotein where it fuses with the bridging cysteine upon CO release to form the H-cluster. A mechanism has been proposed that includes guidance by charged amino acids,⁴⁴⁵ not unlike the channel proposal for the maturation of nitrogenase (Section 2.2). It has been suggested that the role of the rSAM enzyme HydE is the synthesis of the ADT ligand.^545–547

In vitro, the rather intricate maturation process can be hijacked by loading HydF with diiron compounds and mixing HydF with apo-hydrogenase.⁴⁶⁸ This allowed for inserting different diiron site mimics into otherwise unaltered enzyme, creating “cofactor variants” or semisynthetic enzymes that are new to nature.^480,548,549 Further, it has been shown that even HydF can be avoided in the maturation process: incubation of apo-hydrogenase with diiron compounds lead to functional enzyme.⁴⁶⁹ The protocol established by Happe and co-workers led to a great variety of cofactor variants, including different dithiolate and diatomic ligands,^{418,468–474} chalcogenides,^475,476 and metal ions (Table 5).⁴⁷⁸ For example, replacing the central atom X of the dithiolate ligand from nitrogen (ADT) with carbon (PDT), suppresses any redox change at the diiron site.^468–470 This allowed for studying the protonation and redox chemistry of the [4Fe-4S] cluster in detail (Figure 32).⁴⁸⁸

3.2.3. Direct Interactions between H-Cluster and Protein.

The presence of CO and CN⁻ ligand at the active site cofactor of [FeFe]-hydrogenase was proven by FTIR spectroscopy in 1996.⁵¹⁷ When the H-cluster was modeled into the first electron density of [FeFe]-hydrogenases CpI and DdH, the position of the diatomic ligands was assigned according to second coordination sphere interactions.^413,420 The chemical identity of the CO and CN⁻ ligands could not be deduced from the electron densities alone. In CpI, the proximal CN⁻ ligand may accept a hydrogen bond from the side chain of serine S232 and the secondary amine of the backbone formed by S232 and P231 (Figure 36).²⁶ The latter is strictly conserved in [FeFe]-hydrogenase and most likely is responsible for the correct orientation of the backbone amine toward Fe_p-CN⁻; in CpI and DdH the angle between ligand and backbone is ~170°, indicative of a strong hydrogen bond. Serine S232 of CpI is replaced with an alanine in DdH (A109), which cannot form a bond with the H-cluster. Phylogenetically, the low level of conservation (Table 4) suggests a minor role in the second coordination sphere; however, the effect can be demonstrated. Winkler, Happe, Rüdiger, and co-workers identified siteselective shifts of the Fe_p-CN⁻ IR band by analyzing CpI–S232A and the corresponding variant of CrHydA1, A92S.⁴⁶⁴ The authors even introduced an additional hydrogen bond by changing alanine A94 to serine, which doubled the IR shift observed for A92S (Figure 36). Affecting the electron density at Fe_p and the coupling with the [4Fe-4S] cluster, these variants had a clear influence on the directionality of hydrogen turnover,⁴⁶⁴ which may explain the difference in catalytic performance between wild-type CpI and CrHydA1.

Figure 36.

Hydrogen-bonding environment of the CN⁻ ligands. In proximal (p) position, polar interactions (v) with a backbone amine, S232, and variant A230S have been experimentally verified in CpI and CrHydA1. The hydrophilic pocket including K358, S323, and a P324 backbone contact inspired the assignment of the CN⁻ ligand in distal (d) position in CpI and DdH. The putative hydrogen-bonding environment (v) is shown. Drawn after PDB ID 6NAC. Atomic coloring: Fe, orange; S, yellow; C, gray; O, red; N, blue.

The hydrogen-bonding environment of the distal CN⁻ ligand is difficult to characterize experimentally. Early on, the strictly conserved lysine K358 (CpI) and K237 (DdH) was suggested as hydrogen-bonding donor to Fe_d-CN⁻ (Figure 36).⁵⁵⁰ As the chemical nature of the ADT ligand was still unknown in 1999, Fontecilla-Camps and co-workers even proposed a catalytic proton transfer pathway based on K237 in DdH.⁴²⁰ Site-directed mutagenesis of this lysine, for example, toward alanine⁴⁵⁰ or asparagine,⁴⁴⁵ precluded any cofactor insertion and hinted at a key role in the activation of apo-hydrogenase.⁴³⁰ This renders functional studies of hydrogen bonding difficult. Investigating ¹⁴N-labeled DdH by hyperfine sublevel correlation spectroscopy (HYSCORE),⁴⁸³ Lubitz, Silakov, and co-workers assigned the large quadrupole couplings to a hydrogen bond between Fe_d-CN⁻ and K237.⁴³⁶ On the contrary, similar experiments on the CrHydA1 cofactor variant PDT by the same group resulted in substantially simplified ¹⁴N HYSCORE spectra that showed no coupling with K228 (equivalent K358 or K237).⁵⁵¹ While the role of K358 in hydrogen-bonding remains ambiguous, the sum of weak interactions around Fe_d-CN⁻ results in a certain level of structural rigidity in the oxidized state;⁵⁵² however, this might not preclude ligand rotation upon H-cluster state transitions. For example, DFT calculations based on 16 different¹³CO isotopomers of the H-cluster suggested an apical CN⁻ ligand in the H_ox-CO state,⁴⁹³ stabilized by an “internal” hydrogen bond with the ADT ligand. While this interpretation is not without controversy,⁴⁹² the stability of H_ox-CO was found to be largely reduced upon modification of the dithiolate ligand, following a trend, which cannot be explained with an apical CO ligand.⁴⁷⁹ In contrast to CN^–, carbonyl ligands are poor hydrogen bond acceptors.⁵⁵³ A similar concept of internal hydrogen bonding may play a role in the formation and stabilization of H_hyd.⁴⁵⁸ Moreover, CO/CN⁻ ligand rotation was discussed in the context of the temperature-dependent interconversion of H_ox and H_red/H_sred.^505,506 While the role of K357 in hydrogen-bonding remains ambiguous, other conserved residues in the vicinity of Fe_d have been proposed to interact with the CN⁻ ligand, including S323 as well as a backbone contact involving P234 (Figure 36).²⁶

Steric or hydrophobic interactions of the protein fold with the terminal CO ligand have not yet been explored experimentally, but the influence of methionine M353 on the bridging carbonyl was subject to dedicated analysis. Triggering the formation of H_ox from H_ox-CO by visible light irradiation, Lemon and Peters observed an elongation of the distance between μCO and the sulfur atom of the M353 side chain in CpI.⁴¹⁵ Although both H_ox-CO and H_ox carry a μCO ligand,^485–487 the increased intensity of the μCO band in the CO-inhibited state may be related to the lack of steric pressure and a greater change in dipole moment. Twenty years later, Peters, King, and co-workers revisited CpI by cryogenic XFEL crystallography, assigning similar differences to an oxidized “conformation A” and a reduced “conformation B” (Figure 37).⁵⁵⁴ This includes reorientation of both M353 and sereine S357, the latter which is located close to the [4Fe-4S] cluster in CpI (potential effect are discussed in Section 3.2.4). The presence of a μCO ligand has been recognized as a key requirement for hydrogen turnover by various authors.^503,506,555 Precluding the formation of bridging hydride intermediates,⁵⁰² the orientation of M353 may play a key role in maintaining the CO-bridged, rotated H-cluster geometry. Note that M353 and K358 are part of Loop 2 (Figure 30), which may indicate coupling between Fe_d-CN⁻ and μCO. On the contrary, Table 4 indicates that M353 is moderately conserved at best. [FeFe]-hydrogenases with a glycine or serine residues instead of methionine feature altered catalytic properties (CpIII or TamHydS) or pronounced downshifts of the μCO frequency (CpII and TmHydS).^440,443,554 No such shift was observed in MeHydA, despite the unique variation from methionine to threonine.⁴²² In the case of the sensory hydrogenases TmHydS and TamHydS, multiple variations seem to contribute to the catalytic phenotype, which makes it difficult to pinpoint individual effects. Replacement of methionine with leucine in the standard hydrogenases CpI and CrHydA1 greatly diminished catalytic activity.⁴⁵⁰

Figure 37.

Structural differences between “conformation A” and “conformation B”. The M353 side chain shifts by ~0.4 Å away from μCO and Fe_p-CO in the “reduced” state (equivalent conformation B), which lowers the steric constraints (iv). Concomitantly, the side chain of S357 switches to an orientation toward the [4Fe-4S] cluster and the cysteine ligand C355. This may influence the electron density distribution across the cluster (vi). Drawn after PDB ID 6NAC. Atomic coloring: Fe, orange; S, yellow; C, gray; O, red; N, blue.

Methionine M353 may interact with the μCO ligand while M497 and cysteine C299 are hydrogen-bonded to the ADT ligand (Figure 37). These residues are strictly conserved in Group A [FeFe]-hydrogenases but replaced with glycine, leucine, or sereine in TmHydS and TamHydS (Table 4). The cysteine plays a key role in proton transfer and will be discussed in the next paragraph. Methionine M497 is part of Loop 3, which also bears C499 and C503 (Figure 30). Its function is unclear; it has been discussed as an alternative proton transfer relay or hydrogen-bonding partner that keeps the ADT ligand in position.²⁶⁷ Site-directed mutagenesis of CpI (M497L) and CrHydA1 (M415L) demonstrated the importance of this residue.⁴⁵⁰

3.2.4. Proton Transfer and Proton-Coupled Electron Transfer.

In all hydrogenases, proton transfer is a key determinant of catalytic efficiency.²⁹⁷ Protons enter the enzyme to be consumed in the evolution of H₂ or leave the enzyme upon H₂ oxidation, following a Grotthuß-type mechanism. This demands a network of hydrogen-bonded amino acid residues, water molecules, or hydrophilic cofactors. Proton transfer across the second and outer coordination sphere has been recognized as one of the major differences between biomimetic catalysts and enzymes.^556–558

On the basis of the crystal structure of [FeFe]-hydrogenases CpI and DdH various proton transfer pathways have been proposed.^559–562 As an experimental marker for an involvement of individual amino acids, the analysis was initially restricted to catalytic activity.^454,563 However, when it was discovered that the H-cluster accumulates the H_hyd in variants specifically impaired in proton transfer (Table 5),^456–458 spectroscopic methods could be used to investigate the minutiae of proton uptake and release. X-ray crystallography⁴⁴⁸ and FTIR difference spectra⁵⁶⁴ uncovered a tight hydrogen-bonding network in CrHydA1 and CpI, stretching across >20 Å including a small water cluster W1, and C299, the cysteine residue close to the distal iron ion (Figure 38). On the basis of the first crystal structure of CpI, this pathway was proposed early on.⁴¹³ The reduction and protonation of the H-cluster upon formation of H_red over H_ox has been exploited to trigger proton transfer. The changes in protonation and hydrogen bonding gave rise to distinct FTIR difference spectra, suggesting only subtle changes, for example, the formation of a hydrogen bond between E279 and S319 as well as changes with respect to W1. The lack of difference signals in the SH regime hinted toward a largely preserved hydrogen-bonding environment around C299 and did not support protonation of the ADT ligand, that is, in H_red.⁵⁶⁴ While the residues involved in proton transfer are mostly conserved within Group A, sensory [FeFe]-hydrogenases show drastic variations (Table 4).^440,443 It remains to be understood how protons are transferred in such enzymes; the influence of amino acids replacements on catalytic activity, bias and reversibility²⁹ certainly emphasize the influence of the outer coordination sphere.

Figure 38.

The “catalytic” proton transfer pathway in [FeFe]-hydrogenase CpI. The side chain of M497 may stabilizes the orientation of the ADT ligand by a hydrogen bond (v). Dashed blue lines represent the hydrogen-bonding network (ii) of the proton transfer pathway. Note that a bond between E279 and S319 is unlikely in the oxidized state (dashed gray line) but formed upon reduction and protonation of the cofactor. Drawn after PDB ID 4XDC. Atomic coloring: Fe, orange; S, yellow; C, gray; O, red; N, blue.

Next to the “catalytic” proton transfer pathway, a comparison of several [FeFe]-hydrogenase crystal structures suggested the existence of a second, water-based proton transfer pathway.⁴⁸⁸ Because of the lack of “acidic” moieties in this trajectory (e.g., glutamic acid residues), the difference in apparent pK_a compared the catalytic pathway allows controlling proton transfer to the diiron site and/or the [4Fe-4S] cluster.⁵¹² Figure 39 shows how a chain of water molecules permeates toward the intersection of the F- and H-domains, that is, C193, which coordinates the most proximal [4Fe-4S] cluster and C499, the later which in turn coordinates the H-cluster. Protonation of a cysteine ligand is rather uncommon in nature; however, quantum mechanical calculations of the IR spectra of H_oxH, H_red′, and H_hyd provided excellent agreement with the experimental CO/CN⁻ signature when C499 was protonated in silico.⁵⁶⁵ As demonstrated in Figure 32, redox- and protonation changes in the second coordination sphere of the diiron site (e.g., at the [4Fe-4S] cluster) result in small yet discernible IR band shifts. The spectroscopic characterization is in agreement with the pH-dependent population of H_oxH,⁴⁸⁸ the pH-dependent transition potential of the H_ox/H_red′ couple,⁵⁰¹ and the observation that H_hyd is formed upon H₂ oxidation via H_oxH directly.⁴⁴⁶

Figure 39.

