Identification and characterization of a bacterial core methionine synthase

Methionine synthases are essential enzymes for amino acid and methyl group metabolism in all domains of life. Here, we describe a putatively anciently derived type of methionine synthase yet unknown in bacteria, here referred to as core-MetE. The enzyme appears to represent a minimal MetE form and transfers methyl groups from methylcobalamin instead of methyl-tetrahydrofolate to homocysteine. Accordingly, it does not possess the tetrahydrofolate binding domain described for canonical bacterial MetE proteins. In Dehalococcoides mccartyi strain CBDB1, an obligate anaerobic, mesophilic, slowly growing organohalide-respiring bacterium, it is encoded by the locus cbdbA481. In line with the observation to not accept methyl groups from methyl-tetrahydrofolate, all known genomes of bacteria of the class Dehalococcoidia lack metF encoding for methylene-tetrahydrofolate reductase synthesizing methyl-tetrahydrofolate, but all contain a core-metE gene. We heterologously expressed core-MetECBDB in E. coli and purified the 38 kDa protein. Core-MetECBDB exhibited Michaelis-Menten kinetics with respect to methylcob(III)alamin (KM ≈ 240 µM) and L-homocysteine (KM ≈ 50 µM). Only methylcob(III)alamin was found to be active as methyl donor with a kcat ≈ 60 s−1. Core-MetECBDB did not functionally complement metE-deficient E. coli strain DH5α (ΔmetE::kan) suggesting that core-MetECBDB and the canonical MetE enzyme from E. coli have different enzymatic specificities also in vivo. Core-MetE appears to be similar to a MetE-ancestor evolved before LUCA (last universal common ancestor) using methylated cobalamins as methyl donor whereas the canonical MetE consists of a tandem repeat and might have evolved by duplication of the core-MetE and diversification of the N-terminal part to a tetrahydrofolate-binding domain.

amino acid residues [17][18][19] . We here refer to the canonical E. coli-type MetE as 'tandem-repeat MetE' (tr-MetE). The active site of tr-MetE is located within the C-terminal part, where the zinc ion is coordinated by one histidine, two cysteine and one glutamate residue. The binding site for methyl-THF is in the cleft between the two domains of tr-MetE 19 .
Dehalococcoides mccartyi strain CBDB1 is an obligately anaerobic, mesophilic bacterium belonging to the phylum Chloroflexi, class Dehalococcoidia 20,21 . Dehalococcoides species are well known for their ability to use a wide range of persistent and toxic halogenated organic compounds as terminal electron acceptor in anaerobic respiration ("organohalide respiration") with hydrogen as electron donor 22 . Strain CBDB1 encodes one of the largest numbers of B 12 -dependent proteins in known prokaryotes 23 . The most prominent representatives of B 12 -dependent proteins in strain CBDB1 are reductive dehalogenases that are responsible for the reduction of halogenated pollutants as a terminal electron acceptor 24 . Vitamin B 12 in the medium is essential for the growth of Dehalococcoides strains 25 because Dehalococcoides do not contain many of the genes for de novo biosynthesis of cobalamin 26,27 . However, Dehalococcoides strains encode corrinoid-specific ABC-transporter and enzymes for the late B 12 biosynthesis pathway enabling them to incorporate corrinoid precursors from the environment and to modify them to cobalamin 28,29 . Although none of the known D. mccartyi genomes contains full gene homologs of metE, metH, bhmt or bhmt-2 26 , D. mccartyi strains synthesize methionine de novo 24,30,31 . Zhuang et al. demonstrated that acetyl-CoA donates the C 2 -methyl group for an unconventional methionine biosynthesis pathway independent from methylene-tetrahydrofolate reductase (MTHFR). All D. mccartyi species sequenced so far, lack metF encoding for MTHFR that reduces 5,10-methylene-THF-Glu to 5-methyl-THF-Glu. Furthermore, D. mccartyi strain 195 was not able to incorporate 5-methyl-THF from the environment 32 .
