Methylthio-alkane reductases use nitrogenase metalloclusters for carbon–sulfur bond cleavage

Lago-Maciel, Ana; Soares, Jéssica C.; Zarzycki, Jan; Buchanan, Charles J.; Reif-Trauttmansdorff, Tristan; Schmidt, Frederik V.; Lometto, Stefano; Paczia, Nicole; Schuller, Jan M.; Hansen, D. Flemming; Heller, Gabriella T.; Prinz, Simone; Hochberg, Georg K. A.; Pierik, Antonio J.; Rebelein, Johannes G.

doi:10.1038/s41929-025-01426-2

Download PDF

Article
Open access
Published: 23 October 2025

Methylthio-alkane reductases use nitrogenase metalloclusters for carbon–sulfur bond cleavage

Nature Catalysis volume 8, pages 1086–1099 (2025)Cite this article

9024 Accesses
3 Citations
207 Altmetric
Metrics details

Subjects

Abstract

Methylthio-alkane reductases convert methylated sulfur compounds to methanethiol and small hydrocarbons, a process with important environmental and biotechnological implications. These enzymes are classified as nitrogenase-like enzymes, despite lacking the ability to convert dinitrogen to ammonia, raising fundamental questions about the factors controlling their activity and specificity. Here we present the molecular structure of the methylthio-alkane reductase, which reveals large metalloclusters, including the P-cluster and the [Fe₈S₉C]-cluster, previously found only in nitrogenases. Our findings suggest that distinct metallocluster coordination, surroundings and substrate channels determine the activity of these related metalloenzymes. This study provides new insights into nitrogen fixation, sulfur-compound reduction and hydrocarbon production. We also shed light on the evolutionary history of P-cluster and [Fe₈S₉C]-cluster-containing reductases emerging before nitrogenases.

Main

Sulfur is an essential element in amino acids, vitamins and protein cofactors. Bacteria assimilate inorganic sulfur through the uptake and reduction of sulfate¹. However, under sulfate unavailability, bacteria scavenge organic sulfur via methionine-salvage pathways by recycling metabolic by-products to methionine^2,3. Rhodospirillum rubrum utilizes an anaerobic methionine-salvage pathway to convert (2-methylthio)ethanol (MT-EtOH) to methanethiol, direct precursor to methionine, via methylthio-alkane reductase^4,5. In vivo, this enzyme produces methanethiol from several volatile organic sulfur compounds (VOSCs) besides MT-EtOH, including dimethyl sulfide (DMS) and ethyl methyl sulfide (EMS), releasing in a 1:1 stoichiometry ethylene (C₂H₄), methane (CH₄) or ethane (C₂H₆), respectively^4,5. The release of these small hydrocarbons, which are potent greenhouse gases, to the atmosphere can exacerbate the climate crisis. However, these hydrocarbons also offer a biological, sustainable alternative to fossil fuel-based products, like natural gas and plastic precursors. Specifically in the field of renewable plastic production, the methylthio-alkane reductase stands out as a biogenic source of C₂H₄ (ref. ⁶). Unlike all other C₂H₄-forming enzymes, including the 1-aminocyclopropane-1-carboxylic acid oxidase (ACCO), the ethylene-forming enzyme (EFE) and the NADH:Fe(III)EDTA oxidoreductase, the methylthio-alkane reductase is oxygen independent, requires simple substrates and does not release carbon dioxide (CO₂) as a side product⁶. This makes this enzyme an exceptional option for the production of this important chemical building block, currently commercially derived exclusively from fossil fuels. Moreover, methylthio-alkane reductases explain the observed accumulation of C₂H₄ in waterlogged anoxic soils⁷ and, thus, are potentially involved in agricultural yield losses due to inhibitory effects of C₂H₄ on plant growth⁸.

Methylthio-alkane reductases have been phylogenetically classified as nitrogen fixation-like (Nfl) enzymes of the nitrogenase superfamily^4,9 (Fig. 1). Nitrogenases are the only enzymes known to reduce the stable triple bond of dinitrogen (N₂) to produce bioavailable ammonia (NH₃)¹⁰ and distinguish themselves from Nfl enzymes in their catalytic metal centres¹¹. Nfl enzymes such as the dark-operative protochlorophyllide a oxidoreductase (DPOR, encoded by bchLNB) or the Ni²⁺-sirohydrochlorin a,c-diamide reductase (encoded by cfbCD), involved in carbon–carbon double-bond reductions, contain only [Fe₄S₄]-clusters and tetrapyrrole binding sites¹². By contrast, nitrogenases are known to contain large and complex metalloclusters that allow N₂ fixation, including the P-cluster, an [Fe₈S₇]-cluster and the iron-molybdenum cofactor (FeMoco), a [MoFe₇S₉C-(R)-homocitrate]-cluster located in the active site of the molybdenum (Mo)-nitrogenase (encoded by nifHDK)^11,13,14 (Fig. 1). FeMoco is assembled by the maturase Nif(EN)₂, a nitrogenase family member that converts an Fe-only precursor, the [Fe₈S₉C]-cluster (L-cluster), into FeMoco by replacing a terminal Fe atom with Mo and attaching (R)-homocitrate^15,16 (Fig. 1). Distinct activities of nitrogenases and Nfl enzymes are attributed to these differences in metallocluster content and, thus, there is great interest to identify the metal centres dictating the carbon–sulfur (C–S) bond reduction activity in methylthio-alkane reductases.

**Fig. 1: Structural diversity of the nitrogenase and Nfl protein family.**

Here, we determined the molecular structure of the methylthio-alkane reductase complex by anaerobic single-particle cryogenic electron microscopy (cryo-EM) yielding a global resolution of 2.75 Å. We found that the catalytic component contains P-clusters and [Fe₈S₉C]-clusters, making it the only Nfl enzyme with large nitrogenase cofactors, but unable to reduce N₂. Our findings suggest that the protein scaffold, rather than the metalloclusters alone, dictates the differing catalytic activities of nitrogenases and methylthio-alkane reductases, advancing our understanding of the origin and evolution of N₂ fixation.

Results

Structural characterization of the methylthio-alkane reductase

To reveal the molecular machinery responsible for C–S bond reduction of methylthio-alkane reductases (encoded by marHDK), we heterologously produced subunits MarH, MarD and MarK from R. rubrum and its associated maturation enzyme MarB in an engineered Rhodobacter capsulatus strain^17,18 encoding a disrupted Mo- (ΔnifD::SpR) and iron (Fe)-nitrogenases (ΔanfHDGK::gmR; for more details, see Methods and Supplementary Table 3). The reductase and catalytic components of the methylthio-alkane reductase were purified separately using affinity chromatography (Fig. 2a). We determined the oligomeric state of each component by mass photometry: the reductase component forms a MarH₂ homodimer with a measured mass of 69 kDa (Fig. 2d; theoretical mass 64 kDa), while the catalytic component forms a Mar(DK)₂ heterotetramer with a measured mass of 227 kDa (Fig. 2d; theoretical mass 217 kDa). In conclusion, methylthio-alkane reductase components have the same stoichiometry as the Mo-nitrogenase components.

**Fig. 2: Purification and characterization of the methylthio-alkane reductase.**

The purified enzymes were active for the reduction of MT-EtOH, EMS and DMS as measured by the release of C₂H₄, C₂H₆ and CH₄, respectively (Fig. 2e and Extended Data Fig. 1). While the Mo-nitrogenase is unable to reduce VOSCs, the methylthio-alkane reductase converts MT-EtOH at a rate of 2.3 nmol C₂H₄ × mg⁻¹ Mar(DK)₂× min⁻¹, EMS at 7.9 nmol C₂H₆× mg⁻¹ Mar(DK)₂× min⁻¹ and DMS at 6.5 nmol CH₄× mg⁻¹ Mar(DK)₂× min⁻¹ (Fig. 2e). The methylthio-alkane reductase activity is ATP dependent and requires the addition of both reductase and catalytic components as well as sodium dithionite (DT) as an electron donor (Extended Data Fig. 1e). Intriguingly, we detected the formation of molecular hydrogen (H₂) for all tested substrates and in the absence of substrate under an argon atmosphere (Fig. 2e and Extended Data Fig. 1a–d).

While Mo-nitrogenase cannot reduce VOSCs, methylthio-alkane reductase can reduce the alternative nitrogenase substrate acetylene (C₂H₂) to C₂H₄ (Fig. 2e and Extended Data Fig. 1d). Moreover, methylthio-alkane reductase can fully reduce C₂H₂ to C₂H₆, a characteristic of the alternative nitrogenases, vanadium (V)- and Fe-nitrogenases^19,20, which contain V or Fe in the position of Mo in FeMoco, termed FeVco and FeFeco²¹. This suggests the presence of FeFeco/FeVco-like cofactors in the methylthio-alkane reductase.

To reveal the metalloclusters of Mar(DK)₂ involved in catalysis, we determined the molecular structure of the methylthio-alkane reductase complex by anaerobic single-particle cryo-EM. We were unable to obtain the structure of the catalytic component alone, due to severe preferential orientation bias and susceptibility to denaturation of MarD through contact with the gas–water interface. We overcame this issue by forming a complex of Mar(DK)₂ with MarH₂, using the non-hydrolysable MgATP mimic MgADP-AlF₃, thus protecting the catalytic component from denaturation. Reference-free two-dimensional (2D) class averages revealed the existence of two major oligomeric states: an octameric Mar(DK)₂(H₂)₂-complex, and a hexameric Mar(DK)₂H₂-complex (Fig. 2h). Attempts to resolve either complex in three dimensions were still hampered by severe preferential orientation bias, leading to uninterpretable reconstructions (Extended Data Fig. 2c–f). However, we realized that the preferred orientations of the two oligomeric states were perpendicular to each other. This insight prompted a revised classification strategy, where instead of segregating particles based on oligomeric state, we combined all particles thus resolving structural regions present across both complexes. (Extended Data Fig. 2g,h). This approach led to a substantial improvement in the reconstructed density, but the MarD subunits remained poorly resolved due to particle heterogeneity and the intrinsic flexibility of the N-terminal region. Given our goal to resolve the active site metalloclusters of the methylthio-alkane reductase, we prioritized solving a three-dimensional (3D) class that represented a stable state of one MarD subunit, at the expense of resolving the second copy. By applying strict 3D classification on 25 classes using a local mask on the proximal MarD, we selected the class with the best-resolved N terminus of MarD—specifically, the class in which the active site cluster is most deeply buried within the protein. This enabled a focused refinement of the MarDK₂H₂ fragment of the entire Mar complex, yielding a final reconstruction with a global resolution of 2.75 Å. (Fig. 2g and Extended Data Fig. 3). The resulting map enabled us to build a structural model with well-resolved metalloclusters (Fig. 3a), revealing the molecular basis of the methylthio-alkane reductase specificity.

