Identification of a Novel Cobamide Remodeling Enzyme in the Beneficial Human Gut Bacterium Akkermansia muciniphila

Cobamides, comprising the vitamin B12 family of cobalt-containing cofactors, are required for metabolism in all domains of life, including most bacteria. Cobamides have structural variability in the lower ligand, and selectivity for particular cobamides has been observed in most organisms studied to date.


RESULTS
A. muciniphila strain Muc T salvages Cbi to produce pCbl. A. muciniphila strain Muc T lacks most of the genes required for cobamide synthesis and does not produce cobamides de novo (31), but it is predicted to be capable of Cbi salvaging (34). To test this prediction, we extracted corrinoids from A. muciniphila cultured with and without Cbi and analyzed the corrinoid composition of the samples by high-performance liquid chromatography (HPLC). Cultured without Cbi, no corrinoids were detected in the extractions ( Fig. 2A). However, when Cbi was added to the growth medium, a cobamide with the same retention time and a UV-visible light (UV-Vis) spectrum nearly identical to that of pCbl was detected by HPLC ( Fig. 2A). Mass spectrometry (MS) analysis confirmed that this cobamide is pCbl (see Fig. S1 in the supplemental material).
A. muciniphila strain Muc T does not show cobamide selectivity. Having established that A. muciniphila strain Muc T cannot synthesize cobamides without the addition of a precursor, we next examined which cobamides it is capable of using by measuring growth in the presence of various cobamides under a cobamide-requiring condition. A. muciniphila strain Muc T has four cobamide-dependent enzymes; however, MCM is not required for laboratory growth (60), QueG function does not affect growth in other bacteria (61), and the cobamide-independent ribonucleotide reductase NrdDG likely renders NrdJ nonessential. Because A. muciniphila lacks a homolog of the cobamide-independent methionine synthase MetE (34), growth in the absence of methionine is expected to be dependent on the cobamide-dependent methionine synthase MetH. Thus, growth in methionine-deplete medium is predicted to require cobamide addition. We found this to be the case, as addition of Cbi or any of the seven cobamides tested was necessary to support growth of A. muciniphila (Fig. 2B). Surprisingly, however, A. muciniphila showed essentially no cobamide selectivity, with less than 2fold variations in the cobamide concentrations resulting in half-maximal growth (EC 50 ) (Fig. 2B).
A. muciniphila strain Muc T remodels cobamides to pCbl. The ability of all of the tested cobamides to support nearly identical levels of growth of A. muciniphila could have been due to promiscuity in its cobamide-dependent methionine synthase. Alternatively, A. muciniphila might remodel cobamides, despite the absence of a homolog of cbiZ in its genome. If cobamide remodeling occurs in A. muciniphila, exogenously supplied cobamides would be altered by the bacterium. Therefore, we extracted corrinoids from A. muciniphila cultures supplemented with Cbl, [Cre]Cba, or [MeAde]Cba to determine whether the added cobamides could be recovered. HPLC analysis revealed that none of the Cbl or [Cre]Cba and only half of the [MeAde]Cba remained in the extractions. This loss of the added cobamide coincided with the appearance of a new cobamide that coeluted with pCbl ( Fig. 2C). MS analysis confirmed that this cobamide was indeed pCbl (Fig. S2). These results demonstrate that A. muciniphila remodels cobamides to pCbl.
