Structure, reactivity, and spectroscopy of nitrogenase-related synthetic and biological clusters

The reduction of dinitrogen (N2) is essential for its incorporation into nucleic acids and amino acids, which are vital to life on earth. Nitrogenases convert atmospheric dinitrogen to two ammonia molecules (NH3) under ambient conditions. The catalytic active sites of these enzymes (known as FeM-cofactor clusters, where M = Mo, V, Fe) are the sites of N2 binding and activation and have been a source of great interest for chemists for decades. In this review, recent studies on nitrogenase-related synthetic molecular complexes and biological clusters are discussed, with a focus on their reactivity and spectroscopic characterization. The molecular models that are discussed span from simple mononuclear iron complexes to multinuclear iron complexes and heterometallic iron complexes. In addition, recent work on the extracted biological cofactors is discussed. An emphasis is placed on how these studies have contributed towards our understanding of the electronic structure and mechanism of nitrogenases.


Introduction
Iron sulfur cluster-containing proteins are found in all branches of life and carry out a wide range of essential processes, spanning from electron transfer to DNA repair to catalysis. 1,2 The metallocofactors of these proteins range from simple mononuclear iron centers, dinuclear Fe 2 S 2 clusters, tetranuclear Fe 4 S 4 clusters, the complex P-cluster (Fe 8 S 7 ) and the FeM-cofactors (MFe 7 S 9 C, where M = Mo, V or Fe) of the nitrogenase family of enzymes. 3,4 Understanding how the varied structural complexity of these enzymes enables their diverse functionality has been a subject of ongoing research for decades. [5][6][7][8][9][10][11] Max Planck Institute for Chemical Energy Conversion, Stiftstr. [34][35][36]45470 Mülheim an der Ruhr, Germany. E-mail: serena.debeer@cec.mpg.de Since the earliest structural studies of MoFe protein in the late 1970's, 24,25 nitrogenase has served as a source of inspiration for synthetic chemists. Early studies by Holm 26,27 and Coucouvanis 28,29 focused on the syntheses of Fe 4 S 4 clusters by self-assembly reaction from FeCl 3 /NaSH/NaSR. The addition of [MoS 4 ] 2À to the reaction mixture afforded the heterometallic [MoFe 3 S 4 ] 3+ single cubane clusters and [Mo 2 Fe 6 S 8 ] double cubane clusters. These weak-field Mo-Fe-S clusters were among the first bio-inspired synthetic models of FeMo-cofactor, and interestingly predated the protein crystal structure by more than a decade. At the time, the researchers largely took inspiration from the available extended X-ray absorption fine structure (EXAFS) data on FeMo-cofactor, which provided the first hints for the coordination environment surrounding both the Mo and Fe atoms. 25,30 While the crystal structure of FeMo-cofactor, revealed by Rees and coworkers in 1992, 31 turned out to be more complex than either the single or double cubane models of Holm and Coucouvanis, it remains the case that these early models are among some of the closest structural models of FeMo-cofactor. The redox chemistry and electronic structure of many of these metal clusters have been studied by a range of spectroscopic methods, including Mössbauer, electron paramagnetic resonance (EPR), X-ray absorption (XAS) and X-ray emission (XES) spectroscopies. 10,[32][33][34][35] However, neither the single or double cubane models are able to replicate the reactivity of nitrogenase enzymes and no dinitrogen adduct has been isolated. Interestingly, however, [MoFe 3 S 4 ] 3+ and [VFe 3 S 4 ] 2+ cubanes have been shown to enable the catalytic conversion of hydrazine to ammonia, acetylene to ethylene and dimethyldiazene to methylamine, by utilizing cobaltocene and lutidinium chloride as the source of electrons and protons, respectively. 29 It is noteworthy to mention that most of these reactions cannot be enabled by Fe 4 S 4 clusters, and even in cases where this can be achieved (e.g., for acetylene to ethylene conversion), the Fe 4 S 4 clusters are much slower than the heterometal containing [MoFe 3 S 4 ] 3+ and [VFe 3 S 4 ] 2+ cubanes, suggesting a role of the heterometal in optimizing catalytic activity. This parallels the observation that the Fe-only nitrogenases are by far the least efficient for N 2 reduction. 4 Nevertheless, in recent years mutation studies, high-level spectroscopy, and protein crystallography all appear to have converged on the iron atoms being the site(s) of substrate binding. [36][37][38] This has shifted much of the attention of synthetic chemists in recent years towards understanding how molecular iron models can ultimately reduce N 2 . The approach has been in many ways much more reductionist than early nitrogenaseinspired synthetic chemistry. Rather than trying to build a molecule that looks like FeMo-cofactor, researchers have focused more on trying to understand the complex FeMo-cofactor cluster, by analyzing its fundamental building blocks and determining how one can systematically alter the iron electronic structure and ultimately reactivity. 6,7,27,39 To this end, researchers have incorporated various light atoms to model the central carbide and have also incorporated sulfur ligation. [40][41][42][43][44] They have examined the ability of these clusters to bind nitrogen, form metal hydrides, and reduce a range of N 2 and N 2 -derived related substrates. 45,46 Additionally, these models have also been utilized to understand potential pathways for biosynthesis of FeMocofactor. 47 Herein, we review recent synthetic studies aimed at capturing key aspects of the FeMo-cofactor cluster, which enable its unique reactivity. We first review mononuclear iron model complexes that have been utilized as models for N 2 activation. We then discuss multinuclear iron model complexes, heterometallic model complexes and finally the extracted cofactors from the nitrogenase enzymes. In all cases, we place emphasis on the key role that spectroscopy has played in characterizing these complexes and providing key signatures for studies of the enzyme.