The “regulatory” proton transfer pathway in [FeFe]-hydrogenase CpI. In most crystal structures of CpI, a trajectory of water molecules (red spheres) can be found, putatively connected by hydrogen bonds (ii). This water channel “W2” is located at the intersection of the accessory domain (white) and the catalytic domain (including the most proximal iron–sulfur cluster, [4Fe-4S]_p), leading up to cysteine C499 that coordinates the [4Fe-4S] cluster of the catalytic cofactor. Here, protonation and redox changes influence the diiron site (vi). Drawn after PDB ID 4XDC. Atomic coloring: Fe, orange; S, yellow; C, gray; O, red; N, blue.

Influencing the electron density distribution across the H-cluster, the second proton transfer path has been named “regulatory”.⁴⁸⁸ In general, electrons and protons tend to move in a concerted fashion, minimizing energetic barriers^22–24 and facilitating long-range coupling between electron transfer and proton transfer, an extreme case of outer coordination sphere effects.^447,566,567 In the case of [FeFe]-hydrogenase, PCET shapes the electron density distribution across the H-cluster by locking reducing equivalents at the diiron site upon protonation of the ADT ligand⁴⁹⁸ or formation of a bridging hydride (μH).⁵⁰² Similar ideas have been put forward to explain the pH-dependence of the [4Fe-4S] cluster,^488,501 although this has been disputed.⁵⁶⁸ The existence of two separate proton transfer pathways (Figure 39) complicates the analysis of the pH-dependence and led to conflicting results.⁵¹² These differences have important implications for the understanding of [FeFe]-hydrogenase catalysis, as discussed elsewhere.⁵⁰⁶

3.2.5. Gas Channels in [FeFe]-Hydrogenase.

The mass transport of reactants like H₂, O₂, and CO within [FeFe]-hydrogenase CpI has been studied using molecular dynamics.^{465,569–571} Schulten and co-workers calculated how “the protein’s natural equilibrium dynamic motion on the nano-second time scale can define predetermined pathways for hydrophobic gas transport”, arguing that permanent channels may not be needed.⁵⁶⁹ Figure 40A depicts that this applies to small gases like H₂ in particular. But while CO and O₂ may not access the protein as easily as H₂, Mohammadi and Vashisth suggested defining a “network of pathways” that predicts gas diffusion more accurately⁵⁷⁰ (Figure 40B). This complexity must be met with a statistical approach to site-directed mutagenesis, therefore automated screening protocols^572,573 were employed understanding the influence of “gas filters”, for example, selectively precluding access of O₂ and CO to the active site (Table 5).^452,453 Currently, the role of hydrophobic gas channels in the outer coordination sphere is not well understood. With respect to CO inhibition, there is a remarkable variety that may help identifying certain trends. The rate of CO inhibition among different standard [FeFe]-hydrogenases was found to vary over three orders of magnitude^423,522 while sensory [FeFe]-hydrogenases^440,443 and [FeFe]-hydrogenase complexes show even lower CO affinities.⁴³² Apparently, nonconserved residues remote from the active site tune the reactivty; however, the identity of the responsible amino acids and the underlying mechanistic contributions have yet to be explored. In the second coordination sphere of the H-cluster, the replacement of the V296 and F290 in CrHydA1 decreased the rate of CO inhibition ten-fold, but the effects on O₂ deactivation were not proportional.⁴⁶⁵ The simultaneous variation of two residues located between the H-cluster and the proximal accessory cluster in CpI moderately reduced the rate of inactivation.⁴⁵² As the diiron site has been shown to be the initial target of oxidation, this is an interesting observation that may hint at the role of the [4Fe-4S] cluster in the reaction with “activated” O₂.^529,531,533

Figure 40.

Gas diffusion pathways in [FeFe]-hydrogenase. Representative simulations of 1000 copies of (a) H₂ diffusing out from the H-cluster and (b) of O₂ diffusing out from the H-cluster or from the middle of a previously identified H₂ channel. Possible exits, based on the proximity of the external solution, are highlighted with arrows. Contrary to H₂ diffusion, the O₂ molecules move collectively through the same pathway for a given simulation, though they may employ different pathways for different independent simulations. Reprinted (adapted or reprinted in part) with permission from ref 569. Copyright 2005 Cell Press.

4. FORMATE DEHYDROGENASE

Formate dehydrogenases (FDHs) are a diverse group of enzymes in bacteria, archaea, and eukaryotes that catalyze the reversible two electron and one proton abstraction of formate (HCOO⁻) to produce CO₂ (eq 3).^574,575 In general, these enzymes are involved in diverse metabolic pathways but mostly in the production of CO₂ from formate.⁵⁷⁶ On the basis of the low standard potential of formate oxidation (E° = −420 mV vs SHE) prokaryotes derive energy by coupling it to the reduction to one of several terminal electron acceptors.⁵⁷⁷ Some enzymes, however, were described to act as CO₂ reductases, with a preference for the reaction of reducing CO₂ to formate.⁵⁷⁸

FDHs are divided into two major classes, into the metal-dependent and metal-independent FDH enzymes.⁵⁷⁶ The metal-independent FDHs generally are NAD⁺-dependent and distributed from bacteria to yeast, fungi, and plants.⁵⁷⁹ These enzymes have no redox-active metal centers involved in catalysis, but instead NAD⁺ is used as a cosubstrate and serves as a hydride acceptor for converting formate directly to CO₂.⁵⁸⁰ In these proteins, formate is positioned in proximity to the NAD⁺ cosubstrate to facilitate the hydride transfer step. Since this review is focused on GPMs, only the metal-dependent FDHs will be described in detail.

The metal-dependent FDH enzymes are restricted to bacteria and archaea and contain either molybdenum (Mo) or tungsten (W) as metal ions in the active site.^575,581,582 They have a higher complexity by being composed of several subunits, which contain in addition to the active site bis-metal-binding pterin (MPT) guanine dinucleotide (bis-MGD) cofactor (Figure 41) various additional cofactors, such as diverse Fe–S clusters, FMN, or hemes (Figure 42).⁵⁷⁵

Figure 41.

Bis-metal binding pterin guanine dinucleotide cofactor (bis-MGD) present in formate dehydrogenases. (a) Members of the DMSO reductase family of molybdoenzymes bind the bis-MGD form of the cofactor in the active site. The bis-MGD in formate dehydrogenases coordinates the metal (depicted as M to be Mo or W) by the dithiolene groups from two pterin ligands with the additional ligands being a sulfido group, and a sixth ligand which is either a selenocysteine (Sec) or a cysteine (Cys) from the amino acid backbone. (b) Representative depiction of the bis-MGD cofactor coordination environment in formate dehydrogenases, of E. coli FdhF in the oxidized state (PDB ID 1FDO).

Figure 42.

Schematic overview of cytoplasmic and periplasmic FDHs. Localization of FDHs, their subunit organization and cofactor composition. Shown are the membrane bound respiratory FdnGHI from E. coli, the periplasmic FdhAB and FdhABC from the D. gigas and D. vulgaris Hildenborough, the formate hydrogen-lyase complex formed by (FdhF) and Hyd3 (HycBCDEFG) from E. coli, the cytoplasmic FDH from R. capsulatus, the cytoplasmic CO₂-reducing dimeric FDH (FdhAB) from the acetogen Morella thermoacetica, the HDCR complex from A. woodii (Mo-containing) or from T. kivui (W-containing), and the formylmethanofuran dehydrogenase complex from M. wolfei.

4.1. Metal-Containing Cofactor of FDH

The metal-containing FDHs are members of the dimethyl sulfoxide (DMSO) reductase family of mononuclear molybdenum (Moco) or tungsten cofactor (Wco)-containing enzymes.^583,584 In these enzymes, the active site in the oxidized state comprises a Mo or W ion present in the bis-MGD, which is coordinated by the two dithiolene groups from the two MGD moieties, a protein derived selenocysteine or cysteine ligand and a sixth ligand that is accepted to be a sulfido ligand (Figure 41B).

Selenocysteine-containing enzymes have generally a higher turnover number than the cysteine-containing ones.⁵⁸⁵ The analogous chemical properties of W and Mo, the similar active sites of W- and Mo-containing enzymes, and the fact that W can replace Mo in some enzymes have led to the conclusion that both Mo- and W-containing FDHs have the same reaction mechanism.⁵⁸¹ The question of why in FDH enzymes both metal ions can be found while other enzymes of the DMSO reductase family exclusively function with only one type of metal ion is still not completely understood.^584,586,587 Generally, it was found that W-containing enzymes catalyze reactions with low redox potentials (E° = −420 mV vs SHE), and it was shown that W^IV is a more reducing ion than Mo^IV.⁵⁸⁷ Possibly for these reasons, the conversion of CO₂ to formate was demonstrated to occur preferentially for W-FDH enzymes.^588–590 However, the purified Mo-containing FDHs were also shown to perform CO₂ reduction in the presence of excess reducing equivalents.⁵⁷⁸ In the cell, however, this reaction would thermodynamically not be favored.⁵⁷⁷

4.2. Structural Features of FDH

Overall, there is a great variety in the subunit architecture of metal-containing FDHs. The α-subunit, or the Moco/Wco containing domain, is highly conserved and minimally comprises the Moco/Wco binding site and a [4Fe-4S] cluster proximal to the Moco/Wco cofactor (Figure 42).^575,578 These α-subunits or domains are typically between 80–95 kDa and the overall folds are highly identical. In most cases, the Moco/Wco cofactor and the proximal [4Fe-4S] cluster interface with additional Fe–S clusters, either within the α-subunit or close by in a neighboring subunit.

Metal-dependent FDHs are found either in the cytosol, in the periplasm or are membrane-bound, facing the periplasm (Figure 42).^575,582 The enzymes are generally involved in diverse biochemical pathways and use different electron acceptors/mediators like hemes, ferredoxins, NAD⁺, coenzyme F₄₂₀, or membrane quinols for the reaction of formate oxidation or CO₂ reduction.⁵⁹¹ The formate oxidation activity of FDH is typically followed via the reduction of methyl viologen (MV), benzyl viologen (BV), or NAD⁺. Additionally, the analysis of “catalytic” currents can be followed via protein film electrochemistry (Table 7).^592,593

Table 7.

Oxidation Kinetic Parameters of Structurally Characterized FDH and Formylmethanofuran Dehydrogenase Enzymes^a

organism/type	E. coli FdhF	E. coli FdnGHI	D.gigas FdhAB	D. vulgaris Hildenborough FdhAB	R. capsulatus FdsGBAD	M. wolfeii formylmethanofuran dehydrogenase
catalytic subunit	FdhF	FdnG	FdhA	FdhA	FdsA	FwdB
metal	Mo	Mo	W	W	Mo	W
PDB ID	1FDO, 2IV2	1KQF	1H0H	6SDR, 6SDV	6TGA, 6TG9	5T5I
formate oxidation kinetic parameters	kcat = 2800 s−1	N/A	kcat = 138 s−1	kcat = 3684 s−1	kcat = 36 s−1	kcat = 18 s−1
	KM = 26 mM			KM = 1 μM	KM = 281 μM	KM = 13 μM
	Kd = 100 mM				KM NAD+ = 173 μM	KM MV = 400 μM
	KM BV = 3 mM
ref	594		590	595	596	597

^aReported formate oxidation kinetic parameters for E. coli, D. gigas, and D. vulgaris Hildenborough FDH represent those following the enzymatic reduction of benzyl viologen; formylmethanofuran dehydrogenase follows the enzymatic reduction of methyl viologen, while for R. capsulatus FDH the enzymatic reduction of NAD⁺ is reported.

