Identity and function of an essential nitrogen ligand of the nitrogenase cofactor biosynthesis protein NifB

NifB is a radical S-adenosyl-L-methionine (SAM) enzyme that is essential for nitrogenase cofactor assembly. Previously, a nitrogen ligand was shown to be involved in coupling a pair of [Fe4S4] clusters (designated K1 and K2) concomitant with carbide insertion into an [Fe8S9C] cofactor core (designated L) on NifB. However, the identity and function of this ligand remain elusive. Here, we use combined mutagenesis and pulse electron paramagnetic resonance analyses to establish histidine-43 of Methanosarcina acetivorans NifB (MaNifB) as the nitrogen ligand for K1. Biochemical and continuous wave electron paramagnetic resonance data demonstrate the inability of MaNifB to serve as a source for cofactor maturation upon substitution of histidine-43 with alanine; whereas x-ray absorption spectroscopy/extended x-ray fine structure experiments further suggest formation of an intermediate that lacks the cofactor core arrangement in this MaNifB variant. These results point to dual functions of histidine-43 in structurally assisting the proper coupling between K1 and K2 and concurrently facilitating carbide formation via deprotonation of the initial carbon radical.

N ifB plays a key role in the biosynthesis of the active site of Mo-nitrogenase, an [(R-homocitrate)-MoFe 8 S 9 C] metallocofactor (designated M-cluster) that catalyzes the ambient conversion of dinitrogen (N 2 ) to ammonia (NH 3 ) [1][2][3][4][5][6] . A member of the radical S-adenosyl-L-methionine (SAM) enzyme family 7,8 , NifB is a monomeric protein carrying three [Fe 4 S 4 ] clusters: [9][10][11][12] one, designated the SAM-cluster, is ligated by a canonical CxxxCxxC motif and coordinates the SAM-binding [Fe 4 S 4 ] cluster; the other two, designated the K-cluster, are ligated by a number of conserved ligands and supply the full complement of iron (Fe) atoms for the biosynthesis of an 8Fe precursor to the mature cofactor. Previous studies of the NifB proteins from Azotobacter vinelandii and Methanosarcina acetivorans have led to the proposal that NifB utilizes a unique, radical SAM-dependent mechanism for carbide insertion concomitant with the transformation of the K-cluster to an [Fe 8 S 9 C] cluster (designated the Lcluster), which represents both an 8Fe precursor and an all-Fe core of the M-cluster [13][14][15][16] . This process begins with an S N 2-type transfer of a methyl group from a SAM molecule to the K-cluster (Supplementary Fig. 1a, ➀), and it is followed by hydrogen abstraction of the methyl group by a 5′-deoxyadenosyl (5′-dA•) radical that is derived from the homolytic cleavage of a second SAM molecule ( Supplementary Fig. 1a, ➁). The resultant methylene radical then undergoes further deprotonation/dehydrogenation to yield a μ 6coordinated interstitial carbide concomitant with the radicalchemistry-based coupling and rearrangement of the two [Fe 4 S 4 ] modules of the K-cluster (designated K1-and K2-cluster, respectively) into an [Fe 8 S 9 C] L-cluster, which represents an all-Fe core of the M-cluster except for the substitution of one Fe atom for Mo/ homocitrate at one end of the cluster ( Supplementary Fig. 1a, ➂). Once generated, the L-cluster is passed from NifB onto NifEN, the next biosynthetic scaffold, where it is matured into a fully complemented M-cluster upon insertion of Mo and homocitrate by NifH prior to the delivery of M-cluster to its final binding site in NifDK, the catalytic component of Mo-nitrogenase ( Supplementary  Fig. 1a, ➃) [17][18][19][20][21][22] .
