Investigation of bile salt hydrolase activity in human gut bacteria reveals production of conjugated secondary bile acids

Abstract

Through biochemical transformation of host-derived bile acids, gut bacteria mediate host-microbe crosstalk and function at the interface of nutrition and host metabolic regulation. Bile acids play a crucial role in human health by facilitating the absorption of dietary lipophilic nutrients, interacting with hormone receptors to regulate host physiology, and shaping gut microbiota composition through antimicrobial activity. Bile acids deconjugation by bacterial bile salt hydrolase has long been recognized as the first necessary bile acid modification required before further transformations can occur. Here, we show that bile salt hydrolase activity is common among human gut bacterial isolates spanning seven major phyla. However, we observed variation in both the extent and the specificity of deconjugation of bile acids among the tested taxa. Unexpectedly, we discovered that certain strains were capable of directly dehydrogenating conjugated bile acids via hydroxysteroid dehydrogenases to produce conjugated secondary bile acids both in vitro and in vivo. These results challenge the prevailing notion that deconjugation is a prerequisite for further bile acid modifications and lay a foundation for new hypotheses regarding how bacteria act individually or in concert to diversify the bile acid pool and influence host physiology.

Here, the authors show that bile salt hydrolase activity is common among human gut bacterial isolates spanning seven major phyla and identify strains capable of directly dehydrogenating conjugated bile acids via hydroxysteroid dehydrogenases to produce conjugated secondary bile acids, challenging the notion that deconjugation is a prerequisite for further bile acid modifications.

Introduction

The human gut microbiota plays a pivotal role in health and disease by biochemically transforming host-derived bile acids (BAs). Two primary BAs, cholic acid (CA) and chenodeoxycholic acid (CDCA), are synthesized in the human liver from cholesterol and conjugated to glycine or taurine^1,2. These conjugated BAs are stored in the gallbladder and released into the intestines following a meal, where they reach millimolar concentrations^3,4. A significant portion of conjugated BAs are reabsorbed in the terminal ileum and recirculated to the liver through the portal vein in a process called enterohepatic circulation. The remaining BAs in the small and large intestines are subject to deconjugation and transformation by gut bacteria. The transformed BAs are then either passively reabsorbed across the intestinal wall or excreted in feces.

Bile salt hydrolases (BSH) are bacterial enzymes that catalyze deconjugation by cleaving the amide bond in conjugated BAs to release unconjugated BAs. BSH activity has been linked to both positive and negative health outcomes in both humans and mice¹. BSH-active bacteria are reported to alleviate hypercholesterolemia⁵ and metabolic dysfunction associated steatotic liver disease⁶. Elevated levels of primary BAs resulting from high BSH activity have been shown to stimulate hepatic NKT cell accumulation and antitumor immunity in mice⁷. Additionally, BSH activity has been associated with resistance to Clostridioides difficile infection⁸. In other contexts, positive health outcomes have been associated with limited BSH activity. For example, BSH deficiency in mice has been linked to reduced weight gain on a high-fat diet and increased lipid utilization over carbohydrates for energy⁹. Reduced BSH activity has also been associated with slower progression of colorectal cancer^10,11. Yet still, in other contexts, limited BSH activity may lead to adverse outcomes, as elevated levels of conjugated BAs have been associated with inflammatory bowel disease (IBD)¹², Type 2 diabetes¹³, and cholangiocarcinoma, an often fatal cancer of the biliary tract¹⁴. These varied outcomes highlight the need for a comprehensive, systematic understanding of BSH activity across diverse genera of human gut bacteria.

Deconjugation by BSH has long been considered a gateway reaction¹⁵ that allows unconjugated primary BAs to be further transformed into secondary BAs. Subsequent transformations produce BAs such as deoxycholic acid (DCA) and lithocholic acid (LCA) through dehydroxylation at the C7 position by enzymes encoded by the bai operon. Other transformations that occur are the oxidation of hydroxyl groups by ɑ-hydroxysteroid dehydrogenases (HSDs) that generate position-specific -oxoBAs, and the subsequent epimerization by β-HSDs to produce β-oriented BAs, such as ursodeoxycholic acid (UDCA) and ursocholic acid (UCA)¹⁶. Additionally, BAs can be transformed into microbially conjugated BAs (MCBAs)^17,18 through the recently identified transferase activity of BSH enzymes^19,20.

To investigate the role of BSH activity in the diversification of the BA pool, we screened 77 gut bacterial strains spanning seven major phyla for their ability to deconjugate, transform, and reconjugate human conjugated BAs. Due to its recognition as the rate-limiting BA transformation, we evaluated the ways in which altered BSH activity impacted the development of the BA pool for selected strains in time-course monoculture and coculture experiments. Our systematic evaluations generate knowledge regarding bacterial BA deconjugation and transformations and provide a foundation for developing testable hypotheses that define causal links between the microbiome, BA pool composition, and human health.

Results

Bile salt hydrolase activity is widespread

We evaluated the deconjugation ability of 77 human gut bacterial strains (Supplementary Table 1), spanning seven different phyla and 41 genera, against a mixture of the five most prevalent human conjugated BAs: taurocholic acid (TCA), glycocholic acid (GCA), taurochenodeoxycholic acid (TCDCA), glycochenodeoxycholic acid (GCDCA), and taurodeoxycholic acid (TDCA). These strains are from species commonly found in the human gut. Additionally, a majority are commercially available, and their genomes are sequenced. Each strain was cultured anaerobically in medium supplemented with individual conjugated BAs at 100 µM or 500 µM, yielding pooled concentrations of 500 µM and 2.5 mM, respectively, which are within the physiological range found along the gut^3,4,21. Cultures were grown until they reached stationary phase, when the supernatant was collected for LC-MS/MS analysis.

BA deconjugation activity was widespread, with BSH activity detected in 60 of the 77 tested bacterial strains, spanning the Actinomycetota, Bacillota, Bacteroidota, Fusobacteria, and Pseudomonadota phyla (Fig. 1). A single species from the Lentisphaerota and Verrucomicrobiota phyla was tested, and neither exhibited BSH activity. We observed deconjugating activity on both glycine- and taurine-conjugated BAs, with some bacterial strains showing a preference for one over the other. Patterns in deconjugating activity were similar across the 100 µM and 500 µM concentrations, with few exceptions. Beyond the production of unconjugated primary BAs through deconjugation, several strains performed further transformations to the BA core to produce unconjugated secondary BAs (Fig. 1 and Supplementary Fig. 1) and/or reconjugated BAs to generate MCBAs (Fig. 2). On average, the percent recovery of BAs was 86.2% ± 12.9% for the 100 μM concentration and 87.2% ± 14.0% for the 500 μM concentration (Supplementary Fig. 2). Surprisingly, we identified conjugated secondary BAs in several samples (Fig. 1 and Supplementary Figs. 1, 3, and 4). The identity of these conjugated secondary BAs was confirmed by matching the sample MS/MS fragmentation spectra to predicted signature peaks (Supplementary Fig. 3 and Supplementary Table 3).

Fig. 1. Bile acid deconjugation and transformation measurements across phyla.

The heat maps show quantified bile acids (BAs) for the 100 µM (left) or 500 µM (right) conditions, as measured by LC-MS using external standard calibration curves. Two replicates were averaged for each strain at each concentration. For each heatmap, conjugated BA substrate measurements are depicted in the left column, unconjugated BAs (UBA) in the middle column, and transformed BAs in the right column. Color scale denotes BA concentration using a two-scale color gradient, with values below our detection limit of 0.008 μM in white. Phyla information is indicated by color-coded dashed lines in the phylogenetic tree. See Supplementary Fig. 1 for a depiction of BA transformations, including chemical structures and the genes responsible for each transformation and Supplementary Table 2 for a list of the BA standards. BA abbreviations: TCA taurocholic acid, TCDCA taurochenodeoxycholic acid, TDCA taurodeoxycholic acid, GCA glycocholic acid, GCDCA glycochenodeoxycholic acid, CA cholic acid, CDCA chenodeoxycholic acid, DCA deoxycholic acid, 7-oxoLCA 7-oxolithocholic acid, 7-oxoDCA 7-oxodeoxycholic acid, 3-oxoCDCA 3-oxochenodeoxycholic acid, 3-oxoDCA 3-oxodeoxycholic acid, 3-oxoCA 3-oxocholic acid, 12-oxoLCA 12-oxolithocholic acid, 12-oxoCDCA 12-oxochenodeoxycholic acid, HDCA hyodeoxycholic acid, βMCA beta-muricholic acid, G-7-oxoLCA glyco-7-oxolithocholic acid, G-7-oxoDCA glyco-7-oxodeoxycholic acid, T-7-oxoLCA tauro-7-oxolithocholic acid, T-7-oxoDCA tauro-7-oxodeoxycholic acid, G-3-oxoCDCA glyco-3-oxochenodeoxycholic acid, G-3-oxoCA glyco-3-oxocholic acid.

Fig. 2. Microbially conjugated bile acid production at 100 µM.

The heat map shows relative levels of MCBAs produced, as denoted by z-score, with the mean and standard deviation calculated from the raw signal across samples. Phyla information is indicated by color-coded dashed lines in the phylogenetic tree. Amino acid conjugations are indicated by their single-letter abbreviation across the top of the heat map. Glycine and taurine conjugations could not be measured in this study because these conjugated BAs were provided in the media. Percent deconjugation was calculated by subtracting the remaining conjugated BAs from the total BA pool and was normalized to the total BA pool for each strain. The heat map for BAs at 500 µM can be found in Supplementary Fig. 5 and a compound list containing full MCBA names, formulas, m/z, and retention time can be found in Supplementary Table 4.

