Insights into the stereochemical mechanism of the B12-dependent radical SAM glutamine C- methyltransferase (QCMT) Authors: Thibaut Bourdin, Alain Guillot, Mickaël Mauger, Benjamin Lefranc, Sylvain Gervason, Matthieu Glousieau, Stéphane Grimaldi, Jérôme Leprince, Aurélien Thureau, Alhosna Benjdia, Olivier Berteau Published: 2026 ABSTRACT AbstractMethyl-coenzyme M reductase (MCR) is a crucial enzyme for methanogenesis and harbors several unusual post-translational modifications. Recent studies have identified glutamine C-methyltransferase (QCMT), as a B12-dependent radical SAM enzyme responsible for methylating a glutamine residue within the MCR active site. B12-dependent radical SAM enzymes have the remarkable ability to alkylate unactivated Csp2- and Csp3-atoms in a stereoselective manner. However, the factors influencing the stereo-selectivity and catalytic properties of this emerging superfamily of enzymes remain poorly understood. Here, we report the mechanistic, structural, and biochemical investigation of several QCMTs. Our findings reveal significant differences among them, notably in their ability to bind cobalamin. In addition, our data support that Cα H-atom abstraction and methyl transfer are not concerted but rather independent processes that require motion within the enzyme’s active site. We also demonstrate that QCMT catalyzes novel reactions, including the formation of unnatural C-methylated residues, peptide epimerization, reversible H-atom abstraction, and the direct conversion of glycine into D-alanine. Overall, our data are consistent with QCMT being a unique and versatile biocatalyst allowing for the installation of unnatural post-translational modifications and provide a structural and biochemical rationale for the control of the stereochemistry by B12-dependent radical SAM enzymes.Subject terms: Biocatalysis, Enzymes, Enzyme mechanisms Introduction Radical SAM (S-adenosylmethionine) enzymes, with over 800,000 distinct proteins identified, are considered the most versatile and widespread biocatalysts1–6. These enzymes, among other transformations3,7–9, catalyze a wide range of post-translational modifications, from thioether bond formation10,11 and radical activation9,12–14 to anaerobic oxidation15,16 and methylthiolation17–19. Radical SAM enzymes have evolved to incorporate additional domains, expanding their catalytic repertoire. Among them, with more than 220,000 members, the vitamin B12-dependent radical SAM enzymes1 built on a radical SAM and a vitamin B12 (cobalamin) binding domain, form the largest class1,5,20. These biocatalysts possess the remarkable ability to alkylate various unreactive Csp2- (e.g., TsrM21,22) and Csp3-hybridized atoms (e.g., TokK23, PoyC24, CysS25, Mmp105,26), in addition to performing several unrelated reactions1. Initially, it was proposed that these enzymes were built on an N-terminal B12-binding domain and a C-terminal radical SAM domain5,26,27. However, recent studies revealed a more complex picture1,5,26,28. While B12-dependent radical SAM enzymes can modify a broad range of substrates, from secondary metabolites to proteins, they are believed to use a conserved radical mechanism. Many of these enzymes invert the stereochemistry of the methylated carbon atom, in agreement with the unique architecture of their active site, which features both an activation and an alkylation center surrounding the substrate. However, some enzymes retain the stereochemistry of the substrate. For example, based on the structure of its product, (S)-2-methylglutamine (Me-Gln), QCMT, which methylates a glutamine residue within the active site of the alpha-subunit of a subset of methyl-coenzyme M reductases (MCR)26,29–31, the central enzyme in methane production in methanogenic archaea32–34, is proposed to retain the stereochemistry of the methylated carbon atom35–37. Interestingly, Me-Gln is not strictly conserved in MCR, whereas other post-translational modifications, such as Me-Arg (installed by the methanogenesis marker protein 10, Mmp1026,29), are universal26,29,38–41. The exact role of these post-translational modifications remains unclear, but they are believed to be important for optimal in vivo activity and methane metabolism32. Currently, approximately half of the experimentally investigated B12-dependent radical SAM enzymes such as MoeK42, GenD143, Fom344 and TokK45, have been shown to invert the stereochemistry of their substrates, while others like GenK46, Orf2947 and QCMT, retain it. In the last couple of years, several research groups have focused on QCMT, demonstrating its biological function and elucidating its crystallographic structure30,48. However, no rationale has been provided for how this enzyme controls the stereochemistry of its reaction. To gain further insights into this central question and into B12-dependent radical SAM enzymes in general, we undertook the biochemical and structural characterizations of several QCMT enzymes. Unexpectedly, we discovered that QCMT, owing to its unique mechanism, can catalyze reactions, including reversible hydrogen abstraction, peptide epimerization, and unnatural alkylation reactions, including the direct conversion of glycine into D-alanine. Our study demonstrates the unique potential of this enzyme for applications in synthetic biology, late-stage labeling, and peptide modification. In addition, structural investigation supports the notion that cobalamin binding induces a conformational change and that, following substrate activation, the transfer of the methyl group is likely an independent process, with the Cα hydrogen abstraction being a partially rate-determining step. Finally, we also illustrate the biocatalytic potential of QCMT by showcasing its ability to catalyze unnatural post-translational modifications in peptides, thereby expanding the chemical repertoire of radical SAM enzymes. Results QCMT Efficiently Methylates Short MCR Mimics To investigate the mechanism and substrate promiscuity of QCMT, we cloned several homologs from