Methylthio-alkane reductase contains nitrogenase cofactors

Analysing the structure, we observed a canonical reductase component MarH₂, as detailed in Supplementary Discussion 1, containing one MgADP-AlF₃ moiety per monomer (Fig. 3b and Extended Data Fig. 4) and a single [Fe₄S₄]-cluster in its dimeric interface coordinated by Cys97^MarH and Cys133^MarH (Fig. 3c). Presumably, electrons are transferred from the [Fe₄S₄]-cluster to the P-cluster located at a distance of ~15.0 Å, bridging the proximal MarDK dimer (Fig. 3b). The P-cluster is symmetrically coordinated by Cys32^MarK, Cys56^MarK, Cys121^MarK, Cys66^MarD, Cys91^MarD and Cys158^MarD, featuring a centrally shared sulfide (Fig. 3d) and is most likely present in the reduced state (P^N), as detailed in Supplementary Discussion 2. Most excitingly, we observed strong electron density at the active site of the methylthio-alkane reductase, located ~14.4 Å away from the P-cluster (Figs. 3e and 4b). The density neatly accommodates the [Fe₈S₉C]-cluster from Nif(EN)₂, the FeMoco precursor, with no density corresponding to any organic moiety. In addition, we were unable to detect (R)-homocitrate by high-performance liquid chromatography–mass spectrometry (HPLC–MS) of Mar(DK)₂ (Fig. 2c). To determine the metal composition of these metalloclusters, we performed inductively coupled plasma optical emission spectroscopy (ICP-OES; Fig. 2b). ICP-OES analysis revealed 32.6 ± 5.3 Fe atoms per catalytic component, consistent with the presence of two P-clusters ([Fe₈S₇]-clusters) and two [Fe₈S₉C]-clusters per Mar(DK)₂ heterotetramer, and neither Mo nor V was detected. The reductase component MarH₂ contained 3.5 ± 0.9 Fe atoms, indicating the presence of a labile [Fe₄S₄]-cluster. Thus, methylthio-alkane reductase is the only Nfl enzyme described to harbour the large and complex nitrogenase P- and [Fe₈S₉C]-clusters.

Distinct metallocluster coordination and surrounding

The core of the [Fe₈S₉C]-cluster consists of a trigonal prism formed by Fe2–Fe7, which are bridged by three belt sulfides. Due to the presence of all conserved NifB motifs in MarB and the presence the radical SAM domain responsible for carbide insertion⁴, we assume a similar role of MarB and NifB in the assembly of the nitrogenase precursor cluster²². Thus, an interstitial carbide was placed in the centre of the cluster, although the resolution of the electron density map is insufficient to locate the carbide. The metallocluster is anchored to the protein scaffold by two amino acid residues, Cys270^MarD to Fe1 and His429^MarD to Fe8 (Fig. 4a), analogous to the coordination of Fe1 and Mo of FeMoco by Cys275^AvNifD and His442^AvNifD, respectively²³ (Fig. 4c). However, His429^MarD is located at an extended loop specific to MarD, which is not conserved in bona fide nitrogenases (Loop I in Fig. 4a,f and Extended Data Fig. 5). This explains why previous sequence analysis of the methylthio-alkane reductase gene cluster had failed to identify this ligand⁴. The octahedral coordination of Mo in the Mo-nitrogenase, which involves three sulfides, (R)-homocitrate and His442^AvNifD, is replaced in the [Fe₈S₉C]-cluster of the methylthio-alkane reductase by a tetrahedral coordination due to the absence of (R)-homocitrate. As a result, the anchoring to the protein adopts an axial geometry, and the bond distance between His429^MarD and Fe8 of the [Fe₈S₉C]-cluster is shorter (1.9 Å) compared with the bond distance between His442^AvNifD and the Mo atom of FeMoco (3.1 Å in PDB: 7UTA (ref. ²³)). This is also the case when compared with the bond distance between His423^RcAnfD and Fe8 of FeFeco from the Fe-nitrogenase (2.7 Å in PDB: 8OIE (ref. ¹⁷)). The observed coordination of the methylthio-alkane reductase [Fe₈S₉C]-cluster by just two residues (Cys270^MarD and His429^MarD) indicates that this cluster indeed necessitates a carbide in its centre, because the geometry of an [Fe₈S₉]-cluster with six trigonal Fe atoms would not be possible without the structural support of a shared central ligand.

**Fig. 4: Comparison of active sites from methylthio-alkane reductase and Mo-nitrogenase.**

Beyond the distinct cluster composition, the direct protein environment of methylthio-alkane reductases also deviates from nitrogenases. Substrate binding to nitrogenase metalloclusters has been proposed to occur between Fe2 and Fe6, where the bridging sulfide S2B is replaced by the substrate^{24,25,26,27,28}. The conserved residues Val70^AvNifD, Gln191^AvNifD and His195^AvNifD surround S2B and are essential for N₂ reduction²⁹ (Fig. 4c). The cavity left by the small side chain of Val70^AvNifD allows substrate access to Fe6 of FeMoco³⁰. His195^AvNifD forms a hydrogen bond with S2B in the resting state and possibly donates protons during catalysis, while Gln191^AvNifD undergoes structural rearrangements that allow S2B displacement by the substrate during turnover^26,31. The residues lining the active site of the methylthio-alkane reductase differ substantially (Extended Data Fig. 6a): Cys73^MarD and Gln74^MarD replace Val70^AvNifD, while tryptophan (Trp195^MarD) and phenylalanine (Phe199^MarD) residues substitute Gln191^AvNifD and His195^AvNifD, respectively (Fig. 4a). However, because the S2B site of the [Fe₈S₉C]-cluster of the methylthio-alkane reductase is rotated by ~48° compared with the S2B site of FeMoco taking as the vertex its interstitial carbide, only Trp195^MarD is close enough to S2B to form a hydrogen bond (Fig. 4e,f). Trp195^MarD, His194^MarD and His375^MarD surround the S2B site of the [Fe₈S₉C]-cluster at distances of 3.9–4.7 Å and thus could be involved in substrate recognition and binding as well as proton transfer during turnover (Fig. 4a,f). The [Fe₈S₉C]-cluster in Mar(DK)₂ is also stabilized by Arg428^MarD, which forms a hydrogen bond with the belt sulfur S5A (Fig. 4f). Interestingly, Arg428^MarD is located at the extended Loop I of MarD, adjacent to the [Fe₈S₉C]-cluster coordinating His429^MarD, and replacing Gly424^AvNifD, a residue strictly conserved in all nitrogenases³² (Extended Data Fig. 5). Consequently, Arg428^MarD together with His429^MarD might be a hallmark of methylthio-alkane reductases and distinguish them from nitrogenases. In summary, the methylthio-alkane reductase active site suggests a different substrate recognition and proton transfer to nitrogenases.

Substrate channels determine specificity

Differences between methylthio-alkane reductase and Mo-nitrogenase extend to substrate channels: Mo-nitrogenase substrate channels proposed by Igarashi and Seefeldt (IS)³³ and Morrison et al. (AI)³⁴ are restricted from accessing the proximal pocket of FeMoco due to hydrogen bond gating by Tyr229^AvNifD and Ser278^AvNifD³⁵ (Fig. 4d). By contrast, the two most likely methylthio-alkane reductase substrate channels predicted by CAVER³⁶ do not show steric impediments. Restricted access to the active site of Mo-nitrogenase could explain its inability to utilize VOSCs, despite being more efficient at reducing smaller substrates (Fig. 2e). The most likely substrate channel in the methylthio-alkane reductase (channel I) originates at the interface of MarD and MarK, passes the belt sulfide S5A on the way to S2B and occupies the void left by the absent (R)-homocitrate (Fig. 4b,f). The second channel (channel II) originates at the surface of MarD opposite to channel I and ends at S2B of the [Fe₈S₉C]-cluster (Fig. 4b,f). Nevertheless, given the symmetric geometry of the [Fe₈S₉C]-cluster, substrate binding to any of the three belt sulfurs, S2B, S5A and S3A, cannot be discarded at this point. Further investigations are needed to decipher the binding of methylthio-alkanes to the [Fe₈S₉C]-cluster during catalysis.

The [Fe₈S₉C]-cluster does not reduce N₂

Based on our finding that the methylthio-alkane reductase contains an [Fe₈S₉C]-cluster in conjunction with the recent observations that [Fe₈S₉C]-cluster-containing nitrogenase maturase Nif(EN)₂ can reduce N₂ to NH₃ to a slight extent³⁷, we aimed to determine the N₂ reduction activity of methylthio-alkane reductases. No formation of ¹⁵NH₄⁺ beyond the background could be detected by nuclear magnetic resonance (NMR) in ¹⁵N₂-assays mimicking physiological conditions containing catalytic component Mar(DK)₂, reductase component MarH₂, ATP and DT, in contrast to Mo-nitrogenase assays under the same conditions (Fig. 2f and Extended Data Fig. 7). Assays with just the catalytic component Mar(DK)₂ and the strong reductant europium-II-diethylenetriamine pentaacetic acid (Eu(II)-DTPA) also did not display ¹⁵NH₄⁺ formation above background levels (Fig. 2f and Extended Data Fig. 7). This result contrasts with the observation that Nif(EN)₂ can reduce one to three molecules of N₂ in a reductase component- and Eu(II)-DTPA-dependent manner³⁷. Thus, our results reveal that the presence of an [Fe₈S₉C]-cluster is not the sole determinant for N₂ reduction; rather, the coordination and the immediate surroundings of the [Fe₈S₉C]-cluster may determine its activity.

Spectroscopy and redox states of metalloclusters

Nitrogenases and Nfl enzymes exhibit very distinct EPR spectroscopic features¹³, which define the redox state and electronic structure of the metalloclusters. At 4–20 K, DT-reduced MarH₂ exhibited the typical physical spin mixture of a S = 1/2 (g = 2.03, 1.933 and 1.84) and a S = 3/2 species of the [Fe₄S₄]¹⁺ cluster, characteristic for NifH₂ (ref. ³⁸), of which the relative content of the two spin states is ATP dependent, as detailed in Supplementary Discussion 3 (Fig. 3f and Extended Data Fig. 8b,c). Next, we analysed the catalytic component Mar(DK)₂: upon DT reduction, and in a dye-mediated redox titration with a midpoint potential (E_m) of −390 mV versus normal hydrogen electrode, a relatively sharp S = 1/2 signal (g_average = 1.96) appeared (Fig. 3g and Extended Data Fig. 8a). Although the Mar(DK)₂ properties did not match those of the Mo-nitrogenase Nif(DK)₂ (Supplementary Discussion 4, Extended Data Fig. 9 and Supplementary Table 3), it was very similar to the signal assigned to a reduced P-cluster in the catalytic component of V-nitrogenase³⁹. Because the fully intact DT-reduced P^N cluster is diamagnetic and the fully intact P²⁺ oxidized state has a g = 12 integer spin signal in Nif(DK)₂ (ref. ⁴⁰) and Vnf(HDK)₂ (ref. ⁴¹), this signal probably derives from a subclass of incompletely matured P-clusters containing an [Fe₄S₄]¹⁺ half^41,42. Upon oxidation with 5,5′-indigodisulfonate (IDS) or by dye-mediated redox titration with E_m = −250 mV versus normal hydrogen electrode, a peculiar broad S = 1/2 signal (simulated with g_xyz = 1.967, 1.926 and 1.83) with g_average = 1.91 appeared, totally unlike P¹⁺ and P³⁺ EPR signals^13,40. This EPR signal has previously been observed for the [Fe₈S₉C]-cluster-containing nitrogenase maturase Nif(EN)₂ (ref. ³⁷) (Fig. 3g). Our findings show that the g_average = 1.91 signal is the fingerprint for the active site metallocluster of Mar(DK)₂ and indicate that a carbide is indeed present in the [Fe₈S₉C]-cluster of the methylthio-alkane reductase. By mixing reduced (but DT-free) MarH₂ with oxidized (but IDS-free) Mar(DK)₂ and MgATP, we could demonstrate that the structurally defined electron transfer pathway (Fig. 3a) guides an electron to the species with g_average = 1.91 (Fig. 3h). Concomitantly, the signal assigned to the reduced P-cluster partially appeared, in agreement with its low redox potential.