Here, we found that locus cbdbA481 (NCBI accession number CAI82680) of D. mccartyi strain CBDB1 encodes a 343 amino acid protein that is homologous to the C-terminus of canonical tr-MetE and similar to methylcobalamin:homocysteine methyltransferase (core-MetE MTH ) of the methanogenic archaeum Methanobacterium thermoautotrophicum 33 . In our study we provide biochemical and genetic evidence that the locus cbdbA481 encodes a novel type of bacterial methionine synthase that appears to be an anciently derived MetE-related methionine synthase obtaining its methyl group from an external corrinoid rather than from folate. Together with archaeal methylcobalamin:homocysteine methyltransferases and bacterial homologs the gene product of locus cbdbA481 forms a new group of basal methionine synthases, referred to as core-MetE in the following.

Results
Bioinformatic analysis of locus cbdbA481 in the genome of D. mccartyi strain CBDB1. D. mccartyi strains are able to synthesize methionine de novo, although D. mccartyi genomes do not contain gene homologs of metE, metH, bhmt or bhmt-2 26 . In the KEGG (Kyoto Encyclopedia of Genes and Genomes) database several enzymes of the Dehalococcoides methionine metabolism are annotated [34][35][36] . The loci cbdbA476 and cbdbA477 in the genome of strain CBDB1 are annotated as SAM synthetase (EC 2.5.1.6) and S-adenosyl-Lhomocysteinase (EC 3.3.1.1), respectively. Since methionine biosynthesis and SAM metabolism are biochemically closely linked, we started our search for a methionine synthase gene in the genome of strain CBDB1 by inspecting the direct neighborhood of the loci cbdbA476 and cbdbA477. Locus cbdbA481 located in the same operon encodes a 343 amino acid polypeptide (here referred to as core-MetE CBDB ) with a calculated molecular mass of 38 kDa that has sequence similarity with the C-terminal half of canonical tandem-repeat MetE (tr-MetE) proteins. Accordingly, the calculated mass of core-MetE CBDB is only about half the size of tr-MetE proteins (e.g. tr-MetE Eco of E. coli). Phylogenetic analysis of core-MetE CBDB together with 38 other protein sequences including annotated tr-MetE representatives, archaeal methylcobalamin:homocysteine methyltransferase homologs (core-Me-tE Archaea ) and Chloroflexi sequences with high sequence similarity to core-MetE CBDB indicates that core-MetE CBDB is the prototype of a new bacterial cluster of short MetE sequences (Fig. 1a). This new cluster is phylogenetically well separated from the C-termini of tr-MetE proteins and core-MetE Archaea (Fig. 1a, blue colors) 33 . Core-MetE proteins are widely distributed in microorganisms with strongly conserved ancient traits (archaea, Clostridiales, Dehalococcoidia and Chloroflexia classes). Within the archaea, only Haloquadratum spp. encode the tr-MetE, whereas many archaea encode a core-MetE homolog (Fig. 1a, blue colors). Bacterial and archaeal core-MetE form a paraphyletic group (excluding the tr-MetE sequences), probably evolved before LUCA (last universal common ancestor) and branched into two groups. Tr-MetEs appear to have evolved from archaeal core-MetE (e.g. core-MetE in Clostridium kluyveri and C. oryzae) (Fig. 1a).
The computational structural model of core-MetE CBDB (Fig. 1b, green) resembles the C-terminal domain of annotated tr-MetE proteins, bearing the highest similarity to methionine synthase from Neurospora crassa (C-score = −0.27, identity = 21% and RMSD = 3.52) (PDB No. 4ZTX) 37 . Compared to tr-MetE proteins, core-MetE CBDB lacks the N-terminal domain described to be responsible for 5-methyl-THF binding and the linker region between the C-and N-terminal parts (Fig. 1b, grey). An amino acid alignment of core-MetE CBDB with core-MetE Dehly from Dehalogenimonas lykanthroporepellens, core-MetE MMKA from Methanococcus maripaludis KA1 and the C-terminal halves of the tr-MetE ECDH from E. coli DH10B and tr-MetE Sau from Staphylococcus aureus shows that the zinc binding motif HXCX n C, essential for L-homocysteine binding and activation 13 , is conserved in core-MetE CBDB (Fig. 1c). The computational model indicates that in strain CBDB1, zinc is coordinated in a tetrahedral fashion by His215, Cys217 and Cys312 (Fig. 1c) which are conserved among all MetEs and by Asp236. In contrast, zinc of tr-MetE from N. crassa is bound by histidine, two cysteine and one glutamate residue.