Of the 53 strains previously identified to possess a putative bsh gene based on computational analyses^22,23, 44 exhibited BSH activity in our in vitro analysis. Among the remaining 24 strains without a previously identified bsh, 12 exhibited deconjugating activity. The highest levels of deconjugation were observed in the Actinomycetota, with more variable activity in the Bacillota and the Bacteroidota (Fig. 1). All Bifidobacteria and two out of three Collinsella species tested were able to deconjugate a majority of provided conjugated BAs to produce unconjugated BAs. Among the Bacillota, Enterococcus and Eubacterium species effectively deconjugated all provided conjugated BAs, as did Enterocloster bolteae, Roseburia intestinalis, Anaerobutyricum soehngenii, and Coprococcus comes, consistent with prior studies^24,25. The probiotic genus Lactobacilli also exhibited BSH activity, with Lactobacillus ruminis deconjugating all conjugated BAs, while Lactobacillus reuteri showed a preference for glycine-conjugated BAs. In the Bacteroidota, over 80% of strains demonstrated BSH activity, deconjugating measurable amounts of unconjugated BAs (Fig. 1).

We observed limited BSH activity in the Pseudomonadota. Edswardsiella tarda deconjugated ~16% of conjugated BAs when provided 100 µM conjugated BAs, but less than 2% when provided 500 µM. Proteus penneri deconjugated only ~5% of conjugated BAs at both concentrations (Fig. 1). Fusobacterium varium, our only representative from the Fusobacteriota phylum, exhibited deconjugating activity only on taurine-conjugated BAs. The majority of Bifidobacterium and Enterococcus species are known to possess a bsh gene, while Bacteroides, Collinsella, Lactobacilli, and Streptococcus are known to have genus level variation in BSH activity^{22,23,26–35}.

Unconjugated secondary BAs were produced through a combination of deconjugation and subsequent transformations (Fig. 1). The Actinomycetota, Bacillota, and Bacteroidota exhibited both the highest levels of deconjugation and the highest production of secondary BAs. Collinsella aerofaciens, Flavonifractor plautii, Lachnoclostridium scindens, Bacteroides intestinalis, Bacteroides ovatus, and Bacteroides xylanisolvens showed the highest levels of secondary BA production. However, many species, including Collinsella intestinalis, Holdemania filiformis, Dorea formicigenerans, and Erysipelatoclostridium ramosum, exhibited high BSH activity but limited secondary BA production. Ruminoccocus torques, Lachnoclostridium hylemonae, Collinsella stercoris, and Escherichia fergusonii did not generate expected secondary BAs, likely due to insufficient unconjugated BA substrates. For these species, BSH activity appeared to be the rate-limiting step for further transformations to occur^2,36,37.

Unexpectedly, we identified conjugated 7-oxo and 3-oxoBAs in our samples, sometimes constituting ~15% of the total BA pool (Fig. 1). At 100 µM, 13 strains produced conjugated -oxoBAs, while 11 strains did so at 500 µM (Fig. 1). Production of these BAs could result from either (1) sequential deconjugation, dehydrogenation, and reconjugation of BAs, or (2) direct dehydrogenation of the C3 and C7 hydroxyl groups on conjugated BAs by HSDs. Notably, Bacteroides finegoldii produced conjugated secondary BAs without detectable BSH activity, suggesting that HSDs can act directly on conjugated BAs. This observation challenges the widely accepted view that deconjugation is a prerequisite for further BA modifications. The remaining species exhibited both BSH and HSD activity, preventing us from being able to determine, based on a single time point of data, whether HSD activity occurred on conjugated BAs, unconjugated BAs, or both (Fig. 1).

Deconjugation correlates with MCBA production

In recent years, a new mechanism by which bacteria diversify the BA pool was discovered: the conjugation of BAs to amino acids to produce MCBAs^17,18. More recently, BSHs were identified as the enzymes responsible for this conjugation, or transferase, activity^19,20. Our analysis revealed that when provided 100 μM of conjugated BAs, 44 out of the 77 bacterial strains assessed for BSH activity were capable of conjugating a wide range of amino acids to CA, CDCA, and DCA (Fig. 2). At 500 μM, 35 strains produced MCBAs (Supplementary Fig. 5). In general, strains with high deconjugating activity also exhibited high conjugating activity, resulting in MCBA production. Across both concentrations, amino acids were most frequently conjugated to CDCA, followed by DCA and then CA. Phenylalanine, alanine, glutamate, leucine/isoleucine, and methionine were the most frequently conjugated amino acids. The Bacillota and Actinomycetota phyla produced the highest levels and most diverse MCBAs, while the Pseudomonadota and Bacteroidota produced fewer and less diverse MCBAs (Fig. 2 and Supplementary Fig. 5). Notably, Streptococcus infantarius, Bifidobacterium bifidum, and C. comes produced the highest concentrations and most diverse MCBAs.

There were some exceptions to the correlation between deconjugation activity and MCBA production. Specifically, seven members of Bacteroidota, five of Bacillota, one of Actinomycetota, and one of Fusobacteroidota deconjugated BAs but did not produce MCBAs. In contrast, some species produced MCBAs, but did not exhibit deconjugating activity. These species were P. penneri, Proteus mirabilis, Akkermansia muciniphila, and Clostridium sporogenes (Fig. 2). The two Proteus species exhibited MCBA production at 500 µM, while other species, including Bacteroides coprophilus, Bacteroides plebeius, B. intestinalis, Bacteroides thetaiotaomicron 3731, Bacteroides adolescentis, C. sporogenes, and Clostridium symbiosum lost MCBA production at 500 µM (Supplementary Fig. 5). Finally, several species with newly discovered deconjugating activity also produced MCBAs, including H. filiformis, Subdoligranulum variabile, C. symbiosum, Blautia luti, Eubacterium ramulus, Bacteroides cellulosilyticus, and B. thetaiotaomicron 3731.

Deconjugation specificity follows phylogenetic patterns

Taurine- and glycine-conjugated BAs have variable levels of toxicity, and through BSH deconjugation specificity, bacteria have been shown to competitively colonize the intestines^38,39. Moreover, BA conjugation type modulates interactions with host receptors that impact host physiology^40,41. We determined the BSH deconjugation specificity preferences toward taurine- and glycine-conjugated BAs for all BAs combined, and for each BA core, CDCA or CA (Fig. 3). Strains were considered to have BSH specificity if they deconjugated 10% more of one type over the other. At the 100 µM concentration, 26 strains showed a preference for taurine-conjugated BAs, while 8 preferred glycine-conjugated BAs. At 500 µM, 22 strains preferred taurine-conjugated BAs and 11 preferred glycine-conjugated BAs. In addition, on average, 50% more taurine-conjugated BAs were deconjugated than glycine-conjugated BAs. This pattern was consistent across both BA cores (e.g., CDCA vs. CA).

Fig. 3. BSH specificity for glycine- and taurine-conjugated bile acids across phyla.

Strains are organized by phylogeny as indicated by the colored dashed lines of the tree. The green shaded heat map indicates the proportion of BAs deconjugated for each strain, regardless of specificity. Bioinformatically identified BSH motifs are listed on the right side of the heat map, with orange motifs designating predicted taurine-specificity, blue motifs designating predicted glycine specificity, and black motifs designating a BSH unassociated with either specificity. Percent of glycine deconjugation was subtracted from percent taurine deconjugation for each strain at each concentration and the absolute value was plotted to show total BSH preference (left set of columns). This process was repeated for conjugated BAs with a CDCA core (middle set of columns) and with a CA core (rightmost set of columns). A preference for taurine-conjugated BAs is indicated by an orange shaded background, while a preference for glycine-conjugated BAs is shaded in blue, each corresponding with the predicted specificity motif of the same color. Values for the 100 μM condition are in purple and for the 500 μM condition in pink.

Phylogenetic patterns in deconjugation specificity were apparent in the Bacteroidota and Bacillota. Of the 16 Bacteroides strains that exhibited deconjugation specificity, only one showed a preference for glycine-conjugated BAs. Across both concentrations, Bacteroides strains deconjugated ~47% more taurine-conjugated BAs than glycine-conjugated BAs, consistent with previous reports⁹. In contrast, deconjugation specificity within the Bacillota phylum was split, with a similar number of species showing preference for either taurine- or glycine-conjugated BAs. On average, these species deconjugated approximately 25% more taurine-conjugated BAs at 100 µM and 12% more at 500 µM.

In general, bacteria exhibited more complete deconjugation at 100 µM than at 500 µM (Fig. 3). This pattern applied to species in the Bacillota (Clostridium leptum, C. symbiosum, Mediterraneibacter lactaris, Dorea longicatena, Catenibacillus scindens CG19-1, and Tyzzerella nexilis), the Bacteroidota (Alistipes indistinctus, Bacteroides vulgatus, and Bacteroides stercoris), and the Pseudomonodota phyla (E. tarda). Interestingly, at the 500 µM concentration, many species exhibited increased deconjugation specificity (Fig. 3). For example, E. bolteae, Turicibacter sanguinis, and M. lactaris were able to deconjugate all BAs at 100 µM, but exhibited glycine preference at 500 µM. In addition, at the 500 µM concentration, Bacteroidota strains were more likely to deconjugate TCDCA than GCDCA, while at 100 µM conjugation preference was less pronounced. Conversely, and counterintuitively, Bifidobacterium dentium and L. scindens exhibited increased glycine specificity at 100 µM, yet complete deconjugation, and thus no specificity, at the 500 µM concentration. Altogether, these observations suggest a role of environmental conditions in BSH activity, with differences in concentrations of BAs leading to different levels of activity, potentially as a result of enzyme induction⁴² or saturation⁴³.