Evolution of nitrogenase metalloclusters

Our structural and biophysical data conclusively show that methylthio-alkane reductases contain metalloclusters until now exclusively associated with nitrogenases. This strongly suggests that their last common ancestor probably contained a P-cluster and a large active site cluster, as previously hypothesized⁴³. The fact that we find ligating cysteines in methylthio-alkane reductases that are in homologous positions to equivalent cysteines in nitrogenases strongly suggests that these ligation sites already existed in their last common ancestor. To test this prediction, we first inferred a phylogeny of the subunits assembling the catalytic component of the wider nitrogenase family, composed of nitrogenase-like proteins (NflDK) including the methylthio-alkane reductase (MarDK), Mo-nitrogenase (NifDK), FeMoco maturase (NifEN), dark-operative protochlorophyllide a oxidoreductase (BchNB), chlorophyllide a oxidoreductase (BchYZ) and Ni²⁺-sirohydrochlorin a,c-diamide reductase (CfbD). Nitrogenases and their relatives are heterotetrameric proteins comprising two paralogues—usually called D and K. This arrangement is a result of an ancient gene duplication event of an ancestrally likely homomeric protein, represented today by CfbD₂ (ref. ⁴³). Subsequent operon duplications then produced the different paralogous heterotetramers known in this family. CfbD is an outgroup to all other sequences, which roots the tree between the K and D paralogues of the catalytic components of methylthio-alkane reductase and nitrogenase. BchNB and CfbD (just under 30% of all sequences from our alignment) neither have all cysteines conserved nor contain a P- or [Fe₈S₉C]-cluster. Our initial tree search yielded a maximum-likelihood tree in which all Bch proteins are a sister group to all nitrogenase(-like) proteins including methylthio-alkane reductases (Extended Data Fig. 10e). This implies that the duplication from homomers to paralogous heteromers occurred twice, once in nitrogenase(-like) proteins and once in Bch proteins. The support for this unparsimonious split was low, so we inferred another tree in which we constrained the topology such that Bch and nitrogenase(-like) proteins including methylthio-alkane reductases underwent only one initial duplication from a homomer, followed by subsequent operon duplications (Extended Data Fig. 10a). This constrained tree was not rejected by the approximately unbiased test, and we view it as more plausible. We then used ancestral sequence reconstruction to infer sequences of the two subunits (D and K) that together assembled into the last common ancestor of nitrogenases and methylthio-alkane reductases for the unconstrained and constrained trees. Our reconstructions of these two very deep nodes were not confident enough across their entire sequence to resurrect this ancient protein experimentally (average posterior probability across sites 0.70). Posterior probabilities were much higher, however, around the crucial metallocluster binding sites. We confidently infer the six cysteines that would ligate a P-cluster with a posterior probability of 1 for both trees, which are in positions homologous to their positions in methylthio-alkane reductases and nitrogenases. For the residues ligating the [Fe₈S₉C]-cluster, we confidently infer a cysteine homologous to Cys270^MarD (posterior probability of 0.99), but not a histidine equivalent to His429^MarD or His442^AvNifD, because this ligand is located in different loops in nitrogenases and methylthio-alkane reductases (Extended Data Fig. 10). Our results indicate that this common ancestor had a P-cluster and probably an [Fe₈S₉C]-cluster ligated differently than extant nitrogenases and methylthio-alkane reductases. This ancestral reductase was probably catalytically active and contained a complex metallocluster buried in its active site, rather than being a maturase with a surface-exposed cluster like Nif(EN)₂ (ref. ¹⁵), although its activity is unresolved, so far. The discovery of nitrogenase clusters in the methylthio-alkane reductase provides a new perspective on how the protein scaffold controls the maturation, insertion and reactivity of these great clusters of biology⁴⁴ and strongly suggests that their history predates the origin of biological N₂ fixation by bona fide nitrogenases as we know them today. This is consistent with the recent discovery that [Fe₈S₉C]-clusters are part of the activation complex of methyl-coenzyme M reductase, which probably preceded the evolution of [Fe₈S₉C]-cluster-bearing nitrogenase(-like) enzymes⁴⁵.

Discussion

We aimed to identify the factors determining the unique activities of methylthio-alkane reductases. We revealed that these enzymes contain large nitrogenase metalloclusters, including a P-cluster bridging the MarDK subunits and an [Fe₈S₉C]-cluster in the active site of MarD, instead of an [Fe₄S₄]-cluster and a substrate-binding site. Consequently, methylthio-alkane reductase is the only Nfl enzyme with such metalloclusters that cannot perform N₂ fixation. The absence in methylthio-alkane reductase of conserved nitrogenase amino acid residues Val70^AvNifD, Gln191^AvNifD and His195^AvNifD involved in the N₂-reduction mechanism²⁹ could explain this observation. Moreover, complete nitrogenase activity might require (R)-homocitrate, suggested to be crucial for proton transfer during Mo-nitrogenase catalysis^46,47,48. While Mo-nitrogenase is probably more reactive and efficient in electron transport, it cannot break the weaker C–S bond of VOSCs reduced by methylthio-alkane reductases. Our structural data suggest that the specificity of methylthio-alkane reductases for methylated sulfur compounds is probably due to wider substrate channels in comparison with nitrogenases, allowing unrestricted access of these larger compounds to the active site. The insights gained in this study highlight the catalytic potential of the [Fe₈S₉C]-cluster beyond ammonia formation and open the door to engineering the methylthio-alkane reductase for the biotechnological production of small hydrocarbons such as methane and ethylene.

Methods

Chemicals

Chemicals were acquired from Thermo Fisher Scientific, Carl Roth, Sigma-Aldrich and Honeywell. Gas bottles were purchased from Air Liquide.

Bacterial strains and growth conditions

Modified Rhodobacter capsulatus B10S⁴⁹ strain MM0246¹⁷ (ΔnifD::SpR ΔanfHDGK::gmR ΔdraTG ΔmodABC ΔgtaI) was used for the heterologous expression and purification of the methylthio-alkane reductase from R. rubrum. R. capsulatus MM0246 does not encode for a methylthio-alkane reductase⁴ and was designed for the optimal plasmid-based production of nitrogenase-like enzymes under the Fe-nitrogenase anfH promoter. The advantageous genetic modifications of this production strain include the disruption of native Mo-nitrogenase nifD and Fe-nitrogenase anfHDGK genes with a spectinomycin or a gentamycin resistance cassete, respectively¹⁷, the deletion of the nitrogenase posttranslational modification machinery draTG responsible for the reversible inhibition of the nitrogenase reductase component by ADP ribosylation⁵⁰, the deletion of the modABC transporter responsible for the cellular uptake of Mo, a strong repressor of the anfH promoter⁵¹, and the deletion of the quorum sensing gene gtaI responsible for the formation of the capsule, which improves cell pelleting of R. capsulatus during collection⁵². R. capsulatus was cultured anaerobically under photoheterotrophic conditions (60 W krypton lamps) in modified R. capsulatus Minimal Medium V (RCV) liquid medium¹⁷ with the addition of 10 mM serine and 20 µg ml⁻¹ streptomycin sulfate.

Escherichia coli DH5α (Thermo Fisher Scientific) and E. coli ST18 (ref. ⁵³) were used for expression plasmid cloning and conjugation into R. capsulatus by diparental mating⁵⁴. A list of all strains used in this study is found in Supplementary Table 4.

Molecular cloning

The marBHDK gene cluster (Rru_A0793–Rru_A0796) was amplified from R. rubrum DNA by PCR with Q5 high-fidelity DNA polymerase (New England Biolabs) and cloned under the R. capsulatus anfH promoter via Golden Gate assembly⁵⁵ using NEBridge Golden Gate Assembly Kit (New England Biolabs) into a broad-host range plasmid pOGG024 (ref. ⁵⁶) (pMM0165). To purify the methylthio-alkane reductase subunits separately, affinity tags were added to the sequences via Gibson Assembly⁵⁷ using NEBuilder HiFi DNA Assembly Master Mix (New England Biolabs; pMM0181). A hexahistidine tag was added to the N terminus of MarH and a Strep-tag II to the C terminus of MarD, to purify reductase and catalytic components⁵⁸. The sequence of pMM0165 and pMM0181 were confirmed by whole-plasmid sequencing (Plasmidsaurus). Primers and plasmids used in this study are listed in Supplementary Tables 5 and 6.

Protein purification and characterization

Purification of Strep-tagged Mar(DK)₂ and His-tagged MarH₂ from R. capsulatus MM0422 was started by inoculating at an optical density at 660 nm (OD₆₆₀) of 0.05 modified RCV medium¹⁷ with 10 mM serine and 50 µg ml⁻¹ kanamycin sulfate. Liquid cultures were cultivated anaerobically under illumination in 1.2 atm of 100 % argon atmosphere at 30 °C until reaching a final OD₆₆₀ of ~3.2. Cultures were collected inside an anaerobic COY tent (COY Laboratory Products), with a >95% Ar <5% H₂ atmosphere and were spiked with 2 mM sodium DT. Cell pelleting was performed in a Beckman Coulter centrifuge (Beckman Coulter) at 10,000g, for 30 min at 10 °C. The supernatant was discarded, and the cell pellet was resuspended in binding buffer (500 mM NaCl, 50 mM Tris pH 8.5, 10 % glycerol, 2 mM sodium DT, 0.01% Tween 20, 50 mM arginine–glutamate and 20 mM imidazole) supplemented with 0.2 mg ml⁻¹ bovine pancreatic deoxyribonuclease I and one cOmplete EDTA-free protease inhibitor tablet (Sigma-Aldrich) per 30 ml of pellet. Cells were lysed in a French press FA-078AE (Thermo Fisher Scientific) at 20,000 psi. The soluble fraction was separated by ultracentrifugation at 119.028g for 1 h at 4 °C, and filtered before loading onto a Strep-Tactin XT 4Flow high-capacity column (IBA Lifesciences) and a HisTrap HP column (Cytiva) connected to an ÄKTA pure chromatography system (Cytiva) inside an anaerobic COY tent. Binding buffer containing 50 mM biotin was used for the elution of Strep-tagged Mar(DK)₂, while binding buffer containing 250 mM imidazole was used for the elution of His-tagged MarH₂. Eluted proteins were buffer exchanged in Sephadex G-25 packed PD-10 desalting columns (Cytiva) to desalting buffer (150 mM NaCl, 50 mM Tris pH 7.8, 10% glycerol and 2 mM sodium DT), and concentrated using Amicons Ultra 15 ml filters (Merck) of 30 kDa molecular weight cut-off. Protein yields were determined using Quick Start Bradford 1× Dye Reagent (Bio-Rad Laboratories) following the manufacturer’s instructions. Purified protein was stored in liquid nitrogen until further use. The purity of the proteins was determined by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE), the metal content by ICP-OES and the molecular mass by mass photometry as previously described¹⁷. In short, mass photometry measurements were carried out on microscope coverslips (1.5 H, 24 × 50 mm; Carl Roth) with CultureWell Reusable Gaskets (CW-50R-1.0, 50–3 mm diameter × 1 mm depth). Measurements were set up in gaskets assembled on microscope coverslips on the stage of a TwoMP mass photometer (Refeyn) with immersion oil. Samples were measured in anaerobic measurement buffer (150 mM NaCl, 50 mM Tris (pH 7.8), 10% glycerol and 10 mM sodium DT). Then, 0.5 µl of sample (500 nM stock concentration, 4 mM sodium DT) was removed from an anaerobic vial, quickly added to 19.5 µl of measurement buffer and mixed on the stage of the mass photometer. Measurements were started ~5 s after removing protein from the anaerobic environment. Data were acquired for 60 s at 100 frames per second using AcquireMP v2.3 (Refeyn). Mass photometry data sets were processed and analysed using DiscoverMP v20222 R1 (Refeyn).

