Methane production by anaerobic methane-oxidizing archaea is responsible for two-thirds of global methane emissions1, a large part of which originates from marine sediments1 and the mammalian microbiome2,10. In this process, methyl-coenzyme M reductase (MCR) has a central role by catalysing the reversible interconversion of 2-methylmercaptoethanesulfonate (CoM) and 7-thioheptanoylthreoninephosphate (CoB) to a CoB–CoM heterodisulfide and methane (Fig. 1a). The structure of MCR3 has revealed several distinct features for this 300-kDa (αβγ)2 protein complex, such as a unique F430 cofactor11 and unusual post-translational modifications5, including 5-C-(S)-methylarginine4,5, which tunes the reactivity of its active site6,12. Mmp10, which has been shown to catalyse this key post-translational modification7,8, belongs to an emerging superfamily of B12-dependent radical SAM enzymes13,14,15,16,17,18,19,20,21 that encompasses more than 200,000 proteins ( These enzymes are involved in the biosynthesis of myriad natural products including bacteriochlorophyll and antibiotics9,16,18,23 and catalyse  various reactions such as methyl transfer to sp2- and sp3-hybridized carbon atoms13,14,18,24, P-methylation25, ring contraction and cyclization reactions26,27. However, despite the numerous biochemical and spectroscopic studies available in the literature13,14,15,16,17,18,19,20,21, knowledge of these biological catalysts remains limited. Notably, only two structures of B12-dependent radical SAM enzymes have been solved thus far; however, these studies present some limitations precluding a deep understanding of their catalysis24,27. In addition, no structure of a B12-dependent radical SAM enzyme catalysing methyl transfer to an sp3-hybridized carbon atom has been reported so far, even though these enzymes are the only known biological catalysts capable of such transformation and this reaction is by far the most widespread in this enzyme family. Interestingly, these latter enzymes also have the remarkable property of being able to make dual use of SAM to initiate radical chemistry and to catalyse nucleophilic displacement, which remains poorly understood.

Fig. 1: MCR and Mmp10 activity with overall structure of Mmp10.
figure 1

a, The activity of MCR producing CoB–CoM heterodisulfide and methane is enhanced by the post-translational modification of R285 catalysed by the B12-dependent radical SAM enzyme Mmp10. b, Overall structure of Mmp10 with bound sodium, [4Fe–4S] cluster, MTA, MeCbl and single iron atom cofactors (Protein Data Bank (PDB) accession 7QBT). Although Mmp10 was crystallized with SAM, only electron density for MTA was observed (Extended Data Fig. 1, Extended Data Table 1). c, Magnified view of the [4Fe–4S] cluster coordinated by three cysteine residues and Y115 alongside the modelled MTA molecule not coordinated to the cluster (Extended Data Fig. 1b). d, Iron loop with a single iron atom coordinated by four cysteine residues (Extended Data Fig. 1c). Light blue, radical SAM domain; teal, cobalamin-binding domain; purple, iron loop; green, MTA; magenta, MeCbl. The [4Fe–4S] cluster is shown as yellow and orange spheres; the single iron is presented as an orange sphere; and the sodium atom is shown as a violet sphere. Omit maps (blue mesh) of the [4Fe–4S] cluster, its coordinated Y115 and the uncoordinated MTA (c) or single iron atom (d) contoured at 3σ are depicted.

