A low-oxygen process in gut bacteria that turns a nutrient from l-carnitine into trimethylamine

Only a single cultured bacterium, a human fecal isolate of E. timonensis, has been reported to produce TMA from γbb under anaerobic conditions (13). We began our efforts to identify the genes and enzymes responsible for γbb metabolism by testing the type strain E. timonensis SN18 for this activity. Indeed, anaerobic cultures of E. timonensis SN18 supplemented with γbb demonstrated complete consumption of this substrate and production of a stoichiometric amount of TMA (SI Appendix, Fig. S1↗). When the cultures were supplemented with [N-(CD₃)₃]-γbb, deuterium-labeled TMA was detected by liquid chromatography–tandem mass spectrometry (LC–MS/MS) (SI Appendix, Fig. S1↗), confirming its origin from γbb. Resting cell suspensions of E. timonensis SN18 also fully converted γbb to TMA (Fig. 2A). However, this activity was only observed when the cells had been cultured in medium containing γbb; resting cell suspensions of E. timonensis SN18 cultured in the absence of γbb were unable to consume γbb and did not generate TMA (Fig. 2A). Together, these results suggest that γbb induces expression of the genes involved in its metabolism.

We also evaluated the ability of E. timonensis SN18 to metabolize dl-carnitine and γ-aminobutyric acid (GABA), close structural homologs of γbb. Neither of these substrates was consumed during growth (SI Appendix, Fig. S1↗) nor in resting cell suspensions of cultures supplemented with γbb (SI Appendix, Fig. S1↗). Next, we tested whether these substrate analogs could induce expression of the γbb metabolic enzymes. Resting cell suspensions of cultures supplemented with GABA did not exhibit γbb metabolism; however, resting cell suspensions from cultures supplemented with dl-carnitine did convert γbb to TMA (Fig. 2A). Thus, while dl-carnitine appears to induce expression of the genes responsible for γbb metabolism in E. timonensis, the metabolic enzymes are unable to use dl-carnitine as a substrate for TMA generation.

The inducible metabolism of γbb in E. timonensis SN18 suggested that the genes involved in this pathway could be discovered using RNA-sequencing (RNA-seq). Only two sets of genes were both highly and differentially expressed in γbb-induced cultures compared to control cultures grown in the absence of γbb (Fig. 2B and Dataset S1↗). As expected from our previous cell-based activity experiments, expression of these genes was also up-regulated in cells grown with dl-carnitine, but not with GABA (SI Appendix, Fig. S2↗ and Dataset S1↗). The most differentially up-regulated genes are colocalized in the E. timonensis SN18 genome within a region we designate the γbbutilization (bbu) gene cluster (Fig. 2C). This six-gene cluster encodes a predicted flavin-dependent acyl-CoA dehydrogenase (bbuA), two class III CoA transferases (bbuB and bbuC), a di-[4Fe-4S]-cluster ferredoxin (bbuD), a betaine/carnitine/choline transporter (bbuE), and an acyl-CoA thioesterase (bbuF). The second set of genes up-regulated in γbb-induced cells encodes for predicted riboflavin biosynthesis enzymes (Dataset S1↗), which likely support cofactor biosynthesis for the enzyme encoded by bbuA. The highly specific transcriptional response to γbb suggested that the bbu gene cluster was involved in γbb metabolism. In addition, the presence of a gene encoding a transporter for trimethylammonium-containing compounds (bbuE) was consistent with this hypothesis.

Based on the predicted annotations of the bbu genes, we hypothesized that γbb would be initially activated for subsequent TMA elimination through formation of a CoA thioester. The conversion of a carboxylic acid to a thioester is a common first step in many metabolic pathways as it lowers the pK_a of the C_α-protons, facilitating further reactions (26, 27). Indeed, in resting cell suspensions incubated with γbb, a metabolite was detected by liquid chromatography–mass spectrometry (LC–MS) with a peak retention time and mass-to-charge ratio (895.5 m/z, [M+H]⁺) matching those of a γbb-CoA standard (Fig. 3A and SI Appendix, Fig. S3↗). The origin of this metabolite was confirmed by incubating cell suspensions with deuterium-labeled 2,2,3,3,4,4-D₆-γbb (D₆-γbb), which led to an increase of 6 Da (901.5 m/z, [M+H]⁺) for the assigned γbb-CoA peak (Fig. 3A and SI Appendix, Fig. S3↗). These results support the hypothesis that γbb metabolism proceeds via a CoA thioester intermediate.