Heterologous production and purification of core-MetE CBDB . To study the function of core-MetE CBDB in detail, the recombinant protein was heterologously produced in E. coli and purified. First, production and purification attempts were conducted for a C-terminally Streptavidin-tagged core-MetE CBDB using affinity chromatography for purification. However, native polyacrylamide gel electrophoresis (PAGE) indicated misfolding donor and homocysteine as methyl acceptor. In the presence of core-MetE CBDB , the UV/Vis absorption spectrum of methylcob(III)alamin, exhibiting a characteristic maximum at 524 nm, successively changed over time due to the consumption of methylcob(III)alamin and formation of cob(I)alamin and cob(II)alamin, as indicated by the emergence of absorption features at 681 nm and 474 nm, respectively (Fig. 3a). In the absence of core-MetE CBDB or homocysteine, the UV/Vis spectrum of methylcob(III)alamin remained unchanged (Fig. 3b,c). In order to exclude any methyltransferase activity due to impurities of the protein preparation, E. coli cell-free extract was also tested and did not show any activity (Fig. 3d). Finally, in addition to the photometric measurements, the formation of methionine ([M + H]+ = 150.0583 m/z) during the enzymatic reaction was verified via liquid chromatography-mass spectrometry (LC-MS) (Supplementary Figure 2b). In the following, core-MetE CBDB enzyme activity was monitored by measuring the increase of absorption at 681 nm (Fig. 3e,f) or the decrease of absorption at 524 nm (Supplementary Figure 2a). Kinetic parameters for core-MetE CBDB were determined using an enzyme concentration of 0.1 µM. At a constant D,L-homocysteine concentration of 2 mM and varying methylcob(III)alamin concentrations, methionine was formed with a V max = 1664 ± 50 nkat mg −1 and a K M = 236 ± 3 µM for methylcob(III)alamin (Fig. 3e). When different D,L-homocysteine concentrations were used at a fixed methylcob(III)alamin concentration of 0.5 mM, a V max = 1582 ± 11 nkat mg −1 and a K M = 98 ± 0 µM for D,L-homocysteine were estimated (Fig. 3f). Since methionine synthase is specific for L-homocysteine, the apparent K M for L-homocysteine might be half of that for D,L-homocysteine 33 . The maximum turnover number (k cat ) was calculated to be about 60 s −1 . The substrate specificity of core-MetE CBDB was investigated by replacing homocysteine with 2 mM cysteine, 2 mM glutathione or 2 mM dithiothreitol. Core-MetE CBDB did not show any activity towards these thiol analogs (data not shown).
Additionally, 5-methyl-THF-Glu 3 was tested as a methyl group donor for core-MetE CBDB instead of methylcobalamin (Fig. 4b). In the negative control and also in the presence of core-MetE CBDB , slow demethylation of 5 methyl-THF-Glu 3 occurred abiotically [39][40][41] . The demethylation of 5-methyl-THF-Glu 3 in the negative control and in the presence of core-MetE CBDB was not linked to L methionine formation ( Fig. 4b(I)), while in the presence of tr-MetE Eco , methionine was formed exhibiting a signal at [M + H] + = 150.0583 m/z (Fig. 4b(II)). www.nature.com/scientificreports www.nature.com/scientificreports/ pH-optimum and thermal stability of core-MetE CBDB . Methionine synthase activity of core-MetE CBDB was observed between pH 5.0 and 9.0, with an optimum between pH 6 and 6.5 ( Table 1). The thermal stability of purified tr-MetE Eco and core-MetE CBDB were assessed by recording protein melting curves using nano differential scanning fluorimetry (nanoDSF). For tr-MetE Eco , a melting temperature T m = 55.8 ± 0.2 °C was determined. In contrast, the T m of core-MetE CBDB was at 68.8 ± 0.0 °C (Supplementary Figure 3).

Core-MetE CBDB does not complement tr-metE-deficient E. coli in vivo.