Identification and characterization of a novel cobamide remodeling enzyme in A. muciniphila. The identification of cobamide remodeling activity despite the absence of a cbiZ homolog in the genome suggested that a novel enzyme capable of hydrolyzing cobamides is present in A. muciniphila. We reasoned that the gene encoding this enzyme might be located near the cobamide biosynthesis and salvaging genes cobDQ, cbiB, cobT, cobS, and cobU, some or all of which would be required for completion of the remodeling process. These five genes are found at a single locus in the A. muciniphila genome that also contains an open reading frame (ORF) with unknown function, annotated as Amuc_1679 (GenBank accession no. ACD05497.1) (Fig. 3A). Amuc_1679 is predicted to encode a protein with a conserved (b/a) 8 TIM barrel domain from the apurinic/apyrimidinic (AP) endonuclease 2 superfamily (pfam01261). This superfamily is composed of several enzymes, including endonuclease IV, which hydrolyzes phosphodiester bonds at AP sites in DNA (62). The proximity of Amuc_1679 to genes involved in cobamide biosynthesis and the presence of a phosphodiester bond connecting the lower ligand to the aminopropanol linker suggested that Amuc_1679 might play a role in cobamide biology in A. muciniphila. Further, homologs of this gene in other bacteria are also found in loci containing similar cobamide biosynthesis enzymes ( To determine whether Amuc_1679 encodes an enzyme that can hydrolyze cobamides, we overexpressed and purified Amuc_1679 for analysis of its activity in vitro. Nterminal hexahistidine (His 6 ) and maltose-binding protein (MBP) tags were added to aid in the purification and to increase the solubility of the protein, respectively (Fig. S4A). First, we tested whether a new product was formed when the protein was incubated with coenzyme B 12 (AdoCbl), an active cofactor form of Cbl. We observed complete conversion of AdoCbl to a new corrinoid compound in reactions performed under anaerobic conditions (Fig. 3B). MS analysis showed that two reaction products, cobinamide-phosphate (Cbi-P) and a-ribazole, were formed, indicating hydrolysis of the phosphoribosyl bond of AdoCbl ( Fig. 3C and D). Notably, Amuc_1679 targets a bond distinct from the enzyme CbiZ ( Fig. 1A) (59). Synthesis of the cobamide from Cbi-P might be a more efficient remodeling process as it would bypass the need to rebuild the aminopropanol linker. In keeping with the tradition of naming cobamide biosynthesis and remodeling enzymes with a "Cbi" prefix, we refer to Amuc_1679 as CbiR here.
We were able to monitor CbiR activity continuously by measuring the rate of decrease in absorbance at 534 nm (A 534 ), as the reaction is characterized by a change in the UV-Vis spectrum that reflects the loss of AdoCbl and formation of AdoCbi-P (Fig. 3E). With this method, we found that the reaction proceeded only in the absence of oxygen and, additionally, that the reaction required the reducing agent dithiothreitol (DTT) and was inhibited by the metal chelator EDTA (Fig. S4B). Although other members of the AP endonuclease 2 superfamily require metals (63)(64)(65)(66)(67)(68)(69)(70)(71)(72)(73)(74)(75)(76)(77)(78)(79)(80), CbiR is the only member found to be sensitive to oxygen.
Using the same method, we determined the reaction kinetics of His 6 -MBP-CbiR under steady-state conditions over a range of AdoCbl concentrations (Fig. 3F). On the basis of a fit to the Michaelis-Menten model, the reaction of His 6 -MBP-CbiR with AdoCbl exhibited K M and k cat values for AdoCbl of 194 mM and 6.5 min 21 , respectively. Similarly, His 6 -MBP-CbiR hydrolyzes MeCbl, the active cofactor form used by MetH and other methyltransferases, to MeCbi-P (Fig. S4C), with comparable kinetic parameters (Fig. S4D), indicating that AdoCbl and MeCbl are equally suitable substrates for His 6 -MBP-CbiR.
A. muciniphila CbiR can hydrolyze several different cobamides in vitro. To determine the substrate selectivity of CbiR, His 6 -MBP-CbiR activity was measured in vitro with seven different cobamides. Adenosylated cobamides with purinyl or phenolyl lower ligands do not show UV-Vis spectra distinct from that of AdoCbi-P under the reaction conditions, and thus activity was measured by HPLC. Each cobamide was completely converted to AdoCbi-P following 18 h of incubation, demonstrating that all of the cobamides were substrates for CbiR (Fig. 4A). The specific activities of His 6 -MBP-CbiR with benzimidazolyl and purinyl cobamides were similar, with up to 4-fold differences among them, while higher activities were observed with phenolyl cobamides, though with greater variability between experiments (Fig. 4B). These in vitro results seen with CbiR appear to correlate with the remodeling activity found in vivo in A. muciniphila (Fig. 2C). The specific activities of His 6 -MBP-CbiR are similar to though slightly lower than that previously reported for CbiZ with Ado-pCbl (70 nmol/mg/min [59]), albeit under somewhat different reaction conditions.