The unique radical chemistry that occurs on NifB during the nitrogenase cofactor core assembly makes this protein an attractive subject of investigation, as knowledge in this regard will shed light on the unusual synthetic route to a high-nuclearity metallocofactor that is biologically important and chemically elusive. Previously, we have identified three sets of Cys ligands-three Cys per set-for the three [Fe 4 S 4 ] clusters in M. acetivorans NifB (designated MaNifB), namely, the SAM-cluster (Cys 50 , Cys 54 , and Cys 57 ), the K1-cluster (Cys 30 , Cys 63 , and Cys 129 ) and the K2-cluster (Cys 264 , Cys 274 , and Cys 277 ) ( Supplementary Fig. 1b, c) 11 . Further, we have established the coexistence of SAM-and K2-clusters as a prerequisite for methyltransfer and hydrogen abstraction to occur and pinpointed the K2-cluster as the site for methyl attachment and the subsequent hydrogen abstraction from the methyl group by a 5′-dA• radical 11 . Perhaps even more excitingly, using pulse EPR techniques, we have identified a nitrogen atom from a His residue as the fourth ligand for the K1-cluster that is lost upon coupling between the K1-and K2-clusters into an L-cluster 11 . This observation suggests an important role of the His/nitrogen ligand in generating the nitrogenase cofactor core (i.e., the L-cluster), particularly given the ability of His to undergo protonation/deprotonation, which either provides a release mechanism for the L-cluster onto the next biosynthetic apparatus or allows this residue to participate in the further deprotonation/dehydrogenation of the carbon radical to give rise to a central carbide. As such, it is crucial to identify this nitrogen ligand and probe its function in the process of nitrogenase cofactor core formation.
Here we use a combination of site-directed mutagenesis and pulse EPR spectroscopy to show that the His 43 residue is the specific nitrogen ligand for the K1-cluster of MaNifB. Our biochemical and EPR analyses demonstrate the essential role of His 43 in the formation of the nitrogenase cofactor core structure, although substitution of this residue with Ala does not impact early steps of carbide insertion leading to the initial hydrogen atom abstraction of the K2-associated methyl group. Our XAS/EXAFS analysis further reveals a shortened distance between the K1-and K2-clusters upon substitution of His 43 with Ala, as well as a further processing of these clusters into an intermediate between the K-and L-clusters upon incubation with SAM. These observations point to a dual function of His 43 in positioning/orienting the K1-cluster relative to the K2-cluster for proper coupling and deprotonating the initial carbon radical for carbide formation.

Results
The identity of the histidine ligand of the K1-cluster. Sequence analysis of MaNifB revealed the presence of three highly conserved His residues, His 28 , His 43 and His 219 , which could potentially serve as the nitrogen ligand for the K1-cluster on NifB Ma ( Supplementary  Fig. 1b Fig. 2b). Thus, a loss of the His ligand does not seem to impact the ability of MaNifB to ligate any of the three [Fe 4 S 4 ] clusters, likely due to the 3-Cyscoordination of these clusters that is sufficient to secure them in place.