Mmp10 has a unique architecture

The structure of holo-Mmp10 was solved at an atomic resolution of 1.9 Å, and electron density was obtained for the 411 residues of the protein (Extended Data Table 1). In a unique manner, Mmp10 is composed of two domains and an iron loop (Fig. 1b, Extended Data Fig. 1a). The first domain (residues 1–257) has an unusual radical SAM triosephosphate isomerase (TIM) barrel motif (β7α6). The radical SAM [4Fe–4S] cluster28,29,30,31 is coordinated by three cysteine residues (C15, C19 and C22) and, in contrast to all known radical SAM enzymes, by a strictly conserved tyrosine residue (Y115) located between the third β-strand and third α-helix of the TIM barrel (Fig. 1c, Extended Data Figs. 1b, 2). Thus far, FeS cluster coordination by a tyrosine residue has been reported for only the [FeFe]-hydrogenase maturase HydE32 and one nitrogenase33. With Y115 coordinating the radical SAM cluster, SAM is not able to bind the ‘unique iron’. However, we observed an electron density for the adenine moiety of SAM at its expected location near the top β-barrel sheet (Fig. 1b, Extended Data Fig. 1b) and coordinated by the protein backbone through hydrogen bonds. Notably, despite the atomic resolution, we could not resolve the methionine moiety of SAM, suggesting some flexibility for this cofactor, which was hence modelled as S-methyl-5′-thioadenosine (MTA) in this structure. This flexibility probably occurs because SAM is not coordinated to the [4Fe–4S] cluster and no polar interaction occurs with the glycine-rich motif (GGD), which usually binds the amino group of the methionine moiety of SAM29. Co-crystallization of Mmp10 with S-adenosyl-l-homocysteine (SAH) resulted in electron density consistent with the presence of the full SAH cofactor. However, among the five canonical radical SAM motifs, only a few contacts were observed, with no direct interaction between SAH and the GGD and ribose motifs. In these different structures, the adenine moiety of SAH or MTA is close to the radical SAM cluster (3.8 Å from the nearest ion of the cluster to the adenine moiety), whereas the [4Fe–4S] cluster and cobalamin (vitamin B12) are separated by 12 Å (from the nearest ion of the cluster to the cobalt atom).

The second unique feature of Mmp10 is the presence of a loop coordinating mononuclear iron (Fig. 1d, Extended Data Fig. 1c), which is similar to a rubredoxin iron loop34 although with a distinct orientation. This mononuclear iron is coordinated by a unique and conserved cysteine motif (C35, C38, C45 and C48) within the radical SAM domain (Extended Data Fig. 2). Although the presence of additional FeS clusters is common in non-B12-dependent radical SAM enzymes, no radical SAM enzyme has been shown to contain a mononuclear centre. Furthermore, auxiliary FeS clusters are often located in the C-terminal region and outside the TIM barrel domain, with the notable exception of BioB35, which is built on a complete TIM barrel.

Roles of iron sites in Mmp10

To investigate the properties of the radical SAM cluster and iron loop, we generated an A3 mutant lacking the three cysteine residues from the radical SAM cluster and an A4 mutant lacking the four cysteine residues from the iron loop, along with several mutants with individual alanine substitutions. The activity of these mutants was compared to that of wild-type protein using a synthetic peptide substrate ([M + H]+: 1,496.77) mimicking MCR8 and containing R285 (numbered as in MCR), which is the target of the methylation reaction. Mmp10 efficiently transferred a methyl group to R285 in the presence of SAM and Ti(III) citrate (Extended Data Fig. 3a–c), as shown by a mass shift of Δm = +14 Da ([M + H]+: 1,510.79). One molecule each of 5′-deoxyadenosine (5′-dA) and SAH were produced per methylation reaction (Extended Data Fig. 3b), whereas no SAM cleavage was noted in the absence of peptide, regardless of the reductant used. As expected, the A3 mutant was inactive and unable to cleave SAM (Extended Data Fig. 3d). Mutants of the iron loop (A4 and C38A mutants) produced small amounts of 5′-dA but were unable to transfer a methyl group to the substrate (Extended Data Fig. 3e, f). Finally, when we abrogated the coordination of Y115 to the [4Fe–4S] cluster (Y115A mutant), the cleavage activity towards SAM was severely impaired and the methyltransferase activity was abolished (Extended Data Fig. 3g). Substitution of Y115 with phenylalanine only marginally restored enzyme activity (<1% of wild-type activity), which demonstrates the critical involvement of the hydroxyl group of Y115 in polar interactions following substrate binding (Extended Data Fig. 3h).