Expression of cbiR in E. coli enables expanded cobamide use. Given that the product AdoCbi-P can be used as a precursor for construction of a different cobamide, we hypothesize that CbiR activity enables bacteria to remodel cobamides and therefore to gain access to cobamides in the environment that they otherwise might not be able to use. Because methods for targeted inactivation of genes in A. muciniphila have not been established, we used engineered E. coli strains to test this hypothesis. Like A. muciniphila strain Muc T , E. coli MG1655 cannot synthesize cobamides de novo, but its genome has the cobamide biosynthesis genes cobT, cobS, cobU, and cobC, which should allow E. coli to convert AdoCbi-P into a cobamide (81). We first tested whether A. muciniphila CbiR is functional in E. coli grown under aerobic conditions. Expression of untagged CbiR from a plasmid in a DcobTSU DcobC background did result in the loss of added Cbl, pCbl, [MeAde]Cba, and [Cre]Cba and the formation of two new corrinoid compounds (Fig. 5A). One of the products coeluted with AdoCbi-P (Fig. 5A), and MS analysis confirmed that the dominant ion matches the m/z values expected for Cbi-P (Fig. S5A). The second product gave m/z values consistent with Cbi ( Fig. S5B), which likely formed intracellularly by hydrolysis of the phosphate group of AdoCbi-P. Neither product was detected in an E. coli strain containing the empty vector (Fig. 5A, dashed lines). Therefore, the activity of CbiR that we observed in vitro can be recapitulated in E. coli, even when it is cultured aerobically.
Catabolism of ethanolamine in E. coli requires the cobamide-dependent enzyme ethanolamine ammonia lyase. Because this enzyme uses cobamides in the base-on conformation, in which a nitrogen atom in the lower ligand is coordinated to the cobalt atom ( Fig. 1), it can use Cbl as a cofactor but is not functional with [Cre]Cba (82). We took advantage of this selectivity to design a cobamide remodeling-dependent growth assay in E. coli. In minimal medium supplemented with [Cre]Cba and 5,6-dimethylbenzimidazole (DMB; the lower ligand of Cbl), with ethanolamine as the sole nitrogen source, E. coli should be able to grow only if it can remodel [Cre]Cba to Cbl. Cbl, as expected, promoted growth of E. coli under this condition regardless of whether cbiR was present (Fig. 5B). In contrast, when [Cre]Cba was added, growth was observed only in the strain expressing cbiR, suggesting that CbiR activity enabled E. coli to convert [Cre]Cba into Cbl (Fig. 5B). A cobamide with a retention time and m/z values matching those of Cbl was detected in a corrinoid extraction of E. coli expressing cbiR grown with [Cre]Cba and DMB, confirming that cobamide remodeling to Cbl had occurred (Fig. S6). These results demonstrate that expression of CbiR expands the range of cobamides accessible to E. coli and suggest that cobamide remodeling may serve a similar purpose in A. muciniphila.
CbiR homologs are present in diverse bacteria. Analysis of the sequence of CbiR revealed that it is not similar to that of CbiZ or V. cholerae CobS, the other enzymes known to have cobamide remodeling activity. Instead, as a member of the AP endonuclease 2 superfamily, CbiR is homologous to the enzymes endonuclease IV, 2-ketomyo-inositol dehydratase, and xylose isomerase and to other sugar isomerases and epimerases. A phylogenetic tree of this superfamily shows that the CbiR homologs identified by a BLAST search that are encoded in genomic loci containing cobamide biosynthesis genes form a single, distinct clade (Fig. 6A). Some of the biochemically characterized enzymes in the superfamily require metal cofactors for activity (63)(64)(65)(66)(67)(68)(69)(70)(71)(72)(73)(74)(75)(76)(77)(78)(79)(80), and between one and three metal ions, such as Zn 21 , Fe 21 , Mg 21 , and Mn 21 , are found in nearly all X-ray crystal structures of enzymes from the superfamily (83-97). A metal cofactor may also be required for CbiR function; in addition to the inhibition by the metal chelator EDTA (Fig. S4B), CbiR homologs contain conserved His, Asp, and Glu residues that, in the characterized members of the superfamily, are involved in metal coordination (Fig. 6B) (86,90,93,94,96). Significant losses in activity have been observed in site-directed mutants of metal binding residues in AP endonuclease 2 superfamily members (87,(98)(99)(100)(101)(102)(103)(104)(105)(106). Similarly, E. coli strains expressing cbiR alleles with single mutations in many of these conserved residues had little to no AdoCbl hydrolysis activity (Fig. 6C), though effects of these mutations on enzyme folding or stability cannot be ruled out. These results demonstrate that CbiR shares both sequence and functional features common to the members of the AP endonuclease 2 superfamily and represents a new function within the superfamily.