The examination of enzymatic activities of core-MetE CBDB in vitro has limitations. However, we were not able to conduct in vivo mutagenesis studies with strain CBDB1 because Dehalococcoides species are not yet genetically accessible. In order to obtain insights into the physiological role of the cbdbA481 gene product in vivo, we tested whether a tr-metE-deficient E. coli strain could be complemented by core-MetE CBDB . Therefore, we generated a tr-metE-deficient knockout strain of E. coli DH5α (ΔmetE::kan) that still contained the metH gene for the cobalamin-dependent MetH. This strain was not able to grow in medium without added cyanocobalamin (Supplementary Figure 4, red solid line), but grew when cyanocobalamin was supplemented (Supplementary Figure 4, red dotted line). Next, the growth behavior of the mutant strain carrying different complementation plasmids was investigated. Either the original tr-metE Eco gene or the core-MetE CBDB nucleotide sequence, both under the control of an arabinose promotor, were provided. Growth experiments with these complementation strains showed that neither of the two strains grew without inducing gene expression by arabinose. After induction with arabinose, tr-metE Eco was able to complement the

Discussion
Methionine and SAM have been suggested to belong to the most ancient molecules on earth and might have emerged within or even before the "RNA world" [42][43][44] . Although methionine appears to have a continued central metabolic role for more than three billion years, different routes for its biosynthesis have evolved. The biochemically conserved methionine pathway appears to be the product of an evolutionary patchwork involving diverse methionine synthases 5 . In our study, we identified a novel bacterial MetE-like methionine synthase in D. mccartyi strain CBDB1 that uses methylcobalamin as methyl donor instead of methylated tetrahydrofolate. Our results suggest that this enzyme is the basal form of canonical tandem-repeat MetE (tr-MetE) proteins with roughly half its size and without the domain duplication of canonical MetE proteins evolved to enable tetrahydrofolate binding  19 . Homologs of this short methionine synthase are encoded in the genomes of several deeply-rooting obligate anaerobic microorganisms from both prokaryotic domains, including all Dehalococcoidia and many Clostridia (e.g. Desulfitobacterium metallireducens, C. kluyveri, C. oryzae) as well as almost all archaea sequenced so far (Fig. 1a). We refer to this short monomeric MetE form as "core-MetE", because several lines of  www.nature.com/scientificreports www.nature.com/scientificreports/ evidence hint at its basal descendence including the lack of duplication, the exclusive presence in deeply rooting phylogenetic taxa, and the dependence on corrinoids, which are thought to be ancient cofactors 45,46 , as they participate in fundamental processes such as ribonucleotide reduction 47 , the Wood-Ljungdahl-pathway 48 and methane formation 49 .
Compared with tr-MetE Eco , core-MetE CBDB is more stable towards pH changes 18 and thermal denaturation. The turnover number k cat ≈ 60 s −1 of core-MetE CBDB is very high in comparison to other methionine synthases such as tr-MetE Eco with 0.4 s −1 50 or E. coli MetH with 26 s −1 51 . The activity of MetH is based on domain movements, which could contribute to the lower catalytic rate in comparison to a small monomeric core-MetE. The relatively slow conversion rate of tr-MetE proteins can be due to the poor methylation power of 5-methyl-THF-Glu n (Fig. 4a) and the weak nucleophilicity of homocysteine at physiological pH 52 . In tr-MetE and MetH, 5-methyl-THF must be activated for the nucleophilic attack by protonation at N 5 15 . In both MetE and MetH, the nucleophilicity of homocysteine is enhanced by coordination with Zn 2+ that serves as Lewis acid 13 . While in the "base-on" form the dimethylbenzimidazole (Dmbz) base of methylcob(III)alamin is coordinated to the cobalt center of the corrin ring, in the "base-off " mode Dmbz is dissociated from the cobalt. Stabilization of the transition state of methylcob(III)alamin in the "base-off " or "base-off/His-on" binding mode enable nucleophilic attack of homocysteine by weakening the Co-C bond and by reducing the thermodynamic barrier [53][54][55] . Thus, only binding modes "base-off " or "base-off/His-on" enable methyl transfer from methylcob(III)alamin. However, the "base-off " mode of methylcob(III)alamin which is characterized by strong spectral changes namely, a significant blue shift in the UV/Vis spectrum and reduced intensity of the γ-band 38,56,57 , was not observed in our study (Fig. 2). It is difficult to precisely distinguish between the "base-on" and "base-off/His-on" form because, the UV/Vis spectra of them are very similar 56 . The formation of methionine and cob(I)alamin (Fig. 3a) can only take place if methylcob(III) alamin and homocysteine are bound to core-MetE CBDB in a stable and catalytically favorable configuration. Due to the minor spectral changes, we propose that methylcob(III)alamin is utilized by core-MetE CBDB in the "base-off/ His-on" binding mode. The "base-off/His-on" binding mode is found in many B 12 -dependent proteins with a consensus motif DxHxxG, where His represents the lower axial ligand replacing the Dmbz moiety 58 . In our computational model of core-MetE CBDB , His122 points towards the active site of the protein and probably belongs to a truncated B 12 -binding motif with the sequence HxxG, conserved among all Dehalococcoidia (Fig. 5).