Electron paramagnetic resonance (EPR) analysis of Mmp10 revealed unique spectroscopic signatures (Extended Data Fig. 3i). First, oxidized Mmp10 exhibited a strong high-spin signal (S  = 5/2) at g = 4.30, 4.14 and 9.4, which is characteristic of a mononuclear Fe3+ ion, and showed a signal for [3Fe–4S]+ at g = 2.0. The latter signal corresponds to the oxidized radical SAM cluster, whereas the high-spin S = 5/2 signal is unusual and mirrors those reported in oxidized rubredoxin. After FeS reconstitution and reduction, signals at g = 2.03, 1.93 and 1.88 were noted, which correspond to the radical SAM [4Fe–4S]+ cluster (Extended Data Fig. 3i). In the low-field region, signals at g = 5.4 and 3.1 are characteristic of spin systems of S = 3/2 and are fully consistent with a [4Fe–4S] cluster coordinated by three cysteine residues and a non-cysteine ligation36. The reduced A4 mutant exhibited an EPR spectrum similar to that of the wild-type enzyme although with an altered signal of S = 3/2, leading to the appearance of a signal at g = 1.15 (Extended Data Fig. 3e). By contrast, mutation of the three cysteine residues from the radical SAM motif abrogated signal for the S = 3/2 species (Extended Data Fig. 3d). Finally, the addition of SAM to reduced Mmp10 led to a change in the EPR spectrum, with the development of additional signals at g = 1.89 and 1.80. This result is consistent with direct interaction between SAM and the [4Fe–4S] cluster37 (Extended Data Fig. 3i). Collectively, these data support the idea that both the S = 1/2 and S = 3/2 spin systems originate from the radical SAM [4Fe–4S] cluster and are consistent with the existence in solution of free and Y115-bound forms.

A distinct B12-binding domain

The B12-binding domain (158 residues) is formed by four β-strands and seven α-helices (Figs. 12). This domain comprises most of the polar bonds that hold the cobalamin dimethylbenzimidazole (DMB) tail in place. By contrast, a network of interactions from mainly the iron loop (Y23, F24 and Y47) and the penultimate loop of the radical SAM domain (R210, N217, I220, L221 and N223) coordinate the side chains of the tetrapyrrole ring. Owing to the low number of β-strands and α-helices and an absence of the polar α-helix involved in phosphate binding, this domain is only marginally related to a Rossmann fold. Furthermore, none of the canonical B12-binding motifs such as the His-on (DXHXXG) and SXL motifs38 were identified. A molecule of SAM (or SAH) and Y23 are found between the [4Fe–4S] cluster and cobalamin (Fig. 2a, Extended Data Fig. 4a–e). Y23 interacts with the tetrapyrrole C8 side chain and F24 through a π–π interaction (Fig. 2a), and its hydroxyl group is 4.8 Å from the cobalt atom (upper axial coordination), suggesting a role for Y23 in tuning cobalt reactivity. Although several charged residues have been reported to serve as a lower axial ligand in B12-binding enzymes either directly or through water contact, Mmp10 has a hydrophobic residue (L322) in the lower axial position of cobalamin (Fig. 2a). Because it lacks a lone pair, this residue cannot coordinate the cobalt atom. Its role is hence probably to maintain the pentacoordination of the cobalt centre and to prevent water molecules from interacting with the cobalt atom. In support of this conclusion, despite the high resolution of the structure, we observed no water molecules beneath the cobalamin cofactor, which is shielded by a hydrophobic pocket (Extended Data Fig. 4g, h). This novel binding mode is probably responsible for the atypical planar geometry of the tetrapyrrole ring (Extended Data Fig. 4f).

Fig. 2: Binding of vitamin B12 and S-adenosyl ligands by Mmp10.
figure 2

a, The C8 side chain of MeCbl is shown in interaction with Y23, I220, L221 and N223 within the radical SAM domain, resulting in a planar tetrapyrrole ring. MeCbl has no lower axial ligand because it is pentacoordinated; however, L322, which resides at 3.9 Å from the cobalt atom, is part of a loop of residues forming a hydrophobic environment for the cobalt ion. Y23 appears at 4.8 Å from the cobalt ion. b,Snapshots of S-adenosyl cofactors within distinct Mmp10 structures. The distances between the sulfur atom of SAH and MTA or SAM and the cobalt ion are indicated by dashed lines. Top left, Mmp10 crystallized with SAH in the absence of peptide substrate (1: Mmp10 SAH structure; PDB 7QBV). Bottom left and top right, Mmp10 crystallized in the absence of substrate with SAM. Only the density of MTA was observed, which is labelled accordingly (2: Mmp10 MTA_1; PDB 7QBT; 3: Mmp10 MTA_2; PDB 7QBU). Bottom right, Mmp10 crystallized with SAM and its peptide substrate (4: Mmp10–SAM–peptide structure; PDB 7QBS). Light blue and purple, radical SAM domain residues; teal, cobalamin domain; green, SAM, MTA and SAH; magenta, MeCbl. The [4Fe–4S] cluster is shown as orange and yellow spheres. Omit maps (blue mesh) of ligands contoured at 3σ are depicted. See Extended Data Table 1 and Extended Data Fig. 4 for additional information.