To assess whether a nitrogen ligand is still present in these MaNifB variants, pulse EPR spectroscopy-specifically, three-pulse electron spin echo envelope modulation (3P-ESEEM)-was used to observe modulations to the time domain spectrum and corresponding peaks in the fast Fourier transformed (FFT) spectrum that can be assigned to the coordinating nuclei ( Fig. 1; also see Supplementary  Fig. 3). Consistent with our previous observation 11 , MaNifB wt shows deep modulations in the time domain of the ESEEM spectrum ( Fig. 1a, trace 1) and corresponding intensity between 0 and 6 MHz in the FFT (Fig. 1b, trace 1), which have been previously assigned to the hyperfine and quadrupole couplings of a cluster-ligated 14 N nucleus 11 . Similar modulations and intensities are present in the ESEEM spectrum and FFT of MaNifB K1 , a variant carrying only the K1-cluster but no SAM-and K2-clusters (because of substitutions of the Cys ligands of the SAM-and K2-clusters with Ala 11 ), affirming the previous assignment of the nitrogen ligand to the K1-cluster ( Fig. 1a, b, trace 5). While similarly deep modulations and intensities are observed in the ESEEM spectra and FFTs of MaNifB H28A and MaNifB H219A (Fig. 1a, b, traces 2, 3), these features are clearly absent from the ESEEM spectrum and FFT of MaNifB H43A (Fig. 1a, b, trace 4), suggesting His 43 as the nitrogen ligand for the K1-cluster. To seek further support for this assignment, MaNifB K1-H43Aanother MaNifB variant carrying only the Cys ligands of K1 along with a substitution of His 43 with Ala-was heterologously expressed in E. coli. The purified MaNifB K1-H43A demonstrates the same subunit composition as MaNifB K1 , as well as the same Fe content that is consistent with the presence of one [Fe 4 S 4 ] cluster (i.e., the K1-cluster) per protein upon FeS reconstitution ( Supplementary  Fig. 2a, b). However, contrary to MaNifB K1 (Fig. 1a, b, trace 5), MaNifB K1-H43A does not show deep modulations and intensities in its ESEEM spectrum and FFT (Fig. 1a, b, trace 6), firmly establishing His 43 as the nitrogen ligand that specifically coordinates the K1cluster.
The role of histidine ligand in L-cluster maturation. Continuous wave (CW) EPR analysis provided the first insights into the function of His 43 in the cofactor core assembly process. Consistent with the coupling and rearrangement of the K1-and K2-clusters into an L-cluster, an L-cluster-specific, g = 1.94 signal 9,10,15 is observed in the spectra of MaNifB wt , MaN-ifB H28A , and MaNifB H219A upon incubation of these proteins with SAM (Fig. 2a, traces 1-3). In contrast, the g = 1.94 signal is absent from the spectrum of MaNifB H43A following the same treatment, suggesting a lack of K-to L-cluster transformation on this protein after incubation with SAM (Fig. 2a, trace 4). In support of this assignment, the SAM-treated MaNifB wt , MaN-ifB H28A , and MaNifB H219A can be used as M-cluster sources for the subsequent reconstitution and activation of apo-NifDK in an in vitro assay; whereas the SAM-treated MaNifB H43A cannot support the reconstitution and activation of apo-NifDK in the same assay (Fig. 2b). Interestingly, the activity of MaNifB H28A in this assay is~41% less than those of MaNifB wt and MaNifB H219A , suggesting a possible involvement of His 28 in the K-to L-cluster conversion due to the extremely close location of this residue to one of the Cys ligands (Cys 30 ) of the K1-cluster in the primary sequence (see Supplementary Fig. 1b) and, consequently, the tertiary structure of NifB. More importantly, the abolished activity of in MaNifB H43A in the K-to L-cluster transformation points to a critical role of His 43 in coupling the K1-and K2clusters into an 8Fe L-cluster. To further explore the role of His 43 in this process, high performance liquid chromatography (HPLC) was performed to examine the products generated upon incubation of MaNifB H43A with SAM. Like MaNifB wt , MaNifB H28A , and MaNifB H219A (Fig. 3a, b, traces 1-3), MaNifB H43A (Fig. 3a, b, trace 4) can cleave SAM into S-adenosyl-L-homocysteine (SAH) and 5′deoxyadenosine (5′-dAH). In addition, as observed in the cases of MaNifB wt , MaNifB H28A , and MaNifB H219A (Fig. 3c, traces 1-3), formation of methanethiol is detected upon acid quench of an incubation mixture of MaNifB H43A and SAM (Fig. 3c, trace 4). This observation is not particularly surprising because our previous study demonstrates that SAH, 5′-dAH and methanethiol can be generated as long as both SAM-and K2-clusters are present, even in absence of the K1-cluster. The fact that substitution of the K1-specific ligand, His 43 , with Ala does not impact the reactivities associated with the SAM-and K2-clusters is consistent with this observation and places the perturbation of the K-to L-cluster conversion by this substitution after the hydrogen atom abstraction from the K2-associated methyl group (see Supplementary Fig. 1a).