Homocitrate determination

Semi-quantitative determination of homocitrate was performed using high-resolution electrospray liquid chromatography–mass spectrometry. The chromatographic separation was performed on an Agilent Infinity II 1290 HPLC system (Agilent Technologies), using a Kinetex EVO C18 column (150 × 2.1 mm, 3 μm particle size, 100 Å pore size; Phenomenex) connected to a guard column of similar specificity (20 × 2.1 mm, 3 μm particle size; Phenomenex) at a constant flow rate of 0.2 ml min⁻¹ with mobile phase A being 0.1% formic acid in water and phase B being 0.1% formic acid in methanol at 40 °C. The injection volume was 5 µl. The profile of the mobile phase consisted of the following steps and linear gradients: 0–0.5 min constant at 0% B; 0.5–6 min from 0% to 90% B; 6–7 min constant at 90% B; 7–7.1 min from 90% to 0% B; 7.1–12 min constant at 0% B.

An Agilent 6546 Q-TOF mass spectrometer (Agilent Technologies) was used in negative ionization mode with a dual electrospray ionization source: electrospray ionization spray voltage 2,000 V, nozzle voltage 500 V, sheath gas 300 °C at 11 l min⁻¹, nebulizer pressure 45 psig and drying gas 170 °C, at 5 l min⁻¹. The instrument was used in low molecule size mode with a screening window of 50–1,700 m/z. For online mass calibration, 112.9856 and 1033.9881 were used.

Compounds were identified on the basis of their accurate mass within a mass window of 5 ppm and their retention time compared with standards. Chromatograms were integrated using MassHunter software (Agilent). Relative abundance was determined based on the peak area by applying a mass extraction window of 10 ppm.

Methylthio-alkane reductase in vitro activity assays

Enzymatic activity of methylthio-alkane reductase and Mo-nitrogenase was determined by in vitro activity assays with the substrates MT-EtOH, EMS, DMS, protons (under Ar) and acetylene (C₂H₂, 10% in Ar) measuring production of ethylene (C₂H₄), ethane (C₂H₆), methane (CH₄), molecular hydrogen (H₂) and C₂H₄ as well as C₂H₆, respectively. The reductase component (0.15 mg) was added to the catalytic component (0.1 mg) in 700 µl of reaction mix containing 50 mM Tris pH 7.8, 3.5 mM ATP, 7.87 mM MgCl₂, 44.6 mM creatine phosphate, 0.2 mg ml⁻¹ creatine phosphokinase, 5 mM DT and 15 mM of the corresponding VOSC in 1.2 atm Ar, if indicated. Due to the tendency of MarH₂ to precipitate at high concentrations, all assays were conducted at a catalytic:reductase component molar ratio of 1:5. The reactions took place in 10-ml sealed gas chromatography (GC) vials, at 30 °C and shaking at 200 rpm for 10 min after the addition of the catalytic component. Samples were quenched with 300 µl of 400 mM Na₂EDTA pH 8.0 up to a total volume of 1 ml. Content of the headspace was determined using a Clarus 690 GC system (PerkinElmer), as previously described¹⁷.

NMR ¹⁵NH₄ quantification

In vitro nitrogen fixation activity was determined for methylthio-alkane reductase and Mo-nitrogenase by running physiological assays as described above or Eu(II)-DTPA assays, which were performed in a 1.5-ml volume containing 5 mg of the catalytic component Mar(DK)₂, 20 mM Eu(II)-DTPA, 25 mM Tris pH 8.0 and 2 mM DT. Both assays conditions were performed under 50% ¹⁵N₂ (Merck) and 50% Ar at 2 atm in 10 ml GC vials, at 30 °C, shaking at 250 rpm for 5 h (physiological assays) or 15 h (Eu(II)-DTPA assays) quantifying ¹⁵NH₄ with NMR. Negative controls were prepared under Ar or ¹⁴N₂ atmospheres. Proteins were removed from the samples by ultrafiltration at 3,000g and 30 min using Amicon Ultra filters (3 kDa cut-off, Merck). Subsequently, 500 µl of filtrate were combined with 50 µl of 10 M H₂SO₄. Then, 50 µl of d6-DMSO was added as a locking agent to produce a final volume of 600 µl. Two reference samples of 30 mM NH₄⁺ were prepared using natural abundance ¹⁵N-NH₄Cl in assay buffers.

To quantify the concentration of ¹⁴NH₄⁺ in each sample, one-dimensional ¹H spectra with excitation sculpting water suppression were acquired using the zgesgp sequence from the Bruker standard sequence library on an Oxford Instruments 600 MHz magnet equipped with an Avance III console and a TXO Helium cooled cryoprobe. To quantify the concentration of ¹⁵NH₄⁺, ¹⁵N heteronuclear single quantum coherence (HSQC) one-dimensional spectra were acquired on the same hardware using the hsqcetf3gp pulse sequence. Ninety-degree ¹H pulse times were calibrated on water. The ¹H carrier was set on-resonance at water to ensure optimal water suppression. The ¹⁵N carrier was set on-resonance for the ammonium signal at 21 ppm. The 90° ¹⁵N pulse time was calibrated on the high signal-to-noise Mo-nitrogenase sample and assumed constant for the other experiments where calibration was impossible due to low signal-to-noise ratios.

Spectra with varying numbers of scans were recorded for all samples dependent on the observable signal. The signal was rescaled linearly by the number of scans and the receiver gain. The spectra were then processed with zero-filling, an exponential window function and Fourier transformation using the nmrGlue python package⁵⁹.

The ¹⁵NH₄⁺ peak was then fitted in the ¹⁵N HSQC spectrum using a single Lorentzian function within the predefined grey window in Extended Data Fig. 7. The fitted peak was integrated and scaled against the reference peak to provide a ¹⁵NH₄⁺ concentration. A region containing only noise was defined between 0 ppm and 1 ppm, and the standard deviation of this region was multiplied by the fitted full-width-at-half-maximum parameter to estimate the peak integral error.

The ¹⁴NH₄⁺ peaks were fitted in the water-suppressed spectrum by first finding all local maxima above a spectrum-specific threshold, adding extra manual locations where needed, and fitting the resulting n Lorentzian functions to the data. The ¹⁴NH₄⁺ triplet was identified within these fitted peak locations by identifying the characteristic ¹⁴N–¹H coupling constant of 52.5 Hz. The integrals of the triplet were summed, and the fitting error was estimated as the standard deviation of the individual triplet peak integrals. These values were scaled against the respective buffer reference samples to provide a concentration of ¹⁴NH₄⁺ within the sample.

Two sources of NH₄⁺ signal were assumed: protein degradation and nitrogen fixation. All protein was produced in unenriched media, so it was assumed to contain a natural abundance 996:4 ratio of ¹⁴N and ¹⁵N. Therefore, we assumed that protein degradation contributes a natural abundance ratio of ¹⁴NH₄⁺ and ¹⁵NH₄⁺ molecules. To account for this, we subtracted a 4/996 fraction of each ¹⁴NH₄⁺ concentration from the corresponding ¹⁵NH₄⁺ concentration. The subtracted ¹⁵NH₄⁺ signal was then used to calculate the total turnover number.

Cryo-EM sample preparation, collection and processing

MgADP-AlF₃-trapped Mar(DK)₂(H₂)₂-complex was prepared by starting a 5-ml reaction with 22 nmol of MarH₂ and 3 nmol of Mar(DK)₂ in the presence of 100 mM MOPS pH 7.3, 50 mM Tris pH 7.8, 100 mM NaCl, 5 mM DT, 4 mM NaF, 0.2 mM AlCl₃, 8 mM MgCl₂ and 1 mM ATP. After 3 h incubation at 30 °C and 200 rpm, particles with the appropriate molecular mass were separated via size-exclusion chromatography in a HiLoad 26/600 Superdex 200 pg column (Cytiva) and buffer exchanged to 50 mM Tris pH 7.8, 200 mM NaCl and 5 mM DT. The protein complex with a measured molecular mass of ~368 kDa (Extended Data Fig. 2a,b; theoretical mass 344 kDa) was concentrated in Amicon Ultra filters (30 kDa cut-off, Merck) to 0.3 mg ml⁻¹. Grids were prepared with a Vitrobot Mark IV (Thermo Fisher Scientific) placed inside an anaerobic COY vinyl tent (COY Laboratory Products) with a >95% N₂ <5% H₂ atmosphere, at 26 °C and 19% humidity. Four microlitres of sample were added to freshly glow-discharged 300 mesh 2/1 µm C-flat copper grids (Electron Microscopy Sciences) and blotted for 10 s at blot force 4 at 100% relative humidity and 4 °C before being plunge frozen in a liquid mix of 37% (v/v) ethane and 63% (v/v) propane.

Cryo-EM data collection was carried out on a Titan Krios G3i electron microscope (Thermo Fisher Scientific) operated at 300 keV in EFTEM mode and equipped with a BioQuantum-K3 imaging filter (Gatan). Data were collected in electron counting mode with EPU (version 3.6) software (Thermo Fisher Scientific) at a nominal magnification of 105,000×, corresponding to a calibrated pixel size of 0.837 Å per pixel, with a total dose of 50 e⁻ Å⁻² per image, 50 fractions and 2.6 s exposure time. The energy filter slit width was set to 30 eV, and the defocus range was −1.0 µm to −2.5 µm at increments of 0.3 µm.

Cryo-EM data analysis was carried out in CryoSPARC v4.4.1 (ref. ⁶⁰) (Extended Data Fig. 3). A total of 16,951 micrographs were subjected to patch motion correction and patch contrast transfer function estimation before a template picking using the projections of the Fe-nitrogenase cryo-EM volume as templates (EMD-16890)¹⁷. An initial set of 10,625,194 particles were extracted at a box size of 380 pixels and used for three consecutive rounds of 2D classification into 200 classes. Particles from the best 2D classes displaying features of at least one MarH₂ bound to Mar(DK)₂ were reextracted at a box size of 380 pixels, yielding a total of 3,209,215 particles. These were subjected to ab initio reconstruction into three classes. The best ab initio class with 1,286,797 particles was used for non-uniform refinement, yielding a global reconstruction of 2.45 Å with very poor features around both MarD subunits. Hard 3D classification with a local mask on the proximal MarD subunit and 5 Å target resolution resulted in 25 different classes, including classes completely or partially lacking MarD. Particles of the largest class (116,370 particles) displayed the best alignment with reduced particle orientation bias and were subjected to local refinement with a local mask on the proximal MarD and non-uniform refinement, yielding electron density maps with resolutions of 2.86 Å and 2.77 Å, respectively. The final electron density map was obtained by performing a local refinement using a solvent mask yielding a final resolution of 2.75 Å. The resulting map was autosharpened in Phenix v1.21.1 (ref. ⁶¹). Details on cryo-EM data collection, model building and refinement statistics can be found in Supplementary Table 1.

Model building and refinement

The initial rigid-body fit of an AlphaFold2 (ref. ⁶²) multimer model of the Mar(DK)₂H₂-complex was performed in ChimeraX-1.7.1 (ref. ⁶³). Afterwards, manual fitting and refinement of all ligands was performed in Coot v0.8.9.2 (ref. ⁶⁴) using the restraint files from the Coot ligand database (ADP, AF3 (AlF₃), SF4 ([Fe₄S₄]-cluster), CLF (P-cluster) and S5Q ([Fe₈S₉C]-cluster)) allowing bond length deviation between 0.044 Å and 0.124 Å for the metalloclusters as specified in restraint files. Then, automatic model refinements were performed with phenix.real_space_refine of the Phenix software package (Phenix v1.21.1)⁶¹ using the same restrain files. For regions of the map that were not resolved well enough, the corresponding parts of the model were removed. Additional cycles of manual refinements in Coot as well as automatic refinements with phenix.real_space_refine were performed to obtain the final model, always applying ligand restraints.