Studies of mononuclear Fe sites as models for N 2 activation
While FeMo-cofactor is a complex 7Fe-1Mo cluster, chemists have found great utility in utilizing simple mononuclear iron complexes in order to test hypotheses as to how FeMo-cofactor might function. In this context, the work of Peters and coworkers has been of particular interest as they have reported the first mononuclear iron complex for catalytic N 2 reduction. 48 The model complex utilizes a tripodal phosphine-borane ligand (3P:1B), which takes inspiration from the FeMo-cofactor 3S:1C coordination mode of each iron in FeMo-cofactor (Fig. 1). Peters and coworkers designed their model inspired by the idea that the central carbon of FeMo-cofactor might allow for greater conformational and redox flexibility during catalysis by allowing the Fe modulate the degree of interaction it has with the light atom. In their model, the Fe-B interaction acts as the adjustable ''spring'', allowing a lower coordinate Fe to effectively bind the substrate. 48,49 This strategy allowed N 2 coordination followed by catalytic N 2 reduction with the presence of the reductant. Reduction of dinitrogen to form ammonia was achieved with the presence of both the proton source and reductant under N 2 atmosphere at -78 1C using a mononuclear iron complex. 48   double resonance (ENDOR) spectroscopy was used to determine whether the protons were equivalent. Unlike EXAFS and Mössbauer analysis, which probes all iron-containing species in the sample, ENDOR allows the selective detection of the S = 1/2 species. The ENDOR spectra showed two distinct 1 H doublets, which is consistent with an unsymmetrically protonated [-N-NH 2 ] group. The spectroscopic data revealed that the dinitrogen activation for [(SiP iPr 3 )Fe-N 2 ] À followed a distal-like mechanism since the diazene adduct, which would represent the intermediate formed in an alternating mechanism, was not observed spectroscopically.
The activity of the mononuclear iron catalyst for NH 3 formation from N 2 could be improved by nearly an order of magnitude by increasing acid/reductant loading. To interrogate the mechanism of N 2 activation, 57 Fe Mössbauer spectroscopy was utilized to characterize the iron speciation under turnover conditions. 51  ] was the resting state of the turnover reaction as H 2 evolution was also observed. These studies highlight the important mechanistic insights that can be obtained by performing spectroscopic studies of reaction intermediates.
In addition to boron, ligands with either carbon and silicon atom have also been investigated. 52,53 The isolation of the [(BTp iPr 3 )Fe(QN-NH 2 )] + intermediate was hampered by its high reactivity and the fact that it could not be isolated in a pure form. The use of the ligand with silicon atom allowed the isolation of [(SiP iPr 3 )Fe(QN-NH 2 )] + (compound 2, Fig. 2). Single crystal X-ray diffraction analysis revealed a Fe-N distance of 1.672(2) Å, which could be assigned as a doubly-bonded Fe-N. The vibrational data of the solid [(SiP iPr 3 )Fe(QN-NH 2 )] + displayed shifts from 3207 and 3039 cm À1 to 2380 and 2241 cm À1 upon the use of DOTf as the acid. The vibrational peaks at lower energies (2380 and 2241 cm À1 ) were assigned to N-H and N-D stretching frequencies, respectively, with strong hydrogen bonding interaction with the triflate anion in the solid state. 57  )] + was treated with stoichiometric Cp* 2 Co and the resulting mixture was allowed to warm to room temperature over 10 minutes. After filtration, the organic solvents were removed in vacuo, the residue was dissolved in C 6 D 6 for NMR spectroscopy. After HCl workup, N 2 H 4 and NH 3 could be detected and the ratio of the two products was found to be time-dependent with more N 2 H 4 being converted to NH 3 over time. These spectral data along with reactivity studies evidenced that the conversion from N 2 to NH 3 in this system followed a distal-to-alternating mechanism providing precedence for a possible hybrid mechanism in biological systems. Among these ligands, SiP iPr 3 , BTp iPr 3 and CP iPr 3 (see compounds 1, 2, and 3 in Fig. 2), the most flexible axial bonding, BTp iPr (Fe), showed the highest dinitrogen activation reactivity under a common set of reaction conditions. SiP iPr 3 (Fe) only yielded stoichiometric ammonia and CP iPr 3 (Fe) fell in between in terms of both flexibility (i.e., the Fe-B bond was more flexible than the Fe-C bond when a nitrene intermediate was formed) and reactivity. With similar ligand sets, Co, 54 Ru, 55 and Os 55 analogues were found to exhibit activity for dinitrogen fixation. In addition to the reactivity studies, these model complexes have provided insight into a wide range of intermediates relevant to nitrogen reduction. Through EPR, Mössbauer, XAS, single-crystal X-ray diffraction, and DFT studies, the expansive studies of Peters and coworkers have been key in establishing potential spectroscopic fingerprints for key intermediates corresponding to both distal and alternating pathways for N 2 reduction.
In addition to N 2 -to-NH 3 conversion, reduction of the cyanide to methane and ammonia was studied. 56  3 )Fe-C(NH 2 )] + , which was isolated and characterized with X-ray diffraction. The intermediate [(SiP iPr 3 )Fe-C(NH 2 )] + suggested a mechanism related to the distal mechanism of N 2 reduction. These studies also showed that these mononuclear iron complexes were capable of binding a range of substrates, potentially analogous to the Fe sites of FeMo-cofactor. It is noteworthy to mention that among these intermediates (Fe-NQNR, FeQN-NR 2 , Fe-CNR, and FeRC-NR 2 , where R = H, Me), the bond dissociation energies for the N-H bonds range from 37 to 65 kcal mol À1 . 57 The weakening of the N-H bonds can be attributed to the varying degrees of NRN or CRN bond weakening that occurs concomitant with H-atom addition.