Dual actions of histidine ligand in L-cluster maturation. X-ray absorption spectroscopy (XAS)/extended x-ray absorption fine structure (EXAFS) analysis provided further insights into the role of His 43 in L-cluster maturation. XAS/EXAFS analysis has proven to be a valuable tool for obtaining structural information of the cluster species related to the function and assembly of nitrogenase, and previous studies of the wildtype and variant MaNifB proteins have established XAS/EXAFS parameters that can be used in combination with the EPR and biochemical data to conclusively assign cluster species and monitor cluster transformation in this protein. The x-ray absorption near edge structure (XANES) data reveal a K-edge energy for MaNifB H43A similar to that for MaN-ifB wt before or after incubation with SAM (Fig. 4a), suggesting a similar sulfur-rich environment in all these protein species 16 . The pre-edge feature of MaNifB H43A is also similar in intensity to that of MaNifB wt before incubation with SAM; however, its amplitude does not increase as much as that of MaNifB wt upon incubation with SAM ( Table 1), suggesting that the transition metal center in SAM-treated MaNifB H43A (designated MaNifB H43A /SAM) is less distorted away from centrosymmetry 23,24 than that in SAM-treated MaNifB wt (designated MaNifB wt /SAM). Consistent with this observation, the smoothed second derivative of the pre-edge data of MaNifB wt /SAM transitions from a single inverted peak at 7112.6 eV to two inverted peaks at~7112.6 eV and~7114.5 eV, respectively (Fig. 4b). Such a change has been attributed to the conversion of the K-cluster (with typical tetrahedral Fe-site geometries 25 ) to an L-cluster (with an unusual intermediate geometry between tetrahedral and trigonal pyramidal 25 ) in MaNifB wt upon incubation with SAM 16 . In the case of MaNifB H43A , while a similar peak at~7112.6 eV is observed in the second derivative before and after incubation with SAM, the line-shapes of these plots beyond 7113 eV are different than those of the corresponding MaNifB wt species (Fig. 4b). More importantly, the second peak at~7114.5 eV is absent from the plot of MaNifB H43A /SAM, although MaN-ifB H43A /SAM seems to undergo a transition analogous to that of MaNifB wt /SAM on the basis of the similar line-shapes of their second derivative plots (Fig. 4b).
Extended X-ray absorption fine structure (EXAFS) analysis of the Fe K-edges of MaNifB H43A provided important insights into the structural metrics of its associated cluster species. Prior to SAM treatment, MaNifB H43A and MaNifB wt display two similar features at R + Δ~1.7 and 2.4 Å, respectively, in the Fourier transforms (FT; Fig. 4c) of their EXAFS data (Fig. 4d), although the feature of MaNifB H43A at R + Δ~2.4 Å is much more prominent than that of MaNifB wt . For MaNifB H43A , these FT features can be best fit with Fe-S and Fe•••Fe scatterers at 2.29 and 2.71 Å, respectively; whereas for MaNifB wt , they are best fit with Fe-S scatterers at 2.29 Å and Fe•••Fe scatterers at 2.51 and 2.69 Å, respectively ( Table 2). Upon incubation with SAM, MaNifB H43A /SAM and MaNifB wt /SAM seemingly undergo similar changes, both showing extra FT features at R + Δ~3.0 and~3.5 Å; yet, while MaNifB wt /SAM displays a substantially increased intensity and a clear shift of its FT feature at R + Δ~2.4 Å, the corresponding FT feature of MaNifB H43A /SAM remains largely unchanged (Fig. 4c). The differences between the two SAM-treated MaNifB species are clearly illustrated in the best fits of their EXAFS data: MaNifB H43A /SAM is best modeled with two types of Fe-S scatterers at 2.27 and 3.88 Å, respectively, and one type of Fe•••Fe scatterers at 2.69 Å; whereas MaNifB wt /SAM is best modeled with one type of Fe-S scatterers at 2.23 Å and two types of Fe•••Fe scatterers at 2.64 and 3.70 Å, respectively (Table 2). Most notably, the long-range Fe•••Fe distance at 3.70 Å, which originates from the intercubane scattering between the six carbide-coordinated Fe atoms at the cofactor core 25 Fig. 4). Given the homogeneity of the cluster species on MaNifB H43A /SAM, it is likely that one of these proposed models will be identified through future structural characterization of this protein.