Substrate channel estimation

Substrate access channels to the active site of the methylthio-alkane reductase were calculated using the tool CAVER Analyst 2.0 BETA³⁶. The starting point for channel calculating was the proposed binding site S2B of the [Fe₈S₉C]-cluster. The probe radius, shell radius and shell depth were set to 0.6, 3.0 and 4.0 Å, respectively. The two most likely substrate channels, with a maximum bottleneck radius of 27.5 Å and 25.6 Å, respectively, were selected for analysis (Fig. 4b,f).

EPR spectroscopy

All EPR samples were prepared under anaerobic conditions in a Coy anaerobic glove box. All protein samples were buffer exchanged with Sephadex G-25 packed PD-10 desalting columns (Cytiva) into 500 mM NaCl, 50 mM Tris pH 8.5 and 10% glycerol to remove excess DT before sample preparation.

For the redox titrations, samples were titrated in the presence of a cocktail of 40 µM of each redox mediator through the addition of different volumes of buffered DT or potassium ferricyanide solutions (1.5–100 mM) until the desired redox potential was reached. The mediator mix consisted of 2,6-dichlorophenolindophenol (E₀ = +217 mV), phenazine methosulfate (E₀ = +55 mV), methylene blue (E₀ = +11 mV), resorufin (E₀ = −51 mV), IDS (E₀ = −125 mV), 2-hydroxy-1,4-napthaquinone (E₀ = −152 mV), sodium anthraquinone 2-sulfonate (E₀ = −225 mV), phenosafranin (E₀ = −252 mV), safranin O (E₀ = −289 mV), neutral red (E₀ = −329 mV), benzyl viologen (E₀ = −358 mV) and methyl viologen (E₀ = −449 mV). Upon stabilization of the redox potential, a 300-µl sample was withdrawn, transferred to an EPR quartz tube, capped and shock frozen in liquid nitrogen. The InLab Argenthal microelectrode (Ag/AgCl, E = +207 mV versus H₂/H⁺, with combined Pt counter electrode) was calibrated with a quinhydrone saturated pH 7 reference buffer solution (E = +285 mV versus H₂/H⁺ at 25 °C).

MarH₂ samples were prepared by gentle addition of anaerobic stock solutions of the compounds of interest (2 mM DT, 5 mM MgATP, final concentration) to the protein, incubated for at least 3 min and frozen as for the redox titration. For the turnover experiment Mar(DK)₂ was oxidized by addition of 2 equiv. IDS and incubated for 15 min. DT-reduced MarH₂ and the IDS-oxidized Mar(DK)₂ were desalted. Directly after desalting, 300-µl samples of MarH₂ and oxidized and Mar(DK)₂ were frozen separately. DT-removed MarH₂ and IDS-removed Mar(DK)₂ were mixed in ratio of 1.2:1 reductase:catalytic component with MgATP (5 mM final concentration) incubated for 1 min, concentrated with an Amicon Ultra filter (100 kDa cut-off, Merck) to a final volume of 300 µl and shock frozen. Nif(DK)₂ was purified as described¹⁸, desalted and treated with 2 mM DT, 0.5 mM IDS or 50 µM IDS.

Cw-EPR spectra were recorded with a commercial Bruker spectrometer composed of an X-band E580-10/12 bridge and a 4122HQE resonator, for regular (perpendicular mode, for Fig. 3 and Extended Data Fig. 8) and an ER4116 dual mode cavity for parallel mode measurements (for Nif(DK)₂ and trials to detect g = 12 or g = 16 signals in Mar(DK)₂) measurements. The system was equipped with an Oxford Instruments temperature controller and ESR 900 cryostat, which was cryocooled by a Stinger (Cold Edge Technologies) linked to helium compressor (Sumitomo F-70). Spectral simulations were performed with GeeStrain5 (ref. ⁶⁵) or in Excel (Gaussian curves representing the absorption-shaped highest g-value of the very anisostropic doublets of the S = 3/2 system of MarH₂, or the first derivative of a Gaussian curve for the isotropic g = 4.3 EPR signal of the middle doublet of high spin ferric iron). Except for Fig. 3f and Extended Data Fig. 9, five-point moving averages were used to limit the noise of EPR spectra.

Phylogenetic analysis and ancestor prediction

For the phylogeny in Extended Data Fig. 10, sequences from Methanolobus bombayensis Ni²⁺-sirohydrochlorin a,c-diamide reductive cyclase complex component D (CfbD, WP_209620388.1), R. rubrum methylthio-alkane reductase MarDK operon (MarD, WP_011388552.1; MarK WP_011388551.1), A. vinelandii nitrogenase NifDKEN operon (NifD, WP_012698832.1; NifK, WP_012698833.1; NifE, WP_012698838.1; NifN, WP_012698839.1), R. capsulatus chlorophyllide a oxidoreductase BchYZ operon (BchY, WP_023914767.1; BchZ WP_023914765.1) and DPOR BchNB operon (BchN, WP_023913723.1; BchB WP_013066408.1) were used as queries across major archaeal and bacterial lineages, collecting a total of 996 sequences using the web interface of NCBI pblast. Sequences were collected iteratively between 2021 and 2024, and filtered to retain only those with E-values ≤1 × 10⁻³⁰ and query coverage >70%. Sequences were initially assigned as bona fide nitrogenase components through synteny, for cases where NifDKEN are in synteny, because operons of other relatives of this family contain only DK paralogues. Further classification during dataset assembly was achieved through reciprocal best-blast hits. Ultimately, the classification on our tree is based on branching relationships of known sequences in each clade. Sequences were aligned using Muscle v5 (ref. ⁶⁶) and further refined using clipKIT⁶⁷. We have also realigned our sequences three times each with two different algorithms (geneafpair and globalpair algorithms with option -maxiterate 3000) in MAFFT, and verified that the alignment is robust—specifically, that all reported cysteines are consistently aligned in each case. The resulting phylogenetic trees were constructed using Iqtree2 (ref. ⁶⁸) under the LG + G4 model, with 35,000 ultrafast bootstraps and 1500 SH-aLRT replicates⁶⁹. Model selection for evolution was based on the corrected Akaike information criterion (-m MFP -merit AICc option in Iqtree2). The resulting tree was constrained to have BchY and BchN sister to NifDE and nitrogenase-like proteins (NflD) including MarD, and BchB and BchZ clades sister to NifKN, NflK and MarK. Every other node was polytomized. The resulting maximum likelihood analysis of the constrained tree shows similar likelihood values as for the unconstrained tree and it was not rejected by any statistical tree topology test performed with iqtree. This topology was chosen for analysis owing to the more parsimonious history in terms of gene duplication: all nitrogenase relatives other than CfbD are heterotetramers made from two paralogues. In our constrained topology, the tree implies that a single gene duplication of an ancestral homomer produced this arrangement. Subsequent operon duplications would then produce the different types of nitrogenase-like complex, each also consisting of two paralogues (including the maturases of nitrogenases). In the unconstrained tree, all BchNBYZ paralogues form a monophyletic sister group to all other nitrogenase-like genes. This would imply that Bch separately underwent a duplication of an ancestral homomer, which then underwent an operon duplication to yield BchNB and BchYZ. This implies one additional gene duplication compared with our constrained tree and, thus, less parsimony. Supporting values for this topology were calculated with 500 Felsenstein bootstrap analysis using Iqtree2 and the approximate likelihood ratio test (aLRT) using PhyML. Ancestral sequence reconstruction for the nodes of interest was performed using a wrapper script that uses Iqtree2 for ancestral sequence inference using the LG + G4 model and PAUPv4 for a Fitch parsimony inference of which sites should be gaps in our alignment.

AlphaFold Structure Prediction

The structure of the last common nitrogenase and methylthio-alkane reductase ancestor AncD and AncK was predicted by AlphaFold 2⁶² using ColabFold⁷⁰ with default parameters (without templates, model_type: alphafold2_multimer_v3, num_relax: 1).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

The refined MarDK₂H₂ structure and electron density map were deposited under PDB-ID 9FMG and EMD-50553, respectively. Raw data of SDS–PAGE analysis, ICP-OES and mass photometry measurements, gas chromatography and NMR enzyme activity assays, EPR spectroscopy, protein sequences used in multiple sequence alignments, and phylogenetic analysis (protein sequences identifiers, multiple sequence alignment, uncollapsed trees with bootstrap values, and resurrected ancestral protein sequences with calculated posterior probabilities per site) are deposited on Edmond, the Open Research Data Repository of the Max Planck Society (https://doi.org/10.17617/3.OHDUAZ)⁷¹, which will be released upon acceptance.

Code availability

All NMR data and the Python package created for analysis are available via GitHub at https://github.com/charliebuchanan/nitrogenaseNMR.

