The Conversion of Carbon Monoxide and Carbon Dioxide by Nitrogenases

Abstract Nitrogenases are the only known family of enzymes that catalyze the reduction of molecular nitrogen (N2) to ammonia (NH3). The N2 reduction drives biological nitrogen fixation and the global nitrogen cycle. Besides the conversion of N2, nitrogenases catalyze a whole range of other reductions, including the reduction of the small gaseous substrates carbon monoxide (CO) and carbon dioxide (CO2) to hydrocarbons. However, it remains an open question whether these ‘side reactivities’ play a role under environmental conditions. Nonetheless, these reactivities and particularly the formation of hydrocarbons have spurred the interest in nitrogenases for biotechnological applications. There are three different isozymes of nitrogenase: the molybdenum and the alternative vanadium and iron‐only nitrogenase. The isozymes differ in their metal content, structure, and substrate‐dependent activity, despite their homology. This minireview focuses on the conversion of CO and CO2 to methane and higher hydrocarbons and aims to specify the differences in activity between the three nitrogenase isozymes.


Global nitrogen fixation
Bioavailable nitrogen (N) is essential for all life on Earth to build central metabolites such as nucleotides and amino acids. [1] Although an inexhaustible supply of dinitrogen (N 2 ) is stored in the Earth's atmosphere, N 2 is inaccessible to most organisms due to its kinetic stability. The majority of bioavailable N is derived from the atmosphere through the reduction of N 2 to ammonia (NH 3 ) in a process known as nitrogen fixation. For millions of years, nitrogen fixation was unique to diazotrophic microorganisms.
With the beginning of the 20 th century, the industrial Haber-Bosch process was invented to reduce N 2 to NH 3 , to produce explosives and fertilizers. [2] The process requires the reactants to pass over a heterogeneous iron catalyst at > 500°C and 100 bar for the desired product formation. [3] Due to the harmful effects of the process on the environment, it is imperative to work on a more sustainable N 2 fixation. Remarkably, evolution produced a sophisticated enzymatic reduction machinery that is capable of reducing N 2 under physiological conditions. [4] Nitrogenases fix N 2 by consuming ATP and low potential electrons with the following canonical molybdenum (Mo) nitrogenase stoichiometry [Eq. (1)]: In contrast to the HaberÀ Bosch process, nitrogenases rely on reducing equivalents stored in flavodoxins or ferredoxins instead of electrons from elemental hydrogen (H 2 ); nitrogenases rather produce H 2 as byproduct. Additionally, nitrogenases require hydrolysis of ATP for their electron transfer. [5] The N 2 fixation by nitrogenases reaches a reaction rate of~1 s À 1 which is a staggering acceleration compared to the uncatalyzed reduction of N 2 under physiological conditions. [4a,6]

