Direct Cobamide Remodeling via Additional Function of Cobamide Biosynthesis Protein CobS from Vibrio cholerae

ABSTRACT

Vitamin B₁₂ belongs to a family of structurally diverse cofactors with over a dozen natural analogs, collectively referred to as cobamides. Most bacteria encode cobamide-dependent enzymes, many of which can only utilize a subset of cobamide analogs. Some bacteria employ a mechanism called cobamide remodeling, a process in which cobamides are converted into other analogs to ensure that compatible cobamides are available in the cell. Here, we characterize an additional pathway for cobamide remodeling that is distinct from the previously characterized ones. Cobamide synthase (CobS) is an enzyme required for cobamide biosynthesis that attaches the lower ligand moiety in which the base varies between analogs. In a heterologous model system, we previously showed that Vibrio cholerae CobS (VcCobS) unexpectedly conferred remodeling activity in addition to performing the known cobamide biosynthesis reaction. Here, we show that additional Vibrio species perform the same remodeling reaction, and we further characterize VcCobS-mediated remodeling using bacterial genetics and in vitro assays. We demonstrate that VcCobS acts upon the cobamide pseudocobalamin directly to remodel it, a mechanism which differs from the known remodeling pathways in which cobamides are first cleaved into biosynthetic intermediates. This suggests that some CobS homologs have the additional function of cobamide remodeling, and we propose the term “direct remodeling” for this process. This characterization of yet another pathway for remodeling suggests that cobamide profiles are highly dynamic in polymicrobial environments, with remodeling pathways conferring a competitive advantage.

IMPORTANCE Cobamides are widespread cofactors that mediate metabolic interactions in complex microbial communities. Few studies directly examine cobamide profiles, but several have shown that mammalian gastrointestinal tracts are rich in cobamide analogs. Studies of intestinal bacteria, including beneficial commensals and pathogens, show variation in the ability to produce and utilize different cobamides. Some bacteria can convert imported cobamides into compatible analogs in a process called remodeling. Recent discoveries of additional cobamide remodeling pathways, including this work, suggest that remodeling is an important factor in cobamide dynamics. Characterization of such pathways is critical in understanding cobamide flux and nutrient cross-feeding in polymicrobial communities, and it facilitates the establishment of microbiome manipulation strategies via modulation of cobamide profiles.

INTRODUCTION

Vitamin B₁₂ (cyanocobalamin) is an essential micronutrient required for human health (1). It belongs to a family of compounds called cobamides that differ at the lower ligand (or α-axial) position, which falls into three major groups, benzimidazoles, purines, and phenols. A cobamide with 5,6-dimethylbenzimidazole (DMB) is referred to as cobalamin, which includes the semisynthetic form, vitamin B₁₂, that is stabilized via modification with a cyano group at the upper ligand (or β-axial) position. Cobalamin is the most widely studied cobamide due to its impact on human health, but purinyl cobamides are also common analogs found in nature. They are produced by diverse bacteria (2, 3) including cyanobacteria, which are photosynthetic bacteria that are ubiquitous in soil and aquatic environments. Many cyanobacterial species produce the cobamide pseudocobalamin (4, 5), which contains adenine at the lower ligand position. Mass spectrometry and metagenomics analysis of marine environments show that pseudocobalamin is abundant in surface waters with cyanobacteria as the likely producers (6). Phenolylcobamides are the third class of cobamides, and production of para-cresolylcobamide, with para-cresol as the lower ligand, has been observed in Negativicutes bacteria Sporomusa ovata and Veillonella parvula (7, 8). While few studies of complex environments directly assay for different cobamide analogs, analysis of mammalian gastrointestinal tracts demonstrates that they harbor rich cobamide profiles (9–11). Examination of different species of intestinal bacteria, including beneficial commensals Bacteroides thetaiotaomicron (12) and Akkermansia muciniphila (13) and pathogens Clostridium difficile (14) and Vibrio cholerae (15), demonstrates a wide range in cobamide production and utilization of different analogs. With the prevalence of cobamide import and utilization genes, including within the human gut microbiome, it is clear that cobamides play an integral role in nutritional dependencies and competition between microbes (16–19).

Cobamides are cofactors of important enzymes such as cobamide-dependent methionine synthase, methyl-malonyl coenzyme A mutase, ribonucleotide reductase, and a number of other catabolic and biosynthetic enzymes (20). While human cells in the small intestine specifically import cobalamin, bacterial specificity for cobamides is more complex. In bacteria, the cobamide-dependent enzymes can often utilize multiple analogs, but their degree of specificity varies greatly (19). While many bacterial genomes encode cobamide-dependent enzymes, only certain prokaryotes produce cobamides de novo via a well-characterized ∼30 step pathway (21). Some organisms encode partial biosynthesis pathways in which biosynthesis is completed from imported intermediates (22), such as cobinamide salvaging by Escherichia coli (23). Another pathway for cobamide production is remodeling in which an imported cobamide is cleaved into a biosynthetic intermediate and then rebuilt, often as a different analog. One remodeling pathway is mediated by CbiZ, an amidohydrolase first described in the archaeon Halobacterium (24), further characterized in Rhodobacter sphaeroides (25) and associated with cobamide remodeling in Dehalococcoides mccartyi (26). More recently, phosphodiesterase CbiR was identified as a cobamide-remodeling enzyme in Akkermansia muciniphila (13). Additionally, cobamide remodeling has been observed in two species of eukaryotic algae, but the mechanism remains unknown (4, 27).