Methanoculleus thermophilus, Methanothermococcus thermolithotrophicus, Methanoculleus marisnigri, Methanothermococcus okinawensis and Methanoculleus bourgensis as strep-TAG fusion proteins and expressed them in E. coli. The most active and soluble ones, hereafter referred as QCMT1 and QCMT2, were the ones from M. thermophilus and M. thermolithotrophicus recently identified by the Layer Lab29,30 (see Supplementary Fig. S1). They were thus selected for further biochemical and structural characterization. Analysis by UV-visible spectroscopy and LC-MS revealed that while QCMT1 binds cobalamin as previously reported30, QCMT2 only marginally binds cobalamin (Supplementary Fig. S1). Further characterization of QCMT1, which can be isolated in both B12-bound and free forms, was performed by anaerobic size-exclusion chromatography coupled with small-angle X-ray Scattering (SEC-SAXS). As shown, the SEC-SAXS profile of QCMT1 indicated that it is a monomeric, homogeneous, globular protein, with a radius of gyration (Rg) of 24.35 ± 0.04 Å and an estimated molecular weight of 41 kDa (Fig. 1). Incubation with cobalamin resulted in a significant and reproducible decrease in the Rg (Rg = 24.03 ± 0.04 Å), indicating a conformational change in the protein. Conversely, the addition of SAM alone, or in the presence of the peptide substrate, did not cause any measurable structural changes. Overall, the SEC-SAXS analysis suggests that cobalamin induces small but significant conformational changes upon binding, leading to protein compaction, a change that was not observed in OxsB, the only B12-dependent radical SAM enzyme (albeit not a methyltransferase) structurally characterized in the absence of cobalamin49. SEC-SAXS analysis of QCMT1. SEC-SAXS analysis (a) in the absence or (b) in the presence of OHCbl. The UV–visible signal at λ = 280 nm (blue) and the SAXS intensity at zero angle (orange) are superimposed with the computed Rg from SAXS data (gray circles). The peak eluting at ~11 min corresponds to free OHCbl. SAXS curves of QCMT1 in the absence (c) or the presence (d) of OHCbl. Fit curves using PepsiSAXS75 are indicated by a red trace and χ2 value determined between the experimental data and the simulated model (Table 1). In previous studies30,48, QCMT1 was tested with a 24-mer peptide; however, the size and hydrophobicity of this substrate likely hindered the quantification of all the reaction products formed. Indeed, according to its proposed mechanism (Fig. 2a), QCMT produces both 5’-dA and SAH (resulting from radical cleavage and alkylation mediated by the SAM cofactor, respectively), as well as the methylated peptide/protein. To address this issue, we hypothesized that a shorter synthetic peptide (13-mer) framing the targeted glutamine residue would serve as a better substrate. Indeed, a 13-mer peptide has been shown to be sufficient to interact with Mmp10, the only B12-dependent radical SAM enzyme structurally characterized in complex with its peptide substrate26. Both QCMT1 and QCMT2 were reconstituted using standard protocols in order to obtain holoproteins with their intact [4Fe-4S] clusters. They were assayed under anaerobic conditions, in the presence of hydroxycobalamin (OHCbl), SAM, peptide 1Q (13-mer, [M + 3H]³⁺: 438.89), and titanium (III) citrate (Ti(III)), as a one-electron donor. As shown, we obtained efficient conversion of 1Q into a methylated product, CH3-1Q ([M + 3H]³⁺: 443.56). The location of the methyl group on Gln-7 (corresponding to Gln403 in the α-subunit of MCR from M. thermolithotrophicus) was unequivocally confirmed by LC-MS/MS (Fig. 2b, c, and Supplementary Fig. S2). QCMT biochemical assay on a short synthetic peptide. a Mechanistic hypothesis for QCMT alkylation reaction. One molecule of SAM is reductively cleaved for H-atom abstraction leading to the generation of 5′-dA, while a second molecule of SAM is used for MeCbl regeneration. R: Amino acid side chain (b) LC-MS analysis before (upper trace) and after (lower trace) 30 min incubation of peptide 1Q with QCMT1. c MS fragmentation of peptide 1Q (upper trace) and CH3-1Q (lower trace). Ions with a + 14.015 Da mass, corresponding to a net addition of a CH2, are noted by Δ. See Supplementary Fig. S2 and Tables S1 & S3 for full assignment. d Time course analysis of the reaction catalyzed by QCMT1&2: SAH (triangle symbols), 5´-dA (diamond symbols) and CH3-1Q (square symbols). Experiments were performed in triplicate. QCMT (7.5 µM) was incubated after iron-sulfur reconstitution with OHCbl (0.1 mM), SAM (0.5 mM) and 1Q (150 µM) (panels b and c) or with MeCbl (0.1 mM), SAM (0.8 mM) and 1Q (300 µM) (panel d), under anaerobic and reducing conditions. Reactions were performed at 50 °C and initiated by adding titanium (III) citrate (2 mM). To quantify all the products formed, we developed an LC-MS method and optimized the reaction conditions. Notably, we monitored the effects of reactants’ concentration and incubation temperature. Both enzymes showed enhanced activity at elevated temperatures with maximum activity at 70 °C and 50 °C for QCMT1 and QCMT2, respectively (Supplementary Fig. S3). However, at temperatures > 50 °C, significant 5′-dA and SAH degradation was evidenced, and the activity of QCMT2 was impaired. The temperature of 50°C was thus chosen for the quantification assays. Under these assay conditions, we determined specific activities of 167.2 ± 42.6 µmol·min⁻¹·mg⁻¹ for QCMT1 and 193.8 ± 2.7 µmol·min⁻¹·mg⁻¹ for QCMT2, for the formation of the methylated product (CH3-1Q) (Fig. 2d). With a kcat value of approximately ~0.7 min⁻¹ for methylated peptide, both enzymes exhibited an activity that was 10 to 20 times faster than the one previously reported30 and an apparent turnover of ~40 in 1 h (Fig. 2d). It should be noted that this turnover is likely underestimated, since a large portion of the substrate was converted by the end of the reaction. In contrast to recent studies30,48, we were also able to demonstrate that the formation of 5’-dA (i.e., 122.3 ± 13.1 µmol·min⁻¹·mg⁻¹ for QCMT1 and 102.6 ± 4.5 µmol·min⁻¹·mg⁻¹ for QCMT2) and SAH (139.1 ± 36.7 µmol·min⁻¹·mg⁻¹ for QCMT1 and 117.6 ± 4.2 µmol·min⁻¹·mg⁻¹ for QCMT2) paralleled the formation of CH3-1Q, consistent with the dual role of SAM as both a methyl donor and a radical initiator. As explained above, with the temperature used for the assay, 5’-dA and SAH quantifications were slightly underestimated due to their inherent instability in solution. QCMT alkylates the major groups of amino acid residues Having developed a robust biochemical assay, we were able to investigate the substrate promiscuity of QCMT. A previous study with Mmp10 demonstrated strict substrate specificity for this B12-dependent radical SAM enzyme26. In contrast, all peptides with a modified functionalized residue were used as substrates by QCMT1 and QCMT2, although the efficiency varied based on the nature of the side chain (Fig. 3a, b, Supplementary Figs. S4–S14, and Supplementary Table S1). Notably, the closest homolog to 1Q, the peptide 1N (containing an Asn residue in lieu of Gln) proved to be an efficient substrate, along with peptides substituted with short hydrophobic residues such as Ala and Gly (1A & 1G). More distantly related substrates with bulkier side chains (e.g., Arg in 1R), carboxyl groups (1E, 1D), or aromatic groups (1Y) were surprisingly processed, particularly by QCMT2, with conversion yields > 40%. MS/MS analysis confirmed that only the residue in position 7 was alkylated in each peptide assayed (Supplementary Figs. S6-S14). QCMT1 exhibited a similar activity profile to QCMT2 with negatively charged peptides, such as 1D, being poorly methylated. Overall, these data show that while QCMTs display remarkable substrate promiscuity toward the functionalized residue, the nature of the side chain and the proximity of the functional group to the Cα atom affect catalytic efficiency. Activity of QCMTs on a library of synthetic peptides. a Activity of QCMT1 and b QCMT2 on 13mer peptides with various amino acid substitutions at the seventh position (Q, N, P, R, E, D, Y, A, and G). Percentages are given based on conversion into the methylated (blue squares) or epimerized (orange triangles) forms. QCMT1 (7.5 µM) was incubated after iron-sulfur reconstitution with SAM (0.5 mM), peptide (150 µM), and OHCbl (0.1 mM) under anaerobic and reducing conditions for 30 min. Reactions were performed at 50 °C and initiated by adding titanium citrate (2 mM). Single experiments. (See Supplementary Figs. S4-S14and Supplementary Tables S1–S3for the full assignment). c Structures of the different methylated amino acid residues formed by QCMT1 and QCMT2 during catalysis on 13-mer peptides. Among all the peptides tested, the one containing a proline residue (1P) stood out. With this substrate, we obtained only trace amounts of methylated peptide ([M + 3H]3+: 433.23), suggesting that the constrained environment of the Cα-atom hinders enzyme activity (Fig. 3 and Supplementary Figs. S4, S5). However, this substrate was efficiently converted into another species with a molecular weight (1P*, [M + 3H]3+: 428.56) identical to that of the substrate. This product was formed only in the presence of SAM, OHCbl, and a one-electron donor (Ti(III)), indicating that QCMT catalysed a substrate rearrangement. To determine the nature of this modification, the modified peptide (1P*) was purified, and its amino acid content was analyzed after acid hydrolysis and derivatization with N-α-(2,4-dinitro-5-fluorophenyl)-l-valinamide (L-FDVA) (Fig. 4a). The analysis was in agreement with its expected composition, except for the presence of large amount of D-Pro, as shown by its molecular weight ([M + H]+: 396.15) and retention time compared to an authentic standard (Fig. 4a, lower panel). Further confirmation of the composition of 1P* was obtained by comparison with an authentic peptide standard. Amino acid content of the peptides modified by QCMT1. Analysis of the amino acid content of (a) 1P and (b) 1G after reaction with QCMT1. (upper traces: amino acid analysis, middle and lower traces: standards). Peptides were analyzed by acid hydrolysis, and derivatization with L-FDVA (see methods). Standards of derivatized L and D amino acids are indicated by (a) LA, DA, and (b) LP, DP (middle and lower panels). QCMT1 (7.5 µM) was incubated after iron-sulfur reconstitution with SAM (0.5 mM), 1P or 1G (300 µM), and OHCbl (0.1 mM) under anaerobic and reducing conditions at 50 °C in the presence of titanium citrate (2 mM). A close analysis of the reactions revealed that small amounts of epimerized peptides were formed by both enzymes in the presence of 1N, 1Y, and 1Q, mimicking the authentic substrate (Fig. 3a, b). These epimerized peptides were not further processed; notably, they were not alkylated, even when we performed fresh reactions with synthetic epimerized peptides. Interestingly, no epimerization occurred with 1A, despite this peptide being efficiently methylated (Fig. 3a). Collectively, these results demonstrate that QCMT, in addition to its methylation activity, catalyzes the epimerization of amino acid residues similarly to radical SAM peptide epimerases: EpeE (T-SPASM group)50, PoyD and OpsD (One [4Fe-4S] cluster group)51–54. However, unlike these enzymes, this activity was strictly dependent on the presence of SAM, [4Fe-4S] cluster, and cobalamin. In addition, while radical SAM epimerases are active on short hydrophobic amino acid residues (e.g., L-Leu, L-Ile, L-Val), QCMT has a unique substrate scope that includes amido- and aromatic amino acid residues. QCMT activity is independent of the presence of a side chain on the functionalized residue Another surprising finding emerged from the analysis of the reaction with 1G ([M + 3H]3+: 415.21), in which Gln-7 was substituted by a Gly residue. This peptide proved to be an effective substrate for both QCMT1&2, as evidenced by its rapid conversion into a methylated peptide (CH3-1G, [M + 3H]3+: 419.88) (Fig. 3 and Supplementary Figs. S4, S5). LC-MS analysis of CH3-1G