Microbially catalyzed conjugation of GABA and tyramine to bile acids

B. fragilis P207 was incubated in supplemented brain-heart infusion (BHIS) broth with and without DCA (0.01% wt/vol), and samples were analyzed by ultra-high-performance liquid chromatography tandem high-resolution mass spectrometry (UHPLC-MS²). Fragmentation-based networking of the mass spectrometry data identified derivatives of DCA from a set of candidate conjugates presented in Table S1. This metabolomic approach provided evidence for five amide-linked conjugates of DCA, which were only present in the spent media of cultures that contained both B. fragilis P207 and DCA. Fragmentation patterns consistent with DCA-alanine, DCA-glycine, DCA-phenylalanine, DCA-tyramine, and DCA-GABA were identified (Fig. 1). We aimed to validate the synthesis of bile acid-amine conjugates by B. fragilis P207 and to investigate if this strain could link exogenous amines by adding isotopically labeled variants of these compounds to the culture medium. Specifically, we supplemented broth containing B. fragilis P207 and deoxycholic acid (0.01% wt/vol) with 1 mM of either isotopically labeled 4-aminobutyric acid (¹³C₄, 97%–99%), L-alanine (¹³C₃, 99%), glycine (¹³C₂, 99%; ¹⁵N, 98%+), L-phenylalanine (D₈, 98%), or tyramine:HCl (1,1,2,2-D₄, 98%). Analyses of spent media from each of these conditions revealed the expected intact mass (MS¹) and fragments (MS²) for all five DCA conjugates (<0.48-ppm error for all MS¹ ions and <6-ppm error for all MS² ions representing predicted structures), as well as the expected mass shifts when media was supplemented with the isotopically labeled precursors (Fig. 2; Fig. S1 through S5). Given that bile acid conjugation to GABA and tyramine has not been previously reported, we further validated the identity of these compounds by comparing their liquid chromatography retention times and mass spectrometry (MS) data to synthetic standards. These experiments further confirmed the identity of these bile acid amides produced by B. fragilis P207 as DCA-GABA and DCA-tyramine (Fig. 3; Fig. S6 and S7). Mass spectrometry evidence for production of these conjugates is fully presented in the Supplemental Results.

To assess the chemical specificity of bile acid conjugation by B. fragilis P207, we further tested for production of cholic acid (CA) amides. Mass spectrometry analyses revealed trace levels of DCA in the BHIS growth media and traces of DCA and glycocholic acid (GCA) in the starting CA reagent; GCA levels were enhanced by incubation of CA (0.01% wt/vol) with B. fragilis P207 (Fig. 4A). Simultaneous incubation of CA and DCA (0.01% wt/vol total) with strain P207 showed similar production of the same five DCA conjugates described above. Thus, the addition of CA did not apparently impact P207-dependent production of the five DCA conjugates over this timescale (Fig. 4B; Table S2). We conclude that strain P207 prefers DCA as a conjugation substrate over CA in BHIS medium.

Lucas and colleagues have demonstrated that modification of CA into the secondary bile acid, 7-oxo-DCA, is robust and widespread across several species of Bacteroides, while production of DCA from CA was limited to Bacteroides vulgatus (9). We did not observe increased DCA levels when B. fragilis P207 was incubated with CA, nor did we observe production of any DCA conjugates from CA indicating that B. fragilis P207 does not catalyze production of DCA from CA under these conditions. The relative abundances of bile acid amide conjugates produced across replicate experiments are presented in Fig. S8.

Given that some bacteria can deconjugate glycine and taurine from secreted bile acids, we tested whether B. fragilis P207 is able to produce the DCA conjugates described above using glycodeoxycholate (i.e., GDCA) as an initial substrate. B. fragilis P207 was incubated in BHIS broth with and without GDCA (0.01% wt/vol), and culture samples were analyzed by UHPLC-MS². We observed trace contaminants DCA and GCA in the growth medium and bile acid reagent, but incubation of GDCA with strain P207 resulted in decreased GDCA and increased DCA levels, providing evidence that B. fragilis P207 has GDCA deconjugation activity (Fig. 4C). We further observed production of the DCA conjugates described above (GABA, tyramine, alanine, and phenylalanine) from GDCA when B. fragilis was present in the medium (Fig. 4C). We conclude that B. fragilis P207 can produce all the DCA conjugates described in the section above when either DCA or GDCA is present as an initial substrate.