Motion of SAM in the active site

No notable overall structural change was observed when Mmp10 was co-crystallized with the demethylated SAM product, with a root mean squared deviation (r.m.s.d.) of 0.37 Å over 408 residues; SAM adenine binding remained mostly unaffected. However, the distance between the sulfur atom of SAM (MTA) and the cobalt atom of cobalamin was shortened from 8.9 Å to 5.6 Å in an alternative SAM orientation (Fig. 2b, Extended Data Fig. 4). The methionine moiety is hence free to move and rotate to a distance compatible with direct methyl transfer from SAM to the cobalt atom. Following substrate binding, marked changes occurred, with displacement of Y115 coordination from the [4Fe–4S] cluster by the carboxylate and amino groups of SAM (Fig. 2b). The distance between the sulfur atom of SAM and the cobalt atom increased to 9.4 Å, whereas that between the sulfur of SAM and the [4Fe–4S] cluster shortened to 3.4 Å. These results demonstrate that SAM can adopt various conformations within the active site. Unexpectedly, coordination of the [4Fe–4S] cluster by Y115 enables the enzyme to discriminate between radical and nucleophilic uses of SAM, without requiring two SAM-binding sites.

Peptide binding and recognition

Co-crystallization of Mmp10 with its substrate revealed clear electron density for eight residues, including R285, the target of the modification (Fig. 3a). Following substrate binding, Mmp10 adopted a closed conformation involving displacement of the α1a-helix by 11.6 Å and the α1- to α4-helices of the radical SAM TIM barrel by as much as 3.4 Å (Extended Data Fig. 5a, b). In addition to coordination of the methionine moiety of SAM to the cluster, numerous polar contacts were established involving the ribose and GGD motifs and additional interaction between Mmp10 and cobalamin. Unexpectedly, the C2, C7 and C18 side chains of cobalamin established multiple polar interactions with the peptide backbone (Extended Data Fig. 5c). At the entrance of the active site, the peptide backbone formed a sharp twist assisted by two conserved proline residues and a complex network of polar interactions between charged amino acid side chains and the enzyme backbone (D6, Y56, E54 and G87) (Extended Data Fig. 5c). R285 exhibited an extended side chain that protruded into the enzyme active site, and its Cδ atom is at the perfect distance (3.7 Å) and orientation with respect to the C5′ atom of SAM, for direct hydrogen atom abstraction (Fig. 3b). The 4.2-Å distance between the Cδ atom and the methyl group of cobalamin is also perfectly compatible with direct transfer of the methyl group from methylcobalamin (MeCbl) to the Cδ atom. Notably, the guanidinium moiety of R285 was coordinated not only by polar interaction with the protein (E378) and water contacts but also by the SAM cofactor itself through the adenine and ribose moieties (Extended Data Fig. 5d, e). Finally, Y115 became coordinated via hydrogen bonding to E378, enabling SAM to interact with the fourth iron atom of the [4Fe–4S] cluster and  radical chemistry to take place.

Fig. 3: Structure of Mmp10 in complex with its peptide substrate.
figure 3

a, Overview of Mmp10 in complex with peptide substrate, shown in orange (PDB code 7QBS). b, Close-up of the Mmp10 active site showing SAM in green, peptide substrate in orange and cobalamin in magenta. The distance between the C5′ atom of SAM and the Cδ atom of the arginine peptide substrate R285 (3.7 Å) and that between R285 and the methyl group of MeCbl (4.1 Å) are indicated by dashed lines. The omit map (blue mesh) of peptide is contoured at 3σ. c, Sequences of the peptides used as potential substrates with the substitution of arginine (R285) with isoleucine (Ile), leucine (Leu), ornithine (Orn), lysine (Lys), citrulline (Cit) or homoarginine (HArg). d, UV–visible light analysis of Mmp10 pre-incubated with OHCbl. Green line, OHCbl–Mmp10; blue line, OHCbl–Mmp10 after incubation with Ti(III) citrate; red line, reduced OHCbl–Mmp10 exposed to air. See Extended Data Figs. 46 for additional information.