confirmed that a methyl group was transferred to the seventh residue, as expected (Supplementary Fig. S14). Further investigations revealed that Gly-7 was converted to an Ala residue with D-configuration (Fig. 4b). The exclusive transformation of Gly into D-Ala underscores the strict control of stereochemistry exerted by QCMT, which catalyzes a reaction with a net retention of the configuration of the methylated carbon atom. This result also supports that QCMTs abstract the Gly pro-R hydrogen atom. Finally, similarly to peptides containing a D-Pro or D-Gln residue (i.e., 1P, 1Q), the D-Ala-containing peptide (CH3-1G) was not a substrate for QCMT, with notably no methylated product formed. We then questioned whether the size of the substrate affects the reaction outcome (methylation vs. epimerization) and aimed to identify the minimal peptide size recognized by QCMT. A library of peptides was generated by truncating the N- and C-terminal ends. As shown, reducing the substrate size to as little as 7 residues had dramatic consequences on the reaction outcome (Fig. 5). Notably, while the peptide 2Q remained an effective substrate, epimerization (2Q*, [M + 2H]2+: 361.68) was favored over methylation (CH3-2Q, [M + 2H]2+: 368.69) compared to 1Q (Fig. 5& Supplementary Figs. S15–S17). Here, the epimerized product (2Q*) accounted for approximately ~45%, while 1Q* represented < 5% (Fig. 2b) of the products formed. The nature of these products was confirmed through amino acid analysis and comparison with authentic standards. Like with larger substrates, QCMTs were unable to methylate a D-Gln (Fig. 5, synth-2Q*). Influence of the peptide size (2Q to 5Q) on QCMT1 activity. QCMT1 (7.5 µM) was incubated after iron-sulfur reconstitution with SAM (0.5 mM), peptide (150 µM), and OHCbl (0.1 mM) under anaerobic and reducing conditions. Reactions were performed at 50 °C for 2 hours and initiated by adding T(III) (2 mM). See Supplementary Figs. S15–S17 and Tables S1–S3 for full assignment. Synth-2Q*: synthetic peptide containing a D-Gln. Further truncation of the last C-terminal residue did not significantly change the reaction outcome with incubation of QCMT1 with peptide 3Q, resulting in the formation of an approximately 1:1 ratio of epimerized over methylated peptide. However, when the targeted glutamine residue was at the C-terminal end (i.e., 4Q), it was no longer a substrate. Similarly, removing the N-terminal phenylalanine residue from 2Q (resulting in 5Q, [M + 2H]2+: 288.15), resulted in a loss of activity. Collectively, these results demonstrate that QCMT requires only a few flanking residues for activity, establishing a minimalistic motif of six residues: FGGSQR. Importantly, substituting the N-terminal phenylalanine with an alanine or a glycine residue did not restore enzyme activity, indicating a critical role for this conserved residue, which interacts with the coenzyme F430 in MCR55,56. QCMT catalyzes Cα H-atom abstraction and reversible H-atom abstraction The epimerase activity of QCMT, although likely influenced by the in vitro conditions, is, to the best of our knowledge, novel for a B12-dependent radical enzyme. Known radical SAM peptide epimerases typically require either a single catalytic [4Fe-4S] cluster ([4Fe-4S]CAT) or one [4Fe-4S]CAT and one auxiliary cluster ([4Fe-4S]AUX) for catalysis53,57,58. These epimerases utilize a key cysteine residue to introduce a solvent-derived hydrogen atom, resulting in the formation of D-residues50,51,58,59. To investigate whether QCMT employs a similar mechanism, we first performed the reaction in a deuterated buffer. Under these conditions, the epimerized product 2Q# ([M + 2H]2+: 362.19) exhibited a + 1 Da mass increment on the Gln residue, demonstrating that QCMT introduced a solvent-derived hydrogen atom in the epimerized product, like radical SAM epimerases (Fig. 6a, b, and Supplementary Fig. S18). A similar result was obtained with 1P, leading to labeling of the proline residue (Supplementary Fig. S19). However, structural analysis, based on in silico models and the recently solved structure of MgmA48, does not support the use of an active-site cysteine residue to trap the radical intermediate. We therefore hypothesized that epimerization resulted from an uncontrolled reaction between the radical peptide intermediate and a buffer component. To validate this hypothesis, we performed the reaction under turnover-limiting conditions in a deuterated buffer. If epimerization is indeed the result of a free interaction between the radical intermediate and a solvent component, we would expect equal formation of deuterated D- and L-residues. QCMT catalyzes reversible Cα H-atom abstraction. a LC-MS analysis of the reaction on 2Q in deuterated buffer before (left panel) and after (right panel) 120 min incubation with QCMT1. b Isotopic distribution of peptides 2Q (left panel) and 2Q# (right panel) during incubation with QCMT1 in deuterated buffer at T0 or 30 min, respectively (upper traces) and 120 min (lower traces). c LC-MS analysis of labeled peptides 2A-d3 (deuterated alanine on the methyl group) and 2A-d4 (per-deuterated residue) before (upper trace) and after (lower trace) incubation with QCMT1. d MS analysis of the 5´-dA production with peptide 2A-d3 (left panel) and 2A-d4 (right panel). (e) Time course analysis of the reaction catalyzed by QCMT1 on peptides 2A-d3 and 2A-d4. Formation of the methylated peptides 2A-d3 and 2A-d4 is indicated by circle and square symbols, respectively. Experiments were performed in triplicate. QCMT1 (7.5 µM) was incubated after iron-sulfur reconstitution with SAM (0.5 mM), peptide (150 µM), and OHCbl (0.1 mM) under anaerobic and reducing conditions. Reactions were performed at 50 °C and initiated by adding Ti(III) (2 mM). Under our assay conditions, the epimerized peptide 2Q# was fully deuterated ([M + 2H]2+: 362.19), while other products of the reaction, including 5’-dA and SAH, were not labeled. Strikingly, the [M + 2H]2+: 362.19/361.68 ratio of 2Q increased over