The observation of bile acid conjugation to a distinct set of amines in vitro raised the question about the levels of these amines in the culture medium. To measure the relative levels of specific amine-containing compounds in B. fragilis P207 cultures before and after exposure to DCA, we used gas chromatography-mass spectrometry (GC-MS) to analyze the same culture samples in which the DCA conjugates were identified. Among the five amines conjugated to DCA, alanine and phenylalanine were most abundant in the medium, followed by glycine at levels 15–40 times lower. Tyramine and GABA were present at levels approximately 10 times lower than glycine (Fig. 5). Relative amine abundances did not, therefore, directly correlate with the levels of their respective DCA conjugates produced in vitro (Fig. S8).

Notably, when DCA was added to the culture medium, a significant increase in GABA was observed; glycodeoxycholic acid (GDCA) addition did not have this effect on microbial GABA production. The relative concentrations of other conjugated amines (i.e., alanine, phenylalanine, and tyramine) was unchanged across all treatment conditions (Fig. 5; Fig. S9). Bacteroides species in the human gut are known to produce GABA from either glutamate or glutamine, particularly under pH stress conditions (26). Given the increase in GABA following DCA treatment, we expanded our GC-MS analysis to include glutamine, glutamate, as well as other naturally occurring amino acids. DCA treatment did not impact glutamine levels, but it did lead to a modest reduction in glutamate concentrations (q < 0.01; Fig. S9), which is consistent with a model whereby DCA stimulates B. fragilis P207 to convert glutamate into GABA. However, DCA was recently reported to strongly induce expression of glutamate dehydrogenase (gene locus PTOS_003163) in B. fragilis P207 (27), suggesting that the metabolite flux involving glutamate is likely modulated in multiple ways by DCA. Cultivation of B. fragilis P207 did not greatly affect the levels of amino acids in BHIS overall, though glycine, GABA, glutamine, histidine, methionine, proline and tyrosine increased significantly (q < 0.01) after B. fragilis was cultivated in the medium. Asparagine and aspartate showed the largest changes in the presence of B. fragilis: the concentration of these two amino acids decreased by approximate factors of 50 and 10, respectively (q < 0.001), suggesting they serve as primary nutritional substrates for B. fragilis P207 in BHIS medium (Fig. S9).

To test if the bile acid conjugates produced by B. fragilis P207 in vitro were present in the gut of the human patient from whom the strain was isolated, we prepared extracts of stool samples collected from patient 207 at time points before and after onset of pouchitis and during antimicrobial treatment (23). Metabolomic data from this patient were compared to UHPLC-MS² data from 21 healthy donor stool samples. Patient 207 and healthy donor stool contained a complex mixture of bile acids (Fig. 6; Table S2). Unlike healthy donors, which contained expected high levels of DCA, patient 207 lacked DCA across all time points (Fig. S10; Table S2). This result is congruent with a reported deficiency in secondary bile acids in UC pouches (22). The primary bile acids chenodeoxycholate (CDCA) and CA were abundant across all time points in patient 207, at levels that were ≈60 times higher on average than observed in the 21 healthy donor samples (P < 0.001) (Fig. S11; Table S2). Amine conjugates to CA and CDCA, including putative GABA conjugates, were identified in patient 207 stool (Fig. 6 and 7; Fig. S12 and S13; Table S2). We detected several amines conjugated to a bile acid core with an m/z corresponding to DCA or its isomers (Fig. 6; Fig. S8). Considering the lack of DCA and abundance of CDCA in patient 207, we predicted that these species were CDCA derivatives. Indeed, a product with an elution profile and MS² fragmentation pattern matching a CDCA-GABA synthetic standard was identified in patient 207 stool (Fig. 7; Fig. S12). We further identified products matching CA-GABA and CA-tyramine synthetic standards in patient 207 stool and in healthy human donor stool, respectively (Fig. 7; Fig. S13 and S14). The results provide strong evidence that bile acids conjugated to GABA and tyramine are present in the human gut.