In an effort to more closely capture the iron ligation sphere in FeMo-cofactor, Holland and coworkers synthesized an Fe(II) tris(thiolate) adduct ([L(Fe)-N 2 ] 2À , compound 4, Fig. 2) which modelled the Fe-S/C coordination in FeMo-cofactor. 58,59 The group focused on the hypothesis that one of the Fe-S bonds dissociates upon reduction/protonation to provide an open coordination at the iron site for substrate binding. 37,60 [L(Fe)-N 2 ] 2À could be synthesized by reduction of [L(Fe)-SAr*] À (compound 5, Fig. 2) in an N 2 atmosphere. Similarly, reduction of [L(Fe)] À in an N 2 atmosphere also afforded [L(Fe)-SAr*] À . It is noteworthy to mention that the reduction of [L(Fe)-SAr*] À involves Fe-S dissociation, which is consistent with proposed substrate binding mechanism in FeMo-cofactor iron sites. Single-crystal X-ray diffraction of [L(Fe)-N 2 ] 2À revealed Fe-S bond distances of 2.32 and 2.35 Å and a Fe-C bond distance of 2.04 Å, which were in line with averaged Fe-S and Fe-C bond distances (2.26 Å and 2.01 Å respectively) of the resting state of FeMo-cofactor. Superconducting quantum interference device (SQUID) measurements revealed that [L(Fe)-N 2 ] 2À had a highspin (S = 1) electronic configuration in solid state, which was consistent with DFT calculations. Addition of a weak acid to [L(Fe)-N 2 ] 2À afforded NH 3 and N 2 H 4 in low yields. [L(Fe)-SAr*] À could be reduced by one electron in N 2 or Ar without Fe-S dissociation at -70 1C to afford [L(Fe)-SAr*] 2À . EPR measurements showed that [L(Fe)-SAr*] 2À had two signals with slightly different rhombicities, but both with an S = 3/2 spin state. The reason for the two components in the EPR spectrum was unclear, but it was suggested that two conformers of [L(Fe)-SAr*] 2À were present in the frozen sample which could not be resolved at 80 K at zero field. It is noteworthy to mention that EPR spectroscopy is much more sensitive to small perturbations in zero-field splitting than Mössbauer spectroscopy. 61 suggesting [L(Fe)-SAr*] 2À had a high-spin electronic configuration and [L(Fe)] À was low-spin. Reduction of [L(Fe)-SAr*] À by 2.4 equiv. of KC 8 at À65 1C under argon atmosphere resulted in Fe-S dissociation to afford [L(Fe)] 2À . Single-crystal X-ray diffraction revealed the iron atom in [L(Fe)] 2À adopted a Z 6 coordination to the central arene ring. 57 Fe Mössbauer analysis suggested [L(Fe)] 2À had an S = 1 spin state as a high-spin Fe(0) complex. Dinitrogen binding was observed by Mössbauer spectroscopy when [L(Fe)] 2À (generated in situ from [L(Fe)-SAr*] 2À under argon) was exposed to N 2 atmosphere to afford [L(Fe)(N 2 )] 2À . Further reduction could be achieved by the treatment of [L(Fe)(N 2 )] 2À with 20 equiv. of a variety of proton sources, affording 1% or less of hydrazine and ammonia with alternating mechanism. It is noteworthy to mention that the catalytic reaction was not achieved due to the protonation of the thiolate ligand followed by the degradation of the complex via ligand dissociation. Holland's model chemistry resembles a single Fe-S-C site of FeMo-cofactor, and the spectroscopic methods provide insightful information in terms of the electronic configuration of the intermediates and the dinitrogen activation mechanism.

Studies of multinuclear Fe sites as models for N 2 activation
In the previous section, we focused on mononuclear iron complexes which mimic certain aspects of FeMo-cofactor's geometric and electronic properties and in some cases, also its reactivity. However, modeling the complexity of FeMocofactor with high fidelity clearly requires the synthesis of larger, more complex multinculear clusters, which allow the roles of multiple iron sites, as well as that of the central carbide atom, to be more rigorously evaluated.
In contrast to mononuclear iron complexes, Holland and coworkers isolated a dinitrogen complex K 2 [(L)Fe(N 2 )Fe(L)] (compound 6,    Fig. 3), where L was a diketiminate ligand, in which the R group was varied, as shown in Fig. 3. Single-crystal X-ray diffraction analysis revealed that the N-N bond of dinitrogen was more activated in K 2 [(L)Fe(N 2 )-Fe(L)] compared to that in [(L)Fe(N 2 )Fe(L)] (R = t-Bu, N-N bond distances were 1.233(6) and 1.182(5) Å respectively). While the potassium activated dimer clearly showed a longer N-N bond, it could not be completely cleaved in this complex. In order to achieve N-N bond cleavage, the low-coordinate dimeric iron complex was supported by a less bulky ligand. 67 Under N 2 atmosphere, reduction of [L 0 FeCl] 2 by KC 8 afforded an iron tetramer (compound 9, Fig. 3), where the dinitrogen was activated to form a bis-nitride. Single-crystal X-ray diffraction revealed that one of the nitrides was attached to two potassium cations and two iron atoms as the other nitride coordinated to three iron atoms. The zero-field Mössbauer spectrum revealed that the bis-nitrido complex had three different iron environments. The decreased steric hindrance around the iron atoms allowed the three-electron reduction by three iron atoms to break the triply-bonded dinitrogen, evidenced by the Fe 3 N 2 core. The analysis of the isomer shifts indicated the presence of two equivalent highspin Fe(III) nitrides, a high-spin three coordinate Fe(II) site, and a high-spin four coordinate Fe(II) site. Stoichiometric ammonia formation was observed when the Fe 4 N 2 complex was worked up in acidic conditions. In contrast to the monometallic complexes, in which the formation of the metal nitride requires protonation via a distal mechanism, the end-on bimetallic dinitrogen activation allows the formation of two metal nitrides without protonation, resembling the proposed mechanism in the Haber-Bosch process. 68 In addition to the catalytic insights provided by the Holland group's complexes described above, these complexes also served as invaluable references for the development of valence-to-core X-ray emission spectroscopy (VtC XES) as a probe of dintrogen bond actiavtion. 69 XES is element specific which utilizes high-energy X-rays to generate a 1s core hole on a metal atom and the photons emitted are measured when the core vacancy is filled by electron in higher lying orbitals. 