The determined K M -value of core-MetE CBDB for methylcob(III)alamin of ~ 240 µM is likely much higher than the intracellular concentration of free methylcobalamin 29,59 . Therefore, the physiological methyl donor might not be soluble methyl(III)cobalamin. We hypothesize that the physiological methyl donor is a corrinoid protein that directly interacts with core-MetE CBDB . Needless to say, that inference of physiological characteristics from the determination of enzyme activity in vitro is limited. Examining the role of core-MetE in vivo could shed more light on the essentiality and functionality of the enzyme. However, genetic modification of Dehalococcoides strains is not possible yet.
The methyl group transferred by Dehalococcoides methionine synthase origins from exogenously supplied acetate, as has been shown by Zhuang et al. 32 . Acetate is activated in Dehalococcoides by acetyl-CoA synthetase (ACS) to acetyl-CoA 31 . Acetyl-CoA is then cleaved to free coenzyme A, carbon monoxide (which leaves the cell) and a methyl group originating from the C 2 -atom of acetate. This reaction is catalyzed by acetyl-CoA decarbonylase/synthase (AcsB), an enzyme known mostly for its activity in the opposite direction for carbon fixation via the Wood-Ljungdahl pathway 60 . In D. mccartyi strain CBDB1, AcsB represents the acetyl-CoA decarbonylase and AcsCD a dimeric corrinoid iron-sulfur protein (CoFeSP) to which the methyl group from acetyl-CoA is transferred 61 . Zhuang et al. hypothesized that the methyl group is then transferred from AcsCD to tetrahydrofolate and from there to homocysteine, but Dehalococcoides neither encode the methyltransferases acsE, responsible for the methyl transfer from CoFeSP to tetrahydrofolate nor the classical metE/metH, responsible for methyl transfer from methyl-tetrahydrofolate to homocysteine (Fig. 6a,b) 32 . Our results can now explain these two gaps by hypothesizing that the methyl group from AcsCD/CoFeSP is directly transferred to homocysteine by core-MetE CBDB instead of taking the diversion via tetrahydrofolate (Fig. 6c). This hypothesis would also explain the absence of carbon monoxide dehydrogenase in Dehalococcoides which would be needed if the Wood-Ljungdahl pathway was employed for CO 2 fixation. With the direct transfer of methyl groups from CoFeSP to homocysteine, the cells would be independent from methyl-tetrahydrofolate and indeed metF encoding methylene-tetrahydrofolate reductase (MTHFR) is missing in all Dehalococcoides genomes 27 . This might be an unusual pathway in extant microbiology but in our view could represent a very early evolutionary stage in which methyl metabolism could have been independent from folates. This view is supported by the fact that methionine, SAM, corrinoids and coenzyme A are conserved between archaea and bacteria but tetrahydrofolate/ tetrahydromethanopterin are not 45,46 .  32 . The methyl group is derived most probably from formate with the aid of formyltetrahydrofolate synthase (Fhs, yellow) and methylene-tetrahydrofolate dehydrogenase/cyclohydrolase (FolD, blue). (c) Methylation of L-homocysteine is conducted by core-MetE CBDB encoded by the locus cbdbA481 (purple) in D. mccartyi strain CBDB1. The original source of the methyl group is acetate, which is activated to acetyl-CoA and then cleaved by acetyl-CoA decarbonylase (AcsB, green) into HSCoA, carbon monoxide (CO) and a methyl group. The standard activity of AcsB is to transfer the methyl group to a corrinoid iron-sulfur protein complex (CoFeSP) AcsCD (red). We speculate that the methyl group is directly transferred from the CoFeSP to the core-MetE CBDB (A481) for L-homocysteine methylation (dashed arrow) but this transfer could also be indirect via a yet unidentified participant.