time, with up to > 66% deuterium incorporation at the end of the reaction (Fig. 6b & Supplementary Figs. S20, S21). Collectively, these findings support that the substrate radical intermediate reacted with a water-exchangeable buffer component, leading to uncontrolled hydrogen-atom addition. The radical intermediate, being likely planar, can react with equal probability on both faces leading to the formation of a labeled D- or L-residue. To unequivocally demonstrate that QCMT catalyzes Cα hydrogen atom abstraction, we replaced the targeted Gln residue with a deuterated Ala residue, as has been done with other radical SAM enzymes11,60,61. Based on 2Q, the shortest substrate for which an efficient activity was measured, we synthesized two additional peptides: one containing an Alanine with a deuterated side chain (2A-d3) and the other with a per-deuterated residue (2A-d4). Both peptides proved to be substrates; however, their conversion was significantly lower than that of 2Q (Fig. 6c). Incubation with 2A-d4 resulted in the formation of deuterated 5’-dA ([M + H]+: 253.11), whereas with 2A-d3, only unlabeled 5’-dA ([M + H]+: 252.11), was formed (Fig. 6d). These results unambiguously demonstrate that QCMT catalyzes Cα H-atom abstraction. Taken together, our data revealed three unexpected findings. First, unlike Mmp1026, for which the side chain is critical for peptide interaction and modification, QCMT activity is only marginally affected by the nature of the side chain. Second, by incorporating Cα-methylation on a wide range of residues, QCMT enables the formation of unprecedented and unnatural post-translational modifications, including CαMe-Asn, CαMe-Arg, CαMe-Asp, CαMe-Glu, CαMe-Tyr, and CαMe-Ala. Finally, QCMTs, by converting a glycine residue into D-Alanine, catalyze a reaction currently not accessible by synthetic chemistry. Activity of QCMT with deuterated peptides 2A-d3 and 2A-d4 As of now, the source of the hydrogen atom that is abstracted by B12-dependent radical SAM enzymes has been experimentally studied in five enzymes: GenK46, Fom344, CysS25, TokK45, and Orf2947. In the case of TokK, and to a lesser degree with Orf29, a mixture of labeled and unlabeled 5’-deoxyadenosine (5’-dA) has been obtained. In contrast, GenK and Fom3 predominantly yielded mono-deuterated 5’-dA, with an observed mass of 253.11 Da, like QCMT (Fig. 6d). For Orf29 and GenK, the isotopic distribution of the methylated products showed an unusual pattern characterized by significant deuterium enrichment. This phenomenon is thought to arise from a large kinetic isotope (KIE) and quantum tunneling effects62. Notably, GenK, the only B12-dependent radical SAM enzyme for which the influence of H-atom abstraction on the kinetic parameters has been investigated, displays a large KIE of 18 ± 0.6. We took advantage of 2A-d3 and 2A-d4 to determine whether QCMT exhibited a similar KIE. As shown, with these labelled substrates, we measured only a modest KIE of approximately 1.8, indicating that hydrogen-atom abstraction is a partially rate-determining step (Fig. 6e). Furthermore, the methylated product displayed the expected isotopic distribution without any deuterium atom scrambling, in contrast to other radical SAM enzymes8,46. Discussion Despite their discovery more than a decade ago, our understanding of B12-dependent radical SAM enzymes remains limited1,21,63, particularly regarding their unique ability to install methyl groups on unreactive sp3-hybridized carbon atoms, in a stereoselective manner. Recently, structures of several B12-dependent radical SAM enzymes have been solved1. Although these structures exhibit distinctive folds, they share notable similarities, particularly concerning the relative positions of the radical SAM [4Fe-4S] cluster and the cobalamin cofactor, which sandwich the enzyme’s substrates. Based on these structural insights, it has been proposed that B12-dependent radical SAM enzymes utilize a stepwise “pull-push” radical transfer mechanism, where 5’-dA “pulls” the hydrogen atom from one side of the substrate and MeCbl “pushes” a methyl radical from the opposite side64. In this mechanism, hydrogen-atom abstraction is predicted to be the rate-limiting step, while methyl transfer is exothermic. We demonstrate here that QCMT uses SAM for both radical initiation and methyl transfer, as expected for a B12-dependent radical SAM methyltransferase catalyzing Csp3-alkylation. However, these two reactions can be dissociated, leading to hydrogen atom exchange and epimerization when the alkylation reaction is impaired. Unexpectedly, we found that QCMT exhibits exceptional substrate promiscuity, catalyzing post-translational modifications across all major classes of amino acid residues, including short hydrophobic residues, amido- and aromatic amino acid residues. This provides access to unnatural post-translational modifications, including the direct conversion of glycine residue into D-Ala, an unprecedented reaction. D-Ala is well-known in bacterial physiology and is typically formed by a two-step reaction involving serine dehydration and reduction65. To the best of our knowledge, QCMT is the only biocatalyst capable of converting Gly to D-Ala, making it a unique biochemical tool for engineering peptides and proteins by introducing unnatural post-translational modifications, for example, within short peptide tags. It also opens novel strategies for directly engineering MCR and for assessing the influence of various post-translational modifications on their activity. Moreover, the hydrogen atom exchange and epimerase activity of QCMT is, to the best of our knowledge, novel, opening the possibility of incorporating hydrogen atoms in late-stage synthesis and to epimerize aromatic and amido-amino acid residues. Our study also reveals critical insights into the mechanism of B12-dependent radical SAM enzymes. We show that QCMT abstracts the Cα hydrogen atom, which is supported by labeling experiments and epimerization activity. The abstraction of a Cα hydrogen