70,71 Since the intensity of XES spectra follows the dipole selection rule (the amount of metal np character present in the donor orbital can be expressed by the oscillator strength), transitions from metal 2p orbitals (Ka emission) and 3p orbitals (Kb emission) are both possible and give intense features. Moreover, any filled ligandbased valence orbitals that possess appreciable metal p character will also display emission features (Kb 00 and Kb 2,5 ''valence-to-core'' (VtC) emission), albeit with relatively low intensity. A series of iron-N 2 complexes (compounds 6, 9, [L 0 FeCl] 2 , and [(L)Fe(N 2 )Fe(L)]) were examined by VtC XES to interrogate the degree of dinitrogen activation (Fig. 4). It was found that the energy of the VtC XES feature at B7100 eV was proportional to the to the N-N stretching frequency and inversely proportional to the N-N bond length. These experiments showed that VtC XES is able to probe N-N bond length changes as small as 0.02 Å. Due to the ability of XES be measured in a dispersive mode, it also opens up a new spectroscopic means to detect small molecule activation during transition metal catalysis. 72 In contrast to end-on Fe-N 2 or linear Fe-N 2 -Fe, examples of non-linear Fe-N 2 -Fe complexes are rare and less investigated. 73,74 Tomson and coworkers utilized a macrocycle as the ligand to construct a constrained environment around the iron atoms and interrogated its impacts on the structural and electronic configuration. The dinucleating macrocycle was used to force the geometry of the bimetallic iron complex which was inaccessible by typical ligands. 75  DFT calculations revealed that the nonlinear Fe 2 N 2 unit had an additional interaction between Fe d and N 2 p* compared to linear Fe 2 N 2 unit, but the mixing was too low to activate dinitrogen significantly. This study opens an avenue for potentially regulating the geometry of the M-N 2 -M motif by using a bulky ligand to constrain the distance between the iron sites. Further, the importance of the substrate geometry with respect to the active site metals is highlighted in order to achieve optimal dinitrogen activation.
It has been proposed that the sulfides and/or the carbide of the FeMo-cofactor cluster play essential roles in enabling conformational flexibility during catalysis. 37,52,58,60,76,77 This could occur through actual Fe-ligand bond breaking processes or simply through a weakening of the Fe-ligand interactions upon substrate binding. The large size of the FeMo-cofactor cluster, may help stabilize these processes, by having the transformation localized at a single iron site. Inspired by this idea, Holland and coworkers isolated a [4Fe-3S] 2À cluster (compound 11, Fig. 5) in which three of the irons are fourcoordinate and one of the iron atoms is a planar threecoordinate site, ligated by only sulfides 78,79 with an average Fe-S bond length of 2.17(1) Å. The EPR measurements of the [4Fe-3S] 2À cluster revealed an S = 1/2 ground state from antiferromagnetic coupling between the four paramagnetic iron centers. Zero-field Mössbauer measurements at 80 K of powder samples of [4Fe-3S] 2À cluster displayed three quadrupole doublets in a 2 : 1 : 1 ratio (Fig. 6). Density functional theory calculations were used to locate the Fe(I) atom. The simulated Mössbauer parameters agreed with the experimental data within 0.10 mm s À1 only when the central iron was assigned as Fe(I).
Oxidation of the [4Fe-3S] 2À cluster by AgPF 6 afforded an allferrous [4Fe-3S] À cluster and the calculated ligand-field splitting diagram revealed that the splitting of the [4Fe-3S] À cluster between (d x 2 Ày 2, d xy ) and (d xz , d yz ) of the central iron was relatively large, rationalizing the low-spin configuration. The electronic configuration of [4Fe-3S] À cluster was further probed by Mössbauer measurements, where the spectrum showed a superposition of two doublets in a 3 : 1 ratio, corresponding to peripheral and central iron site respectively (Fig. 6). A characteristic lower isomer shift of 0.37 mm s À1 attributed to the central Fe(II) of the [4Fe-3S] À cluster fell outside of the reported Fe(II) sites in Fe-S clusters (0.6-0.8 mm s À1 ), 80-83 highlighting its low-spin electronic configuration. EPR measurements showed that the all-ferrous [4Fe-3S] À cluster exhibited a featureless perpendicular mode spectrum. The central iron site of the [4Fe-3S] À cluster exhibited biomimetic reactivity for the N-N reduction of hydrazine and the isolated product was an amide-bound iron-sulfur [4Fe-3S] À -NH 2 cluster. Infrared spectrum of [4Fe-3S] À -NH 2 cluster exhibited amide N-H stretch. 1 H NMR measurements further revealed the signal for the N-H proton, which disappeared when the deuterated analogue was used. Zero-field Mössbauer measurements at 80 K of powder samples of the [4Fe-3S] À -NH 2 cluster (containing three Fe(II) and one Fe(III)) displayed two overlapping quadrupole doublets in a 3 : 1 ratio with isomer shifts of 0.76 and 0.74 mm s À1 , respectively (Fig. 7). The observed isomer shifts were inconsistent with the previously reported four-coordinate Fe(III)-S sites (B0.3 mm s À1 ). Thus, the hole was delocalized over the peripheral iron sites (an average of 2.33), as the central iron was assigned as high-spin Fe(II). To investigate and compare the oxidation states of the [4Fe-3S] 2À , [4Fe-3S] À , and [4Fe-3S] À -NH 2 clusters, the first derivative of the Fe K-edge XAS spectra were interrogated and the [4Fe-3S] À -NH 2 cluster was found to have the highest average oxidation state, while the [4Fe-3S] 2À cluster had the lowest oxidation state in the series (Fig. 8). The shoulder at B7115 eV was assigned to 1s to 4p z transition at the central trigonal-planar iron atom, shifting to a higher energy by 0.41 eV upon oxidation of Fe(I)-containing [4Fe-3S] 2À cluster to all-ferrous [4Fe-3S] À cluster. These XAS data were consistent with the oxidation of the central iron atoms assigned by the Mössbauer measurements. The presented spectroscopic signatures are important for the identification of unsaturated iron sites, which may be of relevance to intermediate states in FeMo-cofactor. The biomimetic hydrazine reduction also highlights the reactivity of the unsaturated iron site, providing precedence for a related mechanism in the enzyme.