We have previously described another remodeling pathway while examining cobamide specificity in Vibrio cholerae (15), a marine bacterium and human pathogen. V. cholerae was able to import all three classes of cobamides and the intermediate cobinamide, but showed selectivity in cobamide utilization by cobamide-dependent methionine synthase MetH; pseudocobalamin could not be utilized but was instead remodeled to cobalamin. Using a heterologous E. coli system, we demonstrated that pseudocobalamin remodeling activity was conferred by V. cholerae cobamide synthase (CobS; EC 2.7.8.26; also referred to as cobalamin synthase) but not the E. coli homolog. CobS is an integral membrane protein (28) which attaches ribosylated lower ligands to the aminopropanol arm (29, 30), a necessary step for cobamide biosynthesis. Here, we demonstrate that additional Vibrio species also perform the same remodeling reaction, and genetic analysis and in vitro assays support a remodeling pathway for direct conversion of pseudocobalamin to cobalamin by V. cholerae CobS. The previously characterized remodeling pathways involve cleavage of cobamides into biosynthetic intermediates, but here, we show a distinct remodeling mechanism mediated by cobamide synthase CobS; in addition to the previously known role in cobamide biosynthesis, some CobS homologs can remodel cobamides directly.

RESULTS

Pseudocobalamin is remodeled by diverse Vibrio species.

Over 95% of Vibrio species encode partial biosynthesis pathways (15), and these last eight steps of cobamide biosynthesis are associated with salvaging pathways in which cobamide biosynthesis is completed from imported intermediates. We previously showed that V. cholerae is able to salvage the intermediate cobinamide to produce cobalamin and could also remodel pseudocobalamin to cobalamin, which contains adenine and DMB, respectively, at the lower ligand position (Fig. 1). To examine partial biosynthesis pathways in other Vibrio species, we selected two distantly related Vibrio species for characterization. V. fluvialis is an emerging pathogen isolated from diarrheal patients (31, 32) which encodes genes for ethanolamine utilization, a cobamide-dependent process (33). V. splendidus is the dominant Vibrio species isolated from coastal waters and is pathogenic to a range of marine animals (34, 35). Analysis of V. fluvialis and V. splendidus genomes showed that, like V. cholerae, they lack the genes required for de novo biosynthesis but encode genes for partial biosynthesis pathways (Fig. 2A; see Table S1 in the supplemental material). These gene products are involved in biosynthesis of the aminopropanol arm and nucleotide loop, which contains the lower ligand. All three Vibrio species lack proteins required for the known cobamide remodeling pathways, amidohydrolase CbiZ and phosphodiesterase CbiR.

FIG 1.

Cobamide structures. The corrin ring is shown in black, the aminopropanol arm is in red, and the nucleotide loop is in blue. Lower ligand structures of cobalamin (Cbl) and pseudocobalamin (psCbl) are shown.

FIG 2.

Salvaging and remodeling pathways for cobalamin production in Vibrio species. (A) Number of cobamide biosynthesis and remodeling proteins in Vibrio species. See Table S1 in the supplemental material for details. (B to G) Cobalamin uptake or production in V. cholerae, V. fluvialis, V. splendidus, and E. coli (B to D) and E. coli ΔcobS expressing cobS from Vibrio species and E. coli (E to G). Cultures were supplemented with cobalamin for uptake assays (B and E) or supplemented with cobinamide (Cbi) (C and F) or pseudocobalamin (D and G) for salvaging and remodeling assays. DMB was also supplemented in salvaging and remodeling experiments, except for V. splendidus, which produces DMB. Cobalamin was quantified by LC-MS, and values were normalized to V. cholerae or E. coli expressing V. cholerae cobS. Graphs show averages ± standard errors of the mean from independent experiments performed with triplicate cultures, with n = 2 experiments for panels B to F and n = 3 experiments for panel G.