We next investigated whether two CoA transferases encoded by the bbu gene cluster, BbuB and BbuC, are directly involved in γbb-CoA production. These proteins both belong to the class III CoA transferase family but only share 30% amino acid sequence identity with one another. Indeed, γbb-CoA was detected from incubations of γbb with crude lysate of E. coli cells constitutively expressing the first three genes of the cluster (bbuABC) (Fig. 3B). Interestingly, γbb-CoA was not detected when either of the two CoA transferase genes was expressed alone (Fig. 3B). Production of γbb-CoA was also observed when recombinant, purified BbuB and BbuC were added together with γbb to lysate of E. coli transformed with empty vector (SI Appendix, Fig. S3↗) but, again, only when both proteins were present. The requirement for both CoA transferase proteins was also confirmed in vitro. In these experiments, the BbuB and BbuC proteins together generated γbb-CoA from γbb and an appropriate acyl-CoA donor substrate (Fig. 3C, SI Appendix, Fig. S3↗).

Based on these biochemical activity results, we proposed that the BbuB and BbuC form a functional hetero-oligomeric complex. To test our hypothesis, we conducted a pull-down affinity chromatography experiment using C-terminal streptavidin-tagged BbuB (BbuB-Strep) and N-terminal six histidine-tagged BbuC (BbuC-His). Lysates from separate heterologous expressions of each protein in E. coli were combined and loaded onto Ni²⁺-NTA resin to bind the BbuC-His protein, after which the resin was washed with buffers containing increasing concentrations of imidazole. We predicted that if the BbuB and BbuC proteins form a stable complex, high concentrations of imidazole would be required to displace the BbuB-Strep protein along with the BbuC-His protein. Western blot analysis of elution fractions using a streptavidin-specific antibody demonstrated that the BbuB-Strep protein was present in fractions eluted with high concentrations (250 mM) of imidazole (SI Appendix, Fig. S4↗). In contrast, control experiments without the BbuC-His protein showed BbuB-Strep protein only in the initial wash fractions with low concentrations (25 mM) of imidazole (SI Appendix, Fig. S4↗). These results provide initial evidence that BbuB and BbuC form a stable hetero-oligomeric complex, although the nature of the complex (e.g., stoichiometry, catalytic roles of each monomer) remains to be determined.

We next sought to determine the substrate specificity of the BbuB/C enzyme reaction. Various short-chain fatty acyl-CoAs (acetyl-, butyryl-, propionyl-, and crotonyl-CoA) were tested as CoA-donating substrates. While all compounds resulted in production of γbb-CoA, acetyl-CoA was the preferred cosubstrate (SI Appendix, Fig. S3↗). This preference for acetyl-CoA in vitro, however, does not preclude the use of an alternative donor in vivo depending on availability or requirements for energy conservation (e.g., CoA recycling via butyryl-CoA). We also tested dl-carnitine and GABA as alternative CoA acceptors in vitro with BbuB/C and acetyl-CoA as the donor. However, neither of the presumed products, carnitinyl-CoA or γ-aminobutyryl-CoA, was detected by LC–MS/MS under the reaction conditions tested (SI Appendix, Fig. S3↗). The high selectivity of the CoA transferases BbuB and BbuC for γbb directly connects the bbu gene cluster to γbb metabolism.

The key step in TMA production from γbb is breaking an unactivated C–N bond. None of the bbu genes is predicted to encode homologs of known enzymes that catalyze C–N bond cleavage and release TMA [e.g., choline TMA-lyase (28), glycine betaine reductase (29), ergothionase (30)], indicating that this pathway uses a distinct enzyme for this critical step. We identified the predicted acyl-CoA dehydrogenase-like enzyme BbuA as the most likely candidate for this reaction. Enzymes belonging to the acyl-CoA dehydrogenase family are flavin-dependent oxidoreductases that act on acyl-CoA thioester substrates, typically installing or removing an α,β-unsaturation through two-electron oxidation or reduction reactions, respectively (31). We envisioned that BbuA might instead use a flavin cofactor in a cryptic radical mechanism to activate γbb-CoA for TMA elimination from the C4 position. This proposed reaction resembles the C4 elimination of water from 4-hydroxybutyryl-CoA catalyzed by the flavin adenine dinucleotide (FAD)-dependent enzyme 4-hydroxybutyryl-CoA dehydratase (4HBD) involved in bacterial GABA and succinate metabolism (32). While BbuA and 4HBD are predicted to have a similar structure and harbor a flavin cofactor, they share very minimal sequence identity (∼10%).