B. fragilis strain P207 dominated the pouch ecosystem of patient 207 from 182 to 434 days after surgical functionalization of the ileal J-pouch (i.e., IPAA) (23). Normalized levels of bile acids, including the primary bile acids CA and CDCA, were lowest at 124 and 236 days after IPAA. There was a marked increase in unconjugated CA and CDCA and bile acid conjugates at 355 days post-functionalization, and levels trended upward by 482 days when the patient was diagnosed with pouchitis and initiated a course of antibiotic therapy (ciprofloxacin) (Fig. 7D; Fig. S11). The contribution of B. fragilis strain P207 to the production of particular bile acid conjugates at these time points is not known, but the in vitro conjugation data presented above indicate that CA conjugates (other than glycocholate) are not produced by B. fragilis P207 (Fig. 4). At a follow-up visit 525 days post-IPAA, ciprofloxacin treatment had resulted in a marked reduction of the pouch microbiome census, and pouch inflammation was resolved (23). Bile acid analysis showed that levels of conjugated and unconjugated bile acids were sharply reduced at this time point except for select low abundance conjugates of CA and DCA isomers to glycine and GABA, which increased (Fig. 7D; Fig. S11 and S15; Table S2). Though inflammation was resolved by ciprofloxacin treatment, inflammation returned by 719 days (23). In this period, the dominant species of the pouch shifted from B. fragilis P207 to Bacteroides ovatus (23). This was correlated with a change in the bile acid conjugate profile, with distinct CA and DCA isomer conjugates to GABA and glycine increasing (Table S2; Fig. S11 and S15)

Metabolomic analysis of stool samples from pouchitis patient 207 and healthy human donors revealed a complex mixture of known and previously unreported bile acid conjugates, including CA conjugates to GABA and tyramine (Fig. 6 and 7). To discover bacteria that can produce these novel CA conjugates, we inspected the genomes of a collection of human gut isolates for genes that encode predicted N-terminal nucleophilic cysteine hydrolase (Ntn) enzymes, which includes known choloylglycine hydrolases [Conserved Domain Database accession cd01902 (28)]. Recent studies report that this class of enzymes can function to both deconjugate and conjugate bile acids (29, 30). B. fragilis P207 encodes an Ntn hydrolase (WP_005817456.1), and we predicted that other strains encoding these enzymes may conjugate GABA or tyramine to CA in vitro. We selected a phylogenetically diverse group of human patient isolates that were also predicted to encode an Ntn hydrolase, including Mediterraneibacter gnavus MSK15.77 (WP_004614568.1), Bifidobacterium longum DFI.2.45 (WP_007052221.1), and Bacteroides ovatus MSK22.29 (which encodes three Ntn paralogs: WP_217723859.1, WP_004308262.1, and WP_004323538.1) (Fig. S15). We further identified a strain in our collection that does not encode a predicted Ntn bile salt hydrolase, Lachnoclostridium scindens SL.1.22. All three Ntn-encoding gut isolates produced putative GABA conjugates to CA, and M. gnavus and B. longum cultures contained a product that had an MS² fragmentation pattern consistent with CA-tyramine (Fig. 8; Fig. S16). L. scindens cultures did not contain CA conjugates.

We further investigated if these strains conjugated amines to DCA and detected conjugates of alanine, phenylalanine, and GABA to DCA in the culture extracts of M. gnavus, B. longum, and B. ovatus. Additionally, DCA-tyramine conjugates were identified in M. gnavus and B. ovatus cultures (Fig. 8). Although the L. scindens SL.1.22 genome does not encode a predicted Ntn family/bile salt hydrolase (BSH) enzyme, we observed low levels of DCA-Phe and DCA-Ala conjugates in its culture extract when this strain was incubated with 0.01% (wt/vol) DCA. The specific genes responsible for DCA conjugation in our L. scindens cultures remain unidentified.

B. fragilis P207 produces a suite of DCA amides in vitro, in which the bile acid carboxylate group is conjugated to the amine group of select amino acids and neuroactive amines. Biosynthesis of bile acid amides by gut microbes is a recent discovery (9, 10), and there is now genetic and biochemical evidence that bacterial BSH enzymes can catalyze aminoacyl transfer to bile salts (29, 30). Thus, BSH proteins have significant N-acyl transferase activity in addition to their long-established function as hydrolases (31). While microbial conjugation of bile acids to the α amino group of amino acids has been described, our data provide evidence that an activity encoded by B. fragilis P207 enables conjugation of the primary amine groups of GABA and tyramine to DCA. DCA conjugate biosynthesis was evident when culture broth was supplemented with either unconjugated (DCA) or conjugated (GDCA) bile acids, which supports a model in which P207 first deconjugates GDCA and then generates DCA amides from the deconjugated product.