In organisms that catalyze the oxidation of formate, this reaction is a key process for obtaining energy and reducing equivalents.⁵⁷⁷ The CO₂ formed is usually incorporated predominantly into the Calvin-Benson cycle, as in autotrophic organisms.⁵⁷⁶ Formate is a key metabolite for bacteria, arising as a metabolic product of bacterial fermentations and functioning as a growth substrate for many microorganisms (e.g., methanogens and sulfate-reducing bacteria).^599–601

Formate is also an intermediate in the energy metabolism of several prokaryotes, and formate is used as electron donor during anaerobic respiration in these organisms. Because of the low redox potential of the CO₂/formate couple, formate can be oxidized not only through the aerobic respiratory chain, but it also can serve as an electron donor for the anaerobic reduction of fumarate (E°_{fumarate/succinate} = +30 mV vs SHE) and nitrate (E°nitrate/nitrite = +420 mV vs SHE) or nitrite (E°nitrite/NO = +375 mV vs SHE).⁵⁸¹

Some FDHs are in complex with hydrogenases, like the formate-hydrogen-lyase (FHL) systems from E. coli, Pectobacterium atrosepticum, or Clostridium carboxidovorans.^602–605 Physiologically, the E. coli FHL is a membrane-bound system involved in formate oxidation and H₂ evolution under fermentative growth conditions.⁶⁰⁶ This system contains the formate dehydrogenase subunit FdhF, and a membrane-bound, cytoplasm-faced [NiFe]-hydrogenase (Section 3.1).⁶⁰⁷ Similar coupling of FDH and hydrogenase subunits are observed in some CO₂ reductases that catalyze the reduction of CO₂ to formate with the simultaneous and direct oxidation of H₂. For example, the enzyme from Acetobacter woodii is a tetramer, with one Moco/[4Fe-4S] cluster subunit, two subunits with four [4Fe-4S] clusters each, and one [FeFe]-hydrogenase subunit (Section 3.2).^431,608 A W-containing homologue has been identified in Thermoanaerobacter kivui.²⁶⁶

4.2.1. Carbon Dioxide Reduction and Coupled Catalysis.

A distinct class of FDH-associated enzymes are those found in methanogenic archaea that are involved in hydrogenotrophic CO₂ fixation, en route toward the generation of methane (CH₄).^609,610 The first enzyme involved in reductive CO₂ fixation is the enzyme formylmethanofuran dehydrogenase, which is biased toward the reduction of CO₂ and the coupling of the intermediate product formate to the C₁-carrier methanofuran without the requirement of ATP.^611,612 Furthermore, some organisms that undergo methylotrophic or aceticlastic methanogenesis have an ability to produce CO₂ as a final step, also employing a formylmethanofuran dehydrogenase.^613,614 The basis for this biochemical transformation, until recently has been largely speculative. Similar to FDHs described above, formylmethanofuran dehydrogenases are able to coordinate a bis-MGD cofactor with either Mo or W;^{597,609,615–620} the enzyme also exhibits a distinct dinuclear Zn amidohydrolase motif that is involved in the condensation of formate and methanofuran.⁶¹² The Methanothermobacter strains M. marburgensis and M. wolfeii harbor distinct operons of formylmethanofuran dehydrogenase that depend on the presence of tungstate (WO₄²⁻) or molybdate (MoO₄²⁻).^{597,615–618,621} The former is constitutively expressed in both organisms, whereby a Mo-substituted isoenzyme has been also isolated.^620,622,623 Similarly, the genome of the hyperthermophilic methanogen Methanopyrus kandleri encodes for two W-containing formylmethanofuran dehydrogenases, with one that is also only expressed in the presence of Se.⁶²⁴ Interestingly, distinct molybdopterin dinucleotides have been characterized for formylmethanofuran dehydrogenases from M. marburgensis, such as coordinating a molybdopterin adenine or hypoxanthine dinucleotide.⁶²⁵

4.2.2. Oxygen Sensitivity.

Most FDH enzymes characterized so far are described to be inactivated in the presence of O₂ (referred to as oxygen-sensitive enzymes).^626–629 To prevent damage by O₂ and to stabilize the enzyme, inhibitors like azide (N₃⁻) or nitrate (NO₃⁻) are added during the purification of the enzyme, inhibitors that are known to significantly increase the stability of most FDH enzymes.⁶³⁰ FDHs that exhibit activity in the presence of O₂ with inhibitor present (referred to as oxygen-tolerant enzymes) were described for the Mo-containing enzymes from R. capsulatus and C. necator and the W-containing enzyme from D. vulgaris Hildenborough.^{595,596,631,632} What contributes to the higher stability of some enzymes (e.g., from various Desulfovibrio specimen^{590,633–636} and Methylobacterium extorquens^609,637) in the presence of O₂ without inhibitor present, while other enzymes are extremely sensitive to the exposure of O₂ (e.g., E. coli FdhF⁶²⁶) is not known so far. While hydrogenases act as oxygenases reducing O₂ to reactive oxygen species (Section 3.2) or water (Section 3.1), the O₂ sensitivity of FDH seems to follow other molecular principles. Recently, it was shown the enzyme from R. capsulatus is mainly sensitive to O₂ in the absence of N₃^–.^638,639 By extended X-ray absorption fine structure (EXAFS) spectroscopy, it was revealed that the exposure of the enzyme to O₂ results in a heterogeneous ligandation at the Mo ion and to the exchange of the terminal sulfido ligand by an oxo ligand.^638,639 Azide stabilizes the sulfido ligand, suggesting that N₃⁻ binds in vicinity to the sulfido ligand and hinders thereby the access of O₂ to the active site sulfido group.

4.2.3. Cofactor Biosynthesis and Moco Sulfuration.

The biosynthetic maturation process of FDHs, in conjunction with the number of redox cofactors, is critical to obtain a bis-MGD cofactor with the correct set of ligands at the active site.⁶⁴⁰ FDHs, in conjunction with most Mo- and W-containing enzymes coordinate multiple redox centers in addition to the prosthetic Moco or Wco.⁵⁷⁵ The maturation of bacterial bis-MGD-containing enzymes is a complex process leading to the insertion of the bulky bis-MGD cofactor into the apoenzyme. Most of these enzymes were shown to contain a specific chaperone for the insertion of the bis-MGD cofactor.⁶⁴⁰ FDH together with its molecular chaperone seems to display an exception to this specificity rule, since the E. coli chaperone FdhD has been proven to be involved in the maturation of all three FDH enzymes present in the cell.^641–643 For the FDH FdsGBAD from R. capsulatus the two proteins FdsC and FdsD were identified to be essential for enzyme activity, but FdsC is not a subunit of the mature enzyme.⁵⁹⁶ While FdsD is a subunit of the (FdsGBAD)₂ heterodimer,⁵⁹¹ it has only counterparts in some O₂-tolerant FDHs.⁵⁷⁵ While an enzyme purified in the absence of FdsD was devoid of Moco, the role of this subunit is not clear so far.^596,644 On the other hand, FdsC shares high amino acid sequence identity to E. coli FdhD, the chaperone for the cytosolic E. coli formate dehydrogenase FdhF and the two membrane-bound FDHs FdnGHI and FdoGHI from E. coli.^644,645 FdhD was reported to be involved in the formation of the essential terminal sulfido ligand at the Mo ion facilitating the reaction between the L-cysteine desulfurase IscS⁶⁴² and FdhF-bound bis-MGD, a mechanism which is essential to yield active FdhF.⁶⁴² Initially, it was proposed that the two cysteine residues C121 and C124 in FdhD are involved in the sulfur transfer reaction from IscS further onto the bis-MGD in FdhF. For sulfuration of bis-MGD, FdhD specifically interacts with IscS in E. coli. It has been suggested that IscS transfers the sulfur from l-cysteine to FdhD in form of a persulfide. Located in a conserved CXXC motif of FdhD, cysteine C121 and C124 were proposed to be involved in the sulfur transfer process from IscS to bis-MGD by Magalon and co-workers⁶⁴² as well as Walburger, Arnoux, and co-workers.⁶⁴³ FdhD was cocrystallized in complex with GDP suggesting direct binding of bis-MGD to FdhD; however, it was not possible to prove that the cofactor is bound in an active form.⁶⁴³ A study using R. capsulatus FdsC provided evidence that the cofactor is indeed bound in the active form containing the terminal sulfido ligand, which was revealed by the insertion of MGD cofactor into TMAO reductase,⁶⁴⁴ producing an active enzyme.⁶⁴⁵ Further, a phylogenetic study revealed that the CXXC motif is not conserved among FdhD-like chaperones and by site-directed mutagenesis it was proven that the two cysteines in the CXXC motif are not essential, putting the mechanism of the persulfide-transfer involving FdhD as a sulfurtransferase into question.⁶⁴⁵ It therefore remains possible that IscS directly transfers the persulfide sulfur to the Mo ion in FdhD-bound bis-MGD, notably without the involvement of additional cysteine residues. The source of electrons for the reductive cleavage of the persulfide is also unknown. Directly comparing the chaperones FdhD from E. coli and FdsC from R. capsulatus in the same study revealed that their roles in the maturation of FDH enzymes from different subgroups can be exchanged.⁶⁴⁵ In summary, the binding of sulfido-containing bis-MGD to the chaperone is a common characteristic of FdhD-like proteins and ensures insertion of a competent active site cofactor for catalysis.

4.3. Role of Sulfido- and Amino Acid Ligands

The first coordination sphere at the active site cofactor of FDHs includes the sulfido ligand and the cysteine or selenocysteine ligand (Figure 41A).^627,646 The sulfido group was shown to act as a hydride acceptor from formate in the reaction of formate oxidation.⁶⁴⁷ This reaction step was mainly established by EPR spectroscopy, showing that the C_α hydrogen atom from formate is transferred to a ligand in the first coordination sphere, which resulted in the binding of a strongly coupled and solvent exchangeable proton, with a hyperfine constant of 20–30 MHz in the Mo^V state.^648–650 Similar hyperfine constant values were determined in the Mo-containing enzyme xanthine oxidase, where a hydrogen is transferred from the C₈ hydrogen atom of xanthine to the active site sulfido ligand at the Mo ion.⁶⁵¹ Similar to xanthine oxidase, previous studies showed that FDHs exhibited sensitivity to cyanide (CN⁻).⁶⁵² In a recent XAS study on R. capsulatus FDH it was shown that the sulfido ligand was lost forming thiocyanate (SCN⁻), which represents another similarity with xanthine oxidase.⁶³⁸ The resulting desulfo enzyme harbors an oxo group in place of the sulfido group, which is not able to act as hydride acceptor. However, examples of FDHs that exhibit apparent or partial CN⁻ insensitivity are known, including W-FDHs from D. desulfuricans^648,653 and D. gigas,⁶³⁶ as well as the W-containing isoenzyme of formylmethanofuran dehydrogenase from M. marburgensis.⁶¹⁷

Formate dehydrogenases catalyze formate oxidation and CO₂ reduction at the bis-MGD with the involvement of amino acids in the first and second coordination sphere. Of these, the first amino acid of importance is the cysteine or selenocysteine residue that serves as a covalent linkage between the bis-MGD cofactor and the protein (Figure 41). Given the diversity of FDHs, the preference to encode for a selenocysteine residue has been discussed to reflect an evolutionary advantage. Selenium as being essential for bacteria was discovered by Pinsent in 1954,⁶⁵⁴ whereby the activity of FDH was dependent on the presence of Se, and was later discovered to be in the form of a single SeCys residue encoded by the unique codon UGA.⁶⁵⁵ Selenocysteine offers a lowered pK_a and improved performance as a nucleophile relative to cysteine that results in enzyme exhibiting an increased catalytic reactivity.⁶⁵⁶ These considerations compelled Axley and co-workers in the early 1990s to investigate the essentiality of the residue in E. coli FdhF, by performing site-directed mutagenesis.⁵⁸⁵ They showed that exchange of the active site selenocysteine to cysteine (U140C) in E. coli FdhF resulted in a reduction of k_cat by a factor of 300, and a decrease in K_{M formate} relative to a constant K_M with BV as an electron acceptor (Tables 7 and 8). While the reduction in k_cat is consistent with differences in chemical properties of Se vs S, it also affected the relative affinity of the substrate formate at the active site.⁵⁸⁵

Table 8.

Amino Acid Comparison and Oxidation Kinetic Parameters of Active Site Variants of Structurally Characterized FDH and Formylmethanofuran Dehydrogenase Enzymes^a

organism/type	E. coli FdhF	E. coli FdnGHI	D. gigas FdhAB	D. vulgaris Hildenborough FdhAB	R capsulatus FdsGBAD	M. wolfeii formylmethanofuran dehydrogenase	Function	Formate Oxidation Kinetic Parameters	ref
conserved amino acids	U140	U196	U158	U192	C386	C118	substrate binding, catalytic efficiency	E. coli (FdhF) U140C: kcat = 9 s−1, KM = 9 mM, KD = 5 mM, KM BV = 2 mM	585
	H141	H197	H159	H193	H387	H119	proton shuttling, channel gating	R capsulatus H387M: kcat = 34 s−1, KM = 3.6 mM R capsulatus H387K: kcat = 1 s−1, KM = 28 mM	598
	R333	R446	R407	R441	R587	R288	substrate binding, channel gating	R capsulatus R587K: kcat = 11 s−1, KM = 362 mM	598
	M157	M213	M175	M209	G403	I133	hydrophobic channel direction	N/A
	V338	L450	V412	V446	V592	V293	hydrophobic interaction at bis-MGD	N/A
	K44	K94	K56	K90	K295	I37	electron transfer to Fe–S relay	N/A
	Q335	H448	E409	E443	Q589	H290	gating proton shuttling	N/A
	V145	V201	V198	V197	G391	V123	hydrophobic interaction with active site Arg	N/A
	Q339	Q452	Q413	Q447	Q593	N297	hydrophobic interaction with active site Arg	N/A

^aAmino acid residues in the primary or secondary coordination sphere by which a role has been assigned or postulated are noted. Enzymatic activity of reported active site variants for respective organisms follow the same electron acceptor used as reported for the WT enzyme, as reported in Table 7.