Initial attempts to access soluble BbuA protein via heterologous expression in E. coli were unsuccessful. Soluble BbuA protein was only obtained under the following conditions: 1) coexpression with the native GroEL-ES proteins from E. timonensis SN18, 2) addition of exogenous riboflavin to the growth medium, and 3) addition of excess FAD during cell lysis. Crude lysates of E. coli cells overexpressing the bbuABC genes under these conditions demonstrated TMA production when incubated anaerobically with γbb and acetyl-CoA (SI Appendix, Fig. S5↗). Purified BbuA containing oxidized FAD was then tested for activity in vitro when incubated with γbb, acetyl-CoA, and the CoA transferases BbuBC. Robust TMA production was observed after a 1-h incubation without addition of redox mediators (Fig. 4B), demonstrating that these reaction components are sufficient to catalyze multiple turnovers of γbb. To probe whether the FAD cofactor is essential for BbuA activity, chemically reduced protein was tested for TMA production, as complete exclusion of FAD from the protein preparation led to protein insolubility. When BbuA was reduced with excess sodium dithionite, TMA production was severely diminished (SI Appendix, Fig. S5↗), suggesting that the FAD cofactor is required for activity and serves a redox role in the reaction. Finally, γbb-CoA was tested directly as the substrate for BbuA in vitro. This reaction also resulted in TMA production (Fig. 4C), demonstrating that BbuA alone is sufficient to catalyze TMA elimination in vitro. These results demonstrate that BbuA is the critical TMA-generating enzyme in O₂-independent γbb metabolism by E. timonensis.

Next, we sought to determine the coproduct of the BbuA-catalyzed reaction. We predicted that crotonyl-CoA would be generated upon elimination of TMA from γbb-CoA by analogy to 4-hydroxybutyryl-CoA dehydration in GABA/succinate metabolism (32). Indeed, crotonyl-CoA was detected by LC–MS/MS in both the E. coli crude lysate incubations (SI Appendix, Fig. S5↗) and the in vitro assays with purified BbuA (Fig. 4). Both the retention time and m/z of the detected metabolite were consistent with a crotonyl-CoA standard (SI Appendix, Fig. S5↗). The production of crotonyl-CoA supports the proposed redox-neutral TMA elimination from γbb-CoA (Fig. 4). Together, the gain-of-function experiments in E. coli and in vitro activity assays demonstrate that the bbu gene cluster is responsible for TMA production from γbb and that the three genes bbuABC are necessary and sufficient to confer this function in E. coli and in vitro.

Although crotonyl-CoA was produced in lysate and in vitro reactions containing BbuA, crotonyl-CoA did not accumulate in resting cell suspensions of E. timonensis SN18 incubated with γbb. Instead, we detected the known products of bacterial crotonyl-CoA metabolism: 3-hydroxybutyryl-CoA, acetyl-CoA, and butyryl-CoA (Fig. 5B). Inspection of their mass spectra showed deuterium incorporation when D₆-γbb was used, confirming their origin from γbb (SI Appendix, Fig. S6↗). These results suggest that crotonyl-CoA generated from γbb is further metabolized by E. timonensis. Notably, the bbu gene cluster does not encode any enzymes known to be involved in crotonyl-CoA metabolism (33). However, protein homologs of enoyl-CoA hydratase, 3-hydroxybutyryl-CoA dehydrogenase, and butyryl-CoA dehydrogenase are encoded by constitutively expressed genes located elsewhere in the E. timonensis SN18 genome (Dataset S2↗). To test whether they could account for the observed crotonyl-CoA metabolism, crotonyl-CoA was incubated with cell lysates of E. timonensis SN18 grown in the presence or absence of γbb. In both cases, crotonyl-CoA was converted to 3-hydroxybutyryl-CoA during an hour incubation (SI Appendix, Fig. S7↗). Longer incubations led to accumulation of acetyl-CoA and low levels of butyryl-CoA (SI Appendix, Fig. S7↗). However, these downstream conversions in the assay were likely limited by the cofactor concentrations present in the lysate and thus do not represent the native partition of the end-products from crotonyl-CoA metabolism. These results demonstrate that the final steps of γbb metabolism in E. timonensis SN18 are known transformations of crotonyl-CoA, catalyzed by constitutively produced enzymes encoded outside of the bbu gene cluster (Fig. 5A).