We assessed the ability of V. fluvialis and V. splendidus to import cobalamin and to produce cobalamin through salvaging cobinamide and remodeling pseudocobalamin (Fig. 2B to D). Like V. cholerae, V. fluvialis and V. splendidus do not produce cobamides de novo but readily import cobalamin (81% and 74%, respectively, relative to V. cholerae) and produce it via cobinamide salvaging (105% and 53%, respectively). Additionally, they can remodel pseudocobalamin to produce cobalamin (96% and 81%, respectively). In contrast, E. coli can import cobalamin and produce it via cobinamide salvaging but cannot remodel pseudocobalamin. Like V. cholerae, V. fluvialis also required supplementation with lower ligand DMB, while V. splendidus did not because it produces DMB via BluB (36). We previously found that V. cholerae CobS (VcCobS) mediated pseudocobalamin remodeling activity when expressed in a heterologous system (15). Analysis of >1,000 Vibrio CobS proteins shows that they are highly conserved with ∼55 to 60% similarity at the amino acid level; V. fluvialis and V. splendidus CobS proteins bear 66% and 61% similarity to VcCobS, respectively, and 55% similarity to each other. To determine whether pseudocobalamin remodeling in V. fluvialis and V. splendidus was mediated by CobS activity itself, V. fluvialis and V. splendidus cobS genes were expressed heterologously in E. coli ΔcobS (Fig. 2E to G). E. coli ΔcobS expressing V. fluvialis cobS (Vf cobS) imported cobalamin to the same levels as the strain expressing V. cholerae cobS (Vc cobS). Additionally, Vf cobS was functional for cobinamide salvaging and pseudocobalamin remodeling at levels comparable to Vc cobS (102% and 98%, respectively). E. coli ΔcobS expressing V. splendidus cobS (Vs cobS) could also import cobalamin, and Vs cobS supported cobinamide salvaging and pseudocobalamin remodeling, although at lower levels than the Vc cobS control (52%, 68%, and 66%, respectively). Although there was no growth defect, expression of heterologous CobS proteins in E. coli could be causing membrane stress, as has been previously observed (28); this could be the cause of the lower levels of cobalamin import and biosynthesis in E. coli expressing Vs cobS. Indeed, cobalamin import was 30 to 40% lower in E. coli expressing Vc cobS than expression of E. coli cobS (Fig. 2E) or vector-only control. Despite V. splendidus having lower overall activity compared to V. fluvialis, both strains were capable of remodeling pseudocobalamin to cobalamin (Fig. 2D), and heterologous expression of Vs cobS and Vf cobS in E. coli also supported remodeling (Fig. 2G). In contrast, expression of the E. coli homolog of cobS in the same system fully supported cobalamin import (Fig. 2E) and cobinamide salvaging (Fig. 2F) but was incapable of remodeling pseudocobalamin (Fig. 2G) (15). This analysis of diverse Vibrio species shows that CobS-mediated pseudocobalamin remodeling is not unique to V. cholerae and suggests that it could be common among Vibrio species.

Genetic analysis of cobamide remodeling in V. cholerae.

To further investigate CbiZ- and CbiR-independent remodeling in V. cholerae, we examined other genes in the cobamide partial biosynthesis pathway. The majority of Vibrio species encode eight proteins involved in the last steps of cobamide biosynthesis (Fig. 3A) which are associated with salvaging of cobamide intermediates. To determine whether these proteins are also involved in remodeling pseudocobalamin, we assayed corresponding transposon mutants from an ordered nonredundant transposon library (37). Insertion sites were verified by PCR, and all strains were tested for import of cobalamin, production of cobalamin via salvaging of cobinamide, and remodeling of pseudocobalamin (Fig. 3B to D). CobA adenosylates the cobalt ion at the upper axial position of different substrates, including cobamides and intermediates. The cobA mutant had a slight impairment in cobalamin (Cbl) uptake (75% relative to wild type) (Fig. 3B) but was greatly impaired in both salvaging cobinamide (Fig. 3C) and remodeling pseudocobalamin (Fig. 3D), supporting 10% and 24%, respectively. These results are consistent with the characterized cobinamide salvaging pathway in which the upper ligand position of cobinamide is first converted to the adenosyl (Ado) form. These results also suggest that Ado-pseudocobalamin (Ado-psCbl) is an intermediate of remodeling in V. cholerae. However, some salvaging and remodeling activity remained in the absence of CobA, suggesting that an alternative mechanism for adenosylation exists, as has been suggested for Salmonella (38), or that Vibrio enzymes act upon the cyano forms as well, albeit much less efficiently. CbiP, CobD, and CbiB function to produce the intermediate cobinamide through amidation of the corrin ring and production and attachment of the aminopropanol arm, respectively. The cbiP, cbiB, and cobD transposon mutants were not impaired for uptake of cobalamin (100%, 100%, and 85%, respectively) (Fig. 3B). As expected, they were able to fully salvage cobinamide (Fig. 3C) since their gene products participate in production of the cobinamide intermediate. Further, they were all able to remodel pseudocobalamin to 102%, 111%, and 79% of wild-type levels, respectively (Fig. 3D).

FIG 3.

Genetic analysis of salvaging and remodeling pathways in V. cholerae. (A) Last steps for biosynthesis of cobalamin as shown for the de novo pathway. Salvaging and remodeling pathways share some or all of these steps. Final dephosphorylation of Ado-cobalamin-phosphate by CobC is not pictured. During salvaging, imported cobinamide is adenosylated by CobA and phosphorylated by CobU, producing the Ado-cobinamide-P intermediate. Remodeling via CbiZ produces Ado-cobyric acid and remodeling via CbiR produces Ado-cobinamide-P. (B to H) Cobalamin (Cbl) uptake and production assays via cobinamide (Cbi) salvaging and pseudocobalamin (psCbl) remodeling in V. cholerae transposon mutants (B to D), in-frame deletion mutants (E to G), and cobU complementation for salvaging (H). DMB was also supplemented in salvaging and remodeling experiments. Cobalamin was quantified by LC-MS, and values were normalized to wild-type V. cholerae or vector control. Graphs show averages ± standard errors of the mean from independent experiments performed with triplicate cultures, with n = 2 experiments for all panels except panel D (n = 3 experiments).