We performed an in silico analysis to compare predicted and observed deconjugation specificity. For this analysis we compiled the genomes of all 77 strains used in this study and identified their BSH protein sequences from the RefSeq NCBI database using the keywords “bile salt hydrolase” and “choloylglycine hydrolase”. We then built a hidden Markov model (HMM) using 84 BSHs from a previous study that found taurine-preferring BSH to have the motif G-X-G with X = T/V and glycine-preferring enzymes to have S-R-X with X = G/S⁸. Our analysis identified one BSH in 39 species, two BSHs in nine species, and three BSHs in one species (Fig. 3). The BSHs in these species were predicted as taurine-preferring if they contained the motif G-X-G (with X = A/H/Q/S) and glycine-preferring if they contained the motif S-R-G. Among the identified BSHs, 23 were taurine preferring, spanning four phyla, and 24 were glycine-preferring, all in the Bacillota phylum.

For most phyla, there was agreement between the predicted and observed BSH deconjugation specificity (Fig. 3). In the Bacteroidota and Fusobacteriota, the identified BSHs were predicted to have taurine specificity, which was consistent with our observed in vitro activity. In the Lentisphaera and Verrucomicrobiota, no bsh were identified, and no deconjugation activity was observed. For several members of the Bacillota, although conjugation preference varied, it was accurately predicted. For example, the BSH from S. variabile contained the G-Q-G motif, associated with taurine specificity, while the BSH from D. longicatena had the S-R-G motif, associated with glycine specificity, both of which were consistent with the observed activity. However, T. sanguinis was predicted to have taurine preference in two out of its three BSHs, yet it exhibited glycine preference at 500 µM.

Discrepancies between predicted and observed activity were most notable when bacteria were able to fully deconjugate all BAs, such as those in the Actinomycetota and several species in the Bacillota. Most species in the Actinomycetota were broadly predicted to have taurine specificity, but deconjugated all supplied BAs. Similarly, in the Bacillota, Entercoccus species, along with several others, were predicted to have glycine specificity but deconjugated all BAs, while Eisenbergella tayi and B. hydrogenotrophica were predicted to have taurine specificity but did not deconjugate any BAs (Fig. 3). The D-S-G motif in BSHs from the Pseudomonadota did not appear to confer deconjugating activity. Conversely, in some species a bsh was not identified, yet deconjugation specificity was still observed. For example, D. formicigenerans did not have an identified bsh, but showed strong glycine preference, while C. intestinalis, H. filiformis, E. ramosum, and Blautia hansenii did not have an identified bsh, but exhibited strong taurine preference. In addition, the motif F-S-G may confer taurine specificity, as seen with F. plautii and C. leptum. Interestingly, BSHs with the F-S-G motif clustered more closely with glycine-preferring BSHs in other Bacillota species. These findings provide further support for the association of specific motifs with deconjugation preferences across phyla.

Investigation of BSH dynamics in monoculture

Previous studies indicate that BSHs are intracellular enzymes with activity typically coupled to growth⁴⁴. BSH activity is recognized as a necessary action preceding further BA transformations^2,15,36. To investigate the dynamics of BSH activity in relation to bacterial growth, we selected seven phylogenetically diverse species and monitored them over a period of 72 h. Each species was cultured with conjugated BAs, each at a concentration of 100 µM. Cultures were sampled to quantify BAs and measure optical density. These experiments revealed nuanced deconjugation dynamics across species and identified two distinct patterns of BSH-mediated secondary BA production (Fig. 4 and Supplementary Fig. 6).

Fig. 4. BSH dynamics in monoculture.

A BA transforming activity for four species with complete BSH activity. B BA transforming activity for four species that produce conjugated secondary BAs, either concurrently or without any BSH activity. Strains were grown in triplicate and sampled over 72 h. Error bars represent the standard deviation of each average measurement. Figure legend shows BA measurements presented in each graph. Conjugated BAs are in Rows 1 and 3, unconjugated BAs in Rows 2 and 4, and conjugated secondary BAs are in Row 5. Culture growth was measured by optical density at a wavelength of 600 nm (OD600) and is indicated by a black dashed line.

Timing of BSH activity varies across taxa

While four species, C. comes, B. dentium, B. plebeius, and L. ruminis, exhibited BSH activity coupled with growth and deconjugated all five conjugated BAs during exponential phase (Fig. 4A), the other four species, C. intestinalis, B. ovatus, F. varium, and B. finegoldii, did not follow the same pattern (Fig. 4B). Among species that deconjugated all BAs, substrate preferences could be discerned. For instance, C. comes preferred to deconjugate glycine-conjugated BAs over taurine-conjugated BAs, with an ~ 3-h lag between the deconjugation of each type. B. dentium showed the same preference, but with a shorter gap of ~1.5 h between glycine- and taurine-conjugated BAs. B. plebeius and L. ruminis showed less distinct preferences, but appeared to deconjugate GCDCA first and TCA last, similar to C. comes (Fig. 4A). These data highlight the importance of time-series analyses in determining enzyme specificity, especially when deconjugation goes to completion.

The remaining four species had varied BSH patterns, either not coupling deconjugation to growth, or failing to deconjugate all BAs within 72 h (Fig. 4B). C. intestinalis coupled BSH activity to growth but displayed a marked substrate preference, rapidly deconjugating all taurine-conjugated BAs while leaving glycine-conjugated BAs untouched. B. ovatus deconjugated glycine- and taurine-conjugated CDCAs during exponential growth, but did not deconjugate the remaining BAs until it reached stationary phase at 11 h. F. varium deconjugated approximately half of the provided TCDCA during exponential growth and the remainder during stationary phase. Between 24 and 72 h, it deconjugated the other taurine-conjugated BAs, but it is unclear if glycine-CBAs were deconjugated, as their disappearance may be explained by HSD activity to produce conjugated secondary BAs. B. finegoldii, however, did not deconjugate any BAs, and only produced the conjugated secondary BA, G-7-oxoLCA. These findings highlight that while many bacteria couple BSH activity to growth, others do not. The timing and specificity of BSH activity have significant implications for the network of potential BA transformations by gut bacteria in in vivo systems.

HSD activity produces conjugated-oxoBAs

While HSD activity is known to produce -oxoBAs from unconjugated BAs, we observed that species with slow-acting, incomplete, or no BSH activity generated conjugated-oxoBAs throughout the time-course. The presence of slow-acting or incomplete BSH activity resulted in a mixed pool of unconjugated and conjugated BAs, enabling C. intestinalis, B. ovatus, and F. varium to simultaneously produce both conjugated and unconjugated-oxoBAs via HSD activity (Fig. 4B). The complete lack of BSH activity, but active HSD in B. finegoldii allowed it to produce conjugated secondary BAs in the absence of unconjugated BAs. While deconjugation has long been considered a prerequisite for further transformations, these results indicate that dehydrogenation by HSD activity can occur concurrently or before deconjugating activity.

C. intestinalis rapidly deconjugated taurine-conjugated BAs, but exhibited no activity against glycine-conjugated BAs. As expected, CDCA, DCA, and CA were transformed into 3-oxoCDCA, 3-oxoDCA, and 3-oxoCA, respectively. Concurrent with the production of unconjugated -oxoBA production, G-3-oxoBAs, G-3-oxoCDCA, and G-3-oxoCA appeared (Fig. 4B). Their presence indicated that the 3ɑ-HSD in C. intestinalis was active on both conjugated and unconjugated BAs present in the media. The presence of glycine-conjugated secondary BAs is unlikely to result from reconjugation of 3-oxoBAs, as C. intestinalis BSH activity is specific, and did not deconjugate glycine-conjugated BAs or produce MCBAs (Fig. 2 and Supplementary Fig. 5). Interestingly, all -oxoBAs were transient, peaking at ~15 µM at 6 h but dropping below 5 µM by 12 h of growth. Concentrations increased again after 24 h, suggesting that C. intestinalis HSD activity may be reactivated during stationary phase.

Similarly, BSH activity in B. ovatus and F. varium produced a mixed pool of unconjugated and conjugated BAs (Fig. 4B). B. ovatus BSH deconjugated BAs in a staggered manner, preferring CDCAs, then CAs, and DCA. CDCA accumulated in the media before it was transformed into 7-oxoLCA via 7ɑ-HSD activity. Subsequently, DCA and then CA levels increased, and after 12 h, both 7-oxoDCA and G-7-oxoDCA were produced. The concurrent production of conjugated and unconjugated-oxoBAs suggests that the B. ovatus HSD was active on both conjugated and unconjugated BAs. While reconjugation could explain the production of G-7-oxoDCA, it is unlikely because G-7-oxoLCA was not observed. F. varium followed a similar pattern to B. ovatus, preferentially deconjugating TCDCA to release CDCA throughout its growth. In addition, F. varium HSD activity simultaneously produced 7-oxoBAs from CA and CDCA, as well as G-7-oxoLCA and G-7-oxoDCA from GCDCA and GCA, respectively (Fig. 4B). The modest increase in conjugated BAs between 48 and 72 h could be attributed to reconjugation activity, but could also be explained by reversible HSD activity, reforming conjugated primary BAs from conjugated secondary BAs. Either mechanism broadens the known repertoire of BA transformations performed by gut bacteria.