atom is a common theme across several classes of radical SAM enzymes, including glycyl radical activating enzymes (GRE-AE)66,67, thioether bond synthases11,68, and peptide epimerases50,58. All these enzymes are presumed to generate a Cα peptide radical intermediate, stabilized by the captodative effect. Although the presence of a side chain theoretically influences the planarity and stability of these Cα peptide radicals, recent studies indicate that peptide radicals can be unexpectedly stable9, consistent with QCMT’s hydrogen-atom exchange activity. Our study also demonstrates that QCMT only abstracts the pro-R H-atom, leading exclusively to the formation of D-Ala from Gly. This strict stereospecificity explains why QCMT, despite being largely insensitive to the amino acid side chain’s nature, is inactive toward D-amino acid residues (e.g., D-Gln or D-Ala in synth-2Q* and CH3-1G, respectively). Surprisingly, using deuterated substrates, we observed only a modest kinetic isotope effect, despite recent computational studies suggesting that hydrogen-atom abstraction should be the rate-limiting step64. Our screening of QCMT homologues revealed that, even within the same enzyme family, some members, such as QCMT1, efficiently bind cobalamin, as shown by UV-visible and SEC-SAX analyses, whereas others, such as QCMT2, bind it reversibly. It remains to be clarified whether this observation is due to in vitro conditions or if it has physiological relevance. Indeed, we showed that in solution, cobalamin binding influences the overall shape of the protein, resulting in a more compact and closed conformation. At the mechanistic level, our data indicate that hydrogen-atom abstraction and methyl-transfer reactions are not concerted. We propose that the radical peptide intermediate reorients during catalysis to attack the methyl group in a stereo-controlled, but not concerted, manner. The observation that the amino acid side chain plays a minor role in the reaction rate suggests that there are limited constraints within the active site. This contrasts with Mmp10, where a dense network of interactions stabilizes the substrate’s side chain within the active site26. In conclusion, we have demonstrated that while QCMT catalyzes a strictly controlled stereo- and regio-selective reaction, it also exhibits unexpected catalytic promiscuity, leading to peptide epimerization, hydrogen atom exchange, and alkylation (Fig. 7). QCMT is a promising radical biocatalyst6,69 for accessing peptide structures with unexpected reactivity. Proposed mechanisms for QCMT leading to methylation, epimerization or reversible H-atom abstraction. QCMT catalytic promiscuity: Following Cα H-atom abstraction, a carbon-centered intermediate is formed that either reacts with MeCbl or a solvent-derived H-atom, leading to methylation, labelling (reversible H-atom abstraction), and epimerization reaction. In the absence of a side-chain (R), methylation leads to the conversion of Gly into D-Ala. Materials and Methods Cloning of QCMT homologs The genes of QCMT from: Methanoculleus thermophilus (QCMT1): WP_066957658, Methanothermococcus thermolithotrophicus (QCMT2): WP_018153439, Methanoculleus marisnigri: A0A117MFC8, Methanothermococcus okinawensis: WP_013866670 and Methanoculleus bourgensis: WP_074174763, were produced by Genecust with codon optimization for Escherichia coli expression. Genes were cloned into pET28a(+) plasmid with a N-terminal Strep-tag sequence. Plasmids were then transformed into E. coli BL21 Star (DE3) cells (Life Technologies) for protein expression. Expression and protein purification E. coli cells transformed with the corresponding plasmids were grown in LB medium supplemented with Ampicillin (0.1 ug.mL−1). Cultures were incubated at 37 °C until OD600 reached 0.8 ( ~ 4 h, 160 rpm), then 0.3 mM of ammonium iron(II) citrate was added and protein expression was induced by adding 0.5 mM isopropyl-β-d-1-thiogalactopyranoside (IPTG). The culture was further incubated (16 h, 21 °C, 160 rpm) before cells are harvested by centrifugation (10 min, 4 °C, 5500 x g). The cell pellet was washed with cold buffer (250 mL, 50 mM Tris, 300 mM NaCl, 10% glycerol pH 8) then frozen in liquid nitrogen and stored at −80 °C until further use. Pellet cell was resuspended (300 mL for 9 L of culture volume) in lysis buffer (50 mM Tris, 300 mM NaCl, 10% glycerol pH 8) containing complete ULTRA EDTA-free protease inhibitor tablets (Roche), 0.5% Triton X-100, and 10 mM β-mercaptoethanol. Lysozyme was then added and the suspension (1 mg.mL-1) was incubated for 30 min at room temperature before cells were lysed by sonication at 60% amplitude (8 × 45 s with 60 s cooling intervals on ice). The lysate was ultracentrifuged (45 min, 4 °C, 150,000 × g), and the supernatant was kept on ice. All subsequent steps were carried out at 4 °C. Purification of the crude extract, containing the recombinant protein QCMT1 or QCMT2 was purified by affinity chromatography on Strep-Tactin® Superflow® high-capacity resin (IBA-Lifesciences) at a flow rate of 1 mL min−1 equilibrated with lysis buffer. The crude extract was applied to the Strep-tag® column and the resin was then washed with 100 mL of lysis buffer. The Strep-tagged protein was eluted (1 mL.min−1) with D-desthiobiotin (3 mM) and dithiothreitol (3 mM) in lysis buffer. Fractions containing the protein are pooled and concentrated on a 30 kDa cutoff filter and subsequently frozen in liquid nitrogen and stored at −80 °C for further use. The purity of the QCMT1 or QCMT2 proteins was verified by SDS-PAGE electrophoresis. Concentrations of protein were estimated by using spectrophotometer (QCMT1: Mw = 60868.31 Da, ε = 87320 M−1 cm−1; QCMT2: Mw = 60804.19 Da, ε = 87320 M−1 cm−1). Reconstitution of QCMTs All protein reconstitutions and experiments were performed under anaerobic conditions, in a glovebox. Reconstitutions were performed by adding 3 mM DTT, followed by a six-fold molar excess of (NH4)2Fe(SO4)2 and Na2S. Precipitated iron sulfides were removed by using