Some species of Clostridia can use crotonyl-CoA metabolism for anaerobic respiration via reduction to butyryl-CoA and as a carbon and energy source through conversion to acetyl-CoA (Fig. 5A) (34, 35). We examined whether γbb metabolism via crotonyl-CoA would enhance the growth of E. timonensis SN18 in media conditions with limited electron acceptors or carbon sources. In both culture conditions, the growth yield was markedly enhanced with γbb supplementation (Fig. 5 C and D) and the growth rate was consistent with the rates of γbb depletion and TMA production (SI Appendix, Fig. S8↗). These results support a key role for crotonyl-CoA as an intermediate in γbb metabolism and indicate a physiological role for this activity in E. timonensis.

Due to the prominent role of butyrate in the human gut, we next investigated whether free butyrate was produced from E. timonensis γbb metabolism. Using a carboxylate-derivatization method and LC–MS, elevated levels of butyrate were detected in suspensions of γbb-induced cells incubated with γbb compared to suspensions of noninduced cells (Fig. 5E and SI Appendix, Fig. S9↗). In addition, deuterium incorporation into butyrate was observed when D₆-γbb was used (Fig. 5F). Analysis of the mass spectrum revealed that the D₆-isotopolog was not present; rather the +4 Da species was the primary isotopolog detected (Fig. 5F and SI Appendix, Fig. S9↗). The loss of two deuteria in the butyrate product is consistent with a crotonyl-CoA precursor. To rule out the possibility that the loss of label resulted from exchange of the acidic C_α-deuterons with protons from solution, a control experiment was performed incubating perdeuterated D₇-butyrate with cell suspensions. The mass spectrum of butyrate from this experiment showed minimal deuterium exchange with solvent protons during the same incubation period (SI Appendix, Fig. S9↗). In addition, γbb-CoA detected from incubations with D₆-γbb showed minimal deuterium loss (SI Appendix, Fig. S9↗). These control reactions support the proposal that γbb metabolism in E. timonensis proceeds through the α,β-unsaturated crotonyl-CoA intermediate and generates free butyrate in resting cell suspensions. To further investigate whether butyrate was the major end-product of γbb metabolism in actively growing E. timonensis, we quantified derivatized butyrate by LC–MS from cultures grown with D₆-γbb in rich medium and the two types of minimal media described above. Deuterium-labeled butyrate accounted for only 8.5% of the total butyrate produced in rich media but up to 34% and 40% of the total butyrate produced in minimal media with limited electron acceptors and carbon sources, respectively. However, in each medium, the deuterium-labeled butyrate only amounted to ∼5% of the original γbb, indicating that butyrate is not the major final product of this metabolism and thus is not likely to contribute substantially to overall levels of butyrate produced in the gut.

As noted previously, E. timonesis is the only bacterial species known to metabolize γbb to TMA under anaerobic conditions. Having discovered the gene cluster responsible for this activity, we searched for homologous gene clusters in other sequenced bacterial genomes. We identified bbu-like gene clusters in four host-associated, cultured bacteria belonging to the order Clostridiales. To examine whether the bbu genes are diagnostic of γbb metabolism, two commercially available strains, the human oral isolate Eubacterium minutum American Type Culture Collection (ATCC) 700079 and the feline gut isolate Agathobaculum desmolans ATCC 43058 (previously Eubacterium desmolans or Butyricoccus desmolans), were tested for this activity. Both strains converted γbb to TMA in anaerobic cultures (Fig. 6A), demonstrating that the presence of the bbu gene cluster predicts the capacity for anaerobic γbb metabolism. The rarity of this gene cluster in sequenced genomes of cultured bacteria motivated us to search uncultured bacterial genomes that had been assembled from human metagenomes (i.e., metagenome-assembled genomes [MAGs]) (36, 37). We identified homologous bbu gene clusters in 11 uncultured, human-associated MAGs, all of which belong to the order Clostridiales (Fig. 6B and Dataset S3↗). Two of the uncultured bacterial genomes were assembled from human oral samples, while the remaining genomes harboring the bbu gene cluster were assembled from human stool samples. The presence of the bbu gene cluster in multiple uncultured gut bacteria suggests that γbb metabolism is underrepresented among cultured isolates and that E. timonensis may not be solely responsible for this metabolism in the human gut. Finally, we searched for the presence of the cai genes responsible for metabolizing l-carnitine to γbb in these bbu-containing bacterial genomes. However, none of these genomes possess homologs of known cai genes, indicating that these γbb-metabolizing bacteria are incapable of the full conversion of l-carnitine to TMA.