References

Kredich, N. M. Biosynthesis of cysteine. EcoSal Plus https://doi.org/10.1128/ecosalplus.3.6.1.11 (2008).
Sekowska, A., Ashida, H. & Danchin, A. Revisiting the methionine salvage pathway and its paralogues. Microb. Biotechnol. 12, 77–97 (2019).
Article PubMed CAS Google Scholar
Parveen, N. & Cornell, K. A. Methylthioadenosine/S-adenosylhomocysteine nucleosidase, a critical enzyme for bacterial metabolism. Mol. Microbiol. 79, 7–20 (2011).
Article PubMed CAS Google Scholar
North, J. A. et al. A nitrogenase-like enzyme system catalyzes methionine, ethylene, and methane biogenesis. Science 369, 1094–1098 (2020).
Article PubMed CAS Google Scholar
North, J. A., Miller, A. R., Wildenthal, J. A., Young, S. J. & Robert Tabita, F. Microbial pathway for anaerobic 5′-methylthioadenosine metabolism coupled to ethylene formation. Proc. Natl Acad. Sci. USA 114, E10455–E10464 (2017).
Article PubMed PubMed Central CAS Google Scholar
Hausinger, R. P. et al. Biological formation of ethylene. RSC Chem. Biol. 4, 635–646 (2023).
Article PubMed PubMed Central CAS Google Scholar
Lynch, J. M. Ethylene in soil. Nature 256, 576–577 (1975).
Article CAS Google Scholar
Manik, S. M. N. et al. Soil and crop management practices to minimize the impact of waterlogging on crop productivity. Front. Plant Sci. 10, 1–23 (2019).
Article Google Scholar
Raymond, J., Siefert, J. L., Staples, C. R. & Blankenship, R. E. The natural history of nitrogen fixation. Mol. Biol. Evol. 21, 541–554 (2004).
Article PubMed CAS Google Scholar
Zhang, X., Ward, B. B. & Sigman, D. M. Global nitrogen cycle: critical enzymes, organisms, and processes for nitrogen budgets and dynamics. Chem. Rev. 120, 5308–5351 (2020).
Article PubMed CAS Google Scholar
Einsle, O. & Rees, D. C. Structural enzymology of nitrogenase enzymes. Chem. Rev. 120, 4969–5004 (2020).
Article PubMed PubMed Central CAS Google Scholar
Moser, J. & Layer, G. Enzymatic systems with homology to nitrogenase: biosynthesis of bacteriochlorophyll and coenzyme F430. Met. Methods Mol. Biol. 1876, 25–35 (2019).
Article CAS Google Scholar
Van Stappen, C. et al. The spectroscopy of nitrogenases. Chem. Rev. 120, 5005–5081 (2020).
Article PubMed PubMed Central Google Scholar
Rutledge, H. L. & Tezcan, F. A. Electron transfer in nitrogenase. Chem. Rev. 120, 5158–5193 (2020).
Article PubMed PubMed Central CAS Google Scholar
Kaiser, J. T., Hu, Y., Wiig, J. A., Rees, D. C. & Ribbe, M. W. Structure of precursor-bound NifEN: a nitrogenase FeMo cofactor maturase/insertase. Science 331, 91–94 (2011).
Article PubMed PubMed Central CAS Google Scholar
Ribbe, M. W., Hu, Y., Hodgson, K. O. & Hedman, B. Biosynthesis of nitrogenase metalloclusters. Chem. Rev. 114, 4063–4080 (2014).
Article PubMed CAS Google Scholar
Schmidt, F. V. et al. Structural insights into the iron nitrogenase complex. Nat. Struct. Mol. Biol. 31, 150–158 (2024).
Article PubMed CAS Google Scholar
Oehlmann, N. N., Schmidt, F. V., Herzog, M., Goldman, A. L. & Rebelein, J. G. The iron nitrogenase reduces carbon dioxide to formate and methane under physiological conditions: a route to feedstock chemicals. Sci. Adv. 10, eado7729 (2024).
Article PubMed PubMed Central CAS Google Scholar
Schneider, K., Gollan, U., Dröttboom, M., Selsemeier-Voigt, S. & Müller, A. Comparative biochemical characterization of the iron-only nitrogenase and the molybdenum nitrogenase from Rhodobacter capsulatus. Eur. J. Biochem. 244, 789–800 (1997).
Article PubMed CAS Google Scholar
Dilworth, M. J., Eady, R. R. & Eldridge, M. E. The vanadium nitrogenase of Azotobacter chroococcum. Reduction of acetylene and ethylene to ethane. Biochem. J. 249, 745–751 (1988).
Article PubMed PubMed Central CAS Google Scholar
Jasniewski, A. J., Lee, C. C., Ribbe, M. W., Ribbe, M. W. & Hu, Y. Reactivity, mechanism, and assembly of the alternative nitrogenases. Chem. Rev. 120, 5107–5157 (2020).
Article PubMed PubMed Central CAS Google Scholar
Wiig, J. A., Hu, Y., Lee, C. C. & Ribbe, M. W. Radical SAM-dependent carbon insertion into the nitrogenase M-cluster. Science 337, 1672–1675 (2012).
Article PubMed CAS Google Scholar
Rutledge, H. L., Cook, B. D., Nguyen, H. P. M., Herzik, M. A. Jr & Tezcan, F. A. Structures of the nitrogenase complex prepared under catalytic turnover conditions. Science 377, 865–869 (2022).
Article PubMed PubMed Central CAS Google Scholar
Spatzal, T., Perez, K. A., Einsle, O., Howard, J. B. & Rees, D. C. Ligand binding to the FeMo-cofactor: structures of co-bound and reactivated nitrogenase. Science 345, 1620–1623 (2014).
Article PubMed PubMed Central CAS Google Scholar
Spatzal, T., Perez, K. A., Howard, J. B. & Rees, D. C. Catalysis-dependent selenium incorporation and migration in the nitrogenase active site iron-molybdenum cofactor. eLife 4, e11620 (2015).
Article PubMed PubMed Central Google Scholar
Sippel, D. et al. A bound reaction intermediate sheds light on the mechanism of nitrogenase. Science 359, 1484–1489 (2018).
Article PubMed CAS Google Scholar
Rohde, M., Grunau, K. & Einsle, O. CO binding to the FeV cofactor of CO-reducing vanadium nitrogenase at atomic resolution. Angew. Chem. Int. Ed. 59, 23626–23630 (2020).
Article CAS Google Scholar
Kang, W., Lee, C. C., Jasniewski, A. J., Ribbe, M. W. & Hu, Y. Structural evidence for a dynamic metallocofactor during N₂ reduction by Mo-nitrogenase. Science 368, 1381–1385 (2020).
Article PubMed PubMed Central CAS Google Scholar
Warmack, R. A. & Rees, D. C. Nitrogenase beyond the resting state: a structural perspective. Molecules 28, 7952 (2023).
Article PubMed PubMed Central CAS Google Scholar
Sarma, R. et al. Insights into substrate binding at FeMo-cofactor in nitrogenase from the structure of an α−70^I1e MoFe protein variant. J. Inorg. Biochem. 104, 385–389 (2010).
Article PubMed CAS Google Scholar
Trncik, C., Detemple, F. & Einsle, O. Iron-only Fe-nitrogenase underscores common catalytic principles in biological nitrogen fixation. Nat. Catal. 6, 415–424 (2023).
Article CAS Google Scholar
Garcia, A. K., McShea, H., Kolaczkowski, B. & Kaçar, B. Reconstructing the evolutionary history of nitrogenases: Evidence for ancestral molybdenum-cofactor utilization. Geobiology 18, 394–411 (2020).
Article PubMed PubMed Central Google Scholar
Igarashi, R. Y. & Seefeldt, L. C. Nitrogen fixation: the mechanism of the Mo-dependent nitrogenase. Crit. Rev. Biochem. Mol. Biol. 38, 351–384 (2003).
Article PubMed CAS Google Scholar
Morrison, C. N., Hoy, J. A., Zhang, L., Einsle, O. & Rees, D. C. Substrate pathways in the nitrogenase MoFe protein by experimental identification of small molecule binding sites. Biochemistry 54, 2052–2060 (2015).
Article PubMed CAS Google Scholar
Gee, L. B. et al. Docking and migration of carbon monoxide in nitrogenase: the case for gated pockets from infrared spectroscopy and molecular dynamics. Biochemistry 54, 3314–3319 (2015).
Article PubMed CAS Google Scholar
Chovancova, E. et al. CAVER 3.0: a tool for the analysis of transport pathways in dynamic protein structures. PLoS Comput. Biol. 8, e1002708 (2012).
Article PubMed PubMed Central CAS Google Scholar
Lee, C. C., Górecki, K., Stang, M., Ribbe, M. W. & Hu, Y. Cofactor maturase NifEN: a prototype ancient nitrogenase? Sci. Adv. 10, eado6169 (2024).
Article PubMed PubMed Central CAS Google Scholar
Hagen, W. R., Eady, R. R., Dunham, W. R. & Haaker, H. A novel S = 3/2 EPR signal associated with native Fe-proteins of nitrogenase. FEBS Lett. 189, 250–254 (1985).
Article PubMed CAS Google Scholar
Lee, C. C., Hu, Y. & Ribbe, M. W. Unique features of the nitrogenase VFe protein from Azotobacter vinelandii. Proc. Natl Acad. Sci. USA 106, 9209–9214 (2009).
Article PubMed PubMed Central CAS Google Scholar
Pierik, A. J., Wassink, H., Haaker, H. & Hagen, W. R. Redox properties and EPR spectroscopy of the P clusters of Azotobacter vinelandii MoFe protein. Eur. J. Biochem. 212, 51–61 (1993).
Article PubMed CAS Google Scholar
Wahl, I. M. et al. Understanding the P-cluster of vanadium nitrogenase: an EPR and XAS study of the holo vs. apo forms of the enzyme. ChemBioChem 26, e202400833 (2024).
Article PubMed PubMed Central Google Scholar
Van Stappen, C. et al. A conformational role for NifW in the maturation of molybdenum nitrogenase P-cluster. Chem. Sci. 13, 3489–3500 (2022).
Article PubMed PubMed Central Google Scholar
Boyd, E. S. & Peters, J. W. New insights into the evolutionary history of biological nitrogen fixation. Front. Microbiol. 4, 201 (2013).
Article PubMed PubMed Central Google Scholar
Rees, D. C. Great metalloclusters in enzymology. Annu. Rev. Biochem. 71, 221–246 (2002).
Article PubMed CAS Google Scholar
Ramírez-Amador, F. et al. Structure of the ATP-driven methyl-coenzyme M reductase activation complex. Nature 642, 814–821 (2025).
Warmack, R. A. et al. Structural consequences of turnover-induced homocitrate loss in nitrogenase. Nat. Commun. 14, 1091 (2023).
Article PubMed PubMed Central CAS Google Scholar
Imperial, J., Hoover, T. R., Madden, M. S., Ludden, P. W. & Shah, V. K. Substrate reduction properties of dinitrogenase activated in vitro are dependent upon the presence of homocitrate or its analogues during iron–molybdenum cofactor synthesis. Biochemistry 28, 7796–7799 (1989).
Article PubMed CAS Google Scholar
Yang, K. Y., Haynes, C. A., Spatzal, T., Rees, D. C. & Howard, J. B. Turnover-dependent inactivation of the nitrogenase MoFe-protein at high pH. Biochemistry 53, 333–343 (2014).
Article PubMed CAS Google Scholar
Klipp, W., Masepohl, B. & Pühler, A. Identification and mapping of nitrogen fixation genes of Rhodobacter capsulatus: duplication of a nifA-nifB region. J. Bacteriol. 170, 693–699 (1988).
Article PubMed PubMed Central CAS Google Scholar
Masepohl, B., Krey, R. & Klipp, W. The draTG gene region of Rhodobacter capsulatus is required for post-translational regulation of both the molybdenum and the alternative nitrogenase. J. Gen. Microbiol. 139, 2667–2675 (1993).
Article PubMed CAS Google Scholar
Wang, G., Angermuller, S. & Klipp, W. Characterization of Rhodobacter capsulatus genes encoding a molybdenum transport system and putative molybdenum-pterin-binding proteins. J. Bacteriol. 175, 3031–3042 (1993).
Article PubMed PubMed Central CAS Google Scholar
Brimacombe, C. A. et al. Quorum-sensing regulation of a capsular polysaccharide receptor for the Rhodobacter capsulatus gene transfer agent (RcGTA). Mol. Microbiol. 87, 802–817 (2013).
Article PubMed PubMed Central CAS Google Scholar
Thoma, S. & Schobert, M. An improved Escherichia coli donor strain for diparental mating. FEMS Microbiol. Lett. 294, 127–132 (2009).
Article PubMed CAS Google Scholar
Katzke, N., Bergmann, R., Jaeger, K. E. & Drepper, T. Heterologous high-level gene expression in the photosynthetic bacterium Rhodobacter capsulatus. Methods Mol. Biol. 824, 251–269 (2012).
Article PubMed CAS Google Scholar
Weber, E., Engler, C., Gruetzner, R., Werner, S. & Marillonnet, S. A modular cloning system for standardized assembly of multigene constructs. PLoS ONE 6, e16765 (2011).
Article PubMed PubMed Central CAS Google Scholar
Geddes, B. A., Mendoza-Suárez, M. A. & Poole, P. S. A Bacterial Expression Vector Archive (BEVA) for flexible modular assembly of golden gate-compatible vectors. Front. Microbiol. 9, 3345 (2019).
Article PubMed PubMed Central Google Scholar
Gibson, D. G. et al. Enzymatic assembly of DNA molecules up to several hundred kilobases. Nat. Methods 6, 343–345 (2009).
Article PubMed CAS Google Scholar
Schmidt, T. G. M. & Skerra, A. The Strep-tag system for one-step purification and high-affinity detection or capturing of proteins. Nat. Protoc. 2, 1528–1535 (2007).
Article PubMed CAS Google Scholar
Helmus, J. J. & Jaroniec, C. P. Nmrglue: an open source Python package for the analysis of multidimensional NMR data. J. Biomol. NMR 55, 355–367 (2013).
Article PubMed PubMed Central CAS Google Scholar
Punjani, A., Rubinstein, J. L., Fleet, D. J. & Brubaker, M. A. CryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination. Nat. Methods 14, 290–296 (2017).
Article PubMed CAS Google Scholar
Liebschner, D. et al. Macromolecular structure determination using X-rays, neutrons and electrons: recent developments in Phenix. Acta Crystallogr. Sect. D. 75, 861–877 (2019).
Article CAS Google Scholar
Jumper, J. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589 (2021).
Article PubMed PubMed Central CAS Google Scholar
Pettersen, E. F. et al. UCSF ChimeraX: structure visualization for researchers, educators, and developers. Protein Sci. 30, 70–82 (2021).
Article PubMed CAS Google Scholar
Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development of Coot. Acta Crystallogr. Sect. D 66, 486–501 (2010).
Article CAS Google Scholar
Hagen, W. R. Very low-frequency broadband electron paramagnetic resonance spectroscopy of metalloproteins. J. Phys. Chem. A 125, 3208–3218 (2021).
Article PubMed PubMed Central CAS Google Scholar
Edgar, R. C. Muscle5: High-accuracy alignment ensembles enable unbiased assessments of sequence homology and phylogeny. Nat. Commun. 13, 6968 (2022).
Article PubMed PubMed Central CAS Google Scholar
Steenwyk, J. L., Buida, T. J., Li, Y., Shen, X. X. & Rokas, A. ClipKIT: a multiple sequence alignment trimming software for accurate phylogenomic inference. PLoS Biol. 18, e3001007 (2020).
Article PubMed PubMed Central CAS Google Scholar
Nguyen, L. T., Schmidt, H. A., Von Haeseler, A. & Minh, B. Q. IQ-TREE: a fast and effective stochastic algorithm for estimating maximum-likelihood phylogenies. Mol. Biol. Evol. 32, 268–274 (2015).
Article PubMed CAS Google Scholar
Anisimova, M. & Gascuel, O. Approximate likelihood-ratio test for branches: a fast, accurate, and powerful alternative. Syst. Biol. 55, 539–552 (2006).
Article PubMed Google Scholar
Mirdita, M. et al. ColabFold: making protein folding accessible to all. Nat. Methods 19, 679–682 (2022).
Article PubMed PubMed Central CAS Google Scholar
Lago-Maciel, A. Methylthio-alkane reductases use nitrogenase metalloclusters for carbon–sulfur bond cleavage. Edmond https://doi.org/10.17617/3.OHDUAZ (2025).
Article Google Scholar
Moser, J. et al. Structure of ADP-aluminium fluoride-stabilized protochlorophyllide oxidoreductase complex. Proc. Natl Acad. Sci. USA 110, 2094–2098 (2013).
Article PubMed PubMed Central CAS Google Scholar
Muraki, N. et al. X-ray crystal structure of the light-independent protochlorophyllide reductase. Nature 465, 110–114 (2010).
Article PubMed CAS Google Scholar
Madeira, F. et al. The EMBL-EBI Job Dispatcher sequence analysis tools framework in 2024. Nucleic Acids Res. 52, 521–525 (2024).
Article Google Scholar
Sievers, F. & Higgins, D. G. The Clustal Omega multiple alignment package. Methods Mol. Biol. 2231, 3–16 (2021).
Article PubMed CAS Google Scholar
Hagen, W. R. in Advances in Inorganic Chemistry Vol. 38 (eds. Cammack, R. & Sykes, A. G.) 165–222 (Academic Press, 1992).