The B. fragilis P207 genome (GenBank accession CP114371) encodes a single predicted BSH (gene locus PTOS_003312 ) that shares 80% identity with BT2086, a protein that has demonstrated BSH activity in Bacteroides thetaiotaomicron (32). PTOS_003312 is 99% identical to the BSH of B. fragilis strain 638R (locus BF638R_3310), which promotes deconjugation of primary bile acids in the gut of germ-free mice (33). In light of recent in vitro biochemical data showing that purified Clostridium perfrigens (29) and Bifidobacterium longum (30) BSH enzymes produce bile acid amides from both unconjugated (CA) and conjugated [taurocholic acid (TCA)] bile acids, it seems most likely that PTOS_003312 of B. fragilis P207 catalyzes the production of the five bile acid conjugates reported here, using either GDCA or DCA as substrates, though we cannot rule out the possibility that these products arise from the activity of multiple enzymes. P207 demonstrates a clear preference for bile acid amide production from the secondary bile acid, DCA, over primary bile acid CA, which differ only by a single hydroxyl group at C-7. Based on this result, we infer that the C-7 position of the sterol core is important for bile acid substrate interaction with the transforming enzyme(s) of strain P207.

The detection of DCA conjugation to the primary amine groups of GABA and tyramine raises questions about the mechanism of bile acid conjugation, given the relative difference in nucleophilicity and steric accessibility of the primary amine groups of tyramine and GABA (pKa ≈ 10.5–11.0 in aqueous solvent) and the typical α amino group (pKa ≈ 9–9.5) of amino acids. Considering the diversity in primary structure of BSH enzymes and the selectivity of BSH enzymes for particular steroidal cores across Bacteroides spp. (32), it is likely that structural differences in the active site or other regions of the conjugating enzyme (34) will determine whether primary amine groups, such as those on GABA and tyramine, form amide bonds with bile acid(s). Evidence for microbially catalyzed production of GABA and tyramine BA amides presented herein is congruent with a recent report that a diverse array of amines in the gut are conjugated to bile acids (35).

The benefit of bile acid conjugation to B. fragilis and other microbes studied here, if any, is not known. B. fragilis P207 can deconjugate GDCA and produce a variety of conjugated compounds from the deconjugated bile acid substrate (DCA) (Fig. 4). Unconjugated primary and secondary bile acids are often more toxic in vitro than the amino acid-conjugated forms (13, 36, 37). Indeed, we have shown that B. fragilis P207 is more sensitive to CA and DCA than glycocholate and glycodeoxycholate (Fig. S17). We further observed that the secondary bile acid, DCA, is more toxic to B. fragilis P207 than CA. The ability to conjugate amino acids effectively detoxifies DCA in vitro and may therefore influence the impact of bile salts on B. fragilis physiology in certain settings.

This study defines the temporal bile acid profile of the J-pouch of a pouchitis patient before, during, and after inflammation, including a period of antibiotic therapy, and draws comparisons of this profile to stools of healthy individuals. Healthy human donors presented expected high levels of DCA, while the pouchitis patient was severely DCA deficient. This aligns with previous observations of reduced secondary bile acids in UC pouches (22). As expected for a patient with diminished secondary bile acid production, the primary bile acids, CDCA and CA, were significantly elevated, and unique amine conjugates to CA and CDCA, particularly novel GABA conjugates, were identified in this patient. Our in vitro analyses of B. fragilis P207 and of other human gut microbiome isolates have demonstrated that bile acid conjugates to GABA and tyramine observed in stool can be produced by multiple bacterial genera (Fig. 8). Changes in the levels of bile acids leading up to pouchitis, coupled with the marked influence of ciprofloxacin treatment on the bile acid profile (Fig. S9 and S12), underscore the complex interplay between gut microbiota, host factors, and therapeutic interventions on the bile acid profile of humans.