Motivated by the presence of the bbu gene cluster in uncultured gut bacteria and the implications for γbb metabolism in human health, we assessed the prevalence of the bbu gene cluster in publicly available human metagenomic datasets. Our analysis used the bbuA gene as a representative of the full cluster in stool metagenomes from healthy adult populations. We detected hits for the bbuA gene in samples from every cohort, but the proportion of bbuA-positive samples varied substantially between cohorts, ranging from 15% to 85% positive (Fig. 7A). We also quantified the abundance of the bbuA gene in bbuA-positive samples and found that it was consistent across cohorts at 1:1,000 genes per microbial genome, with a range of 1:10² to 1:10⁴ genes per microbial genome (SI Appendix, Fig. S10↗). Since diet has been reported as an important factor in TMA production capacity of the microbiota, we examined whether bbuA gene presence or abundance was correlated with diet. However, neither bbuA gene presence nor abundance was correlated with self-reported omnivorous or vegetarian donors from the BIO-ML cohort (SI Appendix, Fig. S11↗). Overall, the distribution profile of bbu genes suggests that γbb-metabolizing bacteria are frequently present in the human gut microbiota but are low-abundance members of the community.

To provide further support for the relevance of the bbu pathway in the human gut, we evaluated gene expression using metatranscriptomic data from three cohorts: HMP2 (38), iHMP (39), and MLVS (40). We detected bbuA transcripts in 27% to 56% of samples in these cohorts (Fig. 7B), suggesting that the bbu pathway is expressed in the human gut. Transcripts of the bbuA gene were detected in the same samples that were positive for the bbuA gene but also in a subset (16% to 26%) of bbuA-negative samples (SI Appendix, Fig. S12↗). This observation suggests that the bbuA gene abundance was too low to be detected by the metagenomic sequencing but that the transcript levels were high enough to be captured by the read depth of the metatranscriptomic sequencing. Notably, very few samples in these cohorts (0% to 4%) were positive for cntA transcripts (Fig. 7B and SI Appendix, Fig. S12↗), consistent with previously reported analyses (41). Thus, the metatranscriptomic data suggest that the bbu pathway is more likely to be functionally operant in the human gut than the O₂-dependent pathway mediated by cntA.

Next, we examined three cohorts with corresponding stool metabolomics data [HMP2 (38), BIO-ML (42), and PRISM (43)] to evaluate whether the presence of bbuA gene or transcripts is associated with changes in metabolite levels. In each of the three independent cohorts, the levels of γbb were lower in samples that were positive for the bbuA gene compared to those that were negative (Fig. 7C and SI Appendix, Fig. S13↗). In the HMP2 cohort, the samples positive for bbuA transcripts were also correlated with lower γbb levels (SI Appendix, Fig. S14↗). The bbuA gene, however, was not associated with stool carnitine or butyrate levels (SI Appendix, Fig. S15↗). In contrast, the presence of the cntA gene was correlated with higher or equivalent γbb levels, rather than lower, when compared to cntA-negative samples (Fig. 6C and SI Appendix, Fig. S13↗). These results suggest that the bbu gene cluster may be primarily responsible for γbb consumption in the human gut.