The last steps of cobamide biosynthesis are completed by CobT, CobS, CobU, and CobC, and the corresponding genes are clustered at locus vc1237-vc1240. CobU activates the Ado-cobinamide-phosphate intermediate to produce the Ado-cobinamide-GDP (AdoCbi-GDP) intermediate, CobT activates the lower ligand DMB with ribose-phosphate to produce DMB-ribose-5′-phosphate (DMB-R5P), and CobS then ligates AdoCbi-GDP and DMB-R5P to produce Ado-cobalamin-P, which is subsequently dephosphorylated by CobC. The ability of the corresponding transposon mutants to salvage cobinamide (Fig. 3C) was consistent with the characterized biosynthesis pathway. Mutants of cobT, cobS, and cobU were deficient for production of cobalamin via cobinamide salvaging. Cobalamin phosphate could be desphosphorylated despite the absence of cobC (53% relative to wild type); this has been previously observed and was attributed to activity of other phosphatases or spontaneous dephosphorylation (39). These four transposon mutants were then analyzed for their ability to remodel pseudocobalamin (Fig. 3D). As with cobinamide salvaging, cobT-tn and cobS-tn were unable to remodel pseudocobalamin, while 58% dephosphorylation was observed in the cobC-tn strain. Unexpectedly, cobU-tn was able to remodel pseudocobalamin to 91% of wild-type levels despite its inability to salvage cobinamide. These results were confirmed with an in-frame deletion mutant (Fig. 3E to H); ΔcobU could not salvage cobinamide but fully remodeled pseudocobalamin, and this mutation was fully complemented for salvaging activity. Notably, instead of CobU, the archaeal cobamide biosynthesis pathway utilizes CobY, a nonorthologous protein that activates cobinamide-phosphate to produce cobinamide-GDP (40); however, cobY homologs have not been observed in bacterial genomes, and it was verified to be absent from Vibrio species. Additional in-frame deletion mutants ΔcobS and ΔcobC were examined, and these strains imported cobalamin, salvaged cobinamide, and remodeled pseudocobalamin at levels comparable to their corresponding transposon mutants. This genetic analysis of V. cholerae suggests that salvaging of cobinamide occurs via an AdoCbi-GDP intermediate, as is consistent with the previously characterized biosynthesis pathway (21, 41). However, remodeling of pseudocobalamin is not likely to occur with AdoCbi-GDP as an intermediate because CobU was dispensable for remodeling pseudocobalamin, raising the possibility that VcCobS acts upon a different substrate during remodeling. Taking into account that pseudocobalamin accumulated in strains deficient for cobS and partially accumulated in cobA-tn (Fig. S1), we hypothesized that Ado-pseudocobalamin itself is a substrate for remodeling by VcCobS.

In vitro reactions for VcCobS remodeling.

We previously demonstrated that VcCobS conferred the ability to remodel pseudocobalamin in a heterologous system, while E. coli CobS (EcCobS) did not (15), consistent with activity in the native V. cholerae and E. coli strains. To determine whether VcCobS can remodel pseudocobalamin directly, we examined cobamide production reactions in vitro (Fig. 4). Crude membrane preparations of E. coli ΔcobS expressing empty vector, VcCobS, or EcCobS were analyzed by SDS-PAGE and Western blotting, showing comparable levels of protein (Fig. S2). These membrane fractions were tested for cobinamide salvaging and pseudocobalamin remodeling, as determined by cobalamin production assayed by mass spectrometry. To establish in vitro salvaging and remodeling reactions, V. cholerae CobT, CobU, and CobC proteins and S. enterica CobU were purified, and their enzymatic activities were verified (Fig. S3). Salvaging reactions were performed as a one-pot reaction with purified Salmonella or V. cholerae CobU and V. cholerae CobC phosphatase, as has been previously performed (29), and contained substrates AdoCbi, GTP, and DMB-R5P. In vitro reaction mixtures containing VcCobS or EcCobS produced equal levels of cobalamin, with matching retention times and mass spectra as control reaction mixtures containing spiked cobalamin (Fig. 4A and B). Negative controls containing only membrane fractions or only substrates did not produce cobalamin. These in vitro salvaging assays indicate that the membrane fractions of strains expressing VcCobS or EcCobS contain similar levels of functional CobS protein and that the reaction substrates and products are consistent with those of the characterized CobS proteins from Salmonella and Pseudomonas (29, 30). Next, remodeling reactions were performed with substrates Ado-psCbl and DMB-R5P, as well as CobC phosphatase. Reaction mixtures containing VcCobS were able to produce cobalamin, with matching retention times and mass spectra as salvaging reactions and the spiked Cbl control (Fig. 4A and B). Negative controls and reaction mixtures containing EcCobS did not produce cobalamin. These results are consistent with pseudocobalamin-remodeling activity in the native V. cholerae but not E. coli and pseudocobalamin remodeling via heterologous expression of VcCobS but not EcCobS. Furthermore, these in vitro assays indicate that VcCobS directly acts upon Ado-psCbl as a substrate, catalyzing an enzymatic reaction between it and DMB-R5P, a previously unknown reaction for CobS proteins.