Nitrogenases: isozymes, structures and mechanism
The enzyme family of nitrogenases consists of the Mo nitrogenase, the most abundant and best studied nitrogenase, and the two alternatively expressed nitrogenases: the vanadium (V) and the iron-only (Fe-only) nitrogenase. [4b,7] While all known diazotrophic organisms encode the Mo nitrogenase, some diazotrophs harbor one of the alternative nitrogenases and even less organisms encode all three nitrogenases at the same time. [8] Both alternative nitrogenases likely function as fail-safe enzymes in Mo depleted environments. The V and Fe-only nitrogenases are thought to have arisen from the Mo nitrogenase and all three nitrogenases share a high degree of sequence identity. For example, the Mo nitrogenase of the model organism Azotobacter vinelandii has a sequence identity of 44.0 % and 36.7 % with the V and the Fe-only nitrogenase, respectively. [9] The corresponding genes are named nif, vnf and anf for the Mo, V and Fe-only nitrogenase, respectively. The protein structures of the Mo and V nitrogenase are strikingly similar. [10] Spectroscopic studies and the sequence identity indicate that the structure of the Fe-only nitrogenase closely resembles the other two nitrogenases. [11] All known nitrogenases consist of a catalytic component and a reductase component. The catalytic component is called MFe protein (MoFe protein, VFe protein, FeFe protein, according to the metal content). The reductase component is termed Fe protein, for all three nitrogenases. The MoFe protein is encoded by nifDK and forms a (αβ) 2 -heterotetramer (Figure 1a, 3a). Each αβ-subunit forms an independent catalytic unit harboring two metallic cofactors, the [Fe 8 S 7 ] P-cluster involved in the electron transfer and the M-cluster (also called FeMoco) in the active site of the Mo nitrogenase (Figure 1c). The analogous catalytic components of the V and the Fe-only nitrogenase are encoded by vnfDGK and anfDGK, respectively, forming an (αβγ) 2heterohexamer ( Figure 3b). This hexamer possesses two Pclusters and two active site cofactors (V-cluster and Fe-cluster, respectively). The function of VnfG and AnfG is unknown, but an involvement in V/Fe-only nitrogenase maturation as well as Fe protein and VFe/FeFe protein complex formation and electron transfer have been suggested. [12] The Fe proteins encoded by the nifH, vnfH or anfH gene form homodimers with a subunit bridging [Fe 4 S 4 ] cluster responsible for electron delivery.
The three different active site cofactors (M-, V-, Fe-cluster) are the most complex metallocofactors observed in nature and constitute the most apparent difference between the three isozymes. The M-cluster has the chemical composition of [MoFe 7 S 9 C-(R)-homocitrate]. Six Fe atoms form a trigonal prism around an interstitial μ 6 -C atom that is capped with three μ 2 -S 2À ligands (S2B, S3A, S5A), the belt sulfides ( Figure 1c). [13] The trigonal faces of the prism are both connected to three μ 3 -S 2À , one leading to the terminal Fe atom that is linked to the protein scaffold by C275 (numbers according to NifD of A. vinelandii are used below, except indicated otherwise) and the opposite side is connected to the octahedrally coordinated Mo atom. The Mo atom is further coordinated by the bidentate ligand Rhomocitrate and H442 of NifD. The V-cluster has the composition [VFe 7 S 8 C(CO 3 )-(R)-homocitrate] with the S3A belt sulfide being replaced by a CO 3 À -ion, and V in place of Mo (Figure 3c). [14] The Fe-cluster composition is not verified by X-ray crystallography, but spectroscopic investigations suggest a composition of [Fe 8 S 9 C-(R)-homocitrate] with Fe at the Mo position. [11a] In addition to C275 and H442, the M-cluster is surrounded by highly conserved residues, i. e. V70, R96, Q191 and H195, which participate in the N 2 reduction mechanism. [15] R96, Q191 and H195 are involved in proton transport and substrate binding. [16] V70 regulates substrate access to the cofactor by steric control. [17] The overall catalytic mechanism of nitrogenases has been characterized. [4,18] However, the exact catalytic conversion of substrates at the M-cluster is still under debate. [16a,19] The two processes vital for nitrogenase activity are the electron flux and the electron accumulation and substrate conversion at the Mcluster. They are described by the model of 'deficit-spending' and the Lowe-Thorneley reaction scheme, respectively (Figure 2). [4a,20] Niels N. Oehlmann obtained his BS and MS degrees in Chemistry at the Goethe-Universität Frankfurt. In 2020, he joined the research group of Dr. Johannes G. Rebelein   The electron flow in nitrogenases originates from the Fe protein and proceeds via the P-cluster to the M-cluster (Figure 1b, 2a). The Fe protein is an electron donating enzyme that delivers the required electrons for substrate reduction. For electron transfer, the [Fe 4 S 4 ] cluster of the Fe protein is first reduced from the [Fe 4 S 4 ] 2 + state to the [Fe 4 S 4 ] 1 + state by ferredoxins or flavodoxins. Next, ATP binds to the Fe protein, triggering a structural rearrangement and allowing the reductase component to bind to the catalytic component, forming the nitrogenase complex. Upon complex formation between the Fe and the MoFe protein, the reduction potential of the Pcluster is lowered and one electron is transferred to the Mcluster. [21] This leaves the P-cluster in its oxidized state P 1 + , which is re-reduced by the Fe protein. The reduction of the Mcluster with subsequent re-reduction of the P 1 + -cluster is denoted as 'deficit-spending'. Each electron transfer is accompanied by the hydrolysis of two molecules of ATP and followed by the dissociation of the Fe protein from the catalytic component. [22] Repeated association and dissociation of the Fe protein results in a constant flow of electrons to the M-cluster for substrate reduction.
The Lowe-Thorneley scheme has been proposed on the basis of kinetic and spectroscopic investigations of the Mo nitrogenase. According to this scheme, the electron flux subsequently introduces additional electrons to the M-cluster, starting from the most oxidized E 0 state (Figure 2b). [4a] The subscript number counts the added electrons and protons (supplied via proton channeling). After at least two electrons and protons are added to the M-cluster (E 2 , E 3 , E 4 ), the spontaneous release of H 2 gas can occur, leading from the E n state to the E n-2 state (2 � n � 4). After four electrons and protons are added to reach the E 4 state, an exchange of the two bound hydrides by N 2 occurs via a coupled reductive elimination (re) of the hydrides under release of H 2 and oxidative addition (oa) of N 2 . [23] From this E 4 (2N2H) state, subsequent electron flow leads to the full reduction of N 2 to two equivalents of NH 3 without the release of H 2 . This renders the E 4 state the critical intermediate for N 2 reduction, called Janus-Intermediate. [24] The same kinetic scheme can be used for the alternative nitrogenases, though their optimum N 2 fixation stoichiometry shows an increased ATP consumption and H 2 production compared to the Mo nitrogenase. [7b,25] Nitrogenases can be compared by analyzing their overall electron flux and electron usage (Table 1). For A. vinelandii the Mo and V nitrogenase exhibit a conversion of 11-14 e À × protein À 1 × s À 1 , under high electron flux (Fe protein is present in large excess) .
[23b] While the Mo nitrogenase uses 62 % of the electron flux for N 2 reduction, only 34 % is used for N 2 reduction by the V nitrogenase. [23b] The N 2 reduction activity of the Fe-only nitrogenase from A. vinelandii is further lowered by, both, a nearly halved overall electron flux of 7 e À × protein À 1 × s À 1 and only 31 % of the e À are directed towards N 2 reduction. [23b] Consistent with the decrease in activity from the Mo over the V to the Fe-only nitrogenase, is the distribution of nitrogenases in nature. The Mo nitrogenase is the most abundant nitrogenase encoded by all known diazotrophs. Whereas the V and the Fe-only nitrogenase occur only in 6.4 % and 8.9 % of the sequenced diazotrophs, respectively. [26] Furthermore, the alternative nitrogenases are transcriptionally repressed if Mo is present in the environment. [27] Differences in activity, specificity and product profile are commonly observed for the three nitrogenase isozymes. In the case of N 2 reduction, differences between the Mo, V and Fe-only nitrogenase can likely be attributed to the  different active site cofactors. The terminal transition metals (Mo, V, Fe) in the active site cofactor mainly contribute to the spin-state fine-tuning. [28] It is hypothesized that the terminal transition metals are not directly involved in the substrate binding.
Although the described differences in electron flux and product profiles originate from assays with A. vinelandii nitrogenases, each subset of nitrogenases from other organisms reacts generally in similar ways. [4b] Similar activity patterns of the Mo, V and Fe-only nitrogenase from different organisms indicate that the main difference among these proteins arises from the active site cofactor rather than the highly conserved protein environment.