Despite the preliminary studies investigating the role of selenocysteine U140 in E. coli FdhF, the role of the amino acid ligand in catalysis is a matter of debate and not clear to date. As discussed in detail below, mainly two models exist, one in which the proteinaceous ligand remains bound to the Mo ion during catalysis and therefore has no direct role in the catalytic mechanism.⁶³² In the other model, the amino acid ligand is displaced from the Mo ion, thereby providing a coordination site for the substrate and it temporarily serving as a second coordination sphere ligand.⁶⁴⁶ While in the Mo^V state of E. coli FdhF, it was shown that the Mo ion is hexacoordinated with the amino acid ligand,⁶²⁹ different data exist on the reduced Mo^IV state. In a reinterpretation of the crystallographic data of the formate-reduced E. coli FdhF enzyme it was suggested that the selenocysteine residue is dissociated from the Mo ion and is shifted 12 Å away.⁶⁴⁶ However, a recent crystal structure of the formate-reduced D. vulgaris Hildenborough W-FDH showed a hexacoordinated metal ion.⁵⁹⁹ It needs to be considered that in these structures the oxidation state of the Mo/W ion is speculative (given the presence of multiple redox cofactors) and that none of the structures was solved in a high enough resolution to firmly establish the coordination of the substrate formate. On the contrary, different results have been obtained in XAS studies using different enzyme sources. In a recent study of the R. capsulatus enzyme it was suggested that in the Mo^V and the Mo^VI state the cysteine is a ligand to the Mo ion, while in the formate-reduced Mo^IV state (in the absence of N₃^–), the cysteine was displaced from the metal and instead a bond reflecting a Mo–O distance was present.⁶³⁸ Contrary to this study, EXAFS data of the E. coli and D. desulfuricans enzymes in the oxidized and reduced states were interpreted with a hexacoordinated Mo ion.^657,658 Here, it needs to be pointed out that the data of the E. coli were derived in the presence of high concentrations of the inhibitor N₃⁻, which influence the binding of formate (discussed in more detail below). As for the D. desulfuricans enzyme the presence of the sulfido ligand is not clear, since it was not assigned by the EXAFS study.

To probe the displacement of the amino acid ligand, inhibition studies with iodoacetamide were performed, an alkylating agent that reacts with “free” thiols. For both the E. coli and the R. capsulatus enzymes an alkylation of the selenocysteine or cysteine ligand was identified, notably only in the formate-reduced and NO₃⁻/N₃⁻ inhibited enzymes, but not in the oxidized enzyme.^585,598 However, no alkylation of the selenocysteine residue could be identified in the W-containing D. vulgaris enzyme.⁵⁹⁹ The E. coli FdnGHI enzyme likewise is not alkylated when incubated with formate.⁶³⁰ Heider and Böck speculated that this difference might be correlated to electron transfer from the reduced metal center to the [4Fe-4S] cluster and further to the final electron acceptor, since in some enzymes the Mo^V state is more stable than for other enzymes.⁶⁵⁹ A fast reoxidation of the enzyme from the Mo^IV to the Mo^VI state might protect the enzyme from alkylation, while in enzymes with a long-lived reduced state, the amino acid ligand has a higher probability to be alkylated. In conclusion, while hexacoordination of the Mo ion in its Mo^V and Mo^VI states is established now, the coordination of the Mo^IV state needs to be clarified in future studies.

4.3.1. Amino Acids in the Second Coordination Sphere.

In addition to the active site selenocysteine or cysteine residue, primary components of the second coordination sphere of all metal-containing FDHs include a highly conserved arginine and a histidine residue (Figure 41B) that have been proposed to ensure optimal substrate binding to and proton transfer away from the active site.⁵⁷⁵ Site directed mutagenesis studies on R. capsulatus FDH revealed that in the H387M variant (equivalent to H141M in Figure 41B), the k_cat remained similar to that of wild-type enzyme, suggesting a similar orientation of H387 and M387 (Tables 7 and 8).⁵⁹⁸ However, in this variant, the pH optimum was lowered to 7.5 with a 19-fold increase in K_M^formate. The replacement of H387 with a lysine reduced the activity to 3% of that of the wild-type enzyme, accompanied by an increase in K_M^formate and a shift of the pH optimum to 8.0. When R587 (equivalent to R333 in Figure 41B) was replaced by a lysine, retaining the positive charge of the side chain, ~ 30% of wild-type activity was obtained, however, accompanied by an increase in K_M^formate and a shift in the pH optimum to 8.0 (Tables 7 and 8). These results show that H387 influences the pH optimum of the reaction, while the drastic increase in K_M^formate reveals a role for R587 in substrate binding at the active site.

4.3.2. Substrate Channels in the Outer Coordination Sphere.

The first FDH structure characterized by X-ray diffraction was the monomeric FdhF protein from E. coli, which was solved both in the oxidized (2.8 Å resolution) and formate-reduced (2.3 Å resolution) state (Figure 43A and Figure 44A).⁶²⁷ The active site Mo ion was initially interpreted to carry a hydroxyl ligand; however, later the Mo–OH bond was reinterpreted as being a Mo-SH one.⁶⁴⁶ Two other catalytically important amino acids in the active site were identified to be H141 and R333 (Figure 41B). These two residues are strictly conserved in all Mo- and W-containing FDHs, and their involvement in substrate binding and/or proton abstraction from the substrate has been proposed.⁵⁹⁸ After substrate oxidation, electrons are transferred from Mo^IV to the nearby [4Fe-4S] cluster. Here, a conserved lysine K44 is located between one pterin ring of bis-MGD and the [4Fe-4S] cluster (Figure 41B), and an involvement of this residue in the electron transfer reaction has been suggested (Figure 43A and Figure 44A).⁶⁴⁶

Figure 43.

Structural comparison of FDH undergoing redox change. Structures correspond to (a) E. coli FdhF oxidized (PDB ID: 1FDO) and formate-reduced (PDB ID: 2IV2), (b) D. vulgaris Hildenborough FDH oxidized (PDB ID: 6SDR) and formate-reduced (PDB ID: 6SDV), and (c) R. capsulatus FDH oxidized (PDB ID: 6TGA) and NADH-reduced (PDB ID: 6TG9). Participatory amino acids and bound H₂O and glycerol molecules (where appropriate) in the substrate-binding cleft of FDH are depicted in dark salmon in the oxidized state, and deep teal in the reduced state. The structural effect of specific amino acids accompanying redox changes are depicted with black arrows. The appearance of highlighted H₂O molecules in the reduced state are indicated with red arrows.

Figure 44.

Structural characterization of the active site of Mo/W–FDHs and formylmethanofuran dehydrogenases. Structures correspond to (a) E. coli FdhF (PDB ID 1FDO), (b) E. coli FdnG (PDB ID 1KQF), (c) D. gigas FdhA (PDB ID 1H0H), (d) D. vulgaris Hildenborough FdhA (PDB ID 6SDR), (e) R. capsulatus FdsA (PDB ID 6TGA), and (f) M. wolfeii formylmethanofuran dehydrogenase FwdABCD (PDB ID 5T5I). Key amino acid residues and bound H₂O molecules are depicted, where appropriate. Predicted substrate and gas channels, are noted for each enzyme; the hydrophobic channel predicted for D. vulgaris Hildenborough FDH was predicted from the reduced enzyme (PDB ID 6SDV). Subunits harboring the bis-MGD cofactor are labeled with black text. The hydrophilicity or hydrophobicity of the channel is noted. Channels were constructed via the CAVER plugin in PYMOL.

The second structure of an E. coli FDH enzyme is the one of FdnGHI (EcFDH-N) which was solved as a “mushroom shaped” (αβγ)₃ heterotrimer of trimers, with subunits of 113, 32, and 21 kDa, respectively.⁶⁶⁰ The crystal structure of FDH-N showed that the catalytic subunit of FdnG, composed of 982 amino acids, has an overall fold similar to the structure of FdhF. The substrate access funnel is opening down to the active site Mo ion, constructed of predominantly positively charged residues thus facilitating substrate binding, and likewise appears similar to FdhF (Figure 44B). The Mo ion of oxidized FdnG binds the bis-MGD cofactor, while additionally coordinated by U196. A sixth ligand in the oxidized enzyme was initially modeled as a hydroxy species (at 2.2 Å), but in light of the reinterpretation of the FdhF structure, the sixth ligand is most likely a terminal sulfido species.⁶⁴⁶ Overall the Mo coordination sphere closely resembles that seen in oxidized FdhF (Figure 44A), with an arginine and a histidine present in the active site, thus implying a conserved mechanism involving the same amino acids for formate oxidation.

At the active site of D. gigas FDH, the W ion is bound to two MPT moieties, to a selenocysteine, and to an inorganic S or O ligand (Figure 44C).^634,635 The crystallographic data favor a S atom for the sixth ligand, although the resolution of the data did not permit unambiguous discrimination between O and S. As has been shown for other enzymes from the same family, the bis-MGD cofactor is buried in the protein, stabilized by an extensive network of hydrogen bonding interactions. The conserved lysine K56 bridges the pterin of the bis-MGD cofactor and the proximal [4Fe-4S] cluster. In vicinity of the buried W ion H159 and R407 are found (Figure 44C). Formate access to the active site could be facilitated by a channel of charged residues, including H159, R156, H423, H963, H376, K91, K444, K445, A587, and A407. CO₂ release from the active site might be facilitated by a hydrophobic channel, including V412, H159, W730, and Y428.

The crystal structure of the tungsten and SeCys containing FdhAB from D. vulgaris Hildenborough was solved in the oxidized and reduced form (Figure 43B and Figure 44D).⁵⁹⁹ In the oxidized form, the W ion is hexacoordinated to four S atoms from the two dithiolenes, a Se atom from U192 and an indeterminate sulfur ligand (W-SH or W = S). The catalytic pocket and respective entrance channel are positively charged, and three glycerol molecules haven been modeled in this cleft. In the vicinity of the W ion there are two strictly conserved amino acids: H193, establishing a π-interaction with the Se atom of U192, and R441, which establishes a hydrogen bond with a cocrystallized glycerol molecule. A similar π-interaction is observed in the FDH from D. gigas, among respective residues H159, U158, and R407 (Figure 44C).⁶³⁴ The glycerol closest to the W ion is hydrogen-bonded by S194 and H457, the second glycerol is oriented by R441, T450, and a water molecule oriented by Y462. A third glycerol is found at the channel entrance, oriented by a backbone contact including G207 and water molecules that connect with W492 and R206. These glycerol molecules probably occupy formate binding sites and may shed light on the formate shuttling mechanism.

In the formate-reduced form of FdhAB from D. vulgaris Hildenborough, a conformational rearrangement was observed, involving I191, U192, and H193 (Figure 43B).⁵⁹⁹ The U192 C_α atom shifts up to 1.00 Å from the position occupied in the oxidized form while the Se atom remains coordinated to the W ion. Further, the catalytically relevant H193 moves away from the active site and a water molecule occupies the position of the imidazole ring, establishing a hydrogen bond with G442. Together, H193 and R441 are oriented in a cis conformation, relative to the trans orientation observed in the oxidized structure and other FDH structures. This shift also resulted in assignment of a hydrophobic channel distinct from the hydrophilic one identified in the oxidized state (Figure 44D). The U192 loop shift is propagated to neighboring amino acids from different domains that promotes the tilt of the side chains of W533, F537, and F160. Overall, neither formate nor CO₂ were identified in the reduced structure, leaving the possibility that the reduced structure was solved after CO₂ release from the enzyme.

The cryogenic electron microscopy (cryo-EM) structure of R. capsulatus FDH revealed a 360 kDa dimer of FdsABGD heterotetramers in which the FdsD subunit is bound to the FdsA subunit (Figure 43C and Figure 44E).⁵⁹¹ In the oxidized structure, the active site structurally resembles that of E. coli FdhF with six ligands coordinated to the Mo ion represented by the two dithiolene groups of the bis-MGD molecule, the C386 ligand, and a sulfido ligand that is oriented toward V592 (Figure 44E). Additionally, the NADH-reduced cryo-EM structure was also solved; however, the protocol did not result in a full reduction to the Mo^IV state, instead the enzyme was only partially reduced to the Mo^V state.⁵⁹¹ In the Mo^V state, no structural changes at the bis-MGD pterin and dithiolenes were observed and C386 was present as a ligand to the Mo ion (Figure 43C). The inability of NADH to completely reduce the enzyme might either be dependent on unfavorable redox potentials of the cofactors for the back reaction or is based on the presence of 10 mM N₃⁻ that prevent complete reduction of the enzyme. Interestingly, in none of the structures clear evidence for stochiometric binding of N₃⁻ in any particular location of the EM map could be assigned.