Finally, we evaluated whether differences in the presence or abundance of the bbuA gene are correlated with human disease. Although TMA production is not known to be a factor in inflammatory bowel disease (IBD), we noticed that the bbuA gene was detected in fewer samples from donors with Crohn’s disease and ulcerative colitis compared to healthy controls in both HMP2 (38) and PRISM (43) cohorts (Fig. 7D and SI Appendix, Fig. S16↗). The same trend was observed when analyzing bbuA transcripts in the HMP2 study (Fig. 7D). However, there was no difference in gene abundance between the IBD and healthy cohorts (SI Appendix, Fig. S16↗). Next, we analyzed metagenomes of a CVD cohort from China (44). The proportion of bbuA-positive samples was not different between healthy and CVD groups (Fig. 7E), but the gene abundance in bbuA-positive samples was modestly higher in CVD donors compared to healthy controls (Fig. 7F). A more significant elevation was found in the caiA gene abundance in the CVD cohort compared to the healthy group (Fig. 7F), suggesting that the conversion of l-carnitine to γbb may be a more relevant biomarker for CVD than the bbu gene cluster. Consistent with previous reports (41), we also detected a higher prevalence and abundance of the cntA gene in CVD samples compared with healthy controls from this study (Fig. 7 E and F). However, the lack of correlation between cntA gene presence and expression observed in other cohorts raises questions about the functional relevance of this finding.

E. timonensis SN18 was purchased from Leibniz Institute DSMZ. E. timonensis SN18 culturing was performed in Hungate or Balch tubes (Chemglass Life Sciences) at 37 °C and set up in an anaerobic chamber (Coy Laboratory Products) under an atmosphere of 2% to 4% H₂, 2% to 4% CO₂, and N₂ as the balance. Standard cultures were grown in peptone-yeast-glucose (PYG) medium, modified (medium recipe DSM 104, DSMZ Germany) that was sparged with N₂ after autoclaving. Basal medium lacking electron acceptors was prepared as described previously (57), containing 1 g/L tryptone (trypticase peptone; BD Biosciences), 1 g/L yeast extract (BD Biosciences), 0.4 mM l-cysteine, 2.5 g/L NaHCO₃, 1 g/L NaCl, 0.5 g/L MgCl₂•6H₂O, 0.2 g/L KH₂PO₄, 0.3 g/L NH₄Cl, 0.3 g/L KCl, 0.015 g/L CaCl₂•2H₂O, 0.25 mL/L of 0.1% resazurin, and 1% ATCC vitamins and trace mineral solutions (ATCC). NCE medium lacking carbon sources was prepared as described previously (59), containing 4 g/L KH₂PO₄, 5 g/L K₂HPO₄, 3.5 g/L NaNH₄PO₄ 40 mM sodium fumarate dibasic, 1 mM MgSO₄•7H₂O, 0.1% casamino acids (VWR Life Science), and 1% ATCC vitamin and trace mineral solutions (ATCC). All chemicals were purchased from Sigma-Aldrich unless otherwise indicated.

The following general procedure was used for synthesis of both [N-(CD₃)₃]-γbb and D₆-γbb, based on a published protocol (60). In an oven-dried round bottom flask under nitrogen, 1 molar equivalent of GABA (1 g, 9.7 mmol) or D₆-GABA (92 mg, 0.8 mmol) and 4.6 molar equivalents of K₂CO₃ were dissolved in anhydrous methanol (0.1 M) and stirred at room temperature for 15 min. Next, 5.1 molar equivalents of CD₃I (3.1 mL, 49.4 mmol) or CH₃I (0.42 mL, 4.3 mmol) was added to the solution, which was stirred at room temperature for 3 d. The reaction mixture was then concentrated under vacuum. The residue was resuspended in chloroform. The solid was filtered under vacuum and washed twice with chloroform. The solid was dissolved in 10% aqueous HCl (0.5 M). The resulting solution was concentrated under vacuum and further dried under vacuum. The residue was triturated three times with acetone. The combined organic layers were concentrated under vacuum. Anhydrous tetrahydrofuran (THF; 0.1 M) was added and the solution was cooled to 0 °C. The solid was filtered under vacuum and washed with fresh THF to afford the desired products.

N-(CD₃)₃]-γbb was obtained as a white solid (160 mg, 1 mmol, 10% yield) with the following NMR features: ¹H NMR (400 MHz, D₂O) δ ppm 2.07 to 2.16 (m, 2H), 2.53 (q, 2H, J = 7.3 Hz), 3.37 (td, 2H, J = 8.3, 3.9 Hz); ¹³C NMR (100 MHz, D₂O) δ ppm 75.2, 87.3, 109.8, 122.4, 232.4. The NMR data were in accordance with literature values (60).