Substrate range of nitrogenases
For the conversion of N 2 , the M-cluster accumulates electrons and protons until the Janus-Intermediate E 4 is reached. In this process, each added electron is either stored in the reduced state of the M-cluster (E 1 , E 3 ) or pairwise in hydrides bound to the M-cluster (E 2 , E 4 ). [18,29] The hydrides on the active site cofactor enable nitrogenases to reduce additional small compounds. In fact, nitrogenases have one of the broadest substrate scopes known for enzymes. Besides their natural substrates H + and N 2 , nitrogenases can reduce non-physiological nitrogen compounds (azide [N 3 À ], [30] nitrite [NO 2 À ] [31] ), carbon-nitrogen compounds (cyanide [CN À ] [32] ), (iso)nitriles (HC�N, RC�N, RN�C), [33] unsaturated hydrocarbon compounds (acetylene [C 2 H 2 ], [34] cyclopropene [35] ), and carbon-chalcogen compounds (carbon monoxide [CO], [36] carbon dioxide [CO 2 ]) [37] (see Table 2). This incomplete enumeration illustrates the high reactivity of nitrogenases. The full substrate scope of nitrogenases has recently been reviewed elsewhere. [4b,c,38] Most of the unnatural substrates were studied to investigate the mechanism of nitrogenases. CO has been extensively used to investigate the nitrogenase cofactors beyond the resting state E 0 . The properties of CO allowed the trapping of CO bound intermediates under turnover conditions and their characterization by infrared (IR) and electron paramagnetic resonance (EPR) spectroscopy as well as X-ray crystallography. These investigations led to the elucidation of the substrate binding by the nitrogenase active site cofactor. Nonetheless, the 'side reactivities' of nitrogenases potentially have physiological relevance, e. g. the conversion of cyanide and CO was speculated to contribute to cellular detoxification. [39] Furthermore, the reduction of CO 2 to methane (CH 4 ) by nitrogenases has spiked interest for its potential role in shaping microbial communities. [9b] The direct reduction of CO 2 is particularly noteworthy, as it places nitrogenases in a small group of enzymes with the same ability, i. e. nickel-iron CO dehydrogenases and formate (HCOO À ) dehydrogenases, which catalyze the reversible reduction of CO 2 to CO and HCOO À , respectively. [40] Beyond the biological importance of these reactivities, the conversion of CO 2 into hydrocarbons suggests an industrial use of nitrogenases for carbon capture and the production of biofuels in a green economy. Due to the increasing importance of carbon capture and biofuel production, we review and discuss the current knowledge of the CO and CO 2 conversion by all three nitrogenase isozymes: the Mo, V and Fe-only nitrogenase.

CO Reduction by Nitrogenases
Carbon monoxide is a colorless toxic gas that is isoelectronic to N 2 . In contrast to N 2 , CO is a strong π-acid due to its low energy π*-orbital. This allows CO to easily bind to transition metals by synergistic σ donation/π backdonation between the transition metal d-orbitals and the frontier molecular orbitals of CO. Thus, CO is one of the broadly used ligands in transition metal chemistry and likewise strongly interacts with the metal cofactors of the nitrogenases. CO has further been used as a C 1 building block for the synthesis of hydrocarbons in the Fischer-Tropsch process. [41] In this process, syngas (a mixture of CO and H 2 ) is produced by gasification of coal or steam reforming and passed over heterogeneous iron or cobalt catalysts at 300-350°C to produce hydrocarbons. The nitrogenase catalyzed conversion of CO and H + to hydrocarbons could be a green alternative to the Fischer-Tropsch process. Consequently, nitrogenases parallel two important industrial processes: the Haber-Bosch process for fertilizer production and the Fischer-Tropsch process for fuel production.

CO reduction by the Mo nitrogenase
Early studies showed that CO is a potent noncompetitive inhibitor for the reduction of N 2 , N 3 À , C 2 H 2 , and CN À catalyzed by the Mo nitrogenase. [42] While low partial pressures of CO (pCO = 0.005 atm) fully inhibit the reduction of N 2 , the hydrolysis of ATP and the formation of H 2 remain unaffected in the presence of CO. EPR analysis of the Mo nitrogenase under turnover conditions in the presence of CO revealed two new EPR signals. [43] One signal was observed under low CO partial pressures ([CO] = [MoFe protein]) and was denoted lo-CO Table 2. Selection of substrates reduced by nitrogenases.

Substrates
Reaction Ref.

Nitrogen substrates
Azide (S = 1/2). The second signal, formed at higher CO concentrations, was denoted hi-CO (S = 1/2). Labeling the MoFe protein with 57 Fe led to a signal broadening by nuclear hyperfine interactions, demonstrating that the EPR signal arose from one of the Fe-containing clusters of the MoFe protein. Both, the uninhibited H 2 production of the wild type (WT) Mo nitrogenase and the appearance of distinct EPR signals assigned to the CObound MoFe protein renders the CO inhibition assays a powerful tool for the investigation of nitrogenases. Following this strategy, Lowe et al. investigated the C 2 H 2 reduction mechanism of the Mo nitrogenase under CO. Under high electron flux the formation of H 2 was inhibited by C 2 H 2 . The addition of CO together with C 2 H 2 to the headspace did not restore the uninhibited H 2 formation rate. The authors therefore suggested that the Mo nitrogenase has different binding sites for CO and C 2 H 2 . [44] The pH range of the Mo nitrogenases is determined by two groups at pK a values of 6.3 and 9.0. [45] Turnover under CO shifts the basic pK a from 9.0 to 8.5, leading to an apparent inhibition of H 2 formation under basic conditions (pH > 8.5). Newton and co-workers studied different NifD mutations (H195N, Q191K, H195Q, R96Q, R96K, R359K, R277C, R277H, R96K/R359K) to identify the M-cluster position in the protein and to illuminate important residues for nitrogenase catalysis. [15,46] The effects of the mutations on the M-cluster environment were probed by the CO inhibition of the H 2 formation and changes in the lo-CO and hi-CO EPR signals.