Enzyme specificity

Mmp10 introduces only a single modification in MCR, which suggests a strict specificity contrary to that exhibited by enzymes installing multiple post-translational modifications in ribosomally synthesized and post-translationally modified proteins9,23,39,40,41,42,43. To investigate the substrate promiscuity of Mmp10, we substituted R285 with hydrophobic residues (isoleucine or leucine) or structural analogues (Fig. 3c). None of these peptides were modified by Mmp10 despite having a Cδ atom in the target side chain (Fig. 3c, Extended Data Fig. 6a,  Extended Data Table 2), including a citrulline derivative that differed from the wild-type peptide by only one atom. Competition experiments provided additional support that this analogue does not interact with Mmp10 (Extended Data Fig. 6b), which is consistent with the importance of the guanidinium moiety for interaction with E378 (Extended Data Fig. 5). In addition, E378 coordinated with Y115, which supports the idea that substrate binding acts as a switch for cluster availability. Collectively, a complex network of interactions involving water molecules and SAM, along with protein dynamics, controls the strict specificity of this enzyme. Finally, UV–visible light analysis of the hydroxycobalamin (OHCbl)–enzyme complex showed that, following Ti(III) citrate treatment, a Co(I) intermediate is formed (Fig. 3d), providing a route for MeCbl regeneration.

Proposed mechanism for Mmp10 catalysis

Interaction with the substrate, likely assisted by reduction of the [4Fe–4S] cluster, has a major role in displacement of Y115 from the radical SAM [4Fe–4S] cluster, which enables direct coordination of SAM (Fig. 4). After SAM cleavage, the formed 5′-dA radical abstracts the Cδ hydrogen atom of R285, which is at a perfect distance for direct interaction with the methyl group of MeCbl, and induces methyl transfer to R285. Then, Y115 reverts to coordination of the radical SAM [4Fe–4S] cluster, which prevents binding of a novel SAM molecule. The Co(II) intermediate generated during catalysis must be further reduced to produce the super-nucleophile Co(I) for reaction with a second molecule of SAM and to regenerate MeCbl.  Interestingly, in the absence of a strong reductant, Mmp10 can convert OHCbl into MeCbl, similar to what has been observed for TsrM13, albeit with lower efficiency (Extended Data Fig. 6c). Although the iron loop is ideally located and exposed to solvent, it is unlikely to have a redox potential in the range of the base-off Co(I)–Co(II) redox couple44, even though the potential of cobalamin can largely be influenced by the protein matrix45 (Extended Data Fig. 7a). At present, we favour the involvement of a ferredoxin in the reduction of Co(II).

Fig. 4: Proposed mechanism for Mmp10.
figure 4

Nucleophilic and radical catalysis are highlighted in blue and orange, respectively.

A cation modelled as sodium is present and interacts with the residues holding Y115 in place. Its presence appears to be essential for preventing major backbone reorganization during Y115 motion, as mutation of D156, which makes key interactions in the cation-binding site, severely impairs enzyme activity (Extended Data Fig. 7b, c). The presence of a cation is reminiscent of PFL-AE46 and QueE47. However, in Mmp10, the cation is not located in the active site. Finally, four cis peptide bonds are present in the structure, including a rare non-proline cis peptide bond (Extended Data Fig. 8). These bonds are critical for the interface between the radical SAM and cobalamin-binding domains and are necessary for strict control of catalysis.