Formation of a cluster intermediate on the SAM-treated
MaN-ifB H43A that is distinct from both the K-and L-clusters implies that His 43 serves as a key structural element and/or reaction component during the process of cofactor core formation on MaNifB. The substantially increased intensity of the FT feature of MaNifB H43A at R + Δ~2.4 Å relative to that in MaNifB wt prior to incubation with SAM (see Fig. 4c) points to a much stronger Fe•••Fe scattering that results from a much closer distance between the two [Fe 4 S 4 ] units of the K-cluster in MaNifB H43A than those in MaNifB wt . Such a change in the distance and/or orientation of the K1-cluster relative to that of the K2-cluster apparently presents a challenge for MaNifB H43A to initiate a proper coupling between the two K-cluster units. The His 43 ligand, therefore, could play a steric role in keeping the two K-cluster units in the correct distance/orientation to each other by either indirectly pulling the K1-cluster away from the K2-cluster via its ligand capacity or directly separating the K1-and K2-clusters with its bulky imidazole ring ( Supplementary Fig. 4, ➀). Additionally,  given our observation of a loss of the nitrogen ligand upon conversion of the K-cluster to an L-cluster on MaNifB 11 , His 43 may lose its coordination to the K1-cluster via protonation, thereby freeing up K1 for the subsequent coupling with K2 into an L-cluster, and facilitating the release of the completed L-cluster to the next biosynthetic apparatus for further maturation. As such, His 43 likely functions as a molecular switch via reversible protonation and deprotonation events, securing the cluster in place in its deprotonated state while giving the cluster certain structural flexibility in its protonated state to accommodate the different states required for cluster conversion. A similarly labile nitrogen ligand is also found in mitoNEET, where protonation of a His ligand to a [Fe 2 S 2 ] cluster permits transfer of the cluster to downstream acceptor proteins 26,27 .
In light of this proposal, it is interesting to consider a coupling of the function of His 43 as a molecular switch with another role of this residue in cofactor core formation, one that is involved in the further deprotonation/dehydrogenation of the initial methylene radical to yield the interstitial carbide at the center of the L-cluster. The observation that the two K-cluster units in SAM-treated MaNifB H43A become more aligned with each other but remain largely separate FeS cubanes in character ( Supplementary Fig. 4, ➁) is consistent with an interruption of the carbide formation/insertion process that is required for the coupling/rearrangement of K1 and K2 into the geometry of a cofactor core. Consequently, the cluster is rendered in an unfinished state with the carbon intermediate (CH x ) not fully deprotonated or dehydrogenated and attached to one or both of the K-cluster units ( Supplementary Fig. 4, ➂). It is important to note that, other than His 43 , additional residues may also be involved in processing the carbon intermediate to an interstitial carbide, as substitution of His 28 with Ala apparently reduces the efficiency of L-cluster formation on NifB by 41% (see Fig. 2b). A proton relay mechanism involving multiple histidine (or equivalent) residues may be employed in this case to facilitate efficient deprotonation/dehydrogenation of the initial methylene radical to eventually yield a carbide in the center of the Lcluster. While the role of His 43 and other relevant players in the cofactor core formation process is yet to be elucidated, the results of this study provide an important framework for further investigations into the unique radical chemistry underlying the formation of the core structure of the nitrogenase cofactor. Knowledge obtained from these studies will contribute to a better understanding of the mechanism of nitrogenase and shed important light on the mechanisms of analogous biological systems. Cell growth and protein purification. E. coli strains expressing His-tagged MaNifB wt (strain YM114EE), MaNifB H28A (strain YM242EE), MaNifB H43A (strain YM244EE), MaNifB H219A (strain YM246EE), MaNifB K1 (strain YM165EE) and MaNifB K1-H43A (strain YM307EE) were grown in 10-L batches in Difco LB medium containing 100 mg L -1 ampicillin (BD Biosciences) in a BIOFLO 415 fermenter (New Brunswick Scientific) at 37°C, with 200 rpm agitation and 10 L˄min -1 airflow. When OD 600 reached 0.5, the temperature was lowered to 25°C before expression of the wildtype and variant MaNifB proteins was induced by addition of 25 µM IPTG. Expression of proteins was allowed to continue for 16 h before cells were harvested by centrifugation using a Thermo Fisher Scientific Legend XTR centrifuge. Subsequently, His-tagged MaNifB proteins were purified by immobilized metal affinity chromatography (IMAC) 9,10 .