Carbon-nitrogen substrates
Comparing the H 2 -formation under argon, all mutants exhibited a decreased electron flow relative to the WT MoFe protein. The H 2 formation of the mutants R96K, Q191K, R359K and R96K/ R359K was additionally inhibited by CO. Except for R96Q, none of the investigated mutants exhibited the hi-CO state.
Further spectroscopic studies were conducted to better characterize the lo-CO and hi-CO states. 13 C electron nuclear double resonance (ENDOR) spectroscopy revealed the binding of one or two CO molecules in the lo-CO or hi-CO signal, respectively, at the same metal cluster of the MoFe protein. [47] Both species could also be interconverted by pumping off or adding CO without the occurrence of a redox step. This suggests that the lo-and hi-CO signal originate from the same E n state of the MoFe protein. Selectively labeling the M-cluster with 57 Fe confirmed the M-cluster as the CO binding site and excluded the possible binding of CO to the P-cluster. [48] The hyperfine tensors of the 13 CO-labeled lo-and hi-CO states, derived from orientation-selective 13 C ENDOR spectroscopy, located the binding site of CO in the belt region of the Mcluster. An Fe bridging μ-CO was identified for the lo-CO and two terminally bound CO molecules were observed for the hi-CO state. [49] A new MoFe protein species with two bound CO molecules was observed under high electron flux by EPR spectroscopy (S = 3/2), known as hi(5)-CO. [50] While the lo-CO and hi(5)-CO species are photostable, hi-CO can be converted to lo-CO by light irradiation, which suggests structural differences between the hi-CO and hi(5)-CO species. [51] Moreover, CO causes a strong characteristic signal in IR-spectra, independent of the cofactor's spin state. Thus, CO was extensively used as an IR-probe to investigate the nitrogenase inhibition and the cofactor environment. [17,52] Stopped-flow FT-IR and IR-monitored photolysis implied a more complex ensemble of different species in samples prepared under hi-CO conditions. [52a,c] In the IR-monitored photolysis experiments, a species with difference bands at 1936 cm À 1 and 1858 cm À 1 was observed, which likely corresponds to the hi-CO, S = 1/2 EPR species. [52a] A second species with difference bands at 1936 cm À 1 and 1858 cm À 1 was observed in stored samples. The second species was assumed to correspond to the hi(5)-CO species. The two species were denoted Hi-1 and Hi-2. Some of the CO-bound species were investigated via nuclear resonance vibrational spectroscopy (NRVS) and extended X-ray absorption fine structure (EXAFS) spectroscopy. [53] Both methods indicated a symmetry reduction in the central Fe 6 S 9 C core of the M-cluster upon CO binding. Taken together the spectroscopic studies with density functional theory (DFT) calculations, the Fe-2,3,6,7 face of the Mcluster was identified as a potential binding site for CO and a structure with a μ-CO between Fe(2) and Fe(6) was proposed.
The CO binding was further characterized, by driving the MoFe protein electrochemically using Eu(II)L/Eu(III)L (L = polyaminocarboxylate) complexes as direct electron mediators to replace the Fe protein. [54] Comparing the potentials required for H 2 formation and CO binding to the MoFe protein, which both coincide close to À 900 mV, it was suggested that CO binding occurs in a state generated during the catalytic H + reduction cycle of the MoFe protein. With regard to the Lowe-Thorneley scheme, this state could be E 2 , as previously proposed. [55] Finally, the CO-inhibited MoFe protein structure reported by Spatzal et al. provided new insights into the binding mode of CO to the M-cluster (Figure 3c). [56] The CO-bound MoFe protein was isolated from a CO activity assay and subsequently crystallized. The M-cluster contained a bridging μ-CO ligand, which replaced the S2B belt sulfur between the Fe(2) and Fe(6) atoms. Thus, transfer of electrons and protons proceed the reversible replacement of the S2B sulfide. A potential sulfur storage site, binding the sulfur during turnover, was identified. Both findings imply that the activation of the M-cluster via the exposure of the Fe(2) and Fe(6) face is also a crucial step in the reduction of N 2 . More evidence for this hypothesis is added by the fact that CO binds to the M-cluster only under turnover conditions and cannot interact with the cofactor in the E 0 resting state, although the lo-CO state has the same oxidation level as E 0 . This indicates that the Fe(2)-Fe(6) remains protected and inaccessible for CO in the E 0 resting state. Moreover, the first CO-bound structure of the Mo nitrogenase closely resembles the proposed structure of the lo-CO state. [49] Crystals of the Mo nitrogenases in the hi-CO state were obtained from lo-CO single crystals by placing the lo-CO crystals under CO overpressure. [57] The structure of hi-CO showed an additional CO molecule terminally bound to the Fe(6) atom with an occupancy of q = 0.5 (t-site, Figure 3d), but otherwise remained nearly identical compared to the lo-CO structure. The CO-bound structures represent the lo-CO and hi-CO states as confirmed by EPR spectroscopy. [57] After discovering that the V nitrogenase reduces and couples CO to C 1 � 4 hydrocarbons, the activity of the Mo nitrogenase towards CO reduction was revisited by Ribbe and co-workers. [36a,58] In agreement with the lack of H + reduction inhibition by CO in the Mo nitrogenase, the total specific activity for hydrocarbon formation of the Mo nitrogenase is rather low, only~0.1 % of the activity observed for the V nitrogenase. The hydrocarbons produced by the Mo nitrogenase include C 2 H 4 , C 2 H 6 , C 3 H 6 and C 3 H 8 , but CH 4 was not detected. The CO reduction behavior is further influenced by replacing H 2 O with D 2 O in the buffer. In addition to the previously observed hydrocarbon products, D 2 O stimulated the formation of the longer hydrocarbon chains α-C 4 D 8 and n-C 4 D 10 . Besides optimizing the buffer conditions, mutants with improved CO reduction activity were identified. The V70A mutation in NifD increases the space surrounding Fe(2)-Fe(6) CO-binding site and results in an increased hydrocarbon formation activity (Figure 3). [36b] Changes in CO partial pressure and electron flow as well as additional mutations in the cofactor environment further modulated the CO reduction activity and product distribution.
The formation of hydrocarbons triggered the question whether the interstitial carbon of the M-cluster is involved in the conversion of CO. However, the interstitial carbon of the Mcluster is likely not interchangeable, as 13 C-labeling of the Mcluster did not result in the release of 13 C-containing hydrocarbons under turnover conditions. [59] To summarize this chapter, the interaction of CO with the M-cluster has been known since the 1970s and was characterized by mutational and spectroscopic means. CO was further used as a tool for the investigation of the Mo nitrogenase structure and mechanism. The validity and importance of these CO studies were highlighted by the discovery that the Mo nitrogenase can reduce CO to hydrocarbons. The capability of the Mo nitrogenase to bind and reduce CO suggests that the structurally elucidated sites of CO binding are the general sites of substrate binding in the M-cluster.