The neuroactive amines, GABA and tyramine, are potent regulators of mammalian physiology and both molecules are produced by microbes that comprise the gut microbiome, including Bacteroidetes (26). GABA has a well-recognized role in the regulation of gut physiology (38). GABA-producing Bacteroides are reported to increase steady-state levels of GABA in the intestines of mono-associated germ-free mice (39), and we have shown that bile acid exposure enhances B. fragilis GABA production in vitro (Fig. 5) and provided evidence that exogenous GABA can be conjugated by B. fragilis P207 (Fig. 2). The conjugation and subsequent uptake of bile acid-GABA amides into enterohepatic circulation may impact GABAergic signaling in the gut and possibly at other distal tissue sites. Tyramine is a product of tyrosine catabolism but is also present at high levels in a variety of foods. Steady-state tyramine levels in human tissue are typically low due to the activity of monoamine oxidase A but can be modulated in patients treated with monoamine oxidase inhibitors (40). Trace amine-associated receptors (TAARs), with which tyramine can interact, have been reported in enterochromaffin cells of the human gut (41), though the impact of dietary and microbiome-derived tyramine of TAAR signaling in the gut is not known. Tyramine-bile acid amides in human patients likely vary as a function of diet and other factors and vary geographically along the digestive tract, depending on the amount of microbial tyramine production and the spectrum of bile acid conjugating activity conferred by the host microbiome at particular foci. Tyramine feeding data presented in Fig. 2 suggest that dietary tyramine could impact BA-tyramine levels in the gut.

The impact of specific conjugated bile acids reported here on signaling through bile acid receptors is an interesting future area of investigation. BSH activity of non-enterotoxigenic Bacteroides spp. was recently reported to potentiate obesity-related colorectal cancer progression in a mouse model (42); this effect was attributed in part to enhanced TGR5 signaling as a result of increased levels of deconjugated DCA and lithocholic acid in the colon. Certainly, gut microbe BSH activity can lower the level of deconjugated bile acids in the gut. However, it is important to consider that BSH activity results in a spectrum of unconjugated and conjugated bile acids (29, 30), including compounds reported here. The effect of these compounds on signaling from bile acid receptors may shape host physiology, disease, and health.

B. fragilis P207 (NCBI accession NZ_CP114371) was cultivated in BHIS containing either 0.01% (wt/vol) DCA, CA, or GDCA that was inoculated from a saturated starter culture (≈1.0 OD₆₀₀), back-diluted 1:10. Cultures were grown for 24 h; 1 mL of culture was removed and flash frozen for subsequent metabolomic analysis. Mediterraneibacter gnavus MSK15.77 (NCBI accession NZ_JAAIRR010000000), Bacteroides ovatus MSK22.29 (NCBI accession NZ_JAHOCX010000000), Bifidobacterium longum DFI.2.45 (NCBI accession NZ_JAJCNS010000000), and Lachnoclostridium scindens SL.1.22 (NCBI accession GCA_020555615.1)were also cultivated in BHIS. Briefly, starter cultures were inoculated from frozen glycerol stocks and grown anaerobically at 37°C overnight to ≈1.0 OD₆₀₀. These cultures were diluted to ≈0.005 OD₆₀₀ in 1 mL of plain BHIS or BHIS containing 0.01% (wt/vol) DCA or CA in a 96-well deep well plate. The plate was incubated at 37°C anaerobically for 20 h and then frozen at −80°C for subsequent metabolomic analysis.

Bacteria culture supernatants were lyophilized followed by 2× concentration in 100% methanol (containing internal standards). Samples were then centrifuged at −10°C, 20,000 × g for 15 min to generate supernatants for subsequent metabolomic analysis. All ultra-high-performance-liquid chromatography-mass spectrometry (UHPLC-MS) analyses were performed using a Thermo Scientific Vanquish Flex UHPLC coupled with an IQ-X mass spectrometer (Thermo Fisher Scientific). Reversed-phase chromatography was performed on a Waters CORTECS T3 C18 RP-UHPLC column [100 × 2.1 mm inner diameter, 1.6-µm particle size, 120-Å pore size (1 Å = 0.1 nm)]. Mobile phase A was 5% acetonitrile in water with 0.1% formic acid, and mobile phase B was acetonitrile with 0.1% formic acid. For the initial analysis of B. fragilis P207 cultures (as displayed in Fig. 1), the flow rate was 550 µL/min. The chromatographic method used was an isocratic 20% mobile phase B for 0.2 min, followed by a gradient of 20%–97% mobile phase B for 8.7 min with a wash of 97% mobile phase B for 1.0 min. Analysis of B. fragilis P207 cultures for normalized relative quantification (as displayed in Fig. S8) was performed at a flow rate of 480 µL/min. The chromatographic method used was an isocratic 100% mobile phase A for 0.2 min, followed by a gradient of 0 to 97% mobile phase B for 10.2 min with a wash of 97% mobile phase B for 1.0 min. Optimized chromatography was then developed for all other UHPLC-MS/MS experiments to better separate the numerous isomers of the various bile acid amine conjugates. The chromatographic method used a flow rate of 300 µL/ min with a gradient of 20% to 40% mobile phase B for 0.2 min, followed by a gradient of 40%–70% mobile phase B for 11.8 min, a gradient of 70%–97% mobile phase B for 0.1 min, and a wash of 97% mobile phase B for 1.1 min.