a desalting column (PD 10 Desalting Columns, Cytiva) equilibrated with buffer (50 mM Tris, 300 mM NaCl, 10% glycerol pH 8). The protein was concentrated by ultrafiltration (30 kDa MWCO). Peptide synthesis Peptides 1Q to 5Q (see Table S1) were synthetized by Proteogenix. Peptides 1A-d3 and 1A-d4 were synthetized by PRIMACEN platform, Université de Rouen Normandie France. Long peptides (1Q to 1G) were solubilized in a 0.05% Formic acid solution. Short peptides (2Q to 1A-d4) were solubilized in a 50 mM Tris, 300 mM NaCl pH 8 buffer. The pH of each short peptide solution was measured and adjusted to 7–8. Peptide synthesis of 1A-d3 and 1A-d4 were performed by Fmoc solid phase methodology on a Liberty microwave assisted automated peptide synthesizer (CEM) using the standard manufacturer procedures at 0.1 mmol scale on preloaded Fmoc-Ala-Wang resin. The synthetic peptides were deprotected, cleaved from the resin with a mixture of TFA/TIS/H2O (9.5:0.25:0.25) and precipitated in tert-butylmethyl ether, and then purified by reversed-phase (RP) HPLC (Gilson) on a 22 × 250 mm Vydac 218T1022 C18 (10 µm, 300 Å) column (Grace) using a linear gradient (5–30% over 45 min) of acetonitrile/TFA (99.9:0.1) at 10 mL.min−1. Characterization of the purified peptides was then achieved by MALDI-TOF mass spectrometry on a UltrafleXtreme (Bruker) in the reflector mode using α-cyano-4-hydroxycinnamic acid as a matrix. Analytical RP-HPLC, using a 4.6 × 250 mm Vydac 218T54 C18 (5 µm, 300 Å) column, indicated that a peptide purity > 99.9%. Enzymatic assay All assays were performed under anaerobic conditions in a glovebox and in the dark, at 50 °C. Activity assays were done using freshly prepared reconstituted protein (7.5 μM). All reagents were solubilized in water in the glovebox just before using them. Protein was incubated first with 100 µM of OH-cobalamin at 4 °C, then mixed successively with 100 mM Tris pH8, 1 mM DTT, 500 μM SAM (Sigma-Aldrich), 150 μM peptide, and 2 mM titanium(III) citrate (final concentrations). A sample was collected at T = 0 min before the addition of titanium(III) citrate, and a second at T = 30 min (long peptides) or T = 120 min (short peptides). For enzyme kinetics, the concentration of SAM and peptide was increased to 800 µM and 150 µM, respectively, in order to measure linear activity over a longer period of time. Samples were collected at 5 min, 10 min, 20 min, 30 min and 60 min. All samples were quenched by a 10-fold dilution in 0.3% TFA. LC–MS analysis Peptides modification characterization was done by liquid chromatography on a Vanquish UHPLC systems coupled to a Q Exactive Focus mass spectrometer (Thermo Fisher Scientific). Quenched samples, after 15 times dilution in formic acid 0.1% (buffer A), were injected (2 µL) on a reversed-phase Phenyl-Hexyl column (2.1 mm × 50 mm, 1.8 µM, Agilent). Optimization of peptide analysis was performed in order to have the best separation possible, especially between peptides and epimerized peptides. All compounds were eluted at a flow rate of 0.3 mL.min−1 with a gradient from 0% to 18% (large peptides) or from 0% to 12% (short peptides) of buffer B (acetonitrile 80% and formic acid 0.1%). To characterize the post-translational modifications of each peptide, fragmentation was achieved on the most abundant ion (Tables S2, S3). Collision energie of 22% was used for peptide 1Q to 1G, 20% for peptides 2Q to 1A-d4 and 15% for peptides 3Q to 5Q. LC-MS spectras and Isotopic distributions were generated using the Qual Browser software (Thermo Xcalibur, v2.0.3). All compounds were analyzed by mass spectrometry in positive mod with a mass range between 200 and 1000 uma. LC-MS/MS spectra were manually deconvoluted and annotated using the FreeStyle software with the Xtract tool (ThermoFisher, v1.3). When signal intensity was too low for the deconvolution, Qual Browser software was used to generate the spectra (Figs. S15 and S17). HPLC analysis Quantification of compounds from in vitro assays was performed with an Agilent Technologies 1200 series HPLC using calibration curves. Aliquots (10 µL) of standards and quenched samples were injected onto a LiChrospher® 100 RP-18 endcapped column (250 ×4 mm). Elution was performed using an acetonitrile gradient (18.7–60% solvent B) at 1 mL.min−1. Solvent A was 0.1% TFA; solvent B was 80% acetonitrile + 0.1% TFA. UV detection was at 215 nm for peptides and 260 nm for SAH/5’-dA. Peptide purification and L-FDVA derivatization To determine the amino acid composition of peptides 1P, 1G and 2Q, large-scale reactions (500 µL) were performed. Enzymes were removed by adding 50 µL of 10% TFA, then diluting with 450 µL of 0.1% TFA. After 10 min incubation on ice (between each addition of TFA) and centrifugation, peptides were purified on SepPak® C18 cartridges and eluted with 80% acetonitrile + 0.1% TFA. Eluates were dried in borosilicate glass tubes (Kimble, ASTM type 1) for 1 h in a Savant SPD121P SpeedVac. Hydrolysis was performed by adding 1 mL of 6 M HCl, applying partial vacuum ( < 1 mbar), and incubating at 150 °C for 1 h. Hydrolysates were derivatized with 10 µL NaHCO3 1 M, 10 µL L-FDVA (10 µg µL−1 in acetonitrile), 10 µL acetonitrile, and 10 µL water, incubated at 42 °C for 1 h. Derivatized samples were analyzed by LC–MS using a Hypersil GOLD aQ column (2.1 × 200 mm, 1.9 µm). Compounds were eluted between 30 and 70% buffer B at 0.3 mL min−1. Glutamine is converted to glutamic acid during acid hydrolysis; therefore, L- and D-glutamic acid contents were assessed using synthetic peptide Synth-2Q containing D-glutamine. UV-visible spectroscopy The UV-Visible spectra were recorded with a protein concentration of 1 mg mL−1 in a quartz cuvette with an optical path length of 1 cm under anaerobic conditions on a Jasco V-760 spectrophotometer using a scan rate of 40 nm min−1. UV–visible spectra (215–700 nm) were recorded before and after protein reconstitution and incubation with 200 μM OH-cobalamin. Excess OH-cobalamin was removed using a PD-10 column (Cytiva). SEC-SAXS