The structure of Mmp10, the first, to our knowledge, B12-dependent radical SAM enzyme catalysing protein post-translational modification, reveals the mechanism of action of these enzymes in transferring methyl groups to sp3-hybridized carbon atoms. Structural and spectroscopic analyses showed that Mmp10 contains four metallic centres at interaction distances (12–16 Å) (Extended Data Fig. 8c). We establish that Mmp10 has a C-terminal B12-binding domain, although residues from the whole protein are involved in binding of the B12 cofactor. In addition, this study reports the structure of a B12-dependent radical SAM enzyme in complex with its substrate properly positioned in the active site. This structure provides critical information about the structural and functional diversity of radical SAM enzymes as well as the mechanism of these complex biocatalysts that use both a radical and an SN2 mechanism. Major and unprecedented active site reorganization occurred following substrate binding. EPR spectroscopy and crystallographic snapshots establish that the radical SAM cluster can be transiently coordinated by a tyrosine residue, which enables the enzyme to perform either radical or nucleophilic chemistry. Recently, the structure of the B12-dependent methyltransferase TsrM13,14,24 was solved and shown to contain a [4Fe–4S] cluster coordinated by three cysteines and one glutamate residue24, a coordination encountered in several FeS proteins48,49 and proposed to preclude radical catalysis. Our study demonstrates that such coordination in Mmp10 is an intermediate state enabling dual use of the SAM cofactor.

Mmp10 contains a unique iron loop positioned beneath the B12 cofactor that is probably involved in shuttling electrons from the cobalt atom. In addition, a hydrophobic pocket prevents water from converting the pentacoordinated MeCbl into its more stabilized hexacoordinated counterpart, which is a strategy conserved in other B12-dependent radical SAM enzymes. These enzymes thus appear to have evolved unique structures and mechanisms to alkylate sp2- and sp3-hybridized carbon atoms using the twin catalytic power of the cobalamin and SAM cofactors8,13,14,16,17,18,19,21,50,51. In contrast to catalysis by known radical SAM enzymes, catalysis by Mmp10 requires active site reorganization and SAM flexibility within the active site. Although Mmp10 has a unique architecture among known enzymes, the role of such structural rearrangements has probably been underestimated in radical SAM enzymes, with the current work thereby delineating novel catalytic territories.


Protein purification

The gene for Mmp10 was commercially synthesized with codon optimization for Escherichia coli expression. The Mmp10 gene was cloned into pET28-a and was transformed into E. coli BL21 Star (DE3) cells (Life Technologies) alongside a pRSF plasmid expressing the ISC system. Mutants were generated by gene synthesis or site-directed mutagenesis. Cells were cultured in LB with ampicillin (0.1 mg ml−1) and kanamycin (0.05 mg ml−1). Cultures were incubated at 310 K until the OD600 reached 0.7, at which point (NH4)2Fe(SO4)2 and 0.5 mM isopropyl-β-d-1-thiogalactopyranoside (IPTG) were added to the medium. The cultures were then cooled to 291 K and were incubated for 16 h. Mmp10 was purified by affinity chromatography in 50 mM Tris (pH 8), 400 mM NaCl and 3 mM DTT and was concentrated to ~10 mg ml−1.

Reconstitution of Mmp10 and mutants

Reconstitution and all further sample preparation and experiments were performed in a glovebox in the absence of oxygen. The samples were reconstituted overnight with eightfold molar excess of (NH4)2Fe(SO4)2 and Na2S with 3 mM DTT for all experiments unless stated otherwise. Once reconstituted, the samples were buffer-exchanged using PD-10 columns into 50 mM Tris (pH 8), 400 mM NaCl and 1 mM DTT. For crystallography, EPR and UV–visible light analyses, Mmp10 was purified by size exclusion chromatography on a Superdex 200 Increase 10/300GL column using an AKTA Pure system. Samples reconstituted with hydroxocobalamin had tenfold molar excess added to the Mmp10 and were incubated overnight before being passed through a PD-10 column.


The crystallization conditions for holo-native Mmp10 were identified anaerobically at 294 K. Initial crystals appeared after 24 h by using sitting drop diffusion and a 1:1 mixing of protein (10 mg ml−1 with 2 mM SAM and 200 µM MeCbl) and precipitant solutions (100 mM Tris pH 8, 20% polyethylene glycol (PEG) 8000). Holo-native Mmp10 crystals (SAH binding and alternate SAM conformation) were obtained under similar conditions. Holo-native Mmp10 peptide substrate-binding crystals were optimized using sitting drop vapour diffusion with a 1:1:1 ratio of holo-protein solution with 2 mM substrate peptide (EMLPARRARGPNE) to precipitant solution. The crystals were cryoprotected using 10% PEG 400. All were harvested anaerobically and cryocooled in liquid nitrogen.