Two reactant channel could be assigned in the cryo-EM structures, starting at the Mo ion and separating at the active site residue R587 into different exits (Figure 44E). The pore of the shorter channel is mainly formed by polar and charged residues suggesting channeling of hydrophilic substrate (i.e., formate) from an entry site near FdsD to the active site. The channel-forming residues are similar to those in oxidized FdhF. The second channel bears predominantly hydrophobic residues suggesting the possibility of CO₂ transport to or away from the active site. FdhF contains a similar hydrophobic channel, which is however blocked by V145 and M157 in place of glycine residues in R. capsulatus FDH. On the basis of the positioning of R587 at the cross section of both channels, it has been suggested that this residue might control gate opening to each channel thereby facilitating efficient catalysis.^591,598 Intriguingly, the glycine residues are conserved in NAD⁺-dependent FDHs while other FDHs display larger hydrophobic residues at this position.

The structural characterization of the W-containing formylmethanofuran dehydrogenase from M. wolfeii has provided considerable insight toward the importance of the second and outer coordination spheres in the FDH reaction (Figure 44F).⁶¹² The enzyme, crystallized as a dimer and a tetramer of the FwdABCDFG heterohexamer, respectively, showed a modular composition of the catalytic subunits and 46 iron–sulfur clusters bound in the tetrameric supercomplex. Interestingly, an internal 43 Å-long hydrophilic water cavity separates the bis-MGD and the amidohydrolase active sites (Figure 44F). The interface of the subunits FwdB (housing the bis-MGD) and FwdA (housing the Zn₂-amidohydrolase site) that comprises the hydrophilic channel is distinct from the FdsD–FdsA interface characterized in the cryo-EM structure of RcFDH.⁵⁹¹ However, several of the second coordination sphere residues that constitute this channel are similar, including H119, R288, N113, and N297 (MwFwdA numbering). Interestingly, a H₂O molecule was found in proximity to the active site residue R288 at the interface of the hydrophobic and hydrophilic channels (Figure 44F). The lack of access by which exogenous formate might be expected to bind is consistent with the substantial differences in the substrate affinity among the isoenzymes from M. wolfeii and the Mo-containing enzyme from Methanosarcina barkeri.^616,617

Despite being an enzyme that is biased toward the reduction of CO₂, features of the hydrophobic channel associated with CO₂ binding are not so different relative to structurally characterized FDH counterparts (Figure 44). Similar to RcFDH, DvhFDH, and DgFDH, the hydrophobic channel principally begins at the interface of the active site H387 and R587 residues (RcFDH numbering) as an elaborate cavity characterized crystallographically without H₂O molecules bound.⁶¹² While the cryo-EM map of the RcFDH structure⁵⁹¹ does not allow for a direct comparison to assess the hydrophobicity of the channel, the predicted channel is nevertheless similar to that of the crystallographically characterized DvH FDH enzyme.⁵⁹⁹

The hydrophilic channel was proposed to represent a conduit by which protons could be delivered to and from the bis-MGD cofactor in formylmethanofuran dehydrogenase.⁶¹² This is insightful, with respect to clarifying the distinct, requisite pathways to deliver protons, electrons, and substrates. The proton transfer pathway starts at the active site H119 and C118 residue in vicinity of the distal pterin, whereby the protons are aided by numerous H₂O molecules toward H290 and C91 (MwFwdA numbering, Figure 44F).⁶¹² By comparison, other structurally characterized FDHs also exhibit residues at these positions that could in principle aid in proton shuttling to the active site.^591,599,627

4.4. Inhibition of FDH in the First and Second Coordination Sphere

Inhibition of different FDH enzymes has been studied intensively using various inhibitors, and among them the most prominent one is N₃^–, which is an isoelectronic molecule to CO₂. Azide has been thought to be a transition analogue for the FDH reaction, and is most often used during the purification of FDHs to protect the enzyme from oxidative damage and loss of the sulfido ligand.⁶³⁸ Inhibition of FDH using pseudohalides of varying electron donor strengths (e.g., N₃⁻, OCN⁻, NO₂⁻, and NO₃⁻) have been precursorily characterized in solution-based, in vitro assays.^{594,626,652,661–664} While all inhibitors were more potent to inhibit formate oxidation, N₃⁻ and OCN⁻ are generally reported to be mixed-type inhibitors, revealing two binding sites with one being competitive and the other binding site being noncompetitive. By comparison, NO₂⁻ and NO₃⁻ have been shown to be competitive inhibitors using formate as a substrate.⁶⁶¹ In the crystal structure of E. coli FdhF, NO₂⁻ was proposed to be coordinated to the Mo ion.⁶²⁷

Recently, in a chronoamperometry study of E. coli FdhF these pseudohalide inhibitors were used to study their binding to the reduced and oxidized enzyme.⁶⁶⁵ The inhibition studies revealed that inhibitor binding is oxidation state-dependent, since potential- and formate-dependent results could be ascribed to differences in population of the oxidation state Hirst, Reisner, and co-workers concluded that N₃^–, OCN^–, and NO₃⁻ bind directly to a vacant coordination site of the Mo ion and that these inhibitors bind more tightly to Mo^VI than to Mo^IV.⁶⁶⁵ In addition, it further was suggested that the selenocysteine residue in E. coli FdhF has to dissociate from the Mo ion to generate the vacant position to which the inhibitors can bind.

4.5. Catalytic Mechanism of Formate Oxidation

While metal-dependent FDH enzymes have been studied for several decades and the E. coli FdhF enzyme was among the first molybdoenzymes to be crystallized,⁶²⁷ details of the reaction mechanism, involving the first and second coordination sphere remain poorly understood, and the catalytic mechanism of formate oxidation is still unclear. As will be described below, the numerous mechanisms proposed for FDH reflect a lack in clarity of the coordination environment and oxidation state of the bis-MGD cofactor. Overall, the formate oxidation mechanism is believed to be similar between Mo-containing and W-containing FDHs.

To form CO₂ from formate, FDH enzymes have to catalyze the transfer of one proton and two electrons (eq 3). Currently, it is assumed that this reaction is not an O atom transfer reaction, as characteristic of many Mo/W-bis-MGD family enzymes.⁵⁸³ It has been proposed that the reaction product is CO₂ rather than bicarbonate (HCO₃⁻), the product expected in an O atom transfer mechanism. This assumption was first proposed by Thauer and co-workers in 1975 in the characterization of the FDH from Clostridium pasteurianum⁶⁶⁶ and has been supported in 1998 by Khangulov and co-workers on EcFDH–H by determining the product formation using of ¹³C-labeled formate in ¹⁸O-enriched water.⁶⁴⁷

A recent study by Fourmond and co-workers applied an electrochemical approach on the DvhFDH enzyme to confirm that the substrate of FDHs for the backreaction of CO₂ reduction is CO₂ rather than HCO₃⁻.⁵⁹² The method consisted in monitoring the changes in activity that occur during the slow relaxation of the equilibrium between CO₂ and bicarbonate.⁵⁹² However, as already pointed out in 1968 by Benedict, Cooper, and co-workers,⁶⁶⁷ it still remains possible that these observations are rather a reflection of the binding of the preferred molecule to the active site, rather than what is happening directly at the active site (i.e., catalysis). It therefore might be possible that a charged HCO₃⁻ molecule is hindered from entering the active site via the hydrophobic channel, while CO₂ could enter easily. At the active site cofactor, CO₂ then could react with H₂O to form HCO₃^–, which would provide the substrate for the reaction of CO₂ reduction. This option has so far been neglected in most studies and should be considered, since in some structures a water molecule has been identified in the active site bound in vicinity to the conserved arginine residue, as discussed above (Figure 44). It also needs to be considered that H₂¹⁸O exchange in the active site of the enzyme might be slow and therefore the reason why Khangulov and coworkers did not find¹⁸O-labeled CO₂ in their approach.⁶⁴⁷ In this respect, approaches that factor second and outer coordination sphere effects regarding CO₂ production or formate diffusion would provide significant insight toward clarifying the formate oxidation/CO₂ reduction mechanism.

Overall, on the basis of different experimental data and theoretical calculations, different catalytic mechanisms for FDHs have been proposed, which we want to describe in more detail (Figure 45). As explained above, a central question is whether the amino acid ligand (selenocysteine or cysteine) remains coordinated to the Mo- or W metal during catalysis or whether it dissociates to provide a vacant site for substrate binding. Further questions relate to the involvement of the second coordination sphere residues that also serve a critical role in the reaction mechanism.

Figure 45.

Proposed reaction mechanisms for formate oxidation by FDH enzymes. (a) Mechanism proposed by Heider and Böck in 1993; (b) mechanism proposed by Sun and co-workers in 1997; (c) mechanism proposed by Raaijmakers and Romao in 2006; (d) mechanism proposed by Cerqueira, Gonzales, Moura, and co-workers in 2011; (e) mechanism proposed by Zampella and co-workers in 2012; (f) mechanism proposed by Dong and Ryde in 2018; (g) mechanism proposed by Hille and co-workers in 2016; details are given in the text. Active site residues without a defined role in an individual mechanism are depicted in gray.

The first mechanism was proposed by Heider and Böck in 1993, before the crystal structure of E. coli FdhF was described (Figure 45A).⁶⁵⁹ In this mechanism, the binding of only one pterin ligand is proposed (the existence of two dithiolene MPT groups as ligands to the metal ion were first observed in the crystal structure of the Pyrococcus furiosus W-containing aldehyde-ferredoxin-oxidoreductase in 1995),⁶⁶⁸ one oxo ligand, and the fifth ligand being the selenocysteine residue. The authors suggested that binding of formate to the active site displaces the selenocysteine ligand from the Mo ion and the selenide then initiates formate oxidation by a nucleophilic attack to the Mo-bound formyl group. Carboxyl transfer to the selenocysteine residue would releases CO₂ in the next step. Following the first crystal structure of E. coli FdhF solved by Sun and co-workers in 1997⁶²⁷ and concurrent XAS studies,⁶⁵⁷ it was suggested that the Mo ion is coordinated by a hydroxyl group instead of a terminal sulfur atom (Figure 45B). The subsequent identification of a terminal sulfido group in the first coordination sphere of Mo and the observation that the loop containing U140 was shifted away (12 Å) from the Mo ion in the formate-reduced E. coli FdhF led to a reformulation of that mechanism (Figure 45C).⁶⁴⁶ Romao and Raaijmakers suggested that upon formate binding to the active site, U140 is displaced from the Mo ion. After formate binding to the Mo ion through an O atom, The U140 selenol anion (Se⁻, with R333 as counterion) would deprotonate formate at the C_α atom and two electrons are transferred to the metal center simultaneously to form Mo^IV. Subsequently, the proton is transferred to H141, which changes its conformation in the reduced enzyme. After CO₂ release and rebinding of the selenocysteine ligand, the Mo ion is in a hexacoordinated form, with the sulfido group as the axial ligand.⁶⁵⁷ The catalytic cycle would be completed with the oxidation of Mo^IV to Mo^VI, via intramolecular electron transfer to the [4Fe-4S] center. Theoretical calculations of the activation energy for C–H bond cleavage showed that the proton abstraction is much more favored with an unbound selenocysteine (19 kcal/mol compared to 36 kcal/mol for bound selenocysteine). In this mechanistic proposal, the conserved arginine R133 facilitates formate binding and histidine H141 acts as the final proton acceptor.⁶⁴⁶ The terminal sulfido ligand of the Mo ion, on the contrary, would have no active role in formate oxidation.

Another mechanism was proposed based on theoretical calculations involving selenocysteine dissociation from the Mo ion through a “sulfur shift” (Figure 45D).⁶⁵³ In this mechanism, Cerqueira, Gonzales, Moura, and co-workers proposed that the oxidized enzyme is hexacoordinated by the two MGD molecules plus the selenocysteine and the terminal sulfido group. When formate enters the active site, oriented by the positively charged arginine (e.g., R333 in E. coli FdhF), the repulsive environment generated would trigger a ligand shift to yield a Mo–S–SeCys moiety. This would create a vacant position to which the formate could bind. Subsequently, the S-SeCys bond is cleaved and the selenol anion abstracts the formate C_α proton (analogous to the model in Figure 45C). Carbon dioxide is subsequently released. The catalytic cycle would be completed with the oxidation of Mo^IV to Mo^VI via intramolecular electron transfer to the [4Fe-4S] center and deprotonation of the selenocysteine residue. In the absence of formate, the selenocysteine-containing loop (e.g., V139, U140, and H141 in E. coli FdhF, see Figure 43A) is reoriented, and the Mo–S–SeCys bond reformed. This proposal⁶⁵³ is in agreement with the increased acidity of the selenocysteine side chain (pK_a ≈ 5.2), which allows the existence of a selenol anion at physiological pH values. The alkaline pK_a value of a cysteine side chain (pK_a ≈ 8.2) would prevent a cysteine residue having such a function. The pK_a values of both amino acids have been evoked to explain why the U140C variant of E. coli FdhF shows a 300-fold activity decrease.⁵⁸⁵ In addition, the high activation energy calculated for the proton transfer from formate to selenol (SeH) is in agreement with the isotopic effect studies⁶⁴⁷ that showed that the formate C–H bond cleavage is the rate-limiting step of the catalytic cycle. The role of the sulfido ligand, however, is also unclear in this mechanism.