All samples were analyzed using positive ionization. Flow from the UHPLC was ionized with the heated electrospray ionization source set to 3,400 V; ion transfer tube temperature set to 200°C; vaporizer temperature set to 400°C; and sheath, aux, and sweep gases set to arbitrary values of 40, 5, and 1, respectively. Data for MS¹ were acquired using a maximum inject time of 50 ms, a normalized AGC target of 25%, a 300–1,700 m/z quadrupole scan range, and a resolution of 60,000. All tandem MS² mass spectral data were acquired using a 1.5 m/z quadrupole isolation window, a maximum inject time of 22 ms, a normalized AGC target of 20%, and a resolution of 15,000. Each of the metabolite ions of interest was added to an inclusion list for MS² fragmentation by a normalized higher-energy collisional dissociation energy of 30%. Data analysis was performed using FreeStyle software (version 1.8 SP2, Thermo Scientific), MZmine (version 2.53) (43), Global Natural Products Social Molecular Networking (GNPS) platform tools Feature-Based Molecular Networking (FBMN) release 28.2, and Mass Search Tool release 29 (44 – 46), and GraphPad Prism (version 9.4.1 for Windows; GraphPad Software, San Diego, California, USA; www.graphpad.com). Areas under the curve (AUCs) were normalized to the average of D₄-taurocholate and D₄-taurodeoxycholate areas under the curve and multiplied by 1,000:

Normalized AUC = {peak_AUC / [(D₄ − TC_AUC + D₄ − TDC_AUC) / 2]} × 1,000

Data-dependent mass spectrometry data files were processed using MZmine 2.53 and the Feature-Based Molecular Networking function in the GNPS environment to identify and network spectral features while also matching generated data to publicly available library spectra. MZmine was used to detect MS¹ and MS² peaks, build deconvoluted extracted ion chromatograms, group isotopes, and match the resulting features across samples, accounting for retention time drift across injections. Features within each sample were then normalized by dividing peak area by the average peak area of the internal standards in that sample. Features that are found in solvent and method blank controls were then eliminated. The aligned feature lists were then exported for analysis using the FBMN tool within the GNPS platform.

FBMN was performed with precursor ion mass tolerance and fragment ion mass tolerance both set to 0.02 Da, minimum matched fragment ions set to six, networking cosine score set to >0.7, library cosine score set to >0.7, and minimum library shared peaks set to six. A visualization of the network was constructed in Cytoscape by drawing edges between scan nodes with a cosine similarity of >0.7. The network was manually analyzed to identify ions that occurred only in samples containing BHIS, a bacterial strain, and DCA or CA, with a particular focus on those that occurred within the same molecular families as our bile acid internal standards and other nodes that matched to library entries of known bile acids. Nodes in the molecular network were also checked against a list of predicted m/z values for hypothesized amine-containing conjugates presented in Table S1. Data files were also manually analyzed for MS² scans containing characteristic DCA core fragments at m/z 215.1794 and m/z 339.2682 (observable in Fig. S1 through S5) to identify metabolite ions that may not have been clustered in the molecular network due to falling below networking thresholds.

To test for conjugation of isotopically labeled amines to DCA by B. fragilis P207, 1-mL cultures containing 1 mM (final concentration) of either 4-aminobutyric acid (¹³C4, 97%–99%), L-alanine (¹³C3, 99%), glycine (2-¹³C, 99%; ¹⁵N, 98%+), L-phenylalanine (D₈, 98%), or tyramine:HCl (1,1,2,2-D₄, 98%) (Cambridge Isotope Laboratories, Inc.) were added to BHIS containing 0.1% (wt/vol) DCA (Thermo Fisher Scientific). Control samples contained either no isotopically labeled amines, no DCA, or BHIS media without any additional supplement. Culture supernatants were prepared for metabolomic analysis as described above.