Download references

Acknowledgements

We thank the Central Electron Microscopy Facility at the Max Planck Institute of Biophysics for expertise and access to their instruments. We thank C. Thölken, P. Klemm and M. Lechner for assistance with data transfer and computing cluster access. We thank V. Reitz for her help with molecular cloning. We thank C. Scholz for access and measurements of ICP-OES. We thank L. Ernst for his help with biochemistry work and purification of the Mo-nitrogenase. We thank W. R. Hagen for access to the simulation program GeeStrain5. We thank Y. Hu and M. Ribbe for providing an EPR spectrum of [Fe₈S₉C]-cluster-bound Nif(EN)₂. This work was supported by the German Research Foundation (grant 446841743, J.G.R). A.L.-M., J.Z., C.J.B., N.P., G.K.A.H. and J.G.R. are grateful for generous support from the Max Planck Society. A.L.-M. acknowledges the financial support by the International Max Planck Research School Principles of Microbial Life. The BBSRC (BB/R000255/1) is acknowledged for supporting the NMR facility at University College London. The EPR spectrometer upgrade and closed-cycle cryostat (A.J.P.) was funded by the German Research Foundation (248/320-1, project number 444947649) and the Rhineland-Palatinate government. A.J.P. acknowledges S. Leimkühler and the SPP1927 FeS for Life (grants PI610/2-1 and 2). This research is supported by the UKRI and EPSRC (DFH; EP/X036782/1). G.T.H. was supported by a BBSRC Discovery Fellowship (BB/X009955/1). C.J.B. is grateful for the support of a Human Frontiers Science Programme Long Term Fellowship (LT0018/2024-L). J.M.S. acknowledges the Deutsche Forschungsgemeinschaft for an Emmy Noether grant (SCHU 3364/1-1) and the European Union’s Horizon 2020 research and innovation programme (Two-CO2-One; grant agreement no. 101075992). The views and opinions expressed are those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them. Open Access publishing was enabled and organized by Projekt DEAL.

Funding

Open access funding provided by Max Planck Society.

Author information

Authors and Affiliations

Microbial Metalloenzymes Research Group, Max Planck Institute for Terrestrial Microbiology, Marburg, Germany
Ana Lago-Maciel, Frederik V. Schmidt & Johannes G. Rebelein
Department of Chemistry, RPTU University Kaiserslautern-Landau, Kaiserslautern, Germany
Jéssica C. Soares & Antonio J. Pierik
Department of Biochemistry and Synthetic Metabolism, Max Planck Institute for Terrestrial Microbiology, Marburg, Germany
Jan Zarzycki
Evolutionary Biochemistry Group, Max Planck Institute for Terrestrial Microbiology, Marburg, Germany
Charles J. Buchanan, Stefano Lometto & Georg K. A. Hochberg
Department of Structural and Molecular Biology, Division of Biosciences, University College London, London, UK
Charles J. Buchanan, D. Flemming Hansen & Gabriella T. Heller
Department of Chemistry, Philipps–University Marburg, Marburg, Germany
Tristan Reif-Trauttmansdorff, Jan M. Schuller & Georg K. A. Hochberg
Metabolomics Core Facility, Max Planck Institute for Terrestrial Microbiology, Marburg, Germany
Nicole Paczia
Center for Synthetic Microbiology (SYNMIKRO), Philipps–University Marburg, Marburg, Germany
Jan M. Schuller, Georg K. A. Hochberg & Johannes G. Rebelein
The Francis Crick Institute, London, UK
D. Flemming Hansen
Central Electron Microscopy Facility, Max Planck Institute of Biophysics, Frankfurt am Main, Germany
Simone Prinz

Authors

Ana Lago-Maciel
View author publications
Search author on:PubMed Google Scholar
Jéssica C. Soares
View author publications
Search author on:PubMed Google Scholar
Jan Zarzycki
View author publications
Search author on:PubMed Google Scholar
Charles J. Buchanan
View author publications
Search author on:PubMed Google Scholar
Tristan Reif-Trauttmansdorff
View author publications
Search author on:PubMed Google Scholar
Frederik V. Schmidt
View author publications
Search author on:PubMed Google Scholar
Stefano Lometto
View author publications
Search author on:PubMed Google Scholar
Nicole Paczia
View author publications
Search author on:PubMed Google Scholar
Jan M. Schuller
View author publications
Search author on:PubMed Google Scholar
D. Flemming Hansen
View author publications
Search author on:PubMed Google Scholar
Gabriella T. Heller
View author publications
Search author on:PubMed Google Scholar
Simone Prinz
View author publications
Search author on:PubMed Google Scholar
Georg K. A. Hochberg
View author publications
Search author on:PubMed Google Scholar
Antonio J. Pierik
View author publications
Search author on:PubMed Google Scholar
Johannes G. Rebelein
View author publications
Search author on:PubMed Google Scholar

Contributions

J.G.R. conceived and supervised the project. J.G.R. acquired funding. A.L.-M. and J.G.R. planned and analysed experiments. A.L.-M. performed molecular work, anaerobic protein purification, enzyme biochemistry, mass photometry measurements and analysis of the structural data. F.V.S. performed anaerobic Mo-nitrogenase purification and activity assays. S.L. and G.K.A.H. performed the phylogenetic analysis. N.P. performed homocitrate measurements. C.J.B., G.T.H. and D.F.H. performed NMR measurements and analysis. A.L.-M. and S.P. performed cryo-EM data acquisition. A.L.-M., J.Z. and T.R.-T. processed and refined the cryo-EM structure. A.L.-M., J.C.S. and A.J.P. performed EPR measurements and analysis. A.L.-M., J.G.R., J.C.S. and A.J.P. wrote the original manuscript, which was reviewed and edited by all authors.

Corresponding author

Correspondence to Johannes G. Rebelein.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Catalysis thanks Oliver Einsle and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended Data Fig. 1 Product formation of methythio-alkane reductase and Mo-nitrogenase.

a–d, In vitro activities of methylthio-alkane reductase (Mar) and Mo-nitrogenase (Nif) for the reduction of (2-methylthio)ethanol (MT-EtOH, a), ethyl methyl sulfide (EMS, b), dimethyl sulfide (DMS, c) and acetylene (C₂H₂, d). Specific activities are plotted for the formation of ethylene (C₂H₄, green), ethane (C₂H₆, yellow), methane (CH₄, gray) and hydrogen (H₂, blue) (n = 3). n.d., not detected. e, In vitro C₂H₄ specific activity of the methylthio-alkane reductase for the reduction of MT-EtOH is dependent on the addition of reductase and catalytic components, ATP and DT (n = 3). Error bars depict standard deviation of three independent experiments.

Extended Data Fig. 2 Mar(DK)₂(H₂)₂-complex disassembles into Mar(DK)₂H₂-complex with complementary orientations.

a-b, Isolation of the MgADP-AlF₃-trapped Mar(DK)₂(H₂)₂-complex by size exclusion chromatography. a, Molecular weight (MW, kDa) calibration curve for the size exclusion chromatography column HiLoad 26/600 Superdex 200 pg column (Cytiva, Marlborough, Massachusetts, USA) used to isolate the MgADP-AlF₃-trapped Mar(DK)₂(H₂)₂-complex for cryo-EM grid preparation. Protein standards are colored in black. Measured complex, reductase and catalytic components are colored in yellow, grey and green, respectively. b, Elution profile of the Mar(DK)₂(H₂)₂-complex reaction. The protein fraction used for cryo-EM grid preparation is highlighted in mint. c, Electron density maps of the hexameric Mar(DK)₂H₂-complex, e, the octameric Mar(DK)₂(H₂)₂-complex, and g, a particle from combining both populations during cryo-EM analysis. d,f,h, Conical Fourier Shell Correlation (FSC) Area Ratio (cFAR) plots by CryoSPARC v4.4.1 assessing specimen orientation bias for the electron density maps of panels c, e and g, respectively. The line depicts the mean, the dark blue area the standard deviation and the aquamarine area the minimum and maximum value of the correlations at each special frequency. The octameric (c, d) and hexameric (e, f) complexes separately resulted in electron density maps displaying the smearing effect of preferential orientation in different planes. Combining both populations enables the reconstruction of a particle with an improved electron density map, free from the smearing effect (g, h), which was selected for further analysis (see Extended Data Fig. 3).

Extended Data Fig. 3 Cryo-EM data acquisition and analysis of the methylthio-alkane reductase complex.

a, Schematic of the processing workflow for cryo-EM data using CryoSPARC v4.4.1⁶⁰ and Phenix v1.21.1⁶¹. b, Conical FSC Area Ratio (cFAR) plot by CryoSPARC v4.4.1 assessing specimen orientation bias. The line depicts the mean, the dark blue area the standard deviation and the aquamarine area the minimum and maximum value of the correlations at each special frequency. c, Local resolution estimated by CryoSPARC v4.4.1 mapped onto the final electron density map contoured at level 10. d, Gold-Standard Fourier Shell Correlation (GSFSC) plot by CryoSPARC v4.4.1 with resolution calculated at Fourier shell correlation (FSC) = 0.143. e, Map to atomic model FSC plot determined at FSC = 0.143. f, Histogram of local resolution values calculated at FSC = 0.143 by CryoSPARC v4.4.1.