Crystallographic structure determination

Diffraction data were collected on the PROXIMA-1 beamline at the synchrotron SOLEIL (Saint-Aubin, France)52. A crystal of holo-native Mmp10 with peptide substrate (Mmp10–SAM–peptide structure) belonging to the space group P63 was detected, and diffraction data were collected to 2.4 Å with phases obtained through multiwavelength anomalous diffraction (MAD). High-energy remote data were collected using an X-ray wavelength of 0.97857 Å and were scaled with a dataset collected at the iron absorption peak at 1.72200 Å. Diffraction images were recorded using an EIGER-X 16M detector, processed with XDS using the XDSME package53,54 and corrected for anisotropy using STARANISO55. The experimental phasing searched for one FeS cluster site, treated as a super-atom, and one cobalt site using SHARP/AutoSHARP56. At this stage, another unexpected separate iron site was found and was included in the phasing. Substructure determination was performed in SHELXC/D57 with heavy atom refinement, phasing and completion performed using SHARP58 and density modification using SOLOMON59. The model was built using several rounds of automated building with Buccaneer60. The final round of model building used ARP/WARP61, and manual building was performed within Coot62 with refinement by Refmac5 63 and BUSTER64. The final model included the full-length sequence of the protein with one molecule per asymmetric unit. Subsequent data were phased by molecular replacement using PHASER65 with this model and with subsequent manual rebuilding and refinement as described above. A holo-native Mmp10 crystal (crystallized with SAM) with the space group P21212 diffracted to a resolution of 1.9 Å with four molecules per asymmetric unit (Mmp10 MTA_1 structure). Another crystal (alternate MTA conformation) with the space group P212121 diffracting to a resolution of 2.3 Å with two molecules per asymmetric unit was also solved (Mmp10 MTA_2 structure). Finally, a further holo-native Mmp10 crystal (crystallized with SAH) with the space group P21212 diffracted to a resolution of 2.7 Å and had four molecules per asymmetric unit (Mmp10 SAH structure). Data collection and refinement information can be found in Extended Data Table 1. PyMOL (version 2.0) was used in data analysis and figure generation.

Enzymatic assay with purified enzyme

All reactions were performed anaerobically in the dark. Mmp10 reactions (100–150 µM Mmp10, 3 mM DTT, 200 µM MeCbl, 2 mM SAM, 1 mM peptide substrate, 2 mM Ti(III) citrate) were incubated at 298 K for up to 2 h and were analysed by liquid chromatography–mass spectrometry (LC–MS).

LC–MS analysis

LC–MS analysis was performed using an ultra-high-performance liquid chromatography (UHPLC) instrument (Vanquish Flex, Thermo Scientific) connected by an HESI2 ion source to the MS instrument (Q-Exactive Focus, Thermo Scientific). Samples were diluted 50-fold in buffer A with 2 µl injected onto the column. A reverse-phase column (2.1 mm × 50 mm, 1.7 µm; Eclipse Plus C18, Agilent Technologies) was used for separation. To enhance the retention and resolution of the column, we used heptafluorobutyric acid (HBFA) as an ion-pairing agent with acetonitrile used for buffer B. All compounds eluted between 0% and 50% buffer B during 20 min at a flow rate of 0.3 ml min−1. Buffer A contained 0.2% HBFA in milliQ water; buffer B contained 0.2% HBFA in acetonitrile/MilliQ water at a ratio of 80/20.

EPR spectroscopy

EPR spectra were recorded on a Bruker ElexSys-500 X-band spectrometer equipped with a standard rectangular cavity (ST4102) fitted to an Oxford Instruments liquid helium cryostat (ESR900) and temperature control system. Measurements were conducted at 6 K using a 600-mT or 800-mT field sweep range or at 15 K using a 200-mT field sweep range to focus on the g = 2.0 species with a field modulation amplitude of 1 mT at 100 kHz, microwave power of 10 mW and microwave frequency of ~9.48 GHz.

Reporting summary

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