An additional mechanism was proposed by Zampella and coworkers in a computational study (Figure 45E).⁶⁶⁹ In this mechanism formate first binds to the Mo ion by displacing the selenocysteine group, then Mo abstracts a hydride ion from formate, giving a Mo–H species and releasing CO₂. Upon recovery of the pentacoordinated Mo ion, the sulfido group is protonated at the expense of the hydride species. In this mechanism, also the role of the sulfido group is unclear and a Mo–H bond is very unusual. Following theoretical assumptions, Dong and Ryde recently proposed an additional mechanism (Figure 45F).⁶⁷⁰ They proposed that the sulfido ligand abstracts a hydride ion from the substrate, yielding Mo^IV–SH; however, distinct from any other mechanism, carboxylate was considered binding to the cysteine ligand, forming a thiocarbonate/cysteine zwitterion. Upon charge compensation, CO₂ leaves the enzyme and the Mo ion is oxidized back to the Mo^VI state. Finally, Hille and Niks proposed a mechanism that is based on experimental data^582,632 and considers binding of formate to the second coordination sphere without contacting the Mo ion (Figure 45G).⁶⁷¹ In their proposal, the sulfido ligand abstracts a hydride ion from formate, resulting in a two electron-reduced intermediate Mo^IV–SH, and CO₂ is released. In this mechanism, however, the (seleno-) cysteine would not be involved and the role of the metal ion is only to provide a hydride acceptor by coordinating the sulfido ligand.

4.5.1. Conversion of FDH to a Nitrate Reductase.

The sulfur shift-based mechanism to oxidize formate, described above and depicted in Figure 45A, has also been evoked to explain the nitrate reduction catalyzed by periplasmatic nitrate reductases (NR).⁶⁵³ The FDH and nitrate reductase active sites are surprisingly superimposable,⁵⁹⁸ with the later harboring a Mo ion coordinated by the two characteristic MGD molecules, one terminal sulfido group, and a cysteine sulfur atom.⁶⁷² In addition, the nitrate reductase active site also comprises conserved threonine and methionine residues (arginine and histidine, in FDH). In the oxidized active site of both enzymes, the Mo/W ion is hexacoordinated and a sulfur shift is needed to displace the selenocysteine or cysteine residue to create a vacant position for substrate binding (i.e., formate or nitrate).^589,653

The similarities between FDH and nitrate reductases were highlighted by converting RcFDH to an enzyme with nitrate reductase activity.⁵⁹⁸ This was achieved by exchanging the histidine H387 to a methionine residue and the arginine R587 to a threonine residue, both that are conserved in nitrate reductases. In addition, an additional arginine residue was inserted to the active site of RcFDH and this enzyme variant showed bis-MGD dependent nitrate reductase activity. However, the involvement of the sulfido ligand of this enzyme variant for nitrate reductase activity is not clear, since also an enzyme form containing an oxo ligand instead of the sulfido ligand was able to reduce nitrate.⁵⁹⁸ Nitrate reductases catalyzes the typical O atom transfer mechanism characteristic for all other molybdoenzymes.⁶⁷²

4.5.2. Second and Outer Coordination Sphere Effects on Electron Transfer.

In addition to the active site Mo/W ion, the sulfido ligand, and active site residues involved in CO₂ reduction and formate oxidation of FDH and formylmethanofuran dehydrogenases, additional components are critical toward the catalytic activity. For example, a feature common to all FDHs and formylmethanofuran dehydrogenases is the coordination of a [4Fe-4S] cluster that is within electron channeling distance to the bis-MGD cofactor (Figure 41). For all FDH enzymes identified to date, a [4Fe-4S] cluster is within 12 Å of the Mo/W ion, and within 6 Å of the proximal pterin of the bis-MGD (Figure 44).^{591,599,612,627,634,660} This iron–sulfur cluster is proposed to be the entry point of the electron transfer pathway by which electrons that are generated from formate oxidation or are needed for CO₂ reduction are passed along, respectively. Reduction of this [4Fe-4S]^2+/+ cluster has been observed by EPR spectroscopy, using either formate or dithionite as a reductant. These are typically conditions by which the bis-MGD cofactor can also become reduced to a Mo^V/W^V oxidation state, which complicated the spectral analysis. Early characterization work of FDH from Methanobacterium formicicum to differentiate the [4Fe-4S] cluster from the Mo^V/W^V signal showed that this cluster had g-values of 2.047, 1.948, and 1.911;^649,650 however, it should be noted that this FDH also can bind a [2Fe-2S] cluster whose g-values would overlap.⁵⁷⁵ By comparison, reduction of the proximal [4Fe-4S] cluster in EcFDH–H (an FDH that coordinates only one [4Fe-4S] cluster) gave a similar assignment (g-values of 2.045, 1.957, and 1.840). Similar assignments have been made for the corresponding iron–sulfur clusters for several FDHs by which EPR spectral data is available.⁶⁴⁷

As a participatory component in the redox events catalyzed by the bis-MGD cofactor, perhaps it is not surprising that communication with the proximal [4Fe-4S] cluster is essential for the catalytic reaction. However, there are few examples of FDHs by which concomitant catalytic activity and loss of the proximal [4Fe-4S] cluster has been shown. For example, a EPR spectral characterization of the FDH from C. necator FDH has shown that under conditions by which the [4Fe-4S] cluster is reduced and the bis-MGD is poised in the paramagnetic Mo^V state, electron–electron spin coupling between the two paramagnetic centers occurs, primarily reflected in the loss of sharpness of the Mo(V)-based ¹H hyperfine splitting.⁶³² As EPR spectroscopy cannot characterize the diamagnetic oxidation states of the bis-MGD cofactor and iron–sulfur cluster, a parallel, multifaceted spectroscopic approach is required to identify all oxidation states.

In addition to participation of the proximal [4Fe-4S] cluster in the catalytic activity of the bis-MGD cofactor, an additional, undercharacterized aspect of FDH catalysis relates to the two molybdopterin guanine dinucleotide ligands that coordinate the Mo/W ion. These relatively large, pterin-based ligands are noninnocent and redox active, meaning that they partake in the electron transfer reactions between the Mo/W ion and the proximal [4Fe-4S] cluster immediately preceding or following catalysis. By comparison, the longer [4Fe-4S] cluster–Mo/W distance (13–16 Å) observed relative to the shorter proximal [4Fe-4S]–(proximal) MGD distance (6–10 Å) supports the presumption that the proximal MGD ligand links the FDH redox events at the Mo/W ion with the iron–sulfur cluster electron conduit.^{591,599,612,627,634,660} So far, all bacterial FDHs coordinate two MGD ligands at the active site; however, distinct nucleotides were identified to be bound in place of the guanine in FDH-like enzymes from archaea (e.g., formylmethanofuran dehydrogenase from M. marburgensis).⁶²⁵ The reason for this remains unclear so far.

5. CO DEHYDROGENASE

5.1. Physiological Function and Structure of CODH

CO dehydrogenases (CODHs) are gas-processing metalloenzymes that catalyze the reversible oxidation of CO to CO₂ (eq 4). There are two unrelated classes: one is a Mo-based enzyme exclusively found in aerobic carboxydotrophs.^673,674 These CODHs catalyze the reaction in the direction of CO oxidation (CO₂ release) and will not be discussed here any further. The other is a class of metalloenzymes found mainly in anaerobic bacteria and archaea hosting a Ni- and Fe-containing cofactor, which catalyze both CO oxidation and CO₂ reduction.⁶⁷⁵

Organisms such as Rhodospirillum rubrum (Rr) and Carboxydothermus hydrogenoformans (Ch) use the enzyme CODH to oxidize CO to CO₂.⁶⁷⁶ The reaction occurs at the so-called “C-cluster”, a unique [Ni-3Fe-4S] cluster with an additional Fe ion. The enzyme is a homodimer that houses two identical C-clusters and three electron transferring metal centers: two “B-clusters” of the [4Fe-4S] type and one “D-cluster” located at the interface of the two monomers (Figure 46A). The latter is either a [4Fe-4S] cluster (e.g., in RrCODH) or a [2Fe-2S] cluster (e.g., in CODH from Desulvovibrio vulgaris, Dv).⁶⁷⁷ In Figure 46B, the proposed proton transfer pathway toward the C-cluster is shown (see Section 5.3).

Figure 46.

Structure of CODH. Overall structure of ChCODH II (PDB ID 1SU8). Each monomer of the ChCODH II homodimer is represented with a different color. The spheres represent the inorganic cofactors, labeled cluster B–D. The C-cluster is the catalytic active site cofactor. The purple amino acids correspond to the putative proton transfer pathway. (b) Close up view of the C-cluster and residues putatively involved in proton transfer (PDB ID 3B52). Atomic coloring: Fe, orange; Ni, green; S, yellow; C, gray; O, red; N, blue.

In acetogenic bacteria such as M. thermoacetica (Mt) and methanogenic archaea such as Methanosarcina barkeri (Mb), a homodimeric CODH is present as part of bifunctional enzyme complexes. This may be in association with the enzyme acetyl-CoA synthase (ACS) in the ACS/CODH complex, which allows acetyl-CoA formation using a CO molecule produced by the reduction of CO₂ as part of the Wood–Ljungdahl carbon fixation pathway,⁶⁷⁸ or in an acetyl-CoA decarbonylase/synthase (ACDS) complex that degrades acetyl-CoA to form methane and CO₂.⁶⁷⁹ The ACS/CODH complex is a large α₂β₂ tetramer that associates two β CODH subunits and two α ACS subunits. The CO produced from CO₂ at the C-cluster is combined with a methyl group and coenzyme A (CoA) to form acetyl-CoA at the A-cluster of the ACS (α) subunit; in vitro, acetyl-CoA synthesis can also occur using CO as a substrate.⁶⁸⁰

C. hydrogenoformans illustrates the diversity of possible functions of CODHs, since its genome encodes four different isoenzymes:⁶⁷⁶ ChCODH I, which is likely involved on respiration on CO; ChCODH II, proposed to provide electrons to form NADPH; ChCODH III, which is incorporated in a bifunctional ACS/CODH complex; and ChCODH IV, proposed to be involved in resistance against O₂.⁶⁸¹ Eisen and co-workers originally proposed that the genome encodes a fifth CODH in C. hydrogenoformans,⁶⁷⁶ but the latter is more likely to be a hybrid cluster protein.

5.2. Structure and Placidity of the C-Cluster

The structure of the C-cluster has been heavily debated, but the current consensus is that it comprises a [Ni-3Fe-4S] cluster coordinated to a unique fourth iron ion (Fe_u); the latter was also referred to as the “dangling Fe” or “ferrous component II” (Figure 47A).⁶⁸² The [Ni-3Fe-4S] cluster is coordinated by four cysteines as in classical [4Fe-4S] clusters including C333, C446, C476, and C526 (DvCODH II numbering). The dangling Fe ion is coordinated by cysteine C295 and histidine H261. The catalytic active site is completed by histidine H93 and lysine K563 that are within hydrogen-bonding distances from the O atoms of the bound reactant, CO₂.⁶⁸² Neighboring cysteines C295 and C294 are conserved among CODHs. They are found in all clades except D, C1.3, and A4 of the classification proposed by Sako and co-workers;⁶⁸³ however, no CODH with verified activity lacks this cysteine residue. Mutating C294 into a serine prevented the assembly of the C cluster in the case of Mt ACS/CODH,⁶⁸⁴ and yielded a C-cluster devoid of Ni in the case of DvCODH.⁶⁸⁵ Exposure to air of DvCODH crystals yielded an alternative form of the C-cluster in which the Ni ion occupies the position of the dangling iron, which moved a little further (Figure 47B). In this configuration, cysteine C295 becomes a bridging ligand between the Fe and Ni ions and C294 coordinates the dangler ion. This state is formed reversibly, as reducing the crystals gives back the “classical” coordination of the C-cluster, and may be involved in the high resistance of DvCODH to O₂ damage.^686,687 Additionally, C294 may be involved in the insertion of the active site, which would explain the lack of C-cluster when it is mutated into a serine. This hypothesis is also consistent with the fact that mutation of H261 and C295 yield an enzyme devoid of Ni.⁶⁸⁸ Histidine H261 and C295 are ligands to the dangling Fe ion in the usual form of the C-cluster (Figure 47B), but bind the Ni ion in the oxidized C-cluster.