Finally, we investigated the prevalence of CbiR across sequenced organisms by examining genomes with cbiR homologs. The cbiR gene commonly occurs in the Akkermansia genus, as a search of the 191 available genomes in the NCBI database found that 184 have a cbiR homolog. Additionally, 282 homologs of A. muciniphila CbiR with Expect (E) values below 10 23 were identified by BLAST in the genomes of 275 bacterial taxa and 1 archaeal taxon from diverse habitats, including aquatic environments, sewage, digesters, oil spills, bioreactors, soil, and human and animal hosts (see Table S1 in the supplemental material). While 76% are found in the phyla Chlorobi, Chloroflexi, and Proteobacteria, CbiR homologs are also present in 19 additional phyla ( Fig. S3; see also Table S1), with relatively few in the PVC superphylum to which A. muciniphila belongs. Interestingly, nearly 20% of these taxa contain one or more homologs of cbiZ (Table S1), though it remains unknown whether the CbiR and CbiZ homologs in these strains function in cobamide remodeling. Similarly to A. muciniphila, 80% of all cbiR homologs are located adjacent to genes involved in cobamide biosynthesis (Table S1), suggesting that the majority of cbiR homologs function in cobamide remodeling.

DISCUSSION
Cobamides are considered to be important modulators of mammalian gut ecosystems because they are involved in several metabolic pathways, their production is limited to a subset of prokaryotes, and their diverse structures are differentially accessible to different microbes (54, 107, 108). A. muciniphila has been shown to have Cobamide Remodeling in a Beneficial Gut Bacterium ® positive effects on host metabolism, gut barrier function, and the inflammatory response (15)(16)(17)(18)(19)(20)(21)(22)(23)(24), and yet knowledge of its metabolic and ecological roles in the gut remains incomplete. Previous studies showed that A. muciniphila strain Muc T is unable to produce cobamides de novo (31) but can use pCbl produced by E. hallii or externally supplied Cbl for propionate production (26,31). Here, while investigating the cobamide metabolism of A. muciniphila strain Muc T , we uncovered a novel cobamide remodeling activity and identified and characterized an enzyme capable of initiating this process, CbiR. This discovery adds new complexity to the understanding of the roles of A. muciniphila in the gut. Not only does A. muciniphila degrade mucin to provide nutrients to the gut microbiota (26,27), but it is also capable of altering cobamide structure, potentially changing the cobamide composition of its environment.
As a member of the AP endonuclease 2 superfamily, CbiR likely contains a (b/a) 8 TIM barrel domain (83)(84)(85)(86)(87)(88)(89)(90)(91)(92)(93)(94)(95)(96)(97), unlike the structures predicted for the CbiZ and CobS protein families (109). Thus, not only does CbiR catalyze a unique reaction, but it is also distinct from the other cobamide remodeling enzymes in sequence and likely in FIG 6 CbiR is a newly described member of the AP endonuclease 2 superfamily. (A) Maximum likelihood tree of CbiR homologs and members of the AP endonuclease 2 superfamily that have been experimentally characterized. A. muciniphila CbiR is indicated by an asterisk. CbiR homologs included in the tree were identified in a BLAST search queried with A. muciniphila CbiR with E values lower than 10 214 and encoded adjacent to and in the same orientation as one or more cobamide biosynthesis genes (Table S1). Superfamily member sequences are listed in Table S2. Gray circles overlaid on tree nodes represent bootstrap values of .95%. The scale bar corresponds to the average number of substitutions per site across the alignment. (B) Sequence alignment of regions containing highly conserved His, Asp, and Glu residues in A. muciniphila CbiR and representative sequences of biochemically and structurally characterized enzyme classes in the AP endonuclease 2 superfamily (E. coli endonuclease IV, Streptomyces rubiginosus xylose isomerase, Pseudomonas cichorii D-tagatose 3-epimerase, Bacillus subtilis 2-keto-myo-inositol isomerase, and E. coli L-xylulose-5-P 3-epimerase). Numbers correspond to positions in A. muciniphila CbiR. For CbiR, endonuclease IV, xylose isomerase, and D-tagatose (ketose) 3-epimerase, bolded residues represent conserved amino acids in the enzyme classes. Underlined residues indicate amino acids in the X-ray crystal structures that interact with the metal cofactor(s) and with the substrate in the case of P. cichorii D-tagatose 3-epimerase. (C) Mutational analysis of A. muciniphila CbiR. Corrinoids were extracted from cultures of E. coli DcobTSU DcobC strains carrying pETmini-cbiR (wild type [WT]), pETmini-cbiR with the specified alanine mutations, or the pETmini empty vector grown for 20 h with 75 nM Cbl and analyzed by HPLC. Shown on the y-axis is the combined amount of AdoCbi-P and AdoCbi as a percentage of the total adenosylated corrinoids extracted. Minimal amounts of CNCbl were detected in the mutant extractions, and the corresponding data were excluded from the analysis. The total amounts of intracellular corrinoid were similar between samples except for the WT and D228A