CO reduction by the V nitrogenase
Under CO, the H 2 formation of the V nitrogenase is inhibited by an average of 35 % without a decrease in the ATP consumption rate. [36a] The missing electron flow is redirected into the reduction of CO to hydrocarbons in a Fischer-Tropsch-like reaction, as demonstrated by Ribbe and co-workers. [36a,58] The formed hydrocarbons are CH 4 , C 2 H 4 , C 2 H 6 , C 3 H 6 , C 3 H 8 , α-C 4 H 8 and n-C 4 H 10 . C 2 H 4 is the main product, accounting for 94 % of the total hydrocarbon yield. [58] The activity of the V nitrogenase corresponds to an 800-fold increase in hydrocarbon formation compared to the Mo nitrogenase. The hydrogen source of the produced hydrocarbons was traced to H + that originated from the buffer. [60] While the conversion of CO by the V nitrogenase in D 2 O led to the formation of deuterated hydrocarbons, no incorporation of deuterium into the products was detectable for turnover under D 2 gas. In contrast to the Mo nitrogenase, changing the buffer from H to D only slightly effected the CO reduction and the hydrocarbon product distribution of the V nitrogenase. The different impact of D on the Mo and V nitrogenase catalysis suggests differences in the coupled electron/proton tunnelling of this reaction. [58] Due to the CO reduction rates by the V nitrogenase, the isolation of the V nitrogenase corresponding to lo-CO and hi-CO states is hampered under turnover conditions, i. e. under   [62] The hi-CO state of VnfDGK was later also isolated by exposing VnfDGK to an increased CO pressure and using [Eu(II)-DTPA] as reductant. [63] The second CO molecule bound to the V-cluster was not part of the CÀ C coupling reaction and it was suggested that the second CO binding site is independent from the CO conversion to hydrocarbons.
Recently, the X-ray crystal structures of the lo-CO and hi-CO states of the V nitrogenase have been reported by Einsle and co-workers. These structures gave further insights into the VFe active site (Figure 3e, f). [64] The lo-CO and hi-CO structures of the VFe protein closely resemble the homologous structures of the MoFe protein. All structures exhibit a μ-CO bridging Fe(2) and Fe(6) and a terminal bound CO in the hi-CO state (Figure 3e, h). The residues of the MO nitrogenase shown to affect the lo-CO state upon mutation, exhibit similar conformations in the structures of the CO bound MoFe and VFe protein. However, residue R96 of NifD is replaced by lysine (K83) in VnfD and as described above, one of the belt sulfides in the V-cluster is replaced by a CO 3 À -ligand (Figure 3c). How these changes can account for the large difference in Mo and V nitrogenase activity is not obvious. The origin of the reactivity difference is likely to be more subtle. Single crystals of the VFe protein in the lo-CO state could partially be interconverted to the hi-CO state under CO overpressure, with the occupancy of q = 0.5, as observed for the Mo nitrogenase. The binding and dissociation process of the terminally bound CO was completely reversible within 100 s, as monitored by in situ attenuated total reflectance IR difference spectroscopy.
Beyond the in vitro characterization of the CO binding and activation by nitrogenases, it was demonstrated in the model organism A. vinelandii that the V nitrogenase reduces CO to hydrocarbons also in vivo. [65] After expression of the V nitrogenase, the cultures were incubated under 15 % CO for 8 h. C 2 H 4 , C 2 H 6 and, C 3 H 8 were formed and the turnover number was determined to be~750. The hydrocarbon yield could be further increased by incubating the cultures alternately under an air and a 15 % CO atmosphere. Alternating the environmental gas composition allowed A. vinelandii to recover from phases with CO gas. The achieved specific activity for the in vivo C 2 H 4 formation (1 mmol C 2 H 4 × [g dry cell] À 1 × h À 1 ) was already in the range of other bio-ethylene formation methods without any optimization.
The CO-reducing and CÀ C coupling activity of the V nitrogenase renders this enzyme system as a potential biological alternative to the industrial Fischer-Tropsch process for fuel formation.