Scientific RepoRtS |
(2020) 10:2100 | https://doi.org/10.1038/s41598-020-58873-z www.nature.com/scientificreports www.nature.com/scientificreports/ Enzymes similar to the core-MetE identified in Dehalococcoides were also found in the majority of archaea (Fig. 1a, blue colors). M. thermoautotrophicum and other methanogens are described to encode methylcobalamin:homocysteine methyltransferase (core-MetE Archaea ) 33 , a protein of 308 amino acids that is also homologous to the C-terminal part of tr-MetE proteins. In vitro experiments showed, that core-MetE Archaea utilizes methylated corrinoids for the methylation of homocysteine, similar to what we now found for the core-MetE CBDB . Schröder and Thauer concluded that soluble methylcobalamin is unlikely the physiological methyl group donor and hypothesized that a corrinoid protein with yet unknown function could play this role. The gene products of MTH124 or MTH1156 were proposed as possible candidates. MTH1156 encodes MtrH, a protein with sequence similarity to the 5-methyl-THF-Glu-binding domain of MetH (26% identity) 33 . MtrH is part of the methyl-tetrahydromethanopterin-coenzyme M methyltransferase complex and catalyzes the methylation of cob(I)alamin to methylcob(III)alamin using methyl-tetrahydromethanopterin as methyl group donor 62 . In contrast to methyl-THF biosynthesis in Dehalococcoides strains, methanopterin biosynthesis, a functional equivalent to THF in archaea, is fully encoded in all methanogenic archaea 49,63 . The methyl group of methionine in methanogenic archaea is derived from methyl-tetrahydromethanopterin 49,64 , which might be the primary methyl donor of archaeal methylcobalamin:homocysteine methyltransferases.
In conclusion, our findings show that bacterial core-MetE CBDB homologs together with archaeal core-MetE representatives form a basal group of methionine synthases using methylcobalamin in vitro as co-substrate. Due to the fact that organisms encoding core-MetE enzymes are slowly growing strict anaerobes with strongly conserved ancient traits, we speculate that the core-MetE homologs are similar to an ancient methionine synthase encoded already in the genome of a predecessor of LUCA and therefore basal to both archaea and bacteria. We speculate that such basal methionine synthases were active in the metabolism of ancient microorganisms using methylcobalamin-containing proteins as methyl donor. Core-MetE CBDB is the first biochemically described bacterial representative of these core-MetE proteins resembling the methylcobalamin:homocysteine methyltransferase from M. thermoautotrophicum 33 . Tr-MetE proteins appear to have evolved by duplications of core-MetE and subsequently acquired the capacity to bind folate at the N-terminal part. In our phylogenetic analysis tr-MetE clusters with archaeal core-MetE genes (Fig. 1a).

Materials and Methods
General. All chemicals were purchased from Sigma-Aldrich (Munich, Germany) or Carl Roth (Karlsruhe, Germany). Whenever methionine or homocysteine are mentioned in the text, the L-form is meant. Chemicals used for mass spectrometry were obtained in LC-MS grade from Carl Roth. Pteroyltri-γ-L-glutamic acid (PteGlu 3 ) was acquired from Schircks Laboratories (Jona, Switzerland). Restriction enzymes, DNA polymerase, DNA and protein standards were obtained from New England BioLabs (Frankfurt/Main, Germany). Oligonucleotides and sequencing services were provided by Seqlab (Göttingen, Germany). All oligonucleotide primers, plasmids and strains used in this study are listed in Supplementary Tables 1 and 2. Anaerobic experiments were performed in a COY glovebox (Grass Lake, USA).
Bioinformatics. The structural model of CbdbA481 (core-MetE CBDB ) was calculated using the I-TASSER server 65 . Broadly defined, the server aligns the template protein with proteins of similar folds or with super-secondary structures from the PDB library by LOMETS. The overlay of core-MetE CBDB and tr-MetE from N. crassa was generated with PyMOL 66 . The amino acid sequences of tr-MetEs were trimmed approximately at the position 370. For the multiple sequence alignment and construction of the phylogenetic tree, only the C-termini of truncated tr-MetEs from bacteria, yeast and complete amino acid sequences of core-MetEs from archaea, Chloroflexi and Clostridiales were used. MEGA7 67 was used to calculate multiple amino acid sequence alignments using the implemented MUSCLE algorithm with default settings 68 . The evolutionary relationship between different methionine synthase amino acid sequences was inferred by using the Maximum Likelihood method based on the JTT matrix model 69 . Evolutionary distances were computed using Poisson correction and are expressed as the number of amino acid substitutions per site 70 .