BSH activity correlates with MCBA production

BSHs have been identified as the enzymes responsible for conjugating BAs to amino acids to produce MCBAs^19,20. Our data show that levels of BSH activity directly correlated with MCBA production. The four species exhibiting rapid BSH activity and complete deconjugation of all BAs produced MCBAs, while the four species with slow, incomplete, or no BSH activity did not (Fig. 4 and Supplementary Fig. 6). Estimated MCBA concentrations reached 8 µM for C. comes, 0.4 µM for B. dentium, 1.4 µM for B. plebeius, and 6 µM for L. ruminis, with production peaking concurrently with the levels of unconjugated BAs (Fig. 4A). Consistent with previous studies, the most abundant MCBAs were conjugated to the amino acids glutamate, glutamine, alanine, and asparagine¹⁹. Out of the eight species that we analyzed, those with highest BSH activity lacked other types of BA transforming activity. This observation suggests that the combination of robust BSH activity, coupled with the availability of unconjugated BAs that were not diverted toward secondary BA production, promoted MCBA formation.

Coculture experiments reveal BSH impact on BA pool

Although it is widely accepted that bacteria with distinct BA transforming capabilities perform sequential modifications on BAs, direct experimental evidence for this process remains limited^16,22,45. To examine whether sequential transformations take place and assess the impact of BSH activity on secondary BA production, we cocultured bacteria with varying levels of BSH activity—Bifidobacterium angulatum, C. aerofaciens, S. infantarius, and C. symbiosum—alongside B. thetaiotaomicron VPI-5482, which exhibits limited BSH activity but robust secondary BA production via 7ɑ-HSD activity. Each species was provided with 100 µM of each of the five conjugated BAs and cultured individually and in coculture for 72 h. Cell growth was monitored by optical density, and samples were collected at regular intervals for LC-MS/MS analysis of BA concentrations.

The levels and timing of BSH activity directly influenced secondary BA production. When cultured alone, B. angulatum fully deconjugated all provided conjugated BAs within 10 h, generating CA, CDCA, and DCA (Fig. 5A). In coculture with B. thetaiotaomicron, starting at approximately three hours of growth, CA and CDCA were transformed into 7-oxoDCA and 7-oxoLCA, respectively (Fig. 5C). Together, these two species were able to transform conjugated primary BAs into unconjugated secondary BAs, a process that neither species could achieve independently.

Fig. 5. BSH dynamics in coculture.

A BA transforming activity throughout growth for the four species with BSH activity. B BA transforming activity for B. thetaiotaomicron, which has HSD activity. C BA transforming activity for each coculture, with B. thetaiotaomicron abbreviated to B. theta in panel headings. Strains with differing BA transforming capabilities were grown in triplicate individually and in coculture and sampled over the course of 72 h. Error bars represent the standard deviation of each averaged measurement. Culture growth was measured by optical density at a wavelength of 600 nm (OD600) and is indicated by a black dashed line. Conjugated BAs are in Rows 1 and 3, unconjugated BAs in Rows 2 and 4, and conjugated secondary BAs are in Row 5.

C. aerofaciens deconjugated all conjugated BAs to produce unconjugated BAs and exhibited faster rates of deconjugation for glycine-conjugated BAs compared to taurine-conjugated BAs in both monoculture and coculture (GCDCA > GCA > TCDCA = TDCA > TCA) (Fig. 5). In monoculture, C. aerofaciens also converted unconjugated BAs into 12-oxoBAs and 3-oxoBAs (Fig. 5A). However, in coculture with B. thetaiotaomicron, CA and CDCA were transformed into 7-oxoDCA and 7-oxoLCA by B. thetaiotaomicron, which were then epimerized by C. aerofaciens into the 7β-BAs UCA and UDCA, respectively (Supplementary Fig. 7). This exchange between species demonstrated multiple sequential transformations, where the products of one transformation became substrates for the next. Notably, neither 12-oxoBAs nor 3-oxoBAs were detected in coculture. In this and the previous coculture, the rapid BSH activity of B. angulatum and C. aerofaciens resulted in an accumulation of unconjugated BAs (CA, CDCA, and DCA), which were subsequently transformed into 7-oxoBAs by B. thetaiotaomicron 7ɑ-HSD (Fig. 5C).

S. infantarius displayed very fast-acting BSH activity in both monoculture and coculture, deconjugating all conjugated BAs within six hours and generating unconjugated BAs that initially spiked and steadily decreased throughout the time course (Fig. 5). However, the anticipated production of 7-oxoBA in the S. infantarius-B. thetaiotaomicron coculture was limited. This observation might be explained by a loss of BAs to the S. infantarius bacterial membrane⁴⁶, to a BA transformation not identified by our analysis, or to some other catabolic activity. In both monoculture and coculture, S. infantarius exhibited low-level transient accumulation of the MCBAs CDCA/DCA-Glutamine and CDCA-Glutamate between 3 and 6 h of growth. While BSHs were identified as the enzymes responsible for MCBA production^19,20, our observation of transient MCBA production with S. infantarius suggests that BSHs are also responsible for deconjugation of MCBAs. Further studies will be needed to confirm this possibility.

Surprisingly, when B. thetaiotaomicron was grown in pure culture, it accumulated a large amount, ~100 µM, of G-7-oxoLCA from GCDCA by 72 h, leaving all other conjugated BAs intact (Fig. 5B). When C. symbiosum was grown alone, it selectively deconjugated taurine-conjugated BAs after 24 h to release unconjugated BAs (Fig. 5A). In coculture, the slow-acting BSH activity of C. symbiosum allowed for B. thetaiotaomicron 7ɑ-HSD activity to transform conjugated BAs into glycine- and taurine-conjugated secondary BAs (Fig. 5C and Supplementary Fig. 7). Conjugated secondary BA levels decreased after 24 h, indicating that they were subsequently deconjugated by C. symbiosum BSH to release 7-oxoBAs (Fig. 5C). If BSH activity had preceded HSD activity, we would have expected an accumulation of the unconjugated BAs CA, CDCA, and DCA before the production of the secondary 7-oxoBAs. Based on all time-series data, we concluded that HSD activity is more versatile than previously recognized, acting on both conjugated and unconjugated BAs to produce -oxoBAs. Ultimately, the composition of the resulting BA pool was determined by the combined activity and timing of BSH and HSDs from each bacterium.

B. thetaiotaomicron HSD acts directly on conjugated primary BAs

As previously outlined, there are two potential pathways for the production of conjugated secondary BAs: (1) deconjugation by bile salt hydrolase (BSH), followed by secondary transformation by hydroxysteroid dehydrogenase (HSD) and subsequent reconjugation by BSH, or (2) direct transformation of conjugated BAs by HSD. To determine which pathway was responsible for the production of conjugated secondary BAs, we deleted the 7ɑ-HSD gene (Δhsd) in B. thetaiotaomicron⁴⁷ using allelic exchange⁴⁸. If the multi-step pathway was occurring, knocking out the 7ɑ-HSD gene would result in an accumulation of CDCA due to deconjugation by BSH. However, if B. thetaiotaomicron 7ɑ-HSD acted directly on conjugated BAs, the provided conjugated BAs levels would remain unchanged.

We cultured both WT B. thetaiotaomicron and the Δhsd mutant in monoculture and coculture with C. symbiosum, providing five conjugated BAs at 100 µM each (Fig. 6). In monocultures, the WT B. thetaiotaomicron strain transformed GCDCA to G-7-oxoLCA (Fig. 6B), which we verified using a chemically synthesized and purified standard (Supplementary Fig. 4). In contrast, the Δhsd strain did not perform any transformations, leaving all conjugated BAs intact at 72 h (Fig. 6C). These results indicated that the 7ɑ-HSD in B. thetaiotaomicron directly dehydrogenated the C7 hydroxyl group of the conjugated primary BA, GCDCA, producing the conjugated secondary BA, G-7-oxoLCA. This finding reveals a previously undescribed BA transformation route, challenging the established view that BAs must be first deconjugated before transformations on the BA core can occur (Supplementary Fig. 1).

Fig. 6. B. thetaiotaomicron 7ɑ-HSD makes conjugated secondary BAs.

A BA transforming activity for C. symbiosum in monoculture. B BA transforming activity for WT B. thetaiotaomicron and C. symbiosum coculture. C BA transforming activity for B. thetaiotaomicron Δhsd and C. symbiosum coculture. Species were cultured in triplicate and sampled over the course of 72 h. Error bars represent the standard deviation of each averaged measurement. Culture growth was measured by optical density at a wavelength of 600 nm (OD600) and is indicated by a black dashed line. Conjugated BAs are in Row 1, unconjugated BAs in Row 2, and conjugated secondary BAs are in Row 3.