Synthesis of bile acid-amine conjugates followed methods previously described (10). Cholic acid, deoxycholic acid, and chenodeoxycholic acid were dissolved to 0.125 mM in 5-mL anhydrous THF (tetrahydrofuran), each in two separate 20-mL scintillation vials containing stir bars, and kept on an ice bath for a total of six vials. Next, ethyl chloroformate (0.15 mmol, 7 µL) and triethylamine (0.15 mmol, 10.5 µL) were added to each reaction vessel and stirred for 1.5 h in the ice bath. Basic aqueous solutions of GABA (1.54 eq, 0.193 mM) and tyramine (1.54 eq, 0.193 mM) were prepared using NaOH (1.5 eq, 0.188 mM), with 5 mL of the GABA solution added to the first of each pair of bile acid solutions and 5 mL of the tyramine solution added to the second of each pair of bile acid solutions. The reaction mixtures were stirred for 2 h, centrifuged at −10°C, 20,000 × g for 15 min to remove any remaining insoluble reaction components, and injected on the UHPLC-MS/MS using the optimized bile acid chromatography method.

Amine-containing metabolites were measured using GC-MS with electron impact ionization. Bacterial cultures were extracted using four volumes of 100% methanol. Following brief vortexing and centrifugation at 4°C, 20,000 × g for 15 min, we added 100 µL of extract supernatant to prelabeled mass spec autosampler vials (microliter, 09–1200) and dried down completely under nitrogen stream at 30 L/min (top), 1 L/min (bottom) at 30°C (Biotage SPE Dry 96 Dual, 3579M). To dried samples, 50 µL of freshly prepared 20-mg/mL methoxyamine (Sigma, 226904) in pyridine (Sigma, 270970) was added and incubated in a thermomixer C (Eppendorf) for 90 min at 30°C and 1,400 rpm. After samples are cooled to room temperature (RT), 80 µL of derivatizing reagent (BSTFA + 1% TMCS; Sigma, B-023) and 70 µL of ethyl acetate (Sigma, 439169) were added, and samples were incubated in a thermomixer at 70°C for 1 h and 1,400 rpm. Samples were cooled to RT and 400 µL of ethyl acetate was added to dilute samples. Turbid samples were transferred to microcentrifuge tubes and centrifuged at 4°C, 20,000 × g for 15 min. Supernatants were then added to mass spec vials for GC-MS analysis. Samples were analyzed using a GC-MS (Agilent 7890A GC system, Agilent 5975 MS detector) operating in electron impact ionization mode, using an HP-5MSUI column (30 m × 0.25 mm, 0.25 µm; Agilent Technologies 19091S-433UI) and 1-µL injection. Oven ramp parameters were 1 min hold at 60°C, 16°C/min up to 300°C with a 7 min hold at 300°C. Inlet temperature was 280°C and transfer line was 300°C. Data analysis was performed using MassHunter Quantitative Analysis software (version B.10, Agilent Technologies) and confirmed by comparison to authentic standards. Normalized peak areas were calculated by dividing raw peak areas of targeted analytes by averaged raw peak areas of ¹⁵N,d₇-L-proline and U-¹³C-palmitate internal standards.

Patient 207 feces was liquid and extraction was performed by vortexing and diluting 1:2 (200 µL into 400 µL) with 100% methanol containing internal standards, followed by bath sonication for 10 min. Non-diseased volunteer donor feces were solid and were extracted by adding 80% methanol to 100 mg/mL and stored at −80°C for at least 1 h in beadruptor tubes (Fisherbrand, 15–340-154). Donor samples were then homogenized at 4°C on a Bead Mill 24 homogenizer (Fisher, 15–340-163) set at 1.6 m/s with six 30-s cycles, 5 s off per cycle. All samples were then centrifuged at −10°C, 20,000 × g for 15 min to generate supernatants for subsequent metabolomic analysis. Areas under the curve for culture supernatants were normalized to the average of D₄-taurocholate and D₄-taurodeoxycholate areas under the curve and multiplied by 1,000, while areas under the curve for fecal samples were normalized to the average of D₄-taurocholate and D₄-glycocholate areas under the curve and multiplied by 1,000. All peak assignments in stool samples were verified by comparing the MS² fragmentation spectra to MS² fragmentation spectra established from the in vitro analyses described above.

B. fragilis strain P207 was cultivated in BHIS supplemented with increasing concentrations of DCA, CA, GDCA, or GCA. Concentrations ranged from 0.024 to 6.17 mM of each compound. Strain P207 was grown from a saturated starter culture (~1.0 OD₆₀₀) that was back-diluted 1:100. Cultures were then grown for 24 h, at which point terminal culture density was measured at OD₆₀₀.