Construction of expression and complementation plasmids.
Based on pBAD30, expression and complementation plasmids were generated as described in supplementary information. The resulting plasmids pBAD_MetE and pBAD_CbdbA481 were used for the complementation experiments as well as for the heterologous production and purification of tandem-repeat MetE (tr-MetE Eco ) from E. coli and core-MetE CBDB from D. mccartyi strain CBDB1.

Production and purification of recombinant tr-MetE Eco and core-MetE CBDB .
A preculture of E. coli DH10B containing pBAD30_MetE or pBAD30_CbdbA481 was set up in Luria-Bertani (LB) medium containing 100 µg mL −1 ampicillin and grown overnight at 37 °C and 140 rpm. On the following day, 1% (v/v) of the overnight culture was used to inoculate fresh LB medium containing the appropriate antibiotic. The cultures were grown at 37 °C under agitation at 140 rpm until the OD 600 reached 0.4-0.5. Then, the production of either tr-MetE Eco or core-MetE CBDB was induced by the addition of 0.05% (w/v) L-arabinose. Additionally, the medium was supplemented with 1 mM ZnSO 4 . MetE and CbdbA481 were produced for 5 h at 37 °C and 140 rpm. Then, the cells were harvested by centrifugation and washed with 50 mM Tris/HCl, pH 7.5. Purification of tr-MetE Eco and core-MetE CBDB was performed under anoxic conditions in an anaerobic chamber. Both enzymes were purified by anion exchange chromatography using a MonoQ 5/50 GL column connected to an ÄKTA purifier FPLC system (GE Healthcare Life Sciences) as described in detail in the supplementary information.

SDS-PAGE and native PAGE.
The purity of core-MetE CBDB and tr-MetE Eco protein preparations was evaluated by 10% SDS-PAGE. In addition, the oligomeric state of both proteins was investigated via 10% discontinuous native PAGE 71 .

Scientific RepoRtS |
(2020) 10:2100 | https://doi.org/10.1038/s41598-020-58873-z www.nature.com/scientificreports www.nature.com/scientificreports/ Protein identification from SDS-PAGE by LC-MS/MS. Qualitative identification of the purified core-MetE CBDB and tr-MetE Eco proteins was conducted mass spectrometrically. Therefore, protein bands at the height of 38 kDa and 80 kDa were excised from 10% SDS-PAGE gels. Acetonitrile, 10 mM DTT and 100 mM iodoacetamide were used to destain, to reduce and to alkylate the proteins within the gel slices. Subsequently, the proteins were digested with 0.1 µg trypsin (Promega) at 37 °C for 18 h. The resulting peptides were extracted from the gel matrix with 50% (v/v) acetonitrile and 5% (v/v) formic acid and dried. The peptides were again dissolved in 10 µL 0.1% formic acid and subsequently desalted using C 18 ZipTip Pipette Tips (Merck Millipore) and dried in a vacuum centrifuge. Prior analysis, the peptides were resuspended in 20 µL 0.1% formic acid. Samples were analyzed on an LC-MS/MS system composed of a nano-UPLC system (UltiMate 3000 RSLCnano System, Thermo Fisher Scientific) equipped with an Acclaim PepMap 100 75 µm × 25 cm C 18 column and connected to an Orbitrap Fusion mass spectrometer (Thermo Fisher Scientific) via an electrospray ion source (TriVersa NanoMate, Advion). Sample volumes of 5 µL were injected onto the column and separated applying a flow rate of 0.3 µL min −1 with the aid of a 60 min gradient from 3.2% to 44% acetonitrile in water containing 0.1% formic acid. The mass spectrometer was operated in positive-ionization mode. The spray voltage was set at 2.2 kV and an electron spray ionization source temperature at 220 °C. Full MS1 scans were obtained over a mass range of 300-2000 m/z and the resolution in the Orbitrap was set to 240,000. The most intense ions (threshold ion count above 5.0 × 10 4 ) were selected for fragmentation with the quadrupole, setting the isolation window to 1.6 m/z. Ions were fragmented by ETciD (ETD reaction time 100 ms, CID collision energy 35%). The resulting fragment ion spectra were obtained achieved in the Orbitrap at a resolution of 60,000 and a maximum injection time of 120 ms.