Methods
SDS-PAGE analysis. The purified wildtype and variant MaNifB proteins were subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) analysis on a 4-20% precast Tris-glycine gel (Bio-Rad).
Iron determination. The iron contents of the FeS-reconstituted wildtype and variant MaNifB proteins were determined by inductively coupled plasma optical emission spectroscopy (ICP-OES) using a Thermo Scientific iCAP7000. Stock solutions of elemental iron (1 mg mL −1 , Inorganic Ventures) were diluted to make standard solutions for calibration. Each protein sample was mixed with 100 µL concentrated sulfuric acid (H 2 SO 4 ) and 100 µL concentrated nitric acid (HNO 3 ) and heated at 250°C for 30 min. This procedure was repeated until the solutions became colorless. The solution was then cooled to room temperature and diluted to a total volume of 10 mL with 2% HNO 3 prior to sample analysis.
FeS cluster reconstitution. The purified wildtype or variant MaNifB was treated with 20 mM bathophenanthroline disulfonate, an iron chelator, in a buffer containing 2 mM dithionite (DT; Na 2 S 2 O 4 ), 50 mM Tris-HCl (pH 8.0) and 500 mM NaCl, followed by incubation at room temperature for 1 h to remove the endogenous FeS clusters associated with the protein. Subsequently, this mixture was diluted with a buffer containing 50 mM Tris-HCl (pH 8.0) and loaded on a Q Sepharose column (GE Healthcare). The column was then washed with a buffer containing 2 mM DT, 50 mM Tris-HCl (pH 8.0) and 50 mM NaCl prior to elution of the MaNifB protein with a buffer containing 50 mM Tris-HCl (pH 8.0) and 500 mM NaCl. Dithionite was removed by running protein through a Sephadex G-25 (GE Healthcare) column equilibrated with 50 mM Tris-HCl (pH 8.0) and 10% glycerol. Reconstitution of the wildtype or variant MaNifB protein with synthetic [Fe 4 S 4 ] clusters 15 was carried out by adding a dimethylformamide (DMF) solution of synthetic [Fe 4 S 4 ] cluster dropwise at a molar ratio of 5:1 to the MaNifB protein in a buffer containing 20 mM β-mercaptoethanol and 50 mM Tris-HCl (pH 8.0), with continuous stirring on ice. After incubation on ice for 1 h, the reaction mixture was diluted with a buffer containing 2 mM DT and 50 mM Tris-HCl (pH 8.0) and loaded on a Q Sepharose column. The column was then washed with a buffer containing 2 mM DT, 50 mM Tris-HCl and 50 mM NaCl prior to elution of the reconstituted MaNifB with a buffer containing 2 mM DT, 50 mM Tris-HCl (pH 8.0) and 500 mM NaCl. Reconstituted wildtype and variant MaNifB proteins were subjected to metal determination, activity assays and EPR analysis.  at 25°C for 60 min with intermittent mixing, before they were terminated by filtration through Amicon Ultra 30,000 MWCO centrifugal filters. Samples were then supplemented by trifluoroacetic acid (TFA) to a concentration of 0.14% before being analyzed by a Thermo Scientific Dionex Ultimate 3000 UHPLC system equipped with an Acclaim 120 C18 column (4.6 × 100 mm, 5-µm particle size). The flow rate of buffer was 0.5 mL˄min -1 , and the column was kept at 30°C. The column was equilibrated with 98% buffer A (50 mM KH 2 PO 4 , pH 6.6) and 2% buffer B (100% methanol) before each injection of a 100-µL sample. After sample injection, a linear gradient of 2-60% buffer B was applied over 20 min, followed by 8 min of isocratic flow with 60% buffer B and a linear gradient of 60-2% buffer B over 4 min. Elution of products was monitored at a UV wavelength of 254 nm. After each run, the column was equilibrated for 5 min with 2% buffer B before the injection of the next sample.