The oxidation state of the iron sites are important to substrate binding, as described in the previous text. The exact distribution of oxidation states in FeMo-cofactor has been a longstanding question in nitrogenase research. In recent years, combined Mo/Fe XAS, 84,85 Mössbauer, 86 X-ray magnetic circular dichroism (XMCD), 87,88 and DFT studies, 84,89 converged on a 4Fe(II):3Fe(III) oxidation state distribution in FeMo-cofactor in the E 0 state; however, it is important to note that these data do not provide insight into how the electrons are spatially distributed. To this end, spatially resolved anomalous dispersion refinement (SpReAD) of MoFe protein in the resting form, has shown that the Fe1/3/7 positions are more reduced than the other four irons. 35,90 Crystallographic studies have also shown that CO binds at the Fe2/6 site of FeMo-cofactor, replacing a belt sulfide. This finding is somewhat counterintuitive as the CO binds at what has been shown to be is the more oxidized face of FeMo-cofactor, at least in the resting form. However, it is important to note that no SpReAD data are available on CO-bound FeMo-cofactor, and it is possible that redox reorganization could occur within the cluster. This idea is supported by recent synthetic model studies of Agapie and coworkers in which a series of tetranuclear iron complexes with various oxidation states were synthesized as models to study redox chemistry and concomitant structural change upon CO binding. 91 [LFe 3 O(PhIm) 3 Fe] n+ (n = 3, 2, 1, 0), corresponding to Fe II Fe III 3 through Fe II 4 (compound 12, Fig. 9), were synthesized. Bond lengths were determined by single-crystal X-ray diffraction and the analysis focused on the Fe 4 (m 4 -O) motif, which could be viewed as an apical iron atom and three basal iron atoms. Bond length analysis showed that upon oxidation from [LFe 3 O (PhIm) 3 Fe] 2+ to [LFe 3 O(PhIm) 3 Fe] 3+ , a contraction of one of the basal Fe-O (2.092(2) to 1.983(4) Å) was observed, consistent with the oxidation of the basal tri-iron core. In contrast, reduction of [LFe 3 O(PhIm) 3 Fe] 2+ to [LFe 3 O(PhIm) 3 Fe] + resulted in the elongation of the apical Fe-O (1.813(2) to 1.883(4) Å), consistent with the reduction of the apical Fe. 57 Fe Mössbauer spectrum at 80 K of [LFe 3 O(PhIm) 3 Fe] 2+ revealed four quadrupole doublets, corresponding to four inequivalent iron atoms. The best fit to the Mössbauer spectrum revealed a basal Fe II 2 Fe III core with an apical high-spin Fe II . Compared to the spectrum of [LFe 3 O(PhIm) 3 Fe] 2+ , the signal of the basal core Fe II of the [LFe 3 O(PhIm) 3 Fe] 3+ near 3 mm s À1 decreased in intensity, which was consistent with the oxidation of the triiron core. The parameters showed that the apical iron of [LFe 3 O(PhIm) 3 Fe] 3+ was a high-spin Fe III . Upon the reduction of [LFe 3 O(PhIm) 3 Fe] 2+ to form [LFe 3 O(PhIm) 3 Fe] + , there was no change in the relative intensity of the Lorentzian feature near 3 mm s À1 but an increase in the isomer shift (from 0.19 to 0.89 mm s À1 ) of the quadrupole doublet, which was assigned to the reduction of the apical iron. These Mössbauer data was consistent with the X-ray crystallographic analysis. To provide further support for the high-spin configuration of the four-coordinate apical iron of [LFe 3 O(PhIm) 3 Fe] 2+ and [LFe 3 O(PhIm) 3 Fe] 3+ , variable temperature (VT) magnetic susceptibility and variable temperature-variable field (VTVH) magnetization data were obtained to interrogate the nature of the spin ground state and the exchange coupling with a focus on [LFe 3 O(PhIm) 3 Fe] 2+ , which had the shortest apical Fe-O in the series. The fits revealed that the apical iron was Fe III which afforded an S = 4 ground state (the apical Fe III had a strong antiferromagnetic interaction with basal iron atoms, resulting in ferromagnetic alignment of the spins on basal iron sites at low temperatures, S = (À5/2) (apical Fe III ) + 5/2 (basal Fe III ) + 2 Â 2    92 To better understand the intermediates involved in FeMocofactor chemistry, the synthesis of models relevant to more reduced E states, which may bind hydrides and/or N 2 , is essential.     95 The spectroscopic data and the kinetic profile suggested that the E 4 state might comprise [Fe(m-H)(m-S)Fe] motif. As the formation of the bridging metal hydrides in FeMo-cofactor has been suggested to be an essential part of the mechanism, where reductive elimination of H 2 is believed to occur upon N 2 binding at the E 4 2À , which had been characterized previously. 66,97 The formation of the formally diiron(0)-N 2 complex implied another reduction event occurred during the reaction as the KC 8 was only sufficient to reduce [(L)Fe II (m-H)(m-S)Fe II (L)] À to the diiron(I) complex. Gas chromatography of the headspace of the reaction revealed that H 2 formed during the reaction, suggesting the reductive elimination was possible, at least under this reaction condition. However, other unidentified species formed during this process so it was not possible to conclude that the reductive elimination was the second reduction event. Nevertheless, it could be inferred that [(L)Fe II (m-H)(m-S)Fe II (L)] À bound N 2 and formed H 2 thereafter, consistent with reductive elimination of H 2 upon N 2 binding which is proposed as an essential component in the mechanism of all nitrogenases.