Protein and peptide identification. The raw mass spectrometric data were converted to mgf-files using ProteoWizard MSConvert v3.0 72 . The software SearchGUI (v3.3.5) 73 and the OMSSA search algorithm were used for peptide identification. Mass spectrometric data were searched against the E. coli proteome database obtained from UniProt (Taxon identifier 316385). A precursor ion mass tolerance of 10 ppm was used at the MS1 level and up to two missed cleavages were allowed. The fragment ion mass tolerance was set to 0.2 Da for the Orbitrap MS2 detection. The oxidation of methionine was considered as variable modification and carbamidomethylation on cysteines as fixed modification. The false discovery rate (FDR) in peptide identification was limited to a maximum of 0.01 by using a decoy database. The analyzed data were visualized with the PeptideShaker software (v1.16.27) (CompOmics, Ghent University) 74 .
Photometric analysis of methylcob(III)alamin binding to core-MetE CBDB . The binding of methylcob(III)alamin to core-MetE was followed spectrophotometrically in the range between 250 and 800 nm (0.5 nm steps) in 50 mM Tris/HCl (pH 6.5), 150 mM NaCl and 10% glycerol. First, the UV/Vis spectrum of 10 µM free methylcob(III)alamin was recorded. To assess the binding mode of methylcob(III)alamin to core-MetE CBDB , the protein solution was mixed with methylcob(III)alamin in a 1:1 stoichiometry (10 µM each). The UV/Vis spectra of free and bound methylcob(III)alamin were compared.
The synthesis of 5-methyl-THF-Glu 3 was accomplished from commercially available PteGlu 3 under anoxic conditions following the modified protocol of Yeo and Wagner 76 and as described in detail in the supplementary information. 5-methyl-H 4 PteGlu 3 was stored at −20 °C. Enzyme activity assay with 5-methyl-THF-Glu 3 as methyl group donor. Enzyme assays were performed under strictly anoxic conditions. The standard assay was set up in 25 mM Tris/HCl (pH 7.2) or 50 mM KH 2 PO 4 /K 2 HPO 4 (pH 7.2), 100 µM MgSO 4 , 100 µM ZnSO 4, 10 mM dithiothreitol (DTT), 2 mM D,L-homocysteine and 150 µM 5-methyl-THF-Glu 3 . The reaction was started by the addition of 0.25 µM tr-MetE Eco or core-MetE CBDB . After an incubation time of 60 min at 37 °C, the reactions were stopped by heat denaturation at 80 °C for 10 min, then centrifuged at 15,000 rpm for 5 min (Eppendorf Centrifuge 5424 R) and analyzed by HPLC. A negative control under same conditions without protein was run to evaluate abiotic transformation of 5-methyl-THF-Glu 3 . 5-methyl-THF-Glu 3 , other folate derivatives and PteGlu 3 were analyzed with a JASCO HPLC 2000 series system equipped with an Equisil BDS C 18 column (250 × 4.6 mm, 5 μm; Dr. Maisch HPLC GmbH, Ammerbuch-Entringen/Germany) following the modified protocol of Patring 77 . The identities of PteGlu 3 , 5-methyl-THF-Glu 3 and L-methionine were confirmed via liquid chromatography-mass spectrometry in direct injection mode.
Generation of E. coli knockout strain. The E. coli DH5α (ΔmetE::kan) knockout strain was generated using the Quick & Easy E. coli Gene Deletion Kit (GeneBridges GmbH, Heidelberg, Germany) according to the manufacturer's protocol 78 . Hereby, metE gene in E. coli DH5α was replaced by a linear kanamycin cassette. The introduction of the kanamycin cassette allowed to screen for the knockout strain on 20 µg mL −1 kanamycin agar plates.
The main cultures were then incubated at 37 °C and 140 rpm and growth was monitored by measuring the OD 600 . E. coli DH5α still encodes the arabinose operon. However, arabinose at a concentration of 0.05% (w/v) sufficed to induce the production of core-MetE CBDB and tr-MetE Eco . In our experiments, reciprocal metabolism leads to preferential use of glycerol instead of arabinose.