The coculture of C. symbiosum with WT B. thetaiotaomicron produced CA, DCA, 7-oxoLCA, 7-oxoDCA, and all four types of conjugated 7-oxoBAs (Fig. 6B). Given the diversity and abundance of conjugated secondary BAs, we tested this coculture using the B. thetaiotaomicron Δhsd strain. The absence of B. thetaiotaomicron 7ɑ-HSD in the coculture resulted in an accumulation of CA and a lack of conjugated secondary BAs, again demonstrating the ability of its HSD to directly transform conjugated primary BAs (Fig. 6C). Since conjugated secondary BAs were absent, we confirmed that their production in the WT coculture was due to activity by B. thetaiotaomicron 7ɑ-HSD; however, C. symbiosum increased the diversity of conjugated secondary BAs. It remains unclear why B. thetaiotaomicron in coculture with C. symbiosum was able to produce several types of conjugated secondary BAs, when it only produced one in pure culture. Altogether, these data show that the 7ɑ-HSD of B. thetaiotaomicron is responsible for the production of conjugated secondary BAs in both monoculture from GCDCA and coculture from GCDCA, TCDCA, GCA, and TCA (Fig. 6).

B. thetaiotaomicron HSD acts directly on conjugated primary BAs in vivo

To evaluate the production of conjugated secondary BAs in an in vivo setting, we used gnotobiotic mice colonized with either the WT strain or Δhsd mutant of B. thetaiotaomicron. Because GCDCA, the substrate for this transformation, is not endogenously produced in mice, we supplemented their drinking water with 0.1% GCDCA for 48 h. Fecal pellets were collected prior to euthanasia. BAs were extracted from fecal pellets and analyzed via LC-MS. Mice colonized with the WT strain accumulated an average of 6.29 pmol/mg GCDCA and 4.79 pmol/mg G-7-oxoLCA in their feces (Fig. 7). In contrast, mice colonized with the Δhsd strain accumulated an average of 12.12 pmol/mg GCDCA in their feces and produced no detectable G-7-oxoLCA (Fig. 7).

Fig. 7. The conjugated secondary BA, G-7-oxoLCA, is produced directly from GCDCA in vivo.

Mice were colonized with the WT strain (left) or Δhsd mutant (right) of B. thetaiotaomicron and provided water with 0.1% GCDCA for 48 h. Mouse fecal BAs were measured (pmol/mg) with GCDCA levels in black and G-7-oxoLCA levels in pink. Bar height is the average value for each BA measurement with data from each mouse visualized as an individual point (WT: n = 5, 1 male, 4 female, Δhsd: n = 4, 2 male, 2 female). Error bars represent the standard deviation of each averaged measurement. Statistical significance was determined using a one-sample, two-tailed t-test (p = 0.0289).

Importantly, CDCA and 7-oxoLCA, which could represent potential deconjugated intermediates, were either undetectable or present only at trace levels in both groups. In the feces of mice colonized with the WT strain, CDCA was not detected, and 7-oxoLCA levels remained low (0.2 pmol/mg) (Fig. 7). Similarly, in mice colonized with the Δhsd strain, CDCA was present at only 0.09 pmol/mg, and 7-oxoLCA was not detected (Fig. 7). These findings support the conclusion that B. thetaiotaomicron directly converts GCDCA into the conjugated secondary BA, G-7-oxoLCA, without going through deconjugated intermediates.

Colonization efficiency and the total concentrations of conjugated and unconjugated BA pools did not differ significantly (p > 0.05) between groups (Supplementary Fig. 8). These findings provide further evidence that conjugated secondary BAs are produced in vivo, expanding the network of known BAs and BA transformation pathways resulting from bacterial activity. Additional research is needed to investigate the physiological relevance of these conjugated secondary BAs.

Discussion

Our investigation of bile salt hydrolase (BSH) reveals that BSH activity is highly prevalent, with over 70% of the 77 tested strains exhibiting activity. Of these, 60% present varying substrate specificity for taurine- or glycine-conjugated BAs. Using coculture experiments, we demonstrate sequential transformations between bacterial species, highlighting the interplay between BSH and hydroxysteroid dehydrogenase (HSD) activities in shaping BA pool diversity. We find that rapid and complete BSH activity, followed by HSD activity, drives the production of unconjugated secondary BAs. Conversely, limited unconjugated secondary BA transformation activity coincides with the accumulation of MCBAs. Unexpectedly, delayed or incomplete deconjugation activity allows HSDs to act directly on conjugated BAs, producing conjugated secondary BAs, representing a previously uncharacterized transformation (Supplementary Fig. 1). These conjugated secondary BAs can subsequently undergo deconjugation by BSH, further diversifying the BA pool. In addition, we showed that conjugated secondary BAs are produced in vivo when mice were colonized by B. thetaiotaomicron, providing further evidence of their physiological relevance. Our findings challenge the conventional view of BSH activity as the single gateway reaction preceding other BA transformations, instead revealing its nuanced role in BA metabolism.

We further demonstrate that BSHs exhibit diverse dynamics, specificity, and sensitivity, broadening our understanding of the activity of this enzyme class. Historically, BSH activity has been associated primarily with the exponential growth phase, with a few exceptions noted in Bacteroides species^16,44. However, our time-series analysis shows that over one-quarter of the tested species decouple BSH activity from growth. Notably, we observe stationary phase BSH expression in C. symbiosum (Fig. 5), a phenomenon not previously reported in Clostridium species. This observation indicates that while some BSHs are active during exponential growth or in response to BA exposure, others are active during the stationary phase, likely responding to alternative environmental cues.

In general, bacteria deconjugate a similar percentage of BAs at 100 µM compared to 500 µM, although there are exceptions. While E. tarda deconjugates a higher percentage of BAs at 100 µM, L. scindens and B. dentium, exhibit higher deconjugation efficiency at 500 µM (Fig. 3). For the latter species, this response may reflect a detoxification mechanism triggered by higher BA concentrations in these two species. Bacteria also exhibit stronger substrate preferences at elevated BA concentrations, possibly due to inhibited growth at high BA levels, as observed with T. sanguinis⁴⁹. A similar line of reasoning may explain reductions in MCBA production at higher BA concentrations. However, previous research suggests that BSH activity and reduced bacterial growth due to BA toxicity are unrelated properties, at least in the Lactobacilli⁵⁰. Overall, these findings suggest that enzyme capacity plays a role in BA deconjugation at physiologically relevant concentrations encountered in the gut.

In some species, such as C. intestinalis, H. filiformis, D. formicigenerans, and E. ramosum, secondary BA production by HSDs is limited despite robust BSH activity and the availability of unconjugated BA substrates. Similarly, B. intestinalis, B. ovatus, and B. xylanisolvens exhibit reduced HSD activity at 500 µM compared to 100 µM. The underlying causes of these differences in secondary BA production remain unclear but may result from regulatory mechanisms influenced by media composition or BA concentrations.

Across both concentrations, our methods recovered 86–87% of the administered BAs. The unrecovered fraction may represent BAs associated with the cell pellet, either retained intracellularly or bound to the membrane. Variation in recovery could reflect differences in bacterial membrane composition or cellular transport processes across bacterial species, as well as the chemical properties of individual BAs. Further studies are needed to quantify pellet-associated BAs directly and to clarify the mechanisms underlying incomplete recovery.

We demonstrate that B. thetaiotaomicron oxidizes GCDCA to G-7-oxoLCA in pure culture when provided with conjugated BAs. This ability to produce conjugated secondary BAs is not restricted to B. thetaiotaomicron, as other members of the Bacteroidota, most notably B. finegoldii, as well as F. varium (Fusobacteriodota), C. intestinalis (Actinomycetota), and F. plautii (Bacillota), appear to exhibit similar activity. Bacteroides caccae, B. thetaiotaomicron VPI-5482, and B. thetaiotaomicron 3731 produce conjugated secondary BAs more extensively at 500 µM BA concentrations than at 100 µM (Fig. 1). For these strains, increased HSD activity may play a role in detoxifying higher conjugated BA concentrations. For other species, however, the factors driving HSD activity, whether related to redox balance⁵¹, detoxification⁵², or a combination of both, remain unclear.

The observed production of conjugated secondary BAs aligns with prior studies showing that cell crude extracts or partially purified HSD enzymes from Bacteroides fragilis⁵³, B. thetaiotaomicron^47,52, and Clostridium limosum⁵⁴ have HSD activity on both conjugated and unconjugated BAs. By integrating whole-cell assays, time-dependent LC-MS/MS-based BA measurements, molecular genetics, and mouse models our study expands upon these findings by directly validating HSD activity on conjugated BAs, highlighting the widespread promiscuity of HSDs and the relevance of conjugated secondary BAs in vivo.

Coculture experiments reveal that pairing two species can enable sequential, or additive BA transformations, as observed in cocultures of B. thetaiotaomicron with B. angulatum, S. infantarius, and C. aerofaciens (Fig. 5). Notably, in B. thetaiotaomicron and C. aerofaciens cocultures, the combination of 7ɑ-HSD and 7β-HSD activity leads to the production of urso-BAs. The factors governing HSD activity directionality remain poorly understood. However, our results align with previous findings that C. aerofaciens 7β-HSD switches from reduction to oxidation at ~12 h of growth⁴⁵. Such reversible HSD activity may play a pivotal role in shaping the BA pool by redirecting BAs toward or away from other transformations. For example, the dehydroxylation of BAs to produce DCA and LCA cannot occur on -oxoBAs. These urso-BAs, which are more hydrophilic and less toxic to both the microbiota and the host due to their hydrophilicity⁵⁵, have therapeutic relevance for biliary disorders⁵⁶, heart disease⁵⁷, and various cancers⁵⁸.