D₆-γbb was obtained as a yellow solid (76 mg, 0.4 mmol, 48% yield) with the following NMR features: ¹H NMR (400 MHz, D₂O) δ ppm 3.07 (s, 9H).

E. timonensis SN18 was cultured anaerobically at 37 °C in 10 mL PYG-modified medium supplemented with 1 mM γbb, dl-carnitine, or GABA or with an equivalent volume of 1× phosphate-buffered saline (PBS). When the cultures reached an OD₆₀₀ = 0.5, they were centrifuged at 1,500 × g for 10 min at 4 °C. In an anerobic chamber (Coy Laboratory Products) under an atmosphere of 2% to 4% H₂, 2% to 4% CO₂, and N₂ as the balance, the cell pellets were resuspended in 1 mL anoxic 1× PBS and centrifuged at 1,500 × g for 10 min at 4 °C. The cell pellets were resuspended in 0.5 mL anoxic 1× PBS and incubated at room temperature for 1 h with 1 mM γbb, [N-(CD₃)₃]-γbb, D₆-γbb, dl-carnitine, or GABA or with an equivalent volume of 1× PBS. Reactions were analyzed by LC–MS as described below using three different methods for γbb/TMA, CoA, and fatty acid detection.

E. timonensis SN18 cultures were prepared in an anerobic chamber (Coy Laboratory Products), under an atmosphere of 2% to 4% H₂, 2% to 4% CO₂, and N₂ as the balance, in Balch tubes containing 15 mL PYG-modified medium. Cultures were grown anaerobically at 37 °C to an OD₆₀₀ = 0.7, when 0.15 mL of a 1 M O₂-free stock solution of γbb prepared in 1× PBS was added (1 mM final concentration) or an equivalent volume of O₂-free 1× PBS was added to triplicate cultures for each condition. In a separate experiment (SI Appendix, Fig. S2↗), when cultures reached an OD₆₀₀ = 0.5, 0.15 mL of 1 M O₂-free stock solutions of γbb, dl-carnitine, or GABA prepared in 1× PBS were added (1 mM final concentration) or an equivalent volume of O₂-free 1× PBS was added to triplicate cultures for each condition. Cultures were grown anaerobically at 37 °C for an additional 45 min and then centrifuged at 3,320 × g for 10 min at 4 °C. Keeping the samples cold on ice, the supernatant was decanted and then removed entirely using a micropipette. Cell pellets were immediately resuspended in 0.5 mL cold TRIzol reagent (Thermo Fisher) and flash frozen in liquid N₂. Samples were stored at −80 °C until further processing. RNA isolation, library preparation, sequencing, and analysis are described in the SI Appendix↗.

Construction of plasmids containing the bbu genes used for gain-of-function experiments and heterologous expression are described in detail in the SI Appendix↗ and Dataset S5↗. Preparation of E. timonensis BbuA, BbuB, and BbuC proteins as well as the γbb:CoA ligase S. meliloti BcoAB are described in detail in the SI Appendix↗.

Construction of the plasmid for production of C-terminal streptavidin-tagged BbuB and the affinity pull-down chromatography experiment are described in detail in the SI Appendix↗.

Crotonyl-CoA was synthesized from the acid anhydride according to a published protocol (61) with modifications described in the SI Appendix↗.

A 1 mL solution of 0.02 mM S. meliloti BcoAB, 0.2 M MgCl₂, 0.2 M adenosine triphosphate (ATP), 0.02 M CoA (CoALA Biosciences, Austin, TX), and 0.05 M γbb in 0.5 M Tris⋅HCl (pH 8.0) was incubated at room temperature for 30 min. The reaction was quenched with 1% acetic acid (final concentration) and incubated at −20 °C for 30 min. The solution was centrifuged at 16,000 × g for 30 min. The γbb-CoA product was purified by preparative HPLC using the same method as described for crotonyl-CoA. Fractions with absorption at 260 nm were collected and lyophilized. The product identity was confirmed by mass spectrometry.