CO reduction by the Fe-only nitrogenase
The H 2 formation catalyzed by the Fe-only nitrogenase is progressively inhibited by increasing CO concentrations until it reaches an inhibition of 35 % at a CO partial pressure of 0.4 atm. [66] In contrast to the V nitrogenase, the Fe-only nitrogenase reduces CO mainly to CH 4 with a maximum specific activity at a CO partial pressure of~0.05 atm. The CH 4 formation activity of the Fe-only nitrogenase decreased to 10 % of the maximum value, when the CO partial pressure was further increased. Therefore, the CH 4 formation alone cannot fully account for the observed inhibition of H 2 formation at elevated CO concentrations and either CO inhibits the electron flux of the Fe-only nitrogenase or an undetected product is formed. [66] The differences in the product profile between the V nitrogenase (mainly C 2 H 4 ) and the Fe-only nitrogenase (only CH 4 ) is so far unexplained, as no structural or spectroscopic characterization of the Fe-only nitrogenase under CO has been carried out. However, a similar product shift was observed for the CO reduction by the isolated V-and L-cluster under [Eu(II)-DTPA]. [67] The L-cluster is the cofactor precursor of the M-cluster with the chemical composition [Fe 9 S 9 C]. The L-cluster closely resembles the proposed structure of the Fe-cluster except that (R)-homocitrate is missing. The preferred reduction of CO to CH 4 by the L-cluster indicates that the preference for CH 4 formation of the Fe-only nitrogenase could originate from the intrinsic cofactor properties rather than the protein environment. Further investigations are needed to elucidate the CO conversion by the Fe-only nitrogenase.

Comparison of the Mo, V and Fe-only nitrogenase
The specific activity for CO reduction is increasing from the Mo nitrogenase (traces) through the Fe-only nitrogenase (4.5 nmol CH 4 × [nmol FeFe protein] À 1 × min À 1 ) to the V nitrogenase (7.5 nmol C 2 H 4 × [nmol FeFe protein] À 1 × min À 1 ). [66] The pronounced disparities in the CO reduction activity between the Mo and V nitrogenase raise the question whether these disparities originate from the different protein environments or the different cofactor compositions.
The isolated M-cluster and V-cluster are able to produce hydrocarbons from CO at similar rates by using the strong reductants [Eu(II)-DTPA] or SmI 2 as electron sources. [67,68] Still, the rate of the isolated V-cluster only corresponds to 0.05 % of the activity reached by the V nitrogenase, highlighting the strong influence of the V nitrogenase protein. This was further confirmed by investigating the CO-reduction behavior of a Vcluster containing MoFe protein (NifDK V ) and an M-cluster containing VFe protein (VnfDGK M ). [69] The hydrocarbon formation activity of NifDK V was 640 times smaller as the one observed for VnfDGK. [69a] Similarly, the hydrocarbon formation activity for VnfDGK M increased by 100-fold compared to NifDK. [69b] Nevertheless, the cofactor also has a significant influence on the CO reduction activity, as VnfDGK was found to be~6-fold more active than VnfDGK M .
Despite the differences in reactivity, the Mo and V nitrogenase show similar interactions with CO, forming a lo-and hi-CO state, and for both nitrogenases the states are interconvertible. The same behavior and mechanism seem to apply to the Fe-only nitrogenase. On the basis of the lo-and hi-CO structures of the V nitrogenase, a mechanistic scheme was proposed (Scheme 1). [64b] After reaching the E 2 state, CO binds to the t-site and replaces the hydride at the μ-site ((2)-(4)). The reaction can then proceed either by CO insertion of the hydride in (4) leading to the formyl adduct (7) followed by the reduction of CO to hydrocarbons in a cascade of reductions and a CO addition ((8)-(11)) or by hydride protonation forming the lo-CO state (5). The lo-CO state (5) can bind an additional CO molecule to form the hi-CO state (6). This mechanism explains why the hi-CO state cannot couple both bound CO molecules, as the terminal CO at the t-site of (6) has to be exchanged with a hydride to form intermediate (4) before the CO reduction to hydrocarbons is initiated. The preference of C 2 H 4 formation by the V nitrogenase is explained by the preferential β-H elimination in (11) over the methyl protonation in intermediate (8). On the other hand, the Fe-only nitrogenase produces mainly CH 4 , which could be explained by a facilitated protolysis of intermediate (8) or by the occurrence of a σ-bond metathesis between intermediate (8) and a generated H 2 molecule.
To summarize, the activation of CO by the Mo and V nitrogenases seems to depend on the same CO binding mode to the M-and V-cluster. Nonetheless, the CO activity and product profile between the three nitrogenase isoenzymes differs significantly. The activity differences appear to mainly originate from the different protein scaffolds (NifDK, VnfDGK, AnfDGK) as the introduction of the M-cluster into the V nitrogenase and the reverse exchange, the introduction of the V-cluster into the Mo nitrogenase, suggest. However, the exact influence of the protein scaffold on the activity could not be clarified, yet.

CO 2 Reduction by Nitrogenases
The greenhouse gas CO 2 is the main source of carbon in the atmosphere. The anthropogenic CO 2 emissions of the last century increased the abundance of CO 2 in the earth's atmosphere by 40 %, compared to the pre-industrial era. [70] The increased CO 2 concentration is the main contributor to global warming. [71] Besides having a detrimental impact on the environment, the ubiquity of CO 2 renders it a cheap and accessible source of C 1 building blocks. Hence, carbon capture to reduce the CO 2 concentration and conversion of CO 2 into value-added products are pressing challenges for chemistry and biology. Recently, nitrogenase proteins were discovered to be capable of activating and reducing CO 2 to hydrocarbons. Potentially, this reactivity offers new solutions to the great challenge of carbon capture and utilization.