When HSD activity is limited and BSH activity is high, MCBAs are often produced by the enzyme’s recently described BSH acyltransferase activity^19,20. Consistent with prior studies, higher concentrations of MCBAs correlates with greater diversity¹⁹. Certain species, such as Ruminoccocus gnavus, B. bifidum, and E. bolteae produce robust and diverse MCBAs, whereas others like L. scindens, Holdemanella hathewayi, and C. symbiosum show more specificity and produce fewer MCBAs^17,19,59,60. In general, more species were capable of producing MCBAs from conjugated BAs than from unconjugated BAs based on our prior study¹⁷. Increased production of MCBAs may be due to greater induction of BSH by the presence of conjugated BAs.

Similar to how BSHs exhibit specificity for glycine- or taurine-conjugated BAs, they also demonstrate specificity in the types of MCBAs they produce. Species with BSH specificity for glycine or taurine may exhibit deconjugating activity, yet not produce MCBAs with alternative amino acid conjugations. MCBA conjugation profiles may not only be shaped by BSH specificity but also by autotrophic amino acid production of each species²⁰. While glycine-conjugated MCBAs are known to be prevalent, their production could not be measured in this study because glycine-conjugated BAs (GCA and GCDCA) were substrates in the media.

The relationship between bioinformatically predicted BSH activity and actual BA deconjugation is complex and often inconsistent. Several species exhibit deconjugating activity even though no identifiable BSH homologs are found, suggesting that non-homologous enzymes may be responsible for BA deconjugation in these organisms. This observation is noted for H. filiformis, C. intestinalis, E. ramosum, C. symbiosum, D. formicigenerans, B. hansenii, B. stercoris, B. caccae, B. cellulosilyticus, and B. thetaiotaomicron 3731. Interestingly, all these species, except for D. formicigenerans, preferentially deconjugate taurine-conjugated BAs, providing insights into the evolutionary lineage and specialization of taurine-specific BSHs. In contrast, P. rettgeri, P. penneri, and P. mirabilis of the Pseudomonadota, and C. sporogenes and B. hydrogenotrophica of the Bacillota, possess bioinformatically identified BSHs but do not display measurable activity. This lack of activity may result from the absence of functional transporters for importing conjugated BAs, the presence of genetic alterations that render the enzyme inactive, or the lack of BSH expression under the environmental conditions tested in our study. These findings highlight that while bioinformatic predictions are highly informative, they are not sufficient to fully identify BA transforming activity, particularly when novel enzymes, regulatory differences, or uncharacterized transport systems may be involved.

Bacterial BA transformations have long been recognized for their role in converting a limited set of host-synthesized BAs into hundreds, if not thousands, of modified derivatives⁶¹. These diverse BAs vary in their capacity to facilitate fat absorption, shape the gut microbial community, and interact with host receptors. Traditionally, BA transformation has been described as a linear process in which primary conjugated BAs are first deconjugated, releasing free primary BAs that can subsequently be modified into secondary BAs by gut bacteria. However, the recent discovery that gut bacteria can conjugate BAs to glycine, reforming conjugated primary BAs through microbial action, shifts our understanding of microbial contributions to the BA pool and expands the known activity of BSH.

In this study, the identification of conjugated secondary BAs reveals a novel BA transformation and challenges the long-held assumption that deconjugation is a prerequisite for further transformation. Our findings continue to overturn conventional views of the BA transformation process and underscore the need to move beyond the concept of a linear pathway. Instead, we propose a transformation network model that accounts for the timing, specificity, and interconnected nature of BA modifications within a dynamic and ever-changing BA pool. Such a framework will provide a more accurate and relevant understanding of bacterial activity in in vivo systems.

Systematic investigations of in vitro BA transformations by large cohorts of gut bacteria bring us closer to understanding and interpreting in vivo BA pools associated with metabolic diseases, gastrointestinal cancers, and improved health outcomes post-bariatric surgery. Our findings highlight the remarkable diversity of BAs and the transformations that produce them in both in vitro and in vivo environments, emphasizing their potential for manipulation to improve human health.

Further studies are needed to identify the mechanisms that regulate BA-transforming activity in gut bacteria, including the roles of BA transporters, BSH expression levels, and cell death in shaping the dynamic BA network. Systematic, time-course analyses using diverse BA substrates under physiologically relevant conditions are essential to unravel the regulatory pathways and environmental cues that drive these activities. Future studies should build on these findings by examining microbial BA transformations in vivo to better define their physiological relevance. A deeper understanding of these factors is critical if we are to reliably manipulate the BA pool to promote beneficial health outcomes.

Methods

Bacterial strains

All strains are listed in Supplementary Table 1. The 77 bacterial strains used in this study were selected from a curated collection established over several years in the Rey Laboratory^17,62. This collection represents the major phyla of the human gut microbiota and largely overlaps with the Human Microbiome Project reference set, which makes them highly relevant for gut microbiome researchers as they are genome sequenced and commercially available⁶³. Further information for strains isolated in our lab: Strain 1RE7 was isolated from an anaerobic enrichment in a medium supplemented with rutin, inoculated with a human fecal sample. The strain consumes both rutin and quercetin. The sequence of the full-length 16S rRNA gene is 96% identical to that of C. scindens CG19-1. Strain J02 was isolated from an anaerobic enrichment in medium supplemented with rutin, inoculated with a human fecal sample (WLS #82). The sequence of the 16S rRNA gene is >99% identical to that of E. tayi strain B086562 (783/784 bases match). Strain K01 was isolated from an anaerobic enrichment in media supplemented with rutin, inoculated with a human fecal sample. The sequence of the 16S rRNA gene is >99% identical to that of Enterococcus durans JCM8725 (900/901 bases match) and similarly matches many E. faecium strains (902/903 bases match). Strain J01 was isolated from an anaerobic enrichment in media supplemented with quercetin, inoculated with a human fecal sample. The sequence of the 16S rRNA gene is >99% identical to that of several Enterococcus species (lactis, durans, faecium) all with 870/871 bases matching. Strain K02 was isolated from an anaerobic enrichment in medium supplemented with rutin, inoculated with a human fecal sample (WLS #10). The sequence of the 16S rRNA gene is >100% identical to multiple P. mirabilis strains (823/823 bases match). Strain L02 was isolated from an anaerobic enrichment in media supplemented with quercetin, inoculated with a human fecal sample. The sequence of the 16S rRNA gene is >99% identical to that of several S. anginosis strains (854/856 bases match). These lab isolates are not yet genome sequenced. Freezer stocks can be made available upon request.

Bacterial growth conditions

For the systematic BSH analysis, all strains were grown on Colossal Mega Medium (CMM), which was filter-sterilized and stored in a Coy anaerobic chamber (5% H₂, 20% CO₂, and 75% N₂) at least 24 h prior to use. CMM contains (per liter tap distilled water): 100 mL (1 M, pH 7.2) potassium phosphate buffer, 10 × g tryptone peptone, 5 × g yeast extract, 5 × g meat extract, 4 mL (25 mg/100 mL) resazurin, 1.8 × g D-glucose, 0.9 × g D-maltose, 0.86 × g D-cellobiose, 0.46 × g D-fructose, 2 × g CH₃COONa·3 H₂O, 0.02 × g MgSO₄·7 H₂O, 2.1 g NaHCO₃, 0.08 × g NaCl, 1 mL (0.8 × g/100 mL) CaCl₂, 1 mL (1 mg/mL in 100% ethanol) vitamin K₃ (menadione), 1 mL (1.2 mg hematin/mL in 0.2 M histidine, pH 8.0) histidine hematin, 2 mL (25% vol/vol) Tween 80, 10 mL ATCC MD-VS vitamin mix, 10 mL ATCC MD-TMS trace mineral mix, 1 mL (40 mg/100 mL) FeSO₄·7 H₂O, and 0.5 × g L-cysteine·HCl. This specific medium was designed to allow growth for all species in this study. Additions and modifications for specific strains were as follows: For cultures of Akkermansia muciniphila the medium was amended with 1 mg/mL mucin. For cultures of Clostridium orbiscindens the medium was amended with lysine.

For time-course monocultures and cocultures, all strains were grown on Low Yeast Extract (LYE) medium, which was made anaerobic using a triple-vacuumed pressure bottle before being brought into a Coy anaerobic chamber (5% H₂, 20% CO₂, and 75% N₂) and then filter-sterilized. LYE medium contains (500 mL Milli-Q water): 50 mL (1 M, pH 7.0) potassium phosphate buffer, 0.36 × g tricine, 2.0 mL (0.025%) resazurin, 1 × g yeast extract, 0.5 mL (25% [vol/vol]) tween 80, 3.4 × g CH₃COONa·3 H₂O (FW 136), 0.55 × g sodium succinate hexahydrate (FW 270), 1.46 × g sodium chloride (FW 58.44), 0.54 × g ammonium chloride (FW 53.49), 3.6 × g D-glucose (FW 180.16), 1.8 × g D-maltose (FW 360.3), 1.0 mL (0.5 M) potassium sulfate, 1.0 mL (1.0 M) magnesium chloride hexahydrate (MgCl₂·6H₂O), 0.2 mL (1.0 M) calcium chloride dihydrate (CaCl₂·2H₂O), 1.68 × g sodium bicarbonate (FW 84.0), 0.5 mL (1.2 mg hematin/mL in 0.2 M histidine, pH 8.0) histidine hematin solution, 0.125 mL vitamin K1 + K3 solution (used 2× stock), 10 mL ATCC MD-VS vitamin mix, 5 mL 50× trace mineral mix solution [0.29 mL (30 µM) MnCl₂·4H₂O, 0.06 mL (10 µM) ZnCl₂, 0.047 mL (4 µM) CoCl₂·6H₂O, 0.012 mL (1 µM) Na₂MoO₄·2H₂O, 0.008 mL (1 µM) Na₂SeO₃, 0.059 mL (5 µM) NiCl₂·6H₂O, 0.016 mL (1 µM) Na₂WO₄·2H₂O, adjust volume to 1 L, store under N₂, refrigerated], 1 mL ferrous sulfate heptahydrate (FeSO₄·7H₂O), and 0.25 × g L-cysteine HCl. Adjust pH to ∼7.3–7.1.