samples, which had 4-fold-higher and 2.5-fold-higher levels, respectively. Lines show means of results from the two independent experiments. structure. Intriguingly, while CbiR differs in sequence from B 12 -binding domains in cobamide-dependent enzymes, the substrate-binding domains of many cobamide-dependent enzymes are comprised of a (b/a) 8 TIM barrel structure, with the C-terminal face interacting with the cobamide cofactor (110)(111)(112)(113)(114)(115)(116)(117)(118)(119)(120). Given that CbiR is predicted to have a similar fold, it is possible that the cobamide binding mechanisms of CbiR and these cobamide-dependent enzymes share common features. The yet-to-be-discovered enzyme responsible for remodeling in algae may also be unique, as neither a P. lutheri transcriptome (121) nor the C. reinhardtii genome contains homologs of CbiR or CbiZ. It therefore appears that cobamide remodeling mechanisms have independently evolved multiple times. Together with the multiple pathways that exist for cobamide biosynthesis, transport, and precursor salvaging (32,34,35,(122)(123)(124)(125)(126)(127)(128), the addition of CbiR to the growing list of enzymes involved in cobamide metabolism highlights the importance of cobamide physiology in the evolution of bacteria.
In contrast to previously characterized cobamide remodeling enzymes and pathways, CbiR has promiscuous activity, hydrolyzing cobamides irrespective of their lower ligand structure. This activity differs from that of R. sphaeroides CbiZ, which does not hydrolyze Cbl in vitro (59), and V. cholerae CobS, which remodels neither Cbl nor [Cre] Cba (48). In those cases, the cobamide remodeling pathway does not act on a cobamide(s) that can function in its organism's metabolism. In contrast, A. muciniphila CbiR readily hydrolyzes pCbl, which functions as a cofactor for methionine synthesis and propionate metabolism in A. muciniphila and is the product of Cbi salvaging and cobamide remodeling in the bacterium itself. Thus, it is unclear how A. muciniphila prevents CbiR from continuing to hydrolyze pCbl after it is formed via cobamide remodeling. It is possible that pCbl is sequestered intracellularly by binding to MetH or other cobamide-dependent enzymes, such as may have occurred in our ethanolamine-based growth assay in E. coli in which cbiR was constitutively expressed. Alternatively, CbiR activity could be coupled to cobamide uptake, as has been suggested previously for CbiZ (59,129). Indeed, similarly to some cbiZ homologs, 25% of cbiR homologs are located adjacent to genes encoding putative transport proteins, including in A. muciniphila strain Muc T (see Fig. S3 in the supplemental material). Remodeling in D. mccartyi strain 195 shows substrate promiscuity similar to that seen with A. muciniphila with respect to the ability to act on numerous, structurally diverse cobamides (56), but the molecular basis of this promiscuity is unclear because its genome carries seven cbiZ homologs, none of which has been biochemically characterized. Aside from its activity on pCbl, the broad substrate range of CbiR may benefit A. muciniphila by enabling the bacterium to utilize a greater number of the cobamides present in its environment.
The discovery that A. muciniphila remodels cobamides leads us to reexamine its ecological roles in the gut. CbiR is found in all 75 of the recently sequenced A. muciniphila strains from the human and mouse gut, including the 26 strains that contain the de novo cobamide biosynthesis pathway (31,130). Thus, like cobamide-dependent metabolism (31), cobamide remodeling appears to be nearly universal in A. muciniphila. Further, the role of A. muciniphila in the gut may be flexible, ranging from producing cobamides de novo to remodeling cobamides produced by other microbes, depending on which strains inhabit an individual. Notably, the end product of cobamide remodeling in A. muciniphila, pCbl, was the third most abundant corrinoid detected in the human gut in a study of human subjects residing at a single geographic location (42). Interestingly, that study also presented evidence that cobamide remodeling occurs in the human gut, as individuals supplemented with high levels of Cbl showed transiently increased levels of Cbi and of the specific purinyl and phenolyl cobamides that were present in the gut prior to Cbl supplementation. It is possible that A. muciniphila is involved in this remodeling activity and contributes to the pool of pCbl in the gut. This, in turn, could modulate the growth or metabolism of other cobamide-requiring bacteria that rely on particular cobamides for their metabolic needs. CbiR may therefore not only expand access to the cobamides available to A. muciniphila but also affect those accessible to other bacteria in the gut. Further, homologs of CbiR are found in at least 276 other microbial taxa and may function similarly in those microbes that inhabit diverse environments. The addition of CbiR to the cobamide remodeling enzymes that have been characterized to date-CbiZ, certain CobS homologs, and the enzyme(s) responsible for remodeling in algae-suggests that cobamide remodeling is more widespread than previously thought.