CO 2 reduction by the Mo nitrogenase
Seelfeldt et al. first reported the reduction of CO 2 to CO by the Mo nitrogenase. [37a] Hemoglobin was used as a CO probe and the formation of CO was monitored by spectrophotometric quantifications of carboxyhemoglobin. This method further prevented the formed CO from inhibiting the Mo nitrogenase during the in vitro assay, but was not suitable for the detection Scheme 1. Proposed mechanism of the CO reduction by nitrogenases. [64b] Upon electron flux, the resting state of the cofactor (1) acquires two e À and two H + to form the hydride bound E 2 state (2). CO then binds to the t-site resulting in adduct (3), which undergoes a rearrangement of CO and the hydride to (4). The lo-CO state (5) of the cofactor in the formal E 0 redox state is then formed by protonation of the hydride in (4). The lo-CO state (5) is subsequently interconverted to the hi-CO state (6) under CO overpressure. The intermediate (4) alternatively undergoes a CO insertion reaction to the formyl adduct (7), which is the starting point of the CO reduction cascade (8)-(11). After the addition of four extra e À /H + and a CO molecule, the ethyl adduct (11) is formed. Next, intermediate (11) undergoes a β-H elimination reaction to form the hydride bound E 2 state (2) and releases C 2 H 4 . of alternative CO 2 reduction products. The formation of H 2 during Mo nitrogenase turnover at elevated CO 2 concentrations (0.45 atm) is inhibited by 16 %. [72] The electrons are redirected towards CO 2 reduction. HCOO À was later identified by 1 H and 13 C NMR spectroscopy as the main product of the CO 2 reduction by the WT Mo nitrogenase and its mutants. HCOO À is formed at a rate of 9.8 nmol × [nmol MoFe protein] À 1 × min À 1 . The formation of CO is much slower and proceeds at a rate of 0.03 nmol × [nmol MoFe protein] À 1 × min À 1 , although it was discovered first. This product profile was corroborated by quantum mechanical calculations. The reaction pathway of direct hydride transfer from the E 2 state to CO 2 , releasing HCOO À , is favored over the alternative pathway of CO 2 activation at the M-cluster with the reduction of the Fe-CO 2 intermediate leading to CO. The product profile of the CO 2 reduction by the Mo nitrogenase could be extended to CH 4 by introducing two mutations to the NifD subunit (V70A and H195Q). [73] The formation of CH 4 likely proceeds over an intermediate CO in a non-dissociative mechanism. The addition of the CO binder deoxyhemoglobin reduced the amount of formed CH 4 by 25 % in the CO 2 reduction assays. The coupling of two or more CO 2 molecules to longer hydrocarbon chains was not detected. Instead, the Mo nitrogenase could couple C 2 H 2 and CO 2 to C 3 H 6 .
Moreover, the CO 2 reduction behavior of the MoFe protein was electrochemically investigated by immobilizing the MoFe protein in a polymer film on an electrode and cobaltocene was used as the electron mediator. [74] The electrochemical and natural Fe protein-driven CO 2 reduction by the MoFe protein showed similar product distributions with 8 % and 9 % of electron flux directed towards the formation of HCOO À , respectively. The CO 2 reduction activity of the Mo nitrogenase double mutant (V75A and H201Q) was further demonstrated in vivo using the anoxygenic phototroph Rhodopseudomonas palustris. [75] The in vivo formation of CH 4 under photochemotrophic growth is dependent on light for the ATP generation and requires an electron source like acetate (H 3 CCOO À ) or thiosulfate (S 2 O 3 2À ). The yield of CH 4 was further increased by directing metabolic electrons to the nitrogenase proteins. This was partially achieved, either, by preventing further bacterial growth via resuspension of the cells in a nitrogen-depleted medium or by deletion of the Calvin-Benson-Bassham (CBB) cycle from the organism, which is an alternative electron accepting pathway for anoxygenic phototrophs. [76] In short, the Mo nitrogenase converts CO 2 mainly to HCOO À and a little bit of CO. Nonetheless, Mo nitrogenase mutants can also reduce CO 2 via the intermediate CO to CH 4 .

CO 2 reduction by the V nitrogenase
The V nitrogenase can reduce CO 2 to CO at similar rates as the Mo nitrogenase. [37b] Furthermore, Rebelein et al. detected the formation of CD 4 and the CÀ C coupling products C 2 D 4 and C 2 D 6 in sub-stoichiometric amounts by changing the medium from H 2 O to D 2 O. In contrast to the V nitrogenase, no CÀ C coupling products were observed for the Mo nitrogenase. The carbon source for the formed hydrocarbons was traced to CO 2 by 13 C-labeling. Moreover, it was shown that the formation of C 2 D 4 and C 2 D 6 proceeds independently from the produced CO, while CD 4 partially originated from CO. The product distribution of the CO 2 reduction reaction is modulated, when the reductant [Eu(II)-DTPA] is used as the electron source. [77] In this [Eu(II)-DTPA] system, the formation of CH 4 and the CÀ C coupling products occur in H 2 O. Furthermore, the product profile is extended from C 2 H 4 and C 2 H 6 to C 3 H 6 , C 3 H 8 , C 4 H 8 and C 4 H 10 . The hydrocarbon formation activity increases dramatically in D 2 O for the V nitrogenase under [Eu(II)-DTPA] and CO 2 , with the exception of CD 4 , which was not detected under these conditions. Both findings, the longer hydrocarbon chains and the increased product formation indicate that D 2 O facilitates CÀ C coupling by the V nitrogenase. The ability of the VFe protein to CÀ C couple CO 2 to C 2 H 4 and C 3 H 6 was also demonstrated in the electrochemically driven CO 2 reduction by the VFe protein, using 1,1'-dicarboxycobaltocenium as the electron mediator. [78] Moreover, the V nitrogenase forms detectable amounts of CH 4 from H 2 CO 3 inside R. palustris. Increased amounts of CH 4 were observed in strains expressing a double mutant of the VFe protein (V57A and H180Q in VnfD). [8] To summarize, also the V-nitrogenase reduces CO 2 to CO. Furthermore, hydrocarbon chains can be produced, either, by replacing H with D in the buffers or by using the strong reductant [Eu(II)-DTPA].