FIG 4.

V. cholerae CobS directly remodels pseudocobalamin to cobalamin. In vitro reactions for cobalamin production by membrane fractions of E. coli ΔcobS harboring empty vector, expressing E. coli CobS (EcCobS) or V. cholerae CobS (VcCobS), or treated with buffer. Salvaging reactions contained Ado-Cbi and DMB-R5P and were performed as a one-pot reaction with Ado-Cbi-activating enzyme CobU (from S. enterica) and GTP. Remodeling reactions contained Ado-psCbl and DMB-R5P. All reactions also contained CobC and were treated with KCN before analysis by mass spectrometry. A mock-treated sample was spiked with Ado-Cbl for comparison. (A) Extracted mass chromatograms of cyanocobalamin. All panels are scaled identically. The two peaks correspond to different −CN orientations after derivatization. (B) Corresponding mass spectra of indicated chromatogram peaks (numbered 1 to 4). (C and D) For comparison, extracted mass chromatograms and mass spectra of cyanocobinamide (C) and cyanopseudocobalamin (D).

We then examined whether the VcCobS-mediated remodeling reaction incorporates DMB-R5P in a similar manner as the characterized salvaging reaction (see Fig. 3A), i.e., the phosphoribosyl moiety is derived from DMB-R5P and not from the remodeled pseudocobalamin substrate. DMB-R5P was modified to contain a monothiophosphate moiety (also referred to as monophosphorothioate [MPS]) (Fig. 5A and Fig. S4). DMB-R5MPS was tested for incorporation by VcCobS-containing membrane fractions using in vitro reactions for salvaging and remodeling performed in the absence of CobC phosphatase. When supplemented with DMB-R5MPS, both salvaging and remodeling reactions produced cobalamin-MPS (Fig. 5B and D). Compared to the cobalamin phosphate produced in control reactions supplemented with DMB-R5P (Fig. 5B and C), the m/z of cobalamin-MPS was shifted by 8 (m = 16, z = +2), and the corresponding liquid chromatography-mass spectrometry (LC-MS) peaks were shifted 0.25 to 0.28 min later. Incorporation of the MPS moiety suggests that VcCobS attaches the complete DMB-R5P structure (as occurs in salvaging reactions) to the cobinamide-phosphate backbone from the pseudocobalamin substrate to produce cobalamin phosphate. Because this VcCobS-mediated pathway for pseudocobalamin remodeling does not rely on formation of a biosynthetic intermediate, and VcCobS instead acts upon pseudocobalamin itself, we propose that this pathway be called “direct remodeling” (Fig. 6).

FIG 5.

VcCobS-mediated salvaging and remodeling reactions incorporate the same nucleotide loop substrate. (A) Structures of nucleotide loop substrates DMB-R5P and DMB-R5MPS, a monothiophosphate analog. (B) Membrane fractions containing VcCobS were used for in vitro salvaging (plus AdoCbi) and remodeling (plus Ado-psCbl) reaction mixtures containing either DMB-R5P or DMB-R5MPS. Reactions were performed in the absence of CobC phosphatase. Samples were treated with KCN before LC-MS analysis. The two peaks correspond to different −CN orientations after derivatization, and extracted mass chromatograms are scaled identically for each analyte. (C and D) Mass spectra of CN-cobalamin-phosphate (C) and CN-cobalamin-MPS (D) reaction products.

FIG 6.

Pathways for cobamide biosynthesis, including direct remodeling mediated by Vibrio CobS. The de novo pathway requires ∼30 enzymatic steps and includes cobyric acid and cobinamide-phosphate intermediates. In salvaging pathways, biosynthesis is completed from imported intermediates such as cobyric acid and cobinamide. In remodeling pathways, imported cobamides are cleaved by CbiZ to produce cobyric acid or CbiR to produce cobinamide-phosphate. In direct remodeling, some CobS proteins (shaded) can act upon cobamides themselves, converting them directly into another analog. Aminopropanol arm is shown in red, and nucleotide loops with different lower ligands are in gray and blue. The adenosine upper ligand is not shown.

DISCUSSION

Previous work has established the CobS-catalyzed reaction required for the de novo biosynthesis of cobalamin; in vitro assays with crude preparations or partially purified CobS from Salmonella and Pseudomonas (29, 30) demonstrate their catalytic activity of attaching the activated DMB ligand (DMB-R5P) to the aminopropanol arm of the activated cobinamide intermediate (AdoCbi-GDP). Recent work with Salmonella CobS showed that activity was significantly improved upon reconstitution in liposomes (42), highlighting the importance of membrane association to CobS and cobamide biosynthesis. Here, we expand our understanding of the biochemistry of CobS proteins through the identification of additional catalytic capabilities of some homologs. It is known that individual bacterial strains can produce more than one cobamide analog (3, 43), and therefore, CobS proteins must be able to accommodate other lower ligand substrates besides DMB-R5P. Here, we show that V. cholerae CobS can also accommodate the cobamide Ado-pseudocobalamin as a substrate for remodeling in addition to the known biosynthetic intermediate AdoCbi-GDP during salvaging. While the corrin ring-containing substrate differs, we demonstrated that V. cholerae CobS similarly incorporates the complete DMB-R5P substrate in both salvaging and remodeling reactions. CobS forms a phosphodiester bond between the 3′ hydroxyl group of the ribose moiety of DMB-R5P and the phosphate from the aminopropanol arm of AdoCbi-GDP during salvaging; our data suggest that the same bond is formed with the aminopropanol phosphate from Ado-pseudocobalamin during direct remodeling.