To understand the impact of electronic structure of carbide on substrate binding and activation, synthetic analogues are of interest. Both Rauchfuss and Rose have recently isolated iron carbonyl clusters with central carbides and sulfides as the ligands. [98][99][100][101] However, these clusters all contain low-valent, low-spin iron and thus are not of direct electronic structural relevance to FeMo-cofactor. While thus far it has not been possible to fuse two iron sulfur cubanes with a carbide (as realized in the biosynthesis of FeMo-cofactor 47,102 ), synthetic chemists have attempted to approach this question in a more stepwise fashion utilizing high-spin iron sites which are electronically analogous to those in FeMo-cofactor and incorporating light atoms into dinuclear and larger clusters. The intensity of the pre-edge at B7111 eV increased upon oxidation, indicating a higher probability for Fe 1s -Fe-based 3d transitions as more holes were presented in a more oxidized complex. In contrast, the intensity of the pre-edge at B7113-7115 eV decreased upon oxidation, suggesting less Fe 1s -3d/carbyne transitions. The intensity loss of this feature indicated a higher Fe-C covalency. The isomer shifts measured by 57  The exact role of the carbide in FeMo-cofactor and how it modulates the electronic structure of this cluster relative to other iron sulfur clusters remain open questions. 14,15,105 This has motivated efforts to synthetically and spectroscopically investigate the effect of the light atom inclusion in Fe-S clusters. Using a combination of XAS spectroscopy and DFT calculations a series of iron sulfur clusters were investigated. 106 These included [Fe 2 S 2 Cl 4 ] 2À , 107 110 Using ligand K-edge XAS (at both the S and Cl K-edges), the pre-edge features were examined to understand the modulation in metal ligand covalency as a function of light atom introduction into the Fe-S clusters. While strong modulations in the electronic structure were observed for the dimers, the tetramers showed only subtle perturbations upon inclusion of a single ligand (ÀN t Bu). The studies were then extended to hypothetical clusters, and it was suggested that strong electron donors, such as a carbide, were required to modulate the electronic configurations of tetramers or larger clusters. On this basis, it was hypothesized that the carbide might play not only a structural role in FeMo-cofactor, but also modulated the electronic structure to enable optimal reactivity.
The biosynthesis of the FeM-cofactors of nitrogenase involves modification of iron-sulfur (Fe-S) precursors which are enabled by the activity of radical S-adenosylmethionine (SAM) enzymes. The central carbide originates from the transfer of a methyl group from SAM to a precursor [Fe 4 S 4 ] cluster. 32,33,47,111 This suggests that [Fe 4 S 4 ]-alkyl clusters are important intermediates in the formation of the FeM-cofactors, as well as in other radical reactions mediated by SAM enzymes. [112][113][114] However, the role of the alkyl has not been studied extensively in model chemistry. To understand how an electron-donating and strong-field ligand (i.e., alkyl) perturbs the electronic configuration of the [Fe 4 S 4 ] clusters, Suess and coworkers isolated the [Fe 4 S 4 ] 2+ -Et cluster (compound 17, Fig. 11) by adding Et 2 Zn to the [Fe 4 S 4 ] 2+ -Cl cluster. [115][116][117] Single-crystal X-ray diffraction revealed the Fe-C bond length in Fe-Et was 2.05 Å, which was comparable to that in a tris(thioether)borate-ligated Fe 2+ -Me complex (2.03 Å). 118 57 Fe Mössbauer spectra of [Fe 4 S 4 ] 2+ -Cl and [Fe 4 S 4 ] 2+ -Et clusters at 90 K and the calculations revealed that the positive charge was more localized on the apical iron site in the [Fe 4 S 4 ] 2+ -Et cluster due to the electron-donating ethyl group on the apical site. Reversible formation of alkyl radicals was reported by Suess and coworkers by using the electron-donating ligands to promote homolysis of [Fe 4 S 4 ] 2+ -alkyl cluster (compound 18, Fig. 11). 119 The mechanistic studies revealed that the homolysis rate was dependent on the substituent of the ligand; homolysis rate was increased when a more electron-donating ligand was used (i.e., reaction with DMAP had a higher rate than that with pyridine), forming the R-R coupled hydrocarbons as the final products. This facile process even generated primary octyl radicals at room temperature. To further understand the possible intermediates related to the alkylated clusters, the synthesis of [Fe 4 S 4 ] 3+ -Et cluster was attempted. However, one-electron oxidation did not afford target compound. Instead, a tris (imidazolinimine)borate ligand was used and the [Fe 4 S 4 ] 3+ -Me cluster (compound 19, Fig. 11) was synthesized by one-electron oxidation of the [Fe 4 S 4 ] 2+ -Me cluster. 120  The belt sulfide (S2B) has been proposed as a potential leaving group to generate a low-coordinate iron site for N 2 binding. Efforts have been made in cluster synthetic chemistry (e.g., the reduction of Fe-Cl to generate a low-valent Fe site) but the unwanted side reactions (e.g., dimerization of two [MoFe 3 S 4 ] clusters) hamper the N 2 binding as the iron site is not protected by the protein. The same group recently utilized the sterically demanding N-heterocyclic carbenes (NHCs) as the protective ligands to outcompete the side reaction and the reduced clusters bound N 2 to form the bridging complex (compound 20, Fig. 11). 121 Single-crystal X-ray diffraction studies revealed that the N-N bond distance (1.145(3) Å) was elongated in the bridging complex. Mössbauer spectra as well as the bond length analysis revealed the increasing cluster covalency upon N 2 binding. Interestingly, the average Fe-S distances for all the iron sites decreased, suggesting the contribution of all the iron sites for N 2 binding. The findings illustrate that the covalent interactions are important for cluster N 2 binding chemistry.