Acid quench experiments. Detection of MaNifB-dependent production of methanethiol was performed 14  Electron paramagnetic resonance (EPR) analysis. Sample preparation was carried out in a Vacuum Atmospheres glove box with less than 1 ppm O 2 and flash frozen in liquid nitrogen before analysis. The dithionite-reduced samples were prepared by incubating 50 μM wildtype or variant MaNifB with 40 mM SAM for 15 min, followed by re-isolation of MaNifB into a buffer containing 50 mM Tris-HCl (pH 8.0), 500 mM NaCl, and 2 mM dithionite (DT). The indigo disulfonate (IDS)-oxidized samples were prepared by incubating the re-isolated MaNifB with excess IDS for 5 min, followed by removal of excess IDS was using a Sephadex G-25 desalting column. The DT-reduced or IDS-oxidized MaNifB sample was then concentrated to 15 mg mL −1 , followed by transfer of the sample into EPR tubes. The samples were then flash frozen and stored in liquid nitrogen. CW EPR spectra were recorded by an ESP 300 E z spectrophotometer (Bruker) interfaced with an ESR-9002 liquid-helium continuous-flow cryostat (Oxford Instruments) using a microwave power of 5 mW (IDS-oxidized samples) or 50 mW (IDT-reduced samples), a gain of 5 × 10 4 , a modulation frequency of 100 kHz, and a modulation amplitude of 5 G. Five scans was recorded for each sample in perpendicular mode at 20 K (IDS-oxidized samples) or 10 K (DT-reduced samples) using a microwave frequency of 9.62 GHz. All pulse EPR studies were carried out at the UC Davis CalEPR center, using a Bruker EleXsys E580 pulse EPR spectrometer equipped with an Oxford-CF935 liquid-helium cryostat and an ITC-503 temperature controller. Pulse data were collected using a Bruker MS5 probe (X-band) or an R.A. Isaacson-designed cylindrical TE011 resonator (Q-band) 28 adapted for pulse EPR in an Oxford Instruments CF935 cryostat. Three-pulse ESEEM spectra were collected using the pulse sequence π/2-τ-π/2-T-π/2-τ-echo where the delay time, T, was increased by 16 ns steps. ESEEM spectra were recorded at 10 K, τ = 128-144 ns (values chosen to minimize proton modulations to the spectra), π/2 = 12 ns, and a microwave frequency of 9.3366 GHz. Spectral processing (subtraction of a background exponential, application of a hamming window, and FFT) were performed using the EasySpin 5.2.27 toolbox in Matlab R2019a 29 .