CO 2 reduction by the Fe-only nitrogenase
The reduction of CO 2 by the Fe-only nitrogenase has been characterized in AnfH/ATP-dependent in vitro assays and electrochemically, using a polymer-coated electrode containing the FeFe protein and relying on cobaltocene as electron mediator. [74] In the AnfH and the electrochemically driven reaction, 31 % or 32 % of the electron flux is directed towards the reduction of CO 2 to HCOO À , respectively. For the AnfH/ATPdriven in vitro assays, the formation of trace amounts of CH 4 were reported, but the possible formation of CO was not mentioned.
The CO 2 reduction activity by the Fe-only nitrogenase was also confirmed in non-growing R. palustris. In contrast to the Mo and the V nitrogenase, strains expressing the wild type Feonly nitrogenase formed already elevated amounts of CH 4 . [26] Surprisingly, introducing the double mutations of the V nitrogenase to the Fe-only nitrogenase (V57A and H180Q in AnfD) stopped the formation of CH 4 in R. palustris.
In a nutshell, the WT Fe-only nitrogenase has the highest inherent activity for the reduction of CO 2 . Remarkably, the same portion of electrons (31 %) are used either for the reduction of N 2 or for the reduction of CO 2 under the according atmosphere.

CO 2 reduction by the Fe protein
In 2017, Yu and co-workers demonstrated that the Fe proteins of the Mo and V nitrogenase (NifH and VnfH) are capable of reducing CO 2 to CO. [79] Sub-stoichiometric amounts of CO were an interesting research field for the development of new homogenous CO 2 reduction catalysts. [83]

Biological Relevance of the Nitrogenase CO and CO 2 Reduction
Organisms that grow anaerobically encounter the problem of accumulating abundant amounts of reducing equivalents in ferredoxins and flavodoxins, if their carbon source is more reduced than their biomass. The enrichment of reduced redox cofactors ultimately prevents growth, if the organism cannot transfer the electrons to an electron acceptor. Three viable mechanisms have been identified in purple non-sulfur bacteria to independently support anaerobic phototrophic growth from such carbon sources. In the CBB cycle electrons are accepted by CO 2 , DMSO reductases use excess electrons to reduce DMSO, and nitrogenases use electrons for N 2 and H + reduction and the formation of NH 3 and H 2 . [76,84] Therefore, nitrogenases serve as redox mediators. In accordance with this additional role of nitrogenases, the H 2 formation by the Mo nitrogenase is increased in vivo upon interruption of the CBB cycle by deletion of RuBisCO. [85] The reduction of CO and CO 2 by the alternative nitrogenases potentially contributes to this mechanism of redox cofactor recycling.
As discussed above, the Mo nitrogenase is strongly inhibited by CO, even under environmental concentrations. To protect the Mo nitrogenase from CO, some diazotrophs express the protein CowN. CowN interacts with the Mo nitrogenase and prevents its inhibition by CO, without decreasing the nitrogenase activity for N 2 and H + reduction. [86] For instance, the expression of CowN allows the Mo nitrogenase dependent growth of Rhodobacter capsulatus for CO concentrations of up to 3.5 %. However, the deletion of CowN did not affect the cell growth sustained by the Fe-only nitrogenase under CO. Interestingly, AnfA, the transcriptional activator of the Fe-only nitrogenase, represses the expression of CowN, indicating that the CO protection is not required for the Fe-only nitrogenase. Therefore, the promiscuous activity for CO reduction by the V and Fe-only nitrogenase could indicate a function of these nitrogenases as detoxifying enzymes. The amount of CH 4 formed from CO 2 by the Fe-only nitrogenase in R. palustris was also sufficient to support the growth of a co-cultivated methanotroph Methylomonas sp. LW13. [8,9b] A methanotroph living of the CH 4 formed by nitrogenases implies that the CO and CO 2 reduction might play a role in the symbiosis of microbial communities, and not only presents a detoxifying function. Our understanding regarding the biological role of the nitrogenases' 'side reactivities' is very fragmented and calls for further investigation.

Summary and Outlook
The in vivo reduction of N 2 , CO and CO 2 by the WT V and Feonly nitrogenases implies a potential evolutionary link between the nitrogen and carbon cycle on earth. [65] The need of all living organisms for an N source has selected nitrogenase ancestors specialized for N 2 reduction. Repurposing the nitrogenase machinery for the reduction of CO and CO 2 to form CÀ C bonds is an exciting path to biotechnological processes and the synthesis of value-added products. Recently, V nitrogenase expressing A. vinelandii cells were explored for the production of C 2 H 4 from CO in 2 l bioreactors. [87] Here, the achieved C 2 H 4 productivity could not compete with industrial methods. Still, this process can be improved by changing the reaction conditions or by protein engineering. The effect of protein engineering was highlighted by introducing double mutations to the MoFe (V75 A and H201Q) and VFe (V57A and H180Q) proteins, increasing the CO 2 reduction and CH 4 formation activity. [8] Naturally, the Fe-only nitrogenase directs a large portion (31 %) of its electron flux towards the reduction of CO 2 and is capable to produce CH 4 inside cells. This activity renders the Fe-only nitrogenase an ideal basis for the development of reduction catalysts for the utilization of the ubiquitous greenhouse gas CO 2 .