MATERIALS AND METHODS
Media and growth conditions. Akkermansia muciniphila strain Muc T (DSM 22959, ATCC BAA-835) was cultivated at 37°C in a vinyl anaerobic chamber (Coy Laboratory Products Inc.) under an atmosphere of approximately 10% CO 2 , 3% H 2 , and 87% N 2 . A synthetic version of a basal mucin-based medium, in which mucin was replaced by soy peptone (16 g/liter), L-threonine (4 g/liter), glucose (2 g/liter), and Nacetylglucosamine (2 g/liter), was supplemented with 1% noble agar and used as a solid medium (21). This synthetic medium also contained L-methionine (125 mg/liter) and omitted rumen fluid. M8 defined medium, developed by Tramontano et al. (131), was used for liquid culturing. We found that the concentration of mucin in this medium (0.5%) was able to abrogate the requirement of methionine addition to the medium for A. muciniphila growth. Lowering the concentration to 0.25% resulted in methioninedeplete conditions for A. muciniphila, and supplementation with methionine or cobamides restored robust growth. This mucin concentration was used for the MetH-dependent growth assays. However, batch-to-batch variations were seen with mucin such that formulations of media that supported robust growth while remaining methionine-deplete could not always be achieved. Cobamides were omitted from all growth media except when specified. For MetH-dependent growth assays, methionine was omitted from M8 medium.
For all in vivo experiments involving corrinoids, culture media were supplemented with cyanylated cobamides or (CN) 2 Cbi, which are adenosylated following uptake into cells.
Genetic and molecular cloning techniques. The entire A. muciniphila cbiR open reading frame, except the start codon, was cloned into a modified pET16b vector (132) with N-terminal His 6 and MBP tags added for protein purification. For analysis of CbiR activity in E. coli, a minimized 3-kb derivative of pET28a (pETmini) containing a kanamycin resistance (Kan r ) marker, pBR322 origin, and rop gene was used. A constitutive promoter (BBa_J23100; iGEM) and ribosome binding site (RBS) (BBa_B0034) were inserted into the vector for expression in E. coli MG1655-based strains (complete sequence, TTGACGGCTAGCTCAGTCCTAGGTACAGTGCTAGCGAATTCATACGACTCACTATAAAAGAGGAGAAA) and A. muciniphila cbiR was cloned downstream. Site-directed mutations were introduced into cbiR by PCR. All cloning was done by Gibson assembly with E. coli XL1-Blue cells (133).
Construction of the E. coli MG1655 DcobTSU DcobC strain was accomplished using l Red-based recombination (134) and phage P1 transduction (135). An MG1655 DcobTSU::Kan r operon deletion was constructed by l Red-based recombination. The DcobC::Kan r allele was transduced into MG1655 via P1 transduction from E. coli strain JW0633-1, which was obtained from the Keio collection (136). Kan r cassettes were removed by recombination of the flanking FLP recombination target (FRT) sites as described previously (134).
A. muciniphila MetH-dependent growth assay. A. muciniphila was precultured for 48 h in M8 medium supplemented with 125 mg/liter methionine. Cells were pelleted by centrifugation, washed twice with phosphate-buffered saline (PBS), and diluted into 80 ml M8 medium to an optical density at 600 nm (OD 600 ) of 0.02 in a 384-well plate (Nunc) with various concentrations of cobamides and Cbi. The wells were sealed (ThermalSeal RTS; Excel Scientific), and the plate was incubated at 37°C in a BioTek Epoch 2 microplate reader. OD 600 values were measured at regular intervals during growth.
E. coli ethanolamine-dependent growth assay. E. coli was precultured 16 h in ethanolamine medium (38) supplemented with 0.02% ammonium chloride. Cells were pelleted by centrifugation, washed three times with 0.85% NaCl, and diluted to an OD 600 of 0.025 in 200 ml ethanolamine medium with the specified cobamide additions in a 96-well plate (Corning). The wells were sealed (Breathe-Easy; Diversified Biotech), and OD 600 was monitored at 37°C in a BioTek Synergy 2 microplate reader with shaking.