X-ray absorption spectroscopy (XAS) analysis. The XAS samples were prepared the same way as the EPR samples (see above). The sample concentrations were 50 mg mL −1 . Fe K-edge X-ray absorption spectra were collected on SSRL beam line 7-3 using a 30-element solid state Ge detector (Canberra) with a SPEAR3 storage ring current of~500 mA at an energy of 3.0 GeV. The BL7-3 optics consists of a flat, bent, harmonic rejection vertically collimating Rh-coated Si M 0 mirror and a liquid nitrogen cooled double crystal Si(220) monochromator. A total of 7 and 6 scans, respectively, were collected for MaNifB H43A before and after incubation with SAM (designated MaNifB H43A and MaNifB H43A /SAM). All scans were taken between 6882 and 8000 eV at~10 K using an Oxford Instruments CF1208 continuous flow liquid-helium cryostat using a closed-cycle cooled He gas loop. An iron foil was placed in the beam pathway prior to the ionization chamber I 0 and scanned concomitantly for an energy calibration, with the first inflection point of the edge assigned to 7112.0 eV. A Soller slit with a 3 μm Mn filter was used to increase the signal-to-noise ratio of the spectra. Photoreduction was monitored by scanning the same spot on the sample twice and comparing the first derivative peaks associated with the edge energy during data collection.
The detector channels from the scans were examined, calibrated and averaged using EXAFSPAK 30 and then processed for EXAFS analysis using PYSPLINE 31 to extract χ(k). PYSPLINE was used to subtract a second-order background from the entire range of data and subsequently generate a spline function to model background absorption through the EXAFS region. A four-region spline was chosen with 2, 3, 3 order polynomials over the post edge region, and the data were normalized to have an edge jump of 1.0 at 7130 eV. Following a modified data analysis protocol 25 , the Fe K-edge EXAFS data for the clusters associated with MaNifB H43A and MaNifB H43A /SAM were generated by subtracting the k-weighted EXAFS data, χ(k), of MaNifB SAM (an MaNifB variant carrying only the SAMcluster 11 ) from the χ(k) of the samples in a 1:2 ratio on the basis of the proportionate iron quantity for each cluster species (i.e., 4 Fe for the SAM-cluster and 8 Fe for the K-clusters). Theoretical phase and amplitude parameters for a given absorber-scatterer pair were calculated using FEFF 8.40 32 and subsequently applied to the nonlinear least squares Opt fitting program of the EXAFSPAK package during curve fitting. Parameters for each species were calculated using an appropriate model derived from either the crystal structure of the M-cluster in NifDK (PDB code 3U7Q) 5 , where the Mo atom was exchanged for an Fe atom, or from the [Fe 4 S 4 ] cluster in NifH (PDB code 1G5P) 33 because there are no available crystal structures of MaNifB. In all analyses, the coordination number of a given shell (N) was a fixed parameter and was varied iteratively in integer steps, whereas the bond lengths (R) and mean-square deviation (σ 2 ) were allowed to freely float. The estimated uncertainties in R, σ 2 , and N are 0.02 Å, 0.1 × 10 -3 Å 2 , and 20%, respectively. The amplitude reduction factor S 0 was fixed at 1.0 for the Fe K-edge data, whereas the edge-shift parameter ΔE 0 was allowed to float as a single value for all shells. Thus, in any given fit, the number of floating parameters was typically equal to 2 × number of shells + 1. The goodness of fit (GOF) parameters were calculated as follows: ð1Þ The Fe K-edge data were analyzed with a k range of 2-11.2 Å −1 (ΔR = 0.17 Å) to allow comparison between previously reported data 16 , although the data could be analyzed with higher resolution with a k range of 2-14 Å −1 for MaNifB H43A and MaNifB H43A /SAM. Pre-edge analysis was performed on the Fe K-edge fluorescence data normalized to have an edge jump of 1.0 at 7130 eV in PYSPLINE. The preedge features were fit as described elsewhere 23 between 7108 and 7117 eV using the Fityk 34 program with pseudo-Voigt functions composed of 50:50 Gaussian/ Lorentzian functions.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.