Studies of heterometallic sites as models for N 2 activation.
In the previous section, we focused on the recent iron model chemistry aimed at understanding FeMo-cofactor. However, as FeMo-cofactor incorporates Mo into the cluster, the potential role of the heterometal in modulating the electronic structure and reactivity remain a question of ongoing interest. As discussed in the introduction, there has been a longstanding interest in the synthesis of heterometallic clusters to model FeMo-cofactor in order to capture its unique electronic and structural properties, beginning with the early work of Holm and Coucouvanis. As some of the synthetic models capture some aspects of FeMo-cofactor, none of them fully resemble FeMo-cofactor in terms of chemical structure and reactivity on dinitrogen reduction.
In the following section, recent progress on this field will be discussed.
To systematically synthesize various heterometallic clusters, ligand metathesis was utilized to incorporate Mo and W into clusters by Holm and coworkers. 44 This approach afforded versatile heterometallic complete and incomplete cubanes which captured many of the key geometric and electronic features of the active sites in the FeMo-cofactor (Fig. 12). Although reactivity of these clusters was lacking, they have provided key structural and spectroscopic reference points for  (Fig. 13). 126 Although the afforded heterometallic clusters do not capture the properties of FeMo-cofactor, it serves as a useful tool to broaden the scope of the synthetic heterometallic clusters. While many synthetic model complexes related to FeMocofactor have been synthesized and characterized, none of them catalyze dinitrogen activation to form ammonia under ambient conditions. In contrast to reducing N 2 under the strong reducing conditions employed in much of the synthetic chemistry described above, photochemical ammonia production has been realized by using ultraviolet light in semiconductor films (e.g., titania, 127-129 diamond, 130,131 and fullerenes 132 ). Recently, light-driven reduction of dinitrogen to form ammonia in aqueous solution under ambient temperatures and pressures was reported by Kanatzidis and coworkers using the solid-state chalcogels composed of [Mo 2 Fe 6 S 8 (SPh) 3 ] 3+ and [Sn 2 S 6 ] 4À clusters (Fig. 14). 133 The immobilized FeMoS active sites were proposed to feature a high spatial density proximal to other active sites, allowing efficient multielectron transformations during chemical events. In addition, Mo-free chalcogels were investigated which contained only Fe 4 S 4 clusters. 134 Interestingly, the Mo-free variants showed even higher N 2 reduction efficiency, suggesting that Mo is not necessary for optimal catalysis, at least in the tested photochemical conversion.

Reactivity of extracted FeMo-cofactor
In the previous sections, we have discussed various synthetic models to capture the electronic properties, structural properties, and reactivity of the FeMo-cofactor cluster of nitrogenase. Despite all the attempts, a cluster that fully models FeMo-cofactor has yet to be realized. To further interrogate the chemical properties of nitrogenase, isolated FeMo-cofactor, FeV-cofactor and FeFe-cofactor have been extracted by strongly coordinating organic solvents (e.g., N-methylformamide (NMF)) from MoFe protein. 11,135,136 It is noteworthy to mention that extracted cofactors have yet to be structurally characterized by single-crystal X-ray diffraction and our understanding of the cofactors geometric and electronic structures have relied heavily on spectroscopy. Among the three extracted cofactors, extracted FeFe-cofactor has been characterized by EXAFS and 57 Fe Mössbauer spectroscopy. 137 Its EPR-silent nature in the resting state limits further characterization of its ground state electronic properties. Extracted FeV-cofactor, in contrast, has been characterized more recently by both EPR and EXAFS, however, a more detailed characterization has yet to be realized. 138 Extracted FeMo-cofactor is by far the most studied and we would like to briefly review its characterization in this section. Extracted FeMo-cofactor has been characterized by EXAFS 139 and EPR studies on extracted FeMo-cofactor reveal a broadened S = 3/2 X-band spectrum, which has been attributed to the increased D-strain in the extracted cofactor relative to that contained within wild-type MoFe protein. 138,140 It has been proposed that slight geometric variations change the distribution of D-strain, and the change will be more pronounced in the solvated form. 25,141 The underlying structural rigidity in MoFe protein may play a significant role in terms of reactivity. Here it is of interest to note that small-angle X-ray scattering (SAXS) results have suggested extracted FeMo-cofactor is not monomeric in DMF, but rather forms oligomers in DMF and the addition of coordinating thiols does not change the aggregation behavior in DMF. 142 However, as FeMo-cofactor can be reinserted into apo  MoFe protein and recover its reactivity, it appears that the structural integrity should be reasonably retained or at least recovered upon incorporation into MoFe protein. 12, 135 Hu, Ribbe and coworkers studied the activity of extracted FeMo-cofactor and FeV-cofactor on CO and CN À reduction with Eu(II) as the reductant. 143 The reactivity studies revealed that both of the extracted cofactors reduced CO and CN À to hydrocarbons with FeMo-cofactor exhibiting slightly better activity. It is noteworthy that there were discrepancies in the CO reduction reactivity between the solvent-extracted/Eu(II)-driven and protein-bound/ATP-driven reactions. The total amount of hydrocarbons formed by the extracted FeMo-cofactor and FeV-cofactor were 67.9% and 0.05% of the totals produced by the protein-bound FeMo-cofactor and FeV-cofactor respectively. 20,143 The drastic decrease in extracted FeV-cofactor's CO reduction activity rendered extracted FeMo-cofactor, which was only 0.1% as active as FeV-cofactor within proteins, slightly more active than extracted FeV-cofactor, highlighting the significance of the protein environment during chemical reactions in nature. The same group revealed that NMF-extracted FeMo-cofactor, FeV-cofactor, and FeFe-cofactor were capable of reducing CO, CN À , and CO 2 to form hydrocarbons with SmI 2 as the reductant. 144 Unfortunately, while hydrocarbon formation by reduction of carbon monoxide or cyanide has been reported, extracted FeMo-cofactor, FeV-cofactor and FeFe-cofactor exhibit no N 2 reduction reactivity. It should also be noted that all the above-mentioned CN À , CO, and CO 2 reductions are sub-stoichiometric and efficient catalytic reduction of these substrates has yet to be achieved.