E. coli MG1655 chemically competent cells were transformed with pET-proD (vector construction described in SI Appendix↗) plasmids containing E. timonensis bbu genes (Dataset S5↗). Transformed cells with kanamycin resistance were grown in 50 mL Luria Broth (LB) medium with 50 mg/L kanamycin at 37 °C with shaking (180 rpm) for 20 h, reaching an OD₆₀₀ of 1 to 1.4. Culture aliquots of 15 mL were harvested by centrifugation at 3,220 × g for 15 min at 4 °C and cell pellets were frozen at −80 °C until further use. Cell pellets were resuspended with a volume of 50 mM potassium phosphate (pH 7.5) buffer, 300 mM NaCl, and 10% glycerol to achieve a normalized OD₆₀₀ of 20. The cells were lysed on ice by sonication for a total of 1 min, with cycles of 2.5 s on and 10 s off. Reactions (0.05 mL total) containing 0.01 mL crude lysate, 10 mM acetyl-CoA (CoALA Biosciences, Austin, TX), and 25 mM γbb in 100 mM Tris⋅HCl (pH 8.0) buffer were incubated at room temperature for 1 h and analyzed by LC–MS using the CoA detection method (SI Appendix↗).

E. coli MG1655 (DE3) chemically competent cells were cotransformed with a pET28a plasmid containing E. timonensis bbu genes and the pACYC-Duet1 plasmid containing the E. timonensis groEL-ES genes (Dataset S5↗). Transformed cells with kanamycin and chloramphenicol resistance were grown in 50 mL LB medium with 50 mg/L kanamycin, 25 mg/L chloramphenicol, and 0.2 mM riboflavin at 37 °C with shaking (180 rpm). When cultures reached an OD₆₀₀ of 0.6, protein expression was induced by addition of isopropylthio-β-galactoside (IPTG) to a final concentration of 0.25 mM. The cultures were then incubated at 15 °C with shaking (180 rpm) for ∼15 h. Cultures aliquots of 15 mL were harvested by centrifugation at 3,220 × g for 15 min at 4 °C, and cell pellets were frozen at −80 °C until further use. Cell pellets were transferred and thawed in an anerobic chamber (Coy Laboratory Products) located in a cold room at 4 °C under an atmosphere of ∼3% H₂ and N₂ as the balance. The pellets were resuspended in O₂-free 50 mM potassium phosphate (pH 7.5) buffer, 300 mM NaCl, and 10% glycerol to achieve an OD₆₀₀ = 10. The cells were lysed by sonication for a total of 1 min, with cycles of 2.5 s on and 10 s off. In an anaerobic chamber (Coy Laboratory Products) under an atmosphere of ∼3% H₂ and N₂ as the balance, reactions (0.05 mL) containing 0.02 mL of crude lysate, 10 mM acetyl-CoA (CoALA Biosciences, Austin, TX), and 10 mM γbb in O₂-free 100 mM Tris⋅HCl (pH 8.0) buffer were incubated at room temperature for 1 h and analyzed by LC–MS using the methods for CoA and TMA detection (SI Appendix↗).

Reactions (50 μL) containing 0.01 mM BbuB, 0.01 mM BbuC, 10 mM acetyl-CoA, and 10 mM γbb in 100 mM Tris⋅HCl (pH 8.0) buffer were incubated at room temperature for 1 h and analyzed by LC–MS using the CoA detection method (SI Appendix↗).

Protein solutions of BbuA, BbuB, and BbuC were deoxygenated on ice by eight rapid cycles of vacuum and N₂ gas for a total of four times. In an anaerobic chamber (Coy Laboratory Products) under an atmosphere of ∼3% H₂ and N₂ as the balance, reactions (0.1 mL) containing 0.2 mM BbuA, 0.01 mM BbuB, 0.01 mM BbuC, 10 mM acetyl-CoA, and 10 mM γbb in O₂-free 100 mM Tris⋅HCl (pH 8.0) buffer were incubated at room temperature for 1 h and analyzed by LC–MS using the methods for CoA and TMA detection (SI Appendix↗). Reduced BbuA was prepared by preincubation of BbuA with 10 mM sodium dithionite prior to reaction initiation. Reactions with 10 mM γbb-CoA substrate excluded BbuB, BbuC, γbb, and acetyl-CoA.

Sample preparation and analytical methods for detection and quantification of γbb, γbb analogs, TMA, butyrate, and CoA-thioester compounds are described in detail in the SI Appendix↗.

Gene expression profiling by high-throughput sequencing data have been deposited in Gene Expression Omnibus accession no. (GSE165976↗). All other study data are included in the article and/or supporting information.