His 6 -MBP-CbiR in vitro reactions. Due to light sensitivity of the compounds, all work involving adenosylated cobamides or MeCbl was performed in the dark or under red or dim white light. Unless specified, the in vitro reactions were performed at 37°C in a vinyl anaerobic chamber with the atmosphere described above. The components of the reaction mixtures were 50 mM Tris buffer, 0.3 mM purified His 6 -MBP-CbiR, 1 mM DTT, and various concentrations of a cobamide. To prepare the Tris buffer, Tris base was dissolved and equilibrated within the anaerobic chamber. Prior to each experiment, the pH was adjusted with NaOH to account for acidification by the CO 2 present in the atmosphere of the chamber. The pH was adjusted to 8.8 to 8.9 at room temperature (approximately 24°C), corresponding to a predicted pH range of 8.45 to 8.55 at 37°C. Protein concentrations were determined by absorbance at 280 nm (A 280 ).
A BioTek Epoch 2 microplate reader and half-area UV-Star 96-well microplates (Greiner Bio-One) were used for assays monitoring the reaction by absorbance. For these assays, separate 2Â solutions of AdoCbl and His 6 -MBP-CbiR were prepared in a buffer containing 50 mM Tris and 1 mM DTT. A frozen aliquot of His 6 -MBP-CbiR was thawed inside the anaerobic chamber prior to dilution. The 2Â AdoCbl solution was preincubated at 37°C for 60 min, while the 2Â CbiR solution was preincubated at 37°C for 20 min. Volumes of 60 ml each of 2Â AdoCbl and 2Â His 6 -MBP-CbiR were then mixed in a 96-well plate, with 100 ml transferred to a new well prior to performing measurements in the plate reader. Assays with MeCbl were prepared similarly.
Absorbances measured over time at 534 and 527 nm were used to monitor the conversion of AdoCbl to AdoCbi-P and of MeCbl to MeCbi-P, respectively. To enable conversion of A 534 values into moles of AdoCbl and of A 527 values into moles of MeCbl, extinction coefficient values corresponding to AdoCbl and MeCbl and to purified AdoCbi-P and MeCbi-P were determined in a buffer containing 50 mM Tris (pH 8. For reactions performed with adenosylated cobamides with different lower ligands monitored by HPLC, cobamides were mixed at 60 mM with 50 mM Tris buffer and 1 mM DTT and equilibrated to 37°C. His 6 -MBP-CbiR was equilibrated to 37°C at 0.6 mM in a buffer containing 50 mM Tris and 1 mM DTT. Each cobamide and His 6 -MBP-CbiR were mixed in equal volumes, and the reaction mixtures were incubated at 37°C. At three different time points, 100 ml of the reaction mixture was removed and mixed with 5 ml of 600 mM EDTA to quench the reaction. The protein was removed from samples using Nanosep 10K centrifugal devices (Pall) prior to injection onto the HPLC system. AdoCbi-P levels in the samples were quantified by HPLC by comparison to a standard curve generated with known quantities of purified AdoCbi-P. For reactions involving incubations of 4 to 18 h, initial equilibration at 37°C was not performed.
Corrinoid extraction. Cbi salvaging and cobamide remodeling were assessed in A. muciniphila by cultivation in M8 medium supplemented with 10 nM Cbi and cobamide, respectively, for 72 h. CbiR cobamide hydrolytic activity with different cobamides in E. coli was monitored using the MG1655 DcobTSU DcobC strain cultivated in LB medium supplemented with 10 nM cobamide for 16 h. Cobamide remodeling in E. coli MG1655 was assessed by cultivation in ethanolamine medium supplemented with 100 nM [Cre]Cba and 1 mM DMB for 94 h. CbiR mutants were analyzed in E. coli MG1655 DcobTSU DcobC by culturing in LB medium supplemented with 75 nM Cbl for 20 h.