Figure 47.

Structure of the catalytic C-cluster of CODH. (a) Close up view of the CO₂-bound active site (ChCODH II, PDB ID 3B52), with a representation of most of the amino acids discussed here. Fe_u refers to the “dangling” iron ion. (b) View of the active site of DvCODH (PDB ID 6B6W) in the oxidized, protected state (C_ox). The orientation is the same as in the left panel. Metal-coordinating residues are bold. Atomic coloring: Ni, green; Fe, orange; S, yellow; C, gray; O, red; N, blue.

The role of the cysteine ligands to the C-cluster was investigated by site-directed mutagenesis on RrCODH. Cysteines C338, C451, C481, and C531 (corresponding to C333, C446, C476, and C526 in ChCODH II) were mutated into serine or alanine residues.⁶⁸⁹ None of the mutations affected the incorporation of Ni; however, most of the mutants lost their activity, except the C451S and C531A variant, which retained ~1.4% and 0.1% of CO oxidation activity, respectively. The exact reason for the decreased activity is not known yet, but redox effects may be involved. While the EPR signature of C_red2 (see below) was observed in the reduced C451S variant, no C_red1 could be detected under the redox conditions in which the wild-type RrCODH usually gives C_red1.⁶⁸⁹

The C-cluster of CODHs may exist in at least three redox states: the fully oxidized, inactive, and EPR-silent “C_ox” state, which is reduced at potentials of −100 mV to the paramagnetic “C_red1” state (g = 2.01, 1.81, and 1.65). Below approximately −500 mV, the “C_red2” state is formed (g = 1.97, 1.87, and 1.75), which is two electrons more reduced than C_red1. An EPR-silent, singly reduced state “C_int” has been postulated to exist as a transition between C_red1 and C_red2, but it was never isolated.⁶⁹⁰ The consensus is that C_red1, C_int, and C_red2 are catalytic intermediates.^675,691 The proposed mechanisms all assume that the structure of the CO₂-bound enzyme (Figure 47A) corresponds to a catalytic intermediate and that the binding of CO₂ is followed by the breaking of the bond between the C and the O coordinated to the dangling Fe ion, yielding a CO-bound Ni ion and a hydroxo ligand on the dangling Fe (Figure 48). The main dispute for the time being is the nature of the C_red2 state. According to the model by Jeoung and Dobbek,⁶⁸² C_red2 corresponds to a Ni⁰ state with a hydroxo ligand on the dangling Fe and no extraneous ligand on the nickel (this state is indicated by C_red2* in Figure 48).⁶⁸² On the basis of DFT computations, Fontecilla-Camps, Amara, and co-workers have proposed instead that the C_red2 state is a Ni²⁺ hydride (this state is indicated by C_red2^† in Figure 48), and that the CO₂-bound structure is the result of the unusual insertion of CO₂ into the Ni–H bond,⁶⁹² even though insertion of CO₂ into a M-H bond usually results in the production of formate.⁶⁹³ Bruschi and coworkers have recently used density functional theory (DFT) calculations to investigate the binding of CO₂ to different redox states of the active site C-cluster; their results appear to be incompatible with the presence of a Ni–H species in C_red2, thus favoring C_red2* over C_red2^†.⁶⁹⁴ They calculated that the binding is governed by the protonation states of histidine H93 and lysine K563 (equivalent H93 and K563 in ChCODH II, Figure 47), and opened the possibility that CO₂ may also bind to the C_int redox state.

Figure 48.

Proposed mechanisms for the oxidation of CO to CO₂ by CODH. The black scheme is the original proposal by Jeoung and Dobbek.^682,691 The blue part is the modification to that scheme proposed by Fontecilla-Camps, Amara, and co-workers.⁶⁹² For the sake of clarity, we have only indicated the arrows in the direction of CO oxidation, but CO₂ reduction is supposed to follow the exact same steps in reverse order.

5.3. Substrate Access and Product Egress

The diffusion of CO and CO₂ has been investigated both in monofunctional CODH and in the bifunctional ACS/CODH complex. In ACS/CODH, the site of CO production and the site of CO utilization are over 70 Å apart. As CO does not leak to the solvent when the ACS/CODH complex produces acetyl-CoA using CO₂ as a substrate, a gastight channel was proposed that would guide the CO molecule from the C-cluster toward the A-cluster, where CO is condensed with a methyl group and coenzyme A.^695,696

The search for cavities in the X-ray structure of ACS/CODH and the identification of Xe binding sites^680,697,698 defines the long hydrophobic channel that connects the four active sites in the tetrameric ACS/CODH complex (Figure 49). This channel is blocked but the A-cluster is accessible to the solvent for methyl transfer in the so-called open conformation of ACS/CODH.^698,699 Support for these channels transporting CO from the C-cluster to the A-cluster comes from the observation that mutations of the conserved alanine residues that line the putative channel in the ACS subunit of Mt ACS/CODH (A110C, A222L, and A265M) only have an effect of the rate of acetyl-CoA production when CO₂ is used as a substrate.^700,701 The mutations have a no significant effect on the CO oxidation activity, and only a moderate effect on the rate of acetyl-CoA production from CO. The acetyl-coA synthesis from CO is strongly inhibited by CO in wild-type ACS/CODH, but not in the three channel variants; this is explained by the possibility that part of the activity in wild-type ACS/CODH is due to CO molecules entering at the C-cluster and crossing the channel toward the A-cluster, combined with CO crowding in the channel regulating the flow of CO from the C-cluster.⁷⁰⁰

Figure 49.

Hydrophobic gas channels in CODH. Structure of the ACS/CODH from M. thermoacetica (PDB ID 2Z8Y), indicating the position of the clusters A–D, and marking the long hydrophobic channel that connects them. Reprinted (adapted or reprinted in part) with permission from ref 702. Copyright 2022 Elsevier.

The question of how CO₂ reaches the C-cluster of ACS/CODH is less clear. It has been suggested that CO₂ may enter the enzyme at the level of the A-cluster (which would involve transport against the opposite flux of CO),⁶⁹⁹ or via the water channel along the ββ subunit interface,⁷⁰³ or along dynamically formed pathways that are not detected as clear cavities in the crystal structure.⁷⁰⁴ The latter hypothesis has been investigated in molecular dynamics (MD) calculations, according to which CO₂ transport is controlled by two strictly conserved histidine residues, H113 and H116 (H93 and H96 in ChCODH II, Figure 47). Upon reduction of CO₂, a hydrogen-bond network in the active site pocket becomes rigid enough to prevent CO from exiting the protein via the CO₂ access pathway, and directs CO from the C-cluster to the A-cluster.⁷⁰⁴

In the CODH subunit of Mt ACS/CODH, histidines H113 and H116, along with H119 and H122, asparagine N284 and lysine K587 were proposed to be involved in proton transfer between the solvent and the C-cluster (H93, H96, H99, H102, N262, and K563 in ChCODH II, see Figure 46B). Variants H116A and H122A did not show catalytic turnover but activity could be restored by introducing a histidine as a neighboring amino acid.⁶⁸⁴ On the basis of a series of single and multiple site-directed mutagenesis replacements, Kim and co-workers proposed that protons first go to histidine H113 or lysine K587, then necessarily H116, then either H119, asparagine N284 or a water molecule, and finally to the solvent-exposed histidine H122.

Additional hydrophobic cavities have been identified through calculations in the monofunctional CODH from C. hydrogenoformans and R. rubrum.^705,706 Structure comparison suggests that bulky residues that are present in ACS/CODH block some of the pathways that are functional in reactant diffusion in ChCODH and RrCODH, and whose presence in the ACS/CODH complex would prevent effective channeling of CO from the C-cluster to the A-cluster.⁶⁹⁸

5.4. Diversity of CO Dehydrogenase

Figure 50 represents a sequence alignment of the main CODHs that have been studied so far, with the exception of the ACS/CODH from M. barkeri. All amino acid residues that have been modified by mutagenesis so far are indicated using red backgrounds, and the amino acids within 8 Å of the Ni ion of the active site are framed in red. Figure 50 shows that the environment of the active site is highly conserved, with only a few positions with nonconserved amino acids, and a few more with small variations, notably at position 293 (with isoleucine, leucine, or methionine), 311 (with asparagine, histidine, or serine), 312 (with tyrosine, serine, histidine or phenylalanine), 331 (with valine or tyrosine), 444 (with valine, alanine, or cysteine), 478 (with alanine, serine, or threonine), 529 (with asparagine or cysteine), 560 (with methionine, glutamine, or tyrosine), 561 (with histidine, serine, or threonine), and 561 with serine, histidine, or threonine). Positions 477, 479, and 559 are not conserved.

Figure 50.

Sequence alignment of the main CODHs studied so far. Conservation is indicated by a blue background. The amino acids with a red background have been modified by site-directed mutagenesis in the literature. The red frames correspond to the amino acids within 8 Å of any atom from inorganic C-cluster. The numbering is that of ChCODH II.

In spite of this high sequence identity, the properties of the CODH appear to change greatly from one CODH to another. Regarding the K_m for CO, the values range from 20 nM for ChCODH IV to 30 μM for Rr CODH, with other values between 0.5–3 μM for DvCODH, ChCODH I, and TcCODH I and II from Thermococcus sp. AM4 (Tc).^{681,707–710} To date, no clear pattern has been noticed relating the sequence and the values of K_m, and no mutation has been published that affects the K_m for CO. K_m values for CO₂ reduction are much higher than those for CO oxidation, and in some cases they could not be determined because it was not possible to saturate the enzyme with CO₂. They range from 0.2 mM for Rr, 0.47 mM for TcCODH I, to values undetermined but larger than 2 mM (ChCODH I) or 7 mM (TcCODH II).^709–711 Similarly to CO, no mutation yet is known to affect the K_m for CO₂.

CODHs have been said to be highly O₂-sensitive,⁶⁷⁵ but in fact the reactivity of CODHs with O₂ varies greatly, from RrCODH being entirely destroyed by a 10 s exposure to O₂⁷¹² to DvCODH, losing only 70% of its activity upon aerobic purification,⁶⁸⁷ and reactivating almost fully upon reduction after exposure to O₂,⁶⁸⁶ a property that is reminiscent of the well-known reductive activation of [NiFe]-hydrogenases (Section 3.1) and certain [FeFe]-hydrogenases (Section 3.2). ChCODH IV requires much higher concentrations of O₂ than the other CODHs to lose 50% of its activity, but unlike DvCODH it does not reactivate upon reduction.⁶⁸¹ This has been attributed to the presence of a bulky phenylalanine residue on the backside of the C-cluster (F307), in the position corresponding to S312 in ChCODH II (Figure 50). However, TcCODH I and II, which also feature a phenylalanine in the same position (F346 and F322, respectively), do not share the exceptionally high resistance of ChCODH IV against O₂.⁷⁰⁹

Perhaps the best understood reactivity with O₂ is that of DvCODH, which harbors a [2Fe-2S] cluster as the D-cluster, unlike the other CODHs characterized so far, which hold a [4Fe-4S] cluster. The [2Fe-2S] type D-cluster of DvCODH is at least partially responsible for the high O₂ resistance of DvCODH, since upon mutation to a “classical” [4Fe-4S] cluster, the enzyme loses the ability to be purified aerobically.⁶⁸⁷ Deletion of the D-cluster also prevents C-cluster assembly,⁶⁸⁵ which shows that the maturation of the active site is also sensitive to either long-distance effects, or long-range electron transfer.

6. CONCLUDING REMARKS

Throughout Sections 2–5, we have seen how the activity of gas-processing metalloenzymes (GPMs) is defined by second and outer coordination sphere effects. As noted in Section 1, the optimized catalysis of GPMs has been afforded through tailored evolution of the active site cofactors from abiotic minerals and peptide “nests” coordinating these inorganic microcompartments.^15–20 Several factors that drive second and outer coordination sphere effects in nitrogenase, hydrogenase, FDH and CODH have been described herein. It will be difficult to understand GPMs without knowledge about the structure and electronic properties of the metallic cofactors; however, we emphasize that bioinorganic chemistry must explore beyond the first coordination sphere. Concluding this article, we will highlight a number of common motifs including mass transport (Section 6.1), redox leveling (Section 6.2), and the stabilization of reactive geometries (Section 6.3). The developed concepts may inspire the design of biomimetic catalysts for the transformation of N₂, H₂, CO₂, and CO as critical feedstock in “green” energy conversion.