Sample handling and experimental conditions

For the systematic investigation reported in Figs. 1–3 and Supplementary Fig. 5, cultures were started from freezer stocks and grown overnight to a high density (O.D. 600 range of 0.349–1.9, measured directly in the tube) in Hungate tubes containing CMM with an atmosphere of 75% N₂, 20% CO_2, 5% H₂ at 37 °C. These cultures were then used to inoculate (1:15 dilution) 3 mL of CMM in Hungate tubes amended with BAs. There were 2 sets of conditions; media contained 100 μM or 500 μM of each of the five conjugated BAs: GCA, GCDCA, TCA, TCDCA, and DCA. BAs were dissolved in water at 20 mM and filter sterilized for use in each experiment. Each concentration condition was tested in duplicate for all 77 strains, totaling 308 samples, and water blanks run between every 50 samples. In addition, we tested for spontaneous BA degradation or transformation in uninoculated controls containing BAs that were run every 50 samples. Once cultures reached stationary phase, 1 mL of culture was collected, spun down at room temperature for 10 min at 10,000 × g, and the supernatant transferred to a fresh tube. Samples were stored at −80 °C until being thawed for analysis. Using HPLC-grade H₂O, the supernatants were diluted 1:200 or 1:1000 for the 100 μM or 500 μM conditions, respectively. After dilution, 100 μL were transferred to an HPLC vial for analysis.

For the monoculture and coculture time course analyses in Figs. 4–6 and Supplementary Fig. 6, individual freezer stocks were inoculated into CMM and grown overnight to a high bacterial density (OD₆₀₀ > 2). The following day, multiple dilutions were made for each culture in LYE medium containing 0.1% yeast extract and allowed to grow overnight. Cultures in exponential phase were then used as inoculum. Growth curves and BA measurements for Fig. 4 and Supplementary Fig. 6 were performed in 125 mL Erlenmeyer flasks containing 30 mL of 0.1% LYE medium. Growth curves and BA measurements for Figs. 5 and 6 were performed in Hungate tubes containing 10 mL of 0.1% LYE medium. Both sets were amended with 100 µM each of GCA, GCDCA, TCA, TCDCA, and TDCA, except for the B. finegoldii monoculture, which lacked TDCA. BAs were dissolved in water at 20 mM and filter sterilized for use in each experiment. Monoculture starting ODs were 0.05 and cocultures were started with an equal proportion of both monocultures (total OD being ~ 0.1). For B. thetaiotaomicron and C. symbiosum, monocultures starting ODs were 0.03 and 0.12, respectively, with the coculture OD at ~ 0.15. Uninoculated controls containing BAs were used to assess spontaneous BA degradation and transformation. All experiments were performed in triplicate. Samples were drawn at multiple timepoints, including the zero time point based on the growth pattern: rigorous sampling was done during exponential phase (7–8 timepoints) and 4–5 timepoints were included in stationary phase. 0.3 mL of culture was collected at each timepoint and spun down and the supernatant was transferred to a new tube. Samples were stored at −80 °C until being thawed for analysis. Samples were diluted at 1:100 using HPLC-grade H₂O for analysis by HPLC-MS. For Fig. 4, a total of 377 samples were analyzed, including pre-inoculum controls in triplicate for each strain. For Fig. 5, a total of 564 samples were analyzed, including uninoculated media controls, in triplicate at each time point. For Fig. 6, a total of 90 samples were analyzed, including uninoculated media controls, in triplicate at each time point.

uHPLC-MS/MS settings

Samples were analyzed using an ultra-high pressure liquid chromatography-tandem mass spectrometry (uHPLC-MS/MS) system consisting of a ThermoScientific Vanquish uHPLC system coupled to a heated electrospray ionization (using negative polarity) and hybrid quadrupole high resolution mass spectrometer (Q Exactive Orbitrap; Thermo Scientific). Settings for the ion source were: auxiliary gas flow rate of 10, sheath gas flow rate of 30, sweep gas flow rate of 1, 2.5 kV spray voltage, 320 °C capillary temperature, 300 °C heater temperature, and S-lens RF level of 50. Nitrogen was used as nebulizing gas by the ion trap source. Liquid chromatography (LC) separation was achieved using a Waters Acquity UPLC BEH C18 column with 1.7 μm particle size, 2.1 × 100 mm in length. Solvent A was water with 10 mM ammonium acetate adjusted to pH 6.0 with acetic acid. Solvent B was 100% methanol. The total run time was 31.5 min with the following gradient: a 0–24 min gradient from 30% solvent B (initial condition) to 100% solvent B; held 5 min at 100% solvent B; dropped to 30% solvent B for 2.5 min re-equilibration to initial condition. The flow rate was 200 μL/min throughout. Other LC parameters were as follows: autosampler temperature, 4 °C; injection volume, 10 μL; column temperature 50 °C. The MS method performed a full MS1 full-scan (290–1000 m/z) together with a series of parallel reaction monitoring (PRM) scans.

Identity of glyco-7-oxolithocholic acid (G-7-oxoLCA) was confirmed using a synthesized standard (See “Methods” section below called “Conjugated secondary bile acid synthesis”). All other conjugated secondary BAs were confirmed through LC-MS/MS fragmentation and predicted shifts in retention time based on known shifts between -oxoBA standards that have the same mass. The MS method performed a full MS1 full scan (290–2000 m/z) together with a series of PRM scans in positive mode. These MS2 scans (selected-ion fragmentation) were centered at m/z values of 448, 464, 498, and 514. Fragmentations were performed at 30 normalized collision energy. All scans used a resolution value of 17,500, an automatic gain control target value of 1E6, and a maximum injection time (IT) of 40 ms. Predicted and observed fragment sizes, along with mass error measurements, are listed in Supplementary Table 3. Spectra are compiled in Supplementary Fig. 3. A total of 8 samples were analyzed to generate the information in Supplementary Fig. 3 and Supplementary Table 3.

Determination of bile acid concentrations

Untargeted experimental MS data were converted to the mzXML format and used for targeted BA identification using El-MAVEN (v0.12.1-beta) and matching sample peaks to standard peaks⁶⁴. BAs were quantified using six-point external standard curves, with each BA ranging from 0.0625 to 2 μM. The detection limit was below 0.01 μM for all BAs. The threshold for reported core BA transformations was 0.008 μM as calculated based on a minimum raw signal cutoff of 10e4. Standards were purchased from Avanti Polar Lipids and dissolved and stored in methanol at −80 °C. See Supplementary Table 2 for BA standard names, structural features, and retention times. When BAs were not reported for any given experiment, it means that they were not produced. For MCBAs, compounds were identified by their exact mass (mass error of less than 2 parts per million) and previously determined retention times (Supplementary Table 4). Values were presented as z-scores to demonstrate relative abundance. The compound list for MCBAs can be found in Supplementary Table 4. Conjugated secondary BA concentrations, besides G-7-oxoLCA, which we measured using our synthesized standard (See “Methods” section called “Conjugated secondary bile acid synthesis”), were estimated using the commercially available glyco-12-oxolithocholanic acid (G-12-oxoLCA) and tauro-12-oxolithocholanic acid (T-12-oxoLCA). To determine what percentage of BAs were recovered through our sampling of the supernatant, we summed all endpoint BAs and compared them to the initial concentration of conjugated BAs (Supplementary Fig. 2). For Supplementary Fig. 2, MCBA concentrations were estimated using the standard curve for GCA. For BA measurements in time series analyses, unconjugated BAs were normalized to conjugated BA measurements in uninoculated controls. BA stock impurities were calculated using uninoculated media control samples. These levels of contamination were noted in Supplementary Table 5, and values below the determined cutoff were removed from our analyses.

In silico analysis

We accessed the genome sequences in NCBI for 77 bacterial strains and obtained all CDS genes from the 77 available genomes (amino acid sequences). We used the 84 curated BSH genes from Foley et al.⁸ as a BSH reference database for BLASTp. We used the same 84 BSH genes to build the hidden Markov model (HMM) to reserve BSH conserved domains when predicting BSH genes. We used the following criteria to determine the BSH from all CDS: gene length between 300 and 400 bp; at least one BLASTp to BSH reference genes (identity > 25%); and hit to BSH HMM (full sequence score > 100). Clustal Omega was used for multiple alignment of predicted BSH genes, taurine- or glycine-preferring BSH was predicted by a 3-residue selectivity loop: a taurine-preferring BSH contains “G-X-G” motif and a glycine-preferring BSH contains “S-R-X” motif.