The direct remodeling pathway differs from the previously characterized cobamide remodeling pathways, which require the activity of hydrolytic enzymes to produce biosynthetic intermediates. The amidohydrolase CbiZ produces the intermediate cobyric acid by cleaving the amide bond connecting the aminopropanol arm to the corrin ring. The phosphodiesterase CbiR cleaves cobamides at the phosphodiester bond that attaches the nucleotide loop, liberating the ribosyl-lower ligand and producing the cobinamide-phosphate intermediate. Interestingly, the likely mechanism for VcCobS remodeling relies on cleavage of the same phosphodiester bond as with CbiR, but the cobinamide-phosphate produced is a catalytic intermediate instead of a biosynthetic intermediate. This discovery of additional catalytic activity by some homologs of CobS underscores how little we know about this enzyme. It is an inner membrane protein with 4 to 6 predicted transmembrane domains (28), but not much else is known about how it functions. With recent breakthroughs in membrane protein purification, further investigations with purified CobS proteins will allow more detailed biochemical characterization of the different reactions that CobS proteins catalyze.

Cobamide profiles have only been examined in several studies, including samples from marine environments (6) and fecal and intestinal samples from human, murine, or bovine sources (9–11). These polymicrobial environments are rife with microbial competition for resources, such as cobamides, and recent discoveries of additional remodeling pathways suggest that cobamide remodeling could contribute to microbial fitness in competitive environments. Here, we show that multiple Vibrio species can remodel pseudocobalamin, likely through the same direct remodeling pathway as V. cholerae. As marine bacteria, Vibrio species are found in coastal waters, an environment rich with cyanobacteria such as Synechococcus species, which have been shown to produce pseudocobalamin. Although Vibrio species make up a small percentage of total marine bacterioplankton, V. splendidus is one of the more abundant (35). Additionally, Vibrio species are often associated with biotic and abiotic surfaces and could have large effects on local cobamide profiles.

Another environment in which purine-containing cobamides are abundant is the human gut (9). Only some bacteria are de novo producers, and analysis of different groups of intestinal bacteria suggests that nutritional dependencies and vitamin cross-feeding are pervasive in the intestinal environment (44). Most gut microbes encode cobamide transporters, and most bacteria contain cobamide-dependent enzymes (12, 22, 45), many of which have specificity for certain cobamides. Humans must acquire vitamin B₁₂ (cobalamin) through diet, and any unabsorbed cobalamin travels to the large intestine, but benzimidazole-containing cobamides are a minor component in the human gut. Cobamide biosynthesis in situ produces a range of cobamide analogs, and cobamide remodeling likely shapes cobamide profiles as well. The recently described remodeling protein CbiR was discovered in keystone species Akkermansia muciniphila, with metabolic shifts occurring when cocultured with pseudocobalamin-producing Eubacterium hallii (13, 46). CbiR was shown to act upon a range of cobamides, supporting remodeling into pseudocobalamin by A. muciniphila. Protein similarity networks cluster CobS proteins from microbiome-associated bacteria with that of Vibrio species (see Fig. S5 in the supplemental material), raising the possibility that direct remodeling could also contribute to the cobamide profile in the human gut environment. Some cobamide-dependent pathways are utilized by intestinal pathogens, such as ethanolamine degradation and propanediol utilization (41, 47). V. cholerae and V. fluvialis are also intestinal pathogens that encode cobamide-dependent enzymes; both encode MetH, while V. fluvialis also encodes the ethanolamine degradation pathway. These pathogens could also utilize cobamides to support their metabolism during colonization and growth since the small intestinal microbiome harbors de novo cobamide producers (48). To fully understand the microbial interactions that occur in the human gut, mediators such as cobamides must be characterized. Cobamide dynamics are complex, with potential specificity for different analogs during import, utilization, and remodeling. Further investigation of remodeling pathways of commensal and pathogenic bacteria will facilitate prediction of cobamide dynamics in complex microbial communities, such as the mammalian gastrointestinal tract.

MATERIALS AND METHODS

Strain cultivation and construction.