6.1. Mass Transport

Under ambient conditions, N₂, H₂, CO₂, and CO are small volatile molecules with mostly hydrophobic properties. Such reactants approach the active site cofactor by molecular diffusion. Wolfenden and co-workers discussed how enzymatic catalysis is occasionally limited by mass transport, demonstrating the efficiency of GPMs like superoxide dismutase and carbonic anhydrase.⁷¹³ At the example of hydrogenase, FDH, and CODH we saw that proteinaceous “gas filters” influence the catalytic rates and substrate selectivity (Sections 3–5). While hydrophobic channels and pockets are routinely identified in xenon- or krypton-pressurized protein crystals,⁷¹⁴ the collective motion of the protein fold and packing defects⁷¹⁵ complicate the identification of individual “gate keepers”, that is, amino acid residues of key importance. From the viewpoint of synthetic chemistry, we note that it may not be necessary to install molecular gas filters. In recent years, redox hydrogels have been demonstrated to facilitate enzymatic turnover in the presence of gaseous inhibitors, for example, by reducing O₂ before it reaches the enzyme.^716–718

Proton transfer is another mass transport phenomenon critical to the performance of GPMs. In Section 1, we compared the basic reactions catalyzed by nitrogenase, hydrogenase, FDH and CODH (eq 1–4). Each reaction includes at least one proton. Protons are transferred via water molecules and the hydrophilic groups of a protein when hydrogen-bonded or in hydrogen-bonding distance (2.7–3.3 Å).⁷¹⁹ These criteria make proton transfer pathways extremely reliant on protein structural dynamics. In AvNifDK of the Mo-nitrogenase, second coordination sphere residues R96, Q191, and H195 (analogous to K83, Q176, and H180 in AvVnfDGK of the V-nitrogenase) are potentially involved in proton transfer and hydrogen bonding to catalytic intermediates that replace the bridging sulfide ligand S2B (Figure 14). Additionally, the homocitrate ligand may be involved,²³² but any outer coordination sphere effects on proton transfer have yet to be unraveled. The cysteine residue that coordinates the active site cofactor of nitrogenase (C275 in AvNifDK, C257 in AvVnfDGK) is not involved in proton transfer, which is a marked difference from hydrogenase where cysteines of the first and second coordination sphere play key roles in proton transfer.²⁹⁷ In [NiFe]-hydrogenase DvMF, nickel-coordinating C546 accepts a proton after heterolytic H₂ splitting. Glutamic acid residue E34 represents a proton relay before the proton is released into bulk water (Figure 22). In [FeFe]-hydrogenase Cp1, cysteine C299 close to the distal iron ion (Fe_d) of the H-cluster functions as a relay site between cofactor and a ~ 20 Å proton transfer pathway including two glutamic acid residues (E279, E282), a serine residue (S319), and a small water cluster (Figure 39). In the primary sequence, C299 is followed by C300, which binds to the [4Fe-4S] part of the H-cluster. A similar motif can be found in EcFDH where Mo-binding selenocysteine U140 is followed by histidine H141 (Figure 42), the later which has been suggested to play a key role in proton transfer. In the outer coordination sphere, R333 may be involved in proton transfer, but the molecular details of this event are yet to be understood.⁵⁷⁵ Similar to FDH and nitrogenase, histidines are the most likely proton relay sites in CODH. On the basis of biochemical experiments with various amino acid variants of ChCODH II, a trajectory between the catalytic C-cluster and bulk solvent including H93, H96, H99, H102, N262, and K563 has been proposed (Figure 48). Hydrogenases, on the contrary, rely on cysteine and glutamic acid residues primarily,²⁹⁷ resulting in an overall pK_a decrease of 2–3 compared to histidine-based proton transfer pathways. This may reflect the necessity for a tightly controlled proton transfer in systems that exclusively catalyze H₂ oxidation and proton reduction.

6.2. Redox Leveling

While proton transfer demands short distances, the probability for electrons to tunnel across distances as far as 18 Å is high enough to enable electron transfer within biologically relevant time scales.⁷²⁰ Conductive wires are formed by iron–sulfur clusters in nitrogenase, hydrogenase, FDH, and CODH, guiding electrons back and forth between the active site cofactor and redox proteins like ferredoxin, flavodoxin, or cytochrome. The “accessory” clusters of GPMs are determined by first and second coordination sphere effects to ensure directional electron transfer. For example, the replacement of a cysteine by a histidine ligand at the distal [4Fe-4S] cluster in [FeFe]-hydrogenase has a strong effect on activity.⁴¹⁸ In special cases, the electronic properties are flexible and controlled by the enzyme. Examples include the P-cluster in nitrogenase that changes its geometry upon reduction (Figure 3) or the proximal iron–sulfur cluster in O₂-tolerant [NiFe]-hydrogenases (Figure 24). In [FeFe]-hydrogenase, a protonation event at the backend of the H-cluster has been shown to affect the redox potential (Figure 39). These special cases demonstrate how enzymes modulate electron transfer in a flexible way.

Compared to the branching chains of metal clusters found in electron bifurcating/confurcating enzymes,^265,266,721 the GPMs discussed in this review are wired in a straightforward fashion. The sophistication, however, lies in the spatiotemporal coupling of redox and protonation events at the active site cofactor, which allows for an accumulation of charges at mild potentials (“redox leveling”). This is clearly illustrated in the Lowe-Thorneley model for N₂ reduction (Figure 6) and holds true for H₂ turnover as well (Figure 21 and Figure 34). The mechanistic details in FDH and CODH are less well understood, but similar proposals have been considered (Figure 45 and Figure 48). How do GPMs facilitate proton-coupled electron transfer? While the electron transfer chains are hard-wired, the distance constraints of proton transfer allow for flexible fine-tuning, for example, by small changes in protein structure as demonstrated in [NiFe]- and [FeFe]-hydrogenase.²⁹⁷ The data discussed in this review now suggest that metalloenzymes initiate PCET by guiding protons to the cofactor. Here, the catalytic redox reaction is induced only upon arrival of the proton (and other reactants like N₂, CO₂, or CO). As tunneling is a quantum mechanical phenomenon, the formal requirements of PCET are fulfilled.^22–24 From the viewpoint of synthetic chemistry, we note that molecular wiring may not be necessary. For example, when biomimetic Ni-complexes were deposited in a conductive hydrogel, redox events could be triggered by external electric fields.⁷²² At variance, the design of proton transfer pathways seems to be imperative for efficient electron transfer and catalysis.⁵⁵⁶

6.3. Stabilization of Reactive Geometries

Second coordination sphere effects are not limited to mass transport and redox leveling. In nitrogenase, hydrogen-bonding between H195 and R96 with bridging ligands in position S2B or between Q191 and terminal ligands (Figure 14) stabilizes the reactive geometry of the active site cofactor. Ligands include N₂, CO, or H. Positions S3A and S5A could be catalytically relevant as well; in AvNifDK of the Mo-nitrogenase, ligands likely derived from the bound N₂ entities have been identified as being hydrogen-bonded to backbone amines of G356 and G357 at position S3A, and the side chain of R96 at position S5A (Figure 12). In hydrogenases, potential hydrogen-bonding contacts inspired the assignment of CO and CN⁻ ligands. This includes S502, P478 and R479 in [NiFe]-hydrogenase DvMF (Figure 19). Arginine 479 may also form a frustrated Lewis pair (FLP) with the Ni–Fe cofactor and has been suggested to be part of the “outer-shell canopy”, an alternative proton transfer pathway.³⁶⁰ The mechanism shown in Figure 22 favors C546 as FLP but the role of R479 in polarizing the bridging ligand cannot be excluded.³⁶¹ Arginine 96 and H195 at the nitrogenase cofactor may have a similar function (Figure 14). In [FeFe]-hydrogenase, the ADT ligand serves as “internal” FLP; however, hydrogen-bonding with the adjacent C299 and M497 residues (Figure 38) was found to be of key importance for efficient proton transfer and catalysis. Another methionine, M353, may destabilize the μCO ligand and contribute to the flexible geometry of the diiron site (Figure 37). Around the proximal CN⁻ ligand, strong hydrogen bonding with P231 and S232 was confirmed whereas any contacts around the distal CN⁻ ligand remain speculation (Figure 36). The catalytic mechanism of FDH is controversial and various proposals coexist (Figure 45). The involvement of R333 and H141 seems to be accepted, which represents a similar mechanistic element to those employed by [NiFe]-hydrogenase and nitrogenase. Histidines play important roles in CODH, across all coordination spheres. In the reduced, CO₂-bound state of ChCODH II, H93 and K563 are in hydrogen-bonding distance to the substrate, stabilizing the reactive geometry (Figure 47). Lysine K563 is turned into a nickel ligand in the oxidized state, switching between the first and second coordination sphere.⁶⁷⁷ This behavior is reminiscent of C367 in [FeFe]-hydrogenase CbA5H (C299 in CpI) binding to Fe_d in the O₂-inhibited state (Figure 35). From the viewpoint of synthetic chemistry, hydrophilic groups in the second coordination sphere should have dual functions. Our data emphasize that histidines, arginines, and cysteines serve as hydrogen-bonding partners to substrates like N₂, H, CO₂, and CO, stabilizing reactive geometries. However, without contact to residues involved in proton transfer in the outer coordination sphere (Section 6.1), efficient catalysis cannot be achieved.

Biographies

Sven T. Stripp graduated in Biology and Biotechnology and received his Ph.D. degree from Ruhr-Universität Bochum, Germany. Afterwards, he moved to Freie Universität Berlin for postdoctoral studies on surface-enhanced vibrational spectroscopy. Since 2015 he has been a group leader in the Department of Physics where he develops in situ FTIR, Raman, and UV–vis spectroscopic approaches to investigate metalloenzymes under biologically relevant conditions. He habilitated in Physical Chemistry in 2021.

Benjamin R. Duffus received his B.A. in chemistry (cum laude) from Concordia College (Moorhead, MN) in 2005. Following postgraduate study in synthetic inorganic chemistry at Texas A&M University, he completed his Ph.D. at Montana State University in 2014, focusing on the biosynthesis of complex iron–sulfur clusters. Since 2014, he is a postdoctoral research associate in Molecular Enzymology at the University of Potsdam, Germany. His interests blend biochemical and spectroscopic approaches toward mechanistic and biosynthetic studies of metalloenzymes, with a focus toward complex, iron–sulfur cluster molybdoenzymes.

Vincent Fourmond received his Ph.D. at the Université Paris Descartes in 2007. After two postdocs in the groups of Christophe Léger and Vincent Artero, he became permanent CNRS research associate in 2011. His interests lie in using kinetics to understand the mechanism of redox metalloenzymes and in the development of the general purpose data analysis software QSoas.

Christophe Léger received his Ph.D. from the University of Bordeaux and was a postdoc in the group of Fraser Armstrong from 1999 to 2002. He is “Directeur de Recherche” at CNRS. His interests lie in kinetic and mechanistic studies of complex metalloenzymes.

Silke Leimkühler received her Ph.D. in 1998 from Ruhr-Universität Bochum, Germany. After postdoctoral studies at Duke University, USA, she returned to Germany as a research group leader to the Technical-University of Braunschweig. In 2004 she accepted a Junior-Professor position at the University of Potsdam, Germany, and since 2009 she is a full professor in Molecular Enzymology at the same institute. Her research interests focus on molybdenum cofactor biosynthesis, molybdoenzyme enzymology, and iron–sulfur cluster assembly.

Shun Hirota received his Ph.D. from the Graduate University for Advanced Studies in Japan. After postdoctoral studies in Japan and the US, he joined Nagoya University as an assistant professor in 1996 and became an associate professor at Kyoto Pharmaceutical University in 2002. He was invited to Nara Institute of Science and Technology as a full professor in 2007. His research interests include structures, functions, and reaction mechanisms of metalloproteins.

Yilin Hu received her Ph.D. in Biochemistry from Loma Linda University, USA. She was a postdoctoral fellow at the University of California, Irvine, and is currently Professor at the same institute. She focuses on studies related to nitrogenase mechanism and assembly, with an emphasis on the genetic manipulation of nitrogenase systems.

Andrew J. Jasniewski received his B.S. degree from the University of Wisconsin—Madison working with Thomas Brunold on functional models of the Mn-dependent superoxide dismutase. He then moved to the University of Minnesota to study the structures and spectroscopy of nonheme diiron enzymes and related model complexes with Lawrence Que Jr., receiving his Ph.D. degree in 2017. He currently works at the University of California, Irvine with Markus Ribbe and Yilin Hu on the biochemistry and spectroscopy of nitrogenase.

Hideaki Ogata received his Ph.D. from Kyoto University, Japan in 2003. After a short period of the postdoctoral research at Himeji Institute of Technology, he joined the Max Planck Institute for Chemical Energy Conversion, Mülheim an der Ruhr, Germany where he worked as a group leader until 2017. He then moved to Hokkaido University and later to Nara Institute of Science and Technology. He is now a professor at University of Hyogo, Japan. His research interests are structural and spectroscopic studies of metalloenzymes.

Markus W. Ribbe received his Ph.D. in Microbiology from the University of Bayreuth, Germany. He was a postdoctoral fellow at the University of California, Irvine, and is now Chancellor’s Professor at the same institute. He focuses on the mechanistic investigation of nitrogenase catalysis and assembly by combined biochemical, spectroscopic, and structural approaches.