To understand the impact of the protein environment on the reactivity, synthetic clusters have been inserted into proteins in order to understand the contribution of the protein environment. In this context, it is of interest to note this strategy has been very successful for hydrogenases but thus far limited for nitrogenases. [145][146][147][148][149][150][151][152] Hu, Ribbe and coworkers inserted a synthetic iron cluster, [Fe 6 S 9 (SEt) 2 ] 4À , into the catalytic component of the nitrogenase (designated as NifDK apo ) with this biohybrid designated as NifDK Fe . 149 The reactivity studies revealed that C 2 H 4 and C 2 H 2 reduction were achieved when NifDK Fe was used in the presence of Eu(II). [Fe 6 S 9 (SEt) 2 ] 4À and NifDK apo exhibited no reactivity, highlighting the importance of both the metal cofactor and the protein environment in terms of reactivity. An extracted cofactor is possibly reinserted into another protein as a ''hybrid protein'' but detailed experiments and characterization have yet to be realized. 111,153 We note, however, that there are many examples in which synthetic complexes have been inserted into proteins (i.e., artificial metalloenzymes) or synthetic scaffolds (e.g., zeolites), and the environment around the complexes has been shown to play an important role in tuning the chemo-, regio-, site-, and enantioselectivity during catalysis. [154][155][156][157][158][159] High-resolution single-crystal X-ray diffraction has revealed that carbon monoxide, a molecule that inhibits FeMo-cofactor, replaces the belt sulfur atom at 2B position (S2B) in FeMocofactor. 37 The lability of the belt sulfide in FeMo-cofactor inspired the idea to substitute S2B with Se due to its spectroscopic properties by the addition of KSeCN under proton-reducing turnover conditions. 60 Selective substitution of the bridging sulfurs by selenium in FeMo-cofactor served as a tool in which the Fe-Se environment could be interrogated by utilizing highenergy resolution fluorescence detected X-ray absorption spectroscopy (HERFD XAS). Our group, together with the Rees group, utilized Se HERFD XAS to reveal that 2B and 3A/5A bridging positions of Se-substituted FeMo-cofactor were electronically distinct. The Fe2/Fe6 edge consistent with an antiferromagnetically coupled diferric pair, while the Fe3/Fe4/Fe5/Fe7 face of the cofactor exhibited more localized ferrous character. These differences were attributed to asymmetry in the electrostatic and hydrogen bonding interactions with the belt sulfides, suggesting a more localized electronic configuration. Measurements of the extracted selenated cofactor would be an ideal way to test this hypothesis, however, thus far has not been possible. Finally, it is also of interest to note that Se HERFD XAS studies of CO-bound selenated FeMo-cofactor, show that the Fe3/Fe4/Fe5/Fe7 face of the CO-bound cofactor appears more oxidized from the perspective of the coordinated Se ligands than the resting state. This is counterintuitive, since the cluster has to be reduced to bind CO. This observation provides support for redox organization occurring within FeMo-cofactor in order to facilitate substrate binding. Interestingly, the potential importance of redox reorganization in FeMo-cofactor was first hypothesized based on model chemistry. 91 Thus, the subsequent discovery of its relevance to FeMo-cofactor further highlights the synergy that exists between the synthetic and enzymatic systems.

Summary and outlook
In this review, we have discussed various synthetic models that have been used to understand the mechanism, potential intermediates, reactivity, and electronic properties of nitrogenases. Studies of these molecular models have provided key spectroscopic fingerprints which have been, and will continue to be, essential in correlated studies with the enzyme. These molecular model studies have also shown that a single iron site is capable of catalytic N 2 reduction, 48,51-53 albeit only at low temperature and under harsh reducing conditions. Interestingly, dinuclear and multinuclear iron model complexes have been shown to activate, and in one case even cleave, the strong triple bond of N 2 . 67,75,93 However, catalytic N 2 reduction has only been observed in multinuclear clusters through photochemical initiation and in contrast to the enzyme, Mo has been found to decrease, rather than increase, activity. 133,134 Hence, it is clear that a significant gap remains between synthetic modelling and the actual FeM-cofactor clusters.
Interestingly, while the FeM-cofactor clusters can be extracted into organic solvent, research in this area remains limited due to the unstable nature of these clusters outside the protein environment. 20,143 It is noteworthy to mention that the extracted FeMo-cofactor can be reinserted to its protein and the reactivity is restored. This indicates that the protein environment plays a prominent role in optimizing the active site for catalysis. In the field of nitrogenase research, it appears there remains ample room to further explore protein engineering and artificial metalloenzyme approaches, which may allow for the present gaps between synthetic modelling and the biological systems to be bridged. While great advances have been made in our understanding of nitrogenase, with substantial contributions from synthetic chemistry, it is clear there remains much to explore in molecular, biological, and biohybrid systems.

Conflicts of interest
The authors declare no competing financial interests.