Standard cultivation of Vibrio cholerae C6706, V. cholerae transposon mutants (37), Vibrio fluvialis strain XJ85003 (31), and E. coli BL21(λDE3), BL21(λDE3)*, and BL21(λDE3)* ΔcobS (15) was performed in LB medium, with cultures grown with shaking overnight at 37°C. Vibrio splendidus (DSM 19640) was grown with shaking overnight at room temperature (25°C) in tryptic soy broth (Difco) modified to contain 2% (wt/vol) NaCl. All cloning was performed using isothermal assembly (NEB HiFi). In-frame deletion mutants were constructed via allelic exchange using suicide vector pWM91. For expression and protein purification, genes were cloned into the NdeI/XhoI sites of pET16b or pET29a vectors, which introduced N-terminal His₁₀ or C-terminal His₆ affinity tags, respectively. V. fluvialis and V. splendidus cobS genes were cloned into pET16b vectors before transformation into BL21(λDE3)*ΔcobS. V. cholerae cobU was cloned into pBAD24 before electroporation into V. cholerae ΔcobU. For protein expression, V. cholerae cobT, cobU, and cobC and Salmonella enterica cobU (STM14_2506 from strain 14028S) were cloned into pET29a vectors and transformed into BL21(λDE3). For production of DMB-R5MPS, human nicotinamide riboside kinase 1 (NRK1) (49) was codon optimized for expression in E. coli, synthesized (Thermo Fisher), and cloned into pET29a before transformation into BL21(λDE3). Supplements were added, if needed, as follows: 100 μg per ml streptomycin, 50 μg per ml kanamycin, 100 μg per ml ampicillin, and 0.01% (wt/vol) arabinose. Strains and mutants used in this study are given in Table 1.

TABLE 1.

Strains and mutants used in this study

Strain or mutant	Description	Source and/or reference no.
Vibrio species strains
C6706	Wild-type V. cholerae	Zhu Lab
XJ85003	Wild-type V. fluvialis	Zhu Lab (31)
DSM 19640	Wild-type V. splendidus	DSMZ
V. cholerae mutants
cobA-tn	Transposon mutant in vc1040	Mekalanos Lab (37)
cbiP-tn	Transposon mutant in vca0727	Mekalanos Lab (37)
cbiB-tn	Transposon mutant in vc2380	Mekalanos Lab (37)
cobD-tn	Transposon mutant in vc1134	Mekalanos Lab (37)
cobT-tn	Transposon mutant in vc1237	Mekalanos Lab (37)
cobS-tn	Transposon mutant in vc1238	Mekalanos Lab (37)
cobU-tn	Transposon mutant in vc1239	Mekalanos Lab (37)
cobC-tn	Transposon mutant in vc1240	Mekalanos Lab (37)
ΔcobS	In-frame deletion mutant	15
ΔcobU	In-frame deletion mutant	This work
ΔcobU vector	ΔcobU pBAD24	This work
ΔcobU + cobU	ΔcobU pBAD24 cobU	This work
ΔcobC	In-frame deletion mutant	This work
E. coli strains
BL21*	BL21(λDE3) btuB+	15
BL21*ΔcobS	BL21(λDE30 btuB+ ΔcobS	15
BL21*ΔcobS VcCobS	pET16b V. cholerae cobS	15
BL21*ΔcobS EcCobS	pET16b E. coli cobS	15
BL21*ΔcobS VfCobS	pET16b V. fluvialis cobS	This work
BL21*ΔcobS VsCobS	pET16b V. splendidus cobS	This work
BL21(λDE3) VcCobT	pET29a V. cholerae cobT	This work
BL21(λDE3) VcCobU	pET29a V. cholerae cobU	This work
BL21(λDE3) SeCobU	pET29a S. enterica cobU	This work
BL21(λDE3) VcCobC	pET29a V. cholerae cobC	This work
BL21(λDE3) NRK1	pET29a NRK1	This work (49)

Bacterial cobamide uptake and biosynthesis assays.

Cobalamin uptake, cobinamide salvaging, and pseudocobalamin-remodeling assays in Vibrio species and BL21(λDE3)-derived strains were performed as described (15). Strains were cultured in M9 minimal medium supplemented with 0.4% (wt/vol) glucose (22.2 mM) with shaking overnight at 37°C with antibiotics if required. For V. splendidus, cobamide assays were performed in M9 minimal medium (Difco) containing 2% (wt/vol) NaCl and grown with shaking overnight at 25°C. M9 minimal medium was supplemented with 5 nM cyanocobalamin (Sigma), dicyanocobinamide (Sigma), or cyanopseudocobalamin purified from Lactobacillus reuteri (15) and with 10 μM 5,6-dimethylbenzimidazole (DMB; Sigma) if required. Lower concentrations (100 nM and 10 nM DMB) also supported full salvaging and remodeling activity (Fig. S1C to D in the supplemental material). After 16 to 18 h of growth, cells were pelleted by centrifugation, extracted with MeOH, treated with KCN, and analyzed by mass spectrometry for cyanocobalamin (m/z 678.2861, z = +2) using a Waters Synapt mass spectrometer equipped with a Waters Acquity column (UPLC BEH; 2.1 by 50 mm) and a Waters Vanguard precolumn (UPLC BEH C₁₈; 2.1 by 5 mm) (15). Cyanocobalamin values were normalized to stated controls, and standard deviations of normalization controls were less than 10% for each experiment. Statistical analysis was performed using one-way analysis of variance (ANOVA) tests with Tukey posttesting (GraphPad Prism version 5.0b). All strains grew to the same final cell density compared to V. cholerae or E. coli controls, except for V. splendidus and V. cholerae cobD-tn, which grew to ∼60% and ∼40% of wild-type V. cholerae, respectively. However, normalizing cobamide uptake and biosynthesis values by cell density resulted in an overcorrection. Control experiments show that cobamide uptake and biosynthesis (at the 5 nM concentration used for experiments) are not saturated in this cell density range.