Cyanation of corrinoids extracted from cells for the experiments whose results are shown in Fig. 2A and C (see also Fig. S1 and S2 in the supplemental material) was performed as previously described (56), with 5,000 corrinoid molar equivalents of KCN added. For extractions of adenosylated corrinoids (Fig. 5A  and 6C; see also Fig. S5 and S6), cell lysis was performed similarly, with KCN omitted; following removal of cellular debris by centrifugation, deionized water was added to the supernatant to decrease the methanol concentration to 10%. Solid-phase extraction of cyanylated and adenosylated corrinoids with Sep-Pak C 18 cartridges (Waters) was performed as described previously (37). Samples were dried, resuspended in 200 ml deionized water (pH 7), and filtered with 0.45-mm-pore-size filters (Millex-HV; Millipore) or Nanosep 10K centrifugal devices prior to analysis by HPLC. For extractions involving adenosylated cobamides, all steps were performed in the dark or under red or dim white light.
HPLC and MS analysis. Corrinoids were analyzed on an Agilent 1200 series HPLC system equipped with a diode array detector. For the experiments whose results are shown in Fig. 2A and C, 3B, and 4A and B, an Agilent Zorbax SB-Aq column (5-mm pore size, 4.6 by 150 mm) was used as previously described (method 2 [58]). For the experiments whose results are shown in Fig. 5A and 6C (see also Fig. S6A), an Agilent Zorbax Eclipse Plus C 18 column (5-mm pore size, 9.4 by 250 mm) was used with the following method: solvent A, 0.1% formic acid-deionized water; solvent B, 0.1% formic acid-methanol; 2 ml/min at 30°C; 18% solvent B for 2.5 min followed by a linear gradient of 18% to 60% solvent B over 28.5 min.
An Agilent 1260 series fraction collector was used for HPLC purification of corrinoids and CbiR reaction products. Purification of CN-pCbl from A. muciniphila was performed using the Zorbax SB-Aq column with the method described above. Purification of AdoCbi-P and a-ribazole from the hydrolysis of AdoCbl by His 6 -MBP-CbiR was performed in two steps. AdoCbi-P and a-ribazole were first separated and collected on a Zorbax Eclipse XDB C 18 column (5-mm pore size, 4.6 by 150 mm) using the following method: solvent A, 10 mM ammonium acetate (pH 6.5); solvent B, 100% methanol; 1 ml/min at 30°C; 0% solvent B for 2 min followed by a linear gradient of 0% to 15% solvent B over 1.5 min, 15% to 35% over 6.5 min, 35% to 70% over 2 min, and 70% to 100% over 2 min. Each compound was subsequently run and collected on the Zorbax SB-Aq column with the method described above. Purification of MeCbi-P from the in vitro hydrolysis of MeCbl was performed using the Zorbax SB-Aq column with the method described above. The purification of adenosylated hydrolysis products of CbiR from E. coli was performed with two rounds of collection; AdoCbi-P and AdoCbi were first separated and collected on the Zorbax Eclipse Plus C 18 column with the method above, and then each compound was run and collected on the Zorbax SB-Aq column using the method described. AdoCbl remodeled from [Cre]Cba in E. coli was purified using the Zorbax Eclipse Plus C 18 column with the method described above. Collected compounds were desalted with Sep-Pak C 18 cartridges.
MS analysis was performed on a Bruker linear quadrupole ion trap coupled to a Fourier transform ion cyclotron (LTQ-FT) mass spectrometer at the QB3/Chemistry MS Facility (University of California, Berkeley [UC Berkeley]). Prior to MS analysis, the purified adenosylated and methylated corrinoids were exposed to light to remove the adenosyl and methyl upper ligands, respectively.
Phylogenetic analysis. A total of 282 homologs of A. muciniphila strain Muc T CbiR were identified by BLAST, with E values lower than 10 23 (see Table S1 in the supplemental material; sequences from other strains of A. muciniphila are excluded). A subset of 203 sequences with E values lower than 10 214 and whose encoding genes were located adjacent to and in the same orientation as a cobamide biosynthesis gene(s) were chosen for phylogenetic analysis performed with the AP endonuclease 2 superfamily (pfam 01261) (Table S1). Sequences were clustered at 0.95 using CD-HIT to reduce the CbiR homolog sequence set by removing subspecies sequence diversity (140). This final set of 178 sequences and A. muciniphila CbiR were used to infer a phylogenetic tree with experimentally characterized members of the AP endonuclease 2 superfamily (Table S2). The sequences were aligned with MAFFT (141), and positions with 95% or greater gaps were removed by the use of trimAl (142). A maximum likelihood tree (presented in Fig. 6A) was inferred from this alignment using IQ-TREE v1.6.12 (143) with 1,500 ultrafast bootstraps and visualized in iTOL (144).

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.