Extended Data Fig. 4 The reductase component MarH₂ belongs to the family of P-loop NTPases.

a, Clustal Omega^74,75 multiple sequence alignment (MSA) of the R. rubrum methylthio-alkane reductase reductase component MarH to the A. vinelandii Mo-, V- and Fe-nitrogenase reductase components NifH, VnfH and AnfH, colored with Boxshade v3.3. The conserved P-loop, switch I and switch II regions involved in nucleotide binding and hydrolysis are colored in green, pink and blue, respectively. Cys97^MarH and Cys133^MarH coordinating the [Fe₄S₄]-cluster are highlighted in yellow. b, Close-up view of the MgADP-AlF₃ moiety bound to one MarH subunit. P-loop, switch I and switch II regions are colored as in (a). MgADP-AlF₃ is represented in ball-and-stick style with carbon atoms colored in dark gray, nitrogen in blue, oxygen in red, phosphorus in orange, aluminium in light gray, fluoride in pale cyan and magnesium in green. Amino acid residues coordinating the ligand by hydrogen bonding are shown in stick representation.

Extended Data Fig. 5 Sequence alignment of D-subunits of methylthio-alkane reductase and nitrogenases.

MSA of the R. rubrum methylthio-alkane MarD subunit and the A. vinelandii Mo-, V- and Fe-nitrogenase NifD, VnfD and AnfD subunits, colored with Boxshade v3.3. Cysteine triads involved in P-cluster coordination are highlighted in yellow and numbered 1-3. Cysteine and histidine ligands responsible for the coordination of the [Fe₈S₉C]-cluster, FeMoco, FeVco and FeFeco are labeled with a yellow and blue circle, respectively. Strictly conserved residues involved in nitrogenase catalytic mechanism and FeMoco, FeVco and FeFeco stability are highlighted in purple, while residues replacing them in the structure of the methylthio-alkane reductase are highlighted in orange. The extended loop I of the methylthio-alkane reductase containing the coordinating histidine His429^MarD is highlighted in pink, while the Mo-nitrogenase loop II containing the coordinating histidine His422^AvNifD is highlighted in green.

Extended Data Fig. 6 Environment of the methylthio-alkane reductase [Fe₈S₉C]-cluster and P-cluster.

a,b, Close-up view of the [Fe₈S₉C]-cluster of the active site of the methylthio-alkane reductase (a) and the P-cluster bridging the MarDK subunits (b). Metalloclusters are shown as ball-and-sticks, with carbon atoms in dark gray, sulfur in yellow, and iron in dark orange. Amino acid residues surrounding the metalloclusters within a 5 Å radius are shown in stick representation. Amino acid residues of MarD and MarK are colored in yellow and green, respectively.

Extended Data Fig. 7 Nitrogen assays with methylthio-alkane reductase and product detection by NMR spectroscopy.

Nitrogenases provide two sources of NH₄⁺: First, by protein degradation, second, by nitrogen fixation. When ¹⁵N₂ is used as the nitrogen source, the nitrogen fixation will be isotopically labelled ¹⁵NH₄⁺, while protein degradation will remain at natural abundance (99.6% ¹⁴NH₄⁺, 0.4% ¹⁵NH₄⁺) (a). We measured the concentration of both ¹⁴NH₄⁺ and ¹⁵NH₄⁺ and used natural abundance ratios to distinguish nitrogen fixation from protein degradation, providing a more precise quantification of Total Turnover than simply measuring ¹⁵NH₄⁺ (a). A single Lorentzian function is fitted (b, red) to each processed NMR spectrum from the ¹⁵N HSQC experiment (b, black), before being integrated (b, pink shaded) and scaled against reference NH₄⁺ samples to provide a concentration. Performing this on all samples (c) shows good agreement between fitted (c, red) and raw (c, black) data. The extracted ¹⁵NH₄⁺ concentrations span the nanomolar to millimolar range (d). n Lorentzian functions are fitted (e, red and individual fits) to the processed water-suppressed ¹H NMR data and the ¹⁴NH₄⁺ peaks are extracted by identifying the fitted peak locations split by the characteristic 52.5 Hz coupling constant (e, shaded). The summed integral of the triplet is then scaled against reference NH₄⁺ samples to provide a concentration. This was performed on all samples and the resulting fits (f, red) show good agreement with raw (f, black) data. The ¹⁴NH₄⁺ concentrations span the 100 µM to 250 µM range for the protein samples (g). As the protein is isotopically non-enriched, we assume the degradation products will have the natural abundance distribution of 996:4. To account for this, we subtract 0.4% of the ¹⁴NH₄⁺ concentration from the ¹⁵NH₄⁺ concentration to produce a subtracted ¹⁵NH₄⁺ concentration that describes the maximum ¹⁵NH₄⁺ concentration originating from nitrogen fixation (h). The data was used to calculate total turnover numbers for Fig. 2f. The error bars in d, g, h show the propagated instrumental error.

Extended Data Fig. 8 EPR spectroscopy of Mar(DK)₂ and MarH₂ metalloclusters.

a, Amplitudes of the two species from Fig. 3g and further redox titrations of Mar(DK)₂ with fits to the Nernst equation for n = 1 with E_m = −390 and −250 mV. b,c, S = 3/2 EPR signals of the [Fe₄S₄]¹⁺ cluster in dithionite (DT)-reduced MarH₂ with (c) and without (b) MgATP. EPR signals recorded at 4, 10 and 20 K in black lines, and with simulations in red lines (parameters in Supplementary Table 2). EPR conditions: microwave frequency 9.35 GHz; modulation frequency, 100 kHz; 1.5 mT modulation amplitude; microwave power, 20 mW. d, e, Energy level diagrams of the S = 3/2 spin system of MarH₂. The calculated g-values for MarH₂ without ATP (d) are in red, those for MarH₂ in the presence of ATP in blue (e). The intensity of the g = 5.06 peak shows Curie law behaviour, indicating that both doublets are almost equally populated. Thus, it could not be determined whether the peak is from the ground or excited state, and the magnitude and sign of the D value could not be estimated. The two possible energy level diagrams are shown. f, g, Rhombograms for the two |±3/2〉 and |±1/2〉 doublets of the S = 3/2 state⁷⁶. Calculated g-values as function of the rhombicity (E/D) are shown.

Extended Data Fig. 9 EPR spectroscopy of the metalloclusters of Rhodobacter capsulatus Nif(DK)₂.

a, Samples of desalted Nif(DK)₂ were treated with indicated oxidants (5,5′-indigodisulfonate, IDS) or reductant (dithionite, DT) and measured at 4 K (or as indicated otherwise). The strong S = 3/2 EPR signal in DT-reduced Nif(DK)₂ (bottom trace, x0.1 in comparison to the other traces) was accompanied by a weak S = 1/2 EPR signal in the measurement with the perpendicular mode of a dual mode cavity. These signals disappear in the parallel mode. However, at 4 K a very sharp g = 15.6 signal of the S = 4 state of the P²⁺ cluster was detected, which increased several fold upon progressive IDS oxidation. The signal is from an excited state, as its intensity decreased with increased temperature (see trace at 20 K). b, Zoom in of (a), with the bottom trace on the same scale as the other traces to reveal a weak derivative-shaped g = 5.0 signal belonging to the middle g-value of the |±1/2〉 doublet in the S = 5/2 state of the P¹⁺ cluster. The corresponding broad g = 6.7 peak could not be detected. c, The S = 1/2 EPR signal at 20 K of the P¹⁺ cluster in 0.5 mM IDS-oxidized Nif(DK)₂, which could be simulated by a mixture of two species. Top traces: experimental spectrum (black line) with the sum of the simulated spectra (red line, parameters in Supplementary Table 3). Bottom traces: individual simulations in red and blue lines, together with their sum (black line). EPR conditions: microwave frequency, 9.63 GHz (perpendicular mode) or 9.34 GHz (parallel mode); modulation frequency, 100 kHz; 0.5 mT modulation amplitude, microwave power, 20 mW.

Extended Data Fig. 10 Phylogenetic relationship between nitrogenase and the nitrogen fixation-like protein families.

a, Maximum likelihood tree of Ni²⁺-sirohydrochlorin a,c-diamide reductive cyclase complex component D (CfbD), dark-operative protochlorophyllide a oxidoreductase (BchNB), chlorophyllide a oxidoreductase (BchYZ), nitrogenase-like proteins including the methylthio-alkane reductase (NflDK and MarDK) and Mo-dependent nitrogenase (NifDKEN) subunits. The tree was rooted with homomeric archaeal CfbD and constrained to have BchYN sister to NifDE, NflD, and MarD and to have BchBZ sister to NifKN, NflK and MarK. The mirroring topology of nitrogenases and nitrogenase-like proteins implies an ancestral gene duplication giving rise to ancestral nitrogenase-like proteins and a second, more recent duplication for the emergence of bona fide Mo-nitrogenases (NifDKEN). The ancestral nodes resurrected correspond to the last common ancestor of MarD and NifD and the last common ancestor of NifK and MarK. The number of sequences per clade is indicated. Supporting values represent the ultrafast bootstraps replicates by Iqtree2⁶⁸ and the approximate likelihood-ratio (aLRT)⁶⁹ statistical test, respectively. Scale bar represents the number of substitutions per site. b,c, Multiple sequence alignments (MSA) of the Rhodospirillum rubrum methylthio-alkane reductase MarK subunit to the A. vinelandii Mo-nitrogenase NifK subunit and the resurrected ancestral node sequence AncK, from both constrained and unconstrained trees (b) and R. rubrum MarD, A. vinelandii NifD and resurrected ancestral node sequence AncD, from both constrained and unconstrained trees (c), showing the binding sites for P-cluster, FeMoco and [Fe₈S₉C]-cluster. Pos. Prob., posterior probability. d, Structure of the AncDK dimer, from both constrained and unconstrained trees, predicted by AlphaFold2⁷⁰ and colored by predicted local distance difference (pLDDT) score. e, Simplified cladogram of the unconstrained tree topology. f, Schematic representation of the applied constrain, P indicates where polytomies were introduced. This topology was chosen for analysis due to the more parsimonious history in terms of gene duplication: all nitrogenase relatives other than CfbD are heterotetramers made from two paralogs. g, h, Close-up of the methylthio-alkane reductase P-cluster (g) and [Fe₈S₉C]-cluster (h) aligned to the predicted structure of the ancestor AncDK dimer. Amino acid residues coordinating the metalloclusters are shown in stick representation and colored in yellow, green, blue and red for MarD, MarK, AncD and AncK, respectively. An uncollapsed phylogeny with sequence identifiers, a separate file with identifiers, as well as the multiple sequence alignment are available as source data files.

Supplementary information

Supplementary Information

Supplementary Discussions 1–4, Tables 1–6 and references.

Reporting Summary

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Lago-Maciel, A., Soares, J.C., Zarzycki, J. et al. Methylthio-alkane reductases use nitrogenase metalloclusters for carbon–sulfur bond cleavage. Nat Catal 8, 1086–1099 (2025). https://doi.org/10.1038/s41929-025-01426-2

Download citation

Received: 10 October 2024
Accepted: 10 September 2025
Published: 23 October 2025
Version of record: 23 October 2025
Issue date: October 2025
DOI: https://doi.org/10.1038/s41929-025-01426-2

This article is cited by

Dynamics driving the precursor in NifEN scaffold during nitrogenase FeMo-cofactor assembly
- Lucía Payá Tormo
- Tu-Quynh Nguyen
- Yvain Nicolet
Nature Chemical Biology (2025)