Protein purification and membrane fraction preparation.

Proteins were purified from BL21(λDE3) harboring pET29a plasmids expressing V. cholerae cobT, V. cholerae cobC, V. cholerae cobU, S. enterica cobU, and human NRK1. Five hundred milliliter LB cultures were grown with shaking at 37°C until an optical density at 600 nm (OD₆₀₀) of 0.6 to 0.8 was reached and then induced with 500 μM IPTG (isopropyl-β-D-thiogalactopyranoside) overnight with shaking at room temperature. Cell pellets were resuspended in lysis buffer (50 mM potassium phosphate [pH 7.4], 500 mM sodium acetate, 0.1 mM EDTA, 20% glycerol, and 1 μg/ml RNase), sonicated, purified with nickel-nitrilotriacetic acid (Ni-NTA) resin, and eluted with 200 mM imidazole. Purified proteins were buffer exchanged via protein concentrator spin columns and stored at −20°C in 50 mM phosphate buffer [pH 7.4], 100 mM NaCl, 2 mM MgCl₂, and 20% glycerol. For isolation of membrane fractions containing CobS proteins, cultures were grown as described above but induced with 100 μM IPTG. Pellets were resuspended in buffer (20 mM sodium phosphate [pH 7.4], 100 mM NaCl, 1 mM MgCl₂, and 1 μg/ml RNase and DNase), lysed via Emulsiflex cell disruptor, and clarified via low-speed spin (8,000 × g). Membrane fractions were pelleted via ultracentrifugation (163,000 × g) and resuspended via homogenizer in 20 mM sodium phosphate (NaH₂PO₄; pH 7.4), 100 mM NaCl, and 2 mM MgCl₂.

Substrate preparation for in vitro assays.

Substrates for in vitro assays were prepared and purified by high-performance liquid chromatography (HPLC) before quantification as described below. Subsequent preparations were purified over Seppak C₁₈ columns and quantified via mass spectrometry, using previously quantified substrates as standards. AdoCbi and Ado-psCbl were obtained through supplementation of V. cholerae ΔcobU or ΔcobS strains, which accumulate the respective Ado substrates. Strains were grown as 500 ml cultures of M9 minimal medium with 25 nM dicyanocobinamide [(CN)₂Cbi], (Sigma) or 25 nM cyanopseudocobalamin (CNpsCbl) isolated from Lactobacillus reuteri (15). DMB-R5P was produced using purified V. cholerae CobT, DMB, and NaMN as described (50) (see Fig. S3B). For production of DMB-R5MPS (see Fig. S4), 0.2 mg/ml purified NRK1 was incubated with 100 μM nicotinamide riboside, 1 mM ATPγS, 5 mM MgCl₂, and 50 mM Tris buffer (pH 7.8) at 37°C for 1 h. Reaction progression was monitored for formation of nicotinamide riboside monothiophosphate (NRMPS) by LC-MS. After 1 h, 0.2 mg/ml CobT and 1 mM DMB were added and incubated at 37°C for 2 h. AdoCbi, Ado-psCbl, DMB-R5P, and DMB-R5MPS substrates were purified using a Varian semipreparative reverse-phase (RP)-HPLC system with a Phenomenex Luna Omega 5-μm polar C₁₈ 250- by 12-mm column using the following gradient: 95:5 to 70:30 water-acetonitrile containing 0.1% formic acid with a 15 ml-per-minute flow rate. Substrates were lyophilized, resuspended in ultrapure water, and frozen at −20°C. Quantification of substrates was performed via UV absorbance using established extinction coefficients (51, 52). Light exposure to AdoCbi and Ado-psCbl was minimized.

In vitro cobamide production assays.

We added 5 μl of membrane fractions or buffer to 10 μl of master mix for salvaging and remodeling reactions. Salvaging reactions contained 10 μM AdoCbi, 10 μM DMB-R5P, 10 μg purified CobC, 10 μg purified CobU, 2 mM GTP, and 10 mM MgCl₂. Remodeling reactions contained 10 μM Ado-psCbl, 10 μM DMB-R5P, 10 μg purified CobC, and 10 mM MgCl₂. Salvaging and remodeling assays with DMB-R5MPS were performed in the absence of CobC, and concentrations of AdoCbi, Ado-psCbl, DMB-R5P, and DMB-R5MPS substrates were reduced to 4 μM. After incubating in vitro reactions for 1 h at 37°C in the dark, 5 μl of 10 mg per ml KCN was added with additional incubation at 37°C for 5 min. Clarified samples were analyzed by mass spectrometry for production of CNCbl, CNCbl-PO₃H₂, and CNCbl-PO₂H₂S as described (15).

Bioinformatics.

Identification of cobamide biosynthesis and remodeling protein homologs in V. fluvialis and V. splendidus were performed by BLASTP, as described (15). See Table S1 for details.