Due to the known sequence similarity of BSH to penicillin V acylase (PVA) enzymes, we wanted to confirm that none of the BSH were misidentified. To distinguish between BSH and PVA candidate genes, the PVA HMM was built using reference PVA genes from O’Flaherty et al. The HMM search score threshold (full sequence score > 429.5) was determined by the minimal search score from PVA positive genes⁶⁵. We searched our BSH candidate genes against this PVA HMM, and the full sequence scores from all genes were less than 429.5.

Generation of B. thetaiotaomicron VPI-5482 Δhsd mutant

An in-frame hydroxysteroid dehydrogenase (hsd) deletion mutant was generated using a counter-selectable allelic exchange principle⁴⁸. Briefly, ~1 kilobase upstream (including the start codon) and downstream (including the stop codon) fragments of hsd open-reading frame were amplified using the high fidelity Herculase polymerase and the primer pairs (purchased from Integrated DNA Technologies) TAAGATTAGCATTATGAGTGGAAAAGAAAAAGTGATCTGG and ATATTTATGACATATATGTTGAGAATTTGATGATTAC; and CAACATATATGTCATAAATATACCCCGGAC and CGAATTCCTGCAGCCCGGGGATATAAGCGTACGAGGTG, respectively. These amplified fragments were cloned into the BamHI site of the suicide vector pLGB13 via Gibson assembly. The resulting construct was transformed into Escherichia coli S17-1 λ pir strain. After cloning, the junction sequence was verified by Sanger sequencing.

This vector was introduced into B. thetaiotaomicron by biparental mating (conjugation) between E. coli and B. thetaiotaomicron and single-crossover events were selected aerobically on CMM-blood agar plates containing gentamicin (20 µg/mL) and erythromycin (10 µg/mL) to enrich exconjugants. All subsequent steps were performed under anaerobic conditions. Resulting colonies were purified twice on the same plates. The cultures from the purified colonies were further plated on plates containing gentamicin and anhydrotetracycline (100 ng/mL), a counter-selection marker. To identify isolates that had lost the gene, colonies derived from a single original colony were screened by PCR. About 25% colonies were devoid of hsd gene, and one such colony was used following the sequence and functional verification.

Gnotobiotic mouse housing, BA dosing, and sample collection

All animal care procedures were approved by the University of Wisconsin–Madison Institutional Animal Care and Use Committee and conformed to NIH guidelines. All germ-free (GF) Mus musculus wild type C57BL/6 mice were maintained in a temperature and humidity controlled environment. The experiment was performed under a strict 12-h light/dark cycle, and the animals were maintained on standard chow diet and sterile water. GF status was validated by plating fecal suspensions in rich media, both aerobically and anaerobically, and by checking for amplification of the 16S rRNA gene using universal primers.

Approximately 8-week-old male and female mice were distributed into biocontainment cages (2–3 mice per cage; 4–5 mice per treatment group) and gavaged with either Bacteroides thetaiotaomicron WT (n = 5, 1 male, 4 female) or the corresponding hsdA knockout mutant (n = 4, 2 male, 2 female). Twenty-four hours later, mice were supplied with filter-sterilized water containing 0.1% GCDCA for another 48 h. Mice were euthanized 3 days post-gavage and feces were collected. Colonization efficiency and purity were determined by plating representative fecal suspensions (1 mouse/cage) on rich media plates (See Supplementary Fig. 8). BAs were extracted using the Abcam lipid extraction kit as per the manufacturer’s instructions and concentrations were determined as mentioned above. Samples for each sex were not analyzed separately, as the primary objective of this experiment was to demonstrate in vivo BA transformation rather than to assess sex-specific effects. For Fig. 7, 12 samples were analyzed, and 3 were removed from the analysis because GCDCA did not accumulate in those mice.

Conjugated secondary bile acid synthesis

All reagents were obtained commercially unless otherwise noted. Reactions were performed using oven-dried glassware under an atmosphere of nitrogen. Air- and moisture sensitive-liquids were transferred via syringe or stainless-steel cannula. Organic solutions were concentrated under reduced pressure ( ~ 15 Torr) by rotary evaporation. Product purification was accomplished using forced flow chromatography on Silicycle ultrapure silica gel (40–63 μm). Thin layer chromatography was performed on Millipore Sigma silica gel 60 F254 plates (250 μm). Visualization of the developed chromatogram was accomplished by irradiation with 254 nm UV light or by staining with ceric ammonium molybdate solution.

Nuclear magnetic resonance (NMR) spectra were acquired on a Bruker Advance 500 spectrometer operating at 500 MHz and 125 for ¹H and ¹³C NMR, respectively (Supplementary Fig. 4). Spectra are referenced internally according to residual solvent signals and were recorded at room temperature. Data for ¹H NMR are recorded as follows: chemical shift (δ, ppm), multiplicity (s, singlet; d, doublet; t, triplet; q, quartet; quint, quintet; m, multiplet; br, broad), coupling constant (Hz), integration. Data for ¹³C NMR are reported in terms of chemical shift (δ, ppm). MS2 data were acquired on a Q Exactive Orbitrap; Thermo Scientific (Supplementary Fig. 4).

Synthesis of glyco-7-oxolithocholic acid (G-7-oxoLCA) was adapted from a previously published method⁸. To a solution of 7-oxolithocholic acid (7-oxoLCA) (50.0 mg, 0.128 mmol) in anhydrous tetrahydrofuran (2.5 mL, 50.0 mM) was added triethylamine (21.4 μL, 0.154 mmol, 1.2 equiv) and ethyl chloroformate (14.6 μL, 0.154 mmol, 1.2 equiv) sequentially at 0 °C. The reaction was stirred at 0 °C for 1.5 h, following which time a cold solution of glycine (14.4 mg, 0.192 mmol, 1.5 equiv) and NaHCO₃ (16.1 mg, 0.192 mmol, 1.5 equiv) in H₂O (2.5 mL, 50.0 mM) was added. The reaction was allowed to warm to room temperature and stirred for 2 h, following which time the tetrahydrofuran was removed in vacuo. Then, 2.0 M HCl was added to acidify the mixture to pH <2. The resulting white precipitate was extracted with ethyl acetate (3 × 10 mL). The combined organic layers were washed with brine (30 mL), dried over MgSO₄, and concentrated in vacuo to a white solid. Purification of this material by flash silica gel chromatography (9:1 CH₂Cl₂/MeOH with 1% acetic acid) afforded G-7-oxoLCA as a white solid (40.5 mg, 71% yield).

Statistics and reproducibility

For the systematic analysis, a large sample size of 77 strains was chosen to assess bacteria spanning seven phyla common in the human gut. Each of these cultures were performed in duplicate. For the monoculture time series experiments, eight species were chosen for further analysis based on their distinct patterns in activity from the systematic assessment as well as based on their genetic diversity. The five coculture species were chosen based on their differential bile salt hydrolase and secondary bile acid producing activity. All monoculture and coculture experiments were performed with three biological replicates. Averages and standard deviations are reported in figures, with individual data points included in the source data.

The number of mice used in this study was estimated in accordance with the 3Rs principle (Replacement, Reduction, and Refinement) to minimize animal use while ensuring scientific validity⁶⁶. Based on extensive prior experience with gnotobiotic mouse models, we targeted group sizes of n = 4–5 animals per condition. This sample size has been shown to be sufficient for the reliable detection and quantification of bacterially derived metabolites in gnotobiotic mice. Similar group sizes have been successfully used in previous studies from our group to detect microbial metabolic products and host metabolic responses in gnotobiotic settings^62,67,68. Data from three mice (WT: 1 female, 1 male, and hsdKO: 1 female) were excluded from these analyses because they did not accumulate GCDCA in their tissues, which was the substrate for conjugated secondary bile acid production. Mice used in this study were randomly assigned to different groups. Statistical significance was determined using one-sample, two-tailed t-tests. The Investigators were not blinded to allocation during experiments and outcome assessment.

Ethical approval

All authors collaborated on this project and approved the submission of this paper.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

L.N.L., J.M., D.A.N., and F.E.R. conceived the project. R.L.K. provided gut bacterial strains and media recipes. L.E.C. performed systematic in vitro screening for BSH activity in all strains. B.G. and L.N.L. performed in vitro monoculture experiments. J.M. performed in vitro coculture experiments. J.M. generated the Δhsd mutant. L.N.L. and J.M. performed the in vivo experiment. Q.Z. wrote scripts for bioinformatic analyses. A.P.C. and H.E.B. synthesized the glyco-7-oxolithocholic acid standard and confirmed its purity. L.N.L. analyzed all data and generated figures. L.N.L., J.M., D.A.N., and F.E.R. interpreted data. D.M.S. provided guidance on LC-MS/MS methods and analyses. L.N.L. wrote the manuscript, D.A.N. substantively revised it, J.M. and F.E.R. provided feedback throughout the writing process, and all authors edited and approved the final manuscript.

Peer review

Peer review information

Nature Communications thanks Robert Quinn, and the other, anonymous, reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

Accession codes for sequenced bacterial strains have been included in Supplementary Table 1 in the Supplementary Information file. Accession codes and hyperlinks are included in the Source Data for Fig. 3. The raw data generated in this study are provided in the Source Data file. The metabolomics data generated in this study are available in the MassIVE database (gnps.ucsd.edu) under MassIVE ID MSV000100317. Source data are provided with this paper.

Code availability

All custom code can be found at: https://github.com/qijunz/Lucas_BSH_paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-68556-4.