analysis Reconstituted and tag-cleaved QCMT1 was purified by size exclusion on a Superdex 200 Increase 10/300 GL column (GE Healthcare) using 50 mM Tris, 300 mM NaCl, 10% glycerol pH 8 buffer containing 1 mM DTT to prevent aggregation. Protein samples ( ≈ 100 µM and 20 µM respectively) were incubated with or without 200 µM OHCbl and analyzed by SEC–SAXS using an in-line Superdex 200 Increase 5/150 GL column (GE Healthcare). Samples were loaded anaerobically onto an HPLC system equilibrated with degassed buffer (50 mM Tris, 300 mM NaCl, 5% glycerol pH 8). SAXS data were collected on the SWING beamline at Synchrotron SOLEIL (Saint-Aubin, France). Acquisition and data-processing parameters are summarized in Table 1. Table 1Report of biomolecular structural modelling of small-angle scattering data of QCMT1a) Sample details.QCMT1QCMT1+ OHCobOrganismMethanoculleus thermophilusUniProt sequence ID (residues in construct)WP_066957658Extinction coefficient [A280, 0.1%(w/v)]0.6750.675Partial specific volume, ν (cm3/g)0.7320.732M from chemical composition (kDa)50.351.6SEC–SAXS column5 × 150 mm Superdex increase S200 Loading concentration (mg mL−1)5 Injection volume (µL)50 Flow rate (mL.min−1)0.3Solvent (solvent blanks taken from SEC flowthrough prior to elution of protein)Tris 50 mM, NaCl 300 mM, Glycerol 5% DTT 1 mMb) SAXS data-collection parameters.Instrument/data processingBioSAXS on the SWING beamline at Synchrotron SOLEIL70Wavelength (Å)1.0332Beam size (µm)500 × 200Camera length (m)2.00q measurement range (Å−1)0.0041–0.5516Absolute scaling methodComparison with scattering from 1 mm pure H2ONormalizationTo transmitted intensity by beam-stop counterMonitoring for radiation damagedata frame-by-frame comparisonExposure timeContinuous 1 s data-frame measurements of SEC elutionSample configurationSEC–SAXS with thermalized quartz capillary (ID 1.5 mm),Nitrogen gas Buffer and AutoSampler saturatedSample temperature (°C)20(c) Software employed for SAXS data reduction, analysis and interpretation.SAXS data reductionI(q) versus q, buffer subtraction & frames selection using Foxtrot 3.10 (https://www.synchrotron-soleil.fr/en/beamlines/swing#paragraphes_menu_left-block-7)Extinction coefficient estimateProtParam (Gasteiger et al., 2005)30Basic analysis: Guinier, P(r), MWRAW 2.2.1 (Hopkins, 2024, Meisburger et al., 2021, Rambo et al. 2013, Svergun, 1992)71–74Three-dimensional graphic model representationsPyMOL v.1.70.0.5 Win64(d) Structural parameters.QCMT1QCMT1+ OHCobGuinier analysis I(0) (cm−1)0.030.03 R g (Å)25.4925.00 qR g max (q min = 0.0066 Å−1)1.11.11 Coefficient of correlation, R 21.01.0 MW from Vc (kDa)47.246.1P(r) analysis I(0) (cm−1)0.030.03 Rg (Å)25.6425.25 dmax (Å)100100 q range (Å−1)0.01-0.560.01-0.56total estimate from GNOM0.8120.814(f) Atomistic modelling.Crystal structuresApo AlphaFold2.0 model aligned with TokK (PDB:7KDX)q range for all modelling0.007–0.49PepsiSAXS75 χ21.001.10(g) SASBDB IDs for data and models.DVY7DVZ7 Report of biomolecular structural modelling of small-angle scattering data of QCMT1 Statistics and Reproducibility All data shown are mean values of at least three independent biological replicates, with error bars denoting the standard deviation, unless indicated otherwise in the figure caption. Calculations were performed in Microsoft Office Excel with built-in shortcuts to process information such as the mean and standard deviation. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Data Reporting Summary Acknowledgements We acknowledge SOLEIL (Saint-Aubin, France) for the provision of synchrotron radiation facilities (proposal 20221418), and we would like to thank the SWING staff for assistance in using the beamline. This work was supported by ANR (Grants: ANR-20-CE44-0005, ANR-21-CE11-0030, ANR-23-CE07-0046). Author contributions Conceptualization: A.B., O.B. Methodology: A.G., B.L., J.L., A. T., A.B., O.B. Investigation: T.B., A.G., M.M., B.L., S.G., M.G., S.G., J.L., A. T., A.B., O.B. Supervision: A.B., O.B. Writing-original draft: T.B., O.B. Writing-review and editing: T.B., A.G., M.M., B.L., S.G., M.G., S.G., J.L., A. T., A.B., O.B. Peer review Peer review information Communications Biology thanks Fumitaka Kudo, Tristan Wagner, and the other, anonymous, reviewer for their contribution to the peer review of this work. Primary Handling Editors: Fumihiro Ishikawa and Laura Rodríguez Pérez. Data availability All data generated or analyzed during this study are included in the text and figures of this article and its Supplementary Materials. The sequence alignments and SDS-PAGE analysis are provided as Supplementary Figs 1 & 22, and the MS analysis as Supplementary Figs 2–21. The source data for all figures is provided as Supplementary Data (see the list in the Supplementary Materials file). Uncropped and unedited SDS-PAGE is shown as Supplementary Fig. S22. The MS data have been deposited in the MetaboLights repository with the study identifier MTBLS14287. All other data are available from the corresponding author upon reasonable request. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Supplementary information The online version contains supplementary material available at 10.1038/s42003-026-10234-8. References Associated Data Supplementary Materials Supplementary Information Description of Additional Supplementary Files Supplementary Data Reporting Summary Data Availability Statement All data generated or analyzed during this study are included in the text and figures of this article and its Supplementary Materials. The sequence alignments and SDS-PAGE analysis are provided as Supplementary Figs 1 & 22, and the MS analysis as Supplementary Figs 2–21. The source data for all figures is provided as Supplementary Data (see the list in the Supplementary Materials file). Uncropped and unedited SDS-PAGE is shown as Supplementary Fig. S22. The MS data have been deposited in the MetaboLights repository with the study identifier MTBLS14287. All other data are available from the corresponding author upon reasonable request.