Characterization of aminoacyl transfer by purified BSH/T
The first BSH/T to be purified and have its hydrolase activity characterized was from Clostridium perfringens (CpBSH/T)13. Due to its known interaction with conjugated bile acids, we investigated whether this enzyme could also exchange the conjugated amino acid. The enzyme-promoted hydrolysis of bile salts occurs via a covalently bound cysteinyl intermediate14 (Fig. 1a, steps 1 and 2) and CpBSH/T is active for hydrolysis over a broad pH range (pH 3-7) (Fig. S1). When CpBSH/T was incubated with an equimolar mix of 20 essential amino acids and TCA, amino acid hydrolysis to cholate (CA) was observed as expected, but molecules of CA acyl-conjugated with a variety of amino acids were also identified (Fig. 1b-d). Indeed, 16 of the 20 amino acids were shown to become linked to CA, with only aspartic acid, methionine, proline, and valine failing to be incorporated under these conditions (Table S1). CpBSH/T may be catalyzing this aminoacyltransfer by reacting the amino acids with the covalent catalytic intermediate (Fig. 1a, steps 1 and 3), using an amino acid as the nucleophile in lieu of a water molecule. The transferase activity exhibits a broad pH range (Fig. 1b), with an optimum at pH 5.3 (Fig. S2, Table S2) based on the relative abundance of peaks following 120-min incubation at 37°C. This value is slightly higher than the reported pH 4.5-4.9 optimum of TCA hydrolysis. The total MCBA abundance was greater than that of the remaining TCA after 120 min of incubation at pH 5.0 (Fig. 1c). Further characterization of CpBSH/T revealed that, at peak activity, the aminoacyltransfer activity reaches 19% of the enzyme’s hydrolysis activity. That is, approximately one amino acid is incorporated for every 5 TCA molecules hydrolyzed, indicating the transfer activity of this previously well-characterized hydrolase is a common property of the enzyme.
We then sought to determine if a similar panel of amino acids would be ligated when the enzyme was provided with Gly-CA. Based on the pH optimum calculated for aminoacyltransfer to TCA, CpBSH/T was incubated with Gly-CA at pH 5.0 and shown to transfer 11 of 19 supplied amino acids (glycine was not included due to its availability following Gly-CA hydrolysis) (Fig. 1e). Surprisingly, CpBSH/T conjugated valine and methionine to CA, a transformation that was not observed when CpBSH/T was supplied TCA. The reduced number of amino acids incorporated when CpBSH/T is supplied Gly-CA, only 11/19 versus 16/20 when starting with TCA, may be a consequence of competition by the larger amounts of glycine released by hydrolysis (Table S1).
In addition to the aminoacyltransfer reactions observed when using TCA or Gly-CA, we demonstrated that CpBSH/T could ligate amino acids directly to CA (Fig. 1f). This reaction is also likely to occur via a covalent intermediate (Fig. 1a, steps 4 and 3). CpBSH/T successfully ligated 12 of 20 amino acids, provided in equimolar concentrations, to CA (Table S1). The CA-supplemented mixture of amino acids and CpBSH/T resulted in the production of valine-conjugated CA (valocholic acid, Val-CA) and methionine-conjugated CA (methionocholic acid, Met-CA), as seen when starting with Gly-CA but unlike the products seen when using TCA. However, consistent with the TCA-supplemented CpBSH/T amino acid study, proline-conjugated CA (prolocholic acid, Pro-CA) and aspartic acid-conjugated CA (aspartatocholic acid, Asp-CA) were not observed. This absence of proline conjugation may be due to its unique secondary amine preventing proper nucleophilic attack, and previous reports have not observed proline conjugations9.
BSH/T sequence variation shapes MCBA conjugation profiles
After identifying BSH/T as an enzyme responsible for MCBA production in vitro, we examined the 16S rRNA genes from 29 different bacterial strains (Fig. 2a, Table S3) to determine if phylogenetic relatedness correlated with conjugation capability. The 29 species analyzed belong to classes Actinomycetia, Verrucomicrobiae, Gammaproteobacteria, Bacilli, and Clostridia, with a particular focus on the latter and members of its Lachnospiraceae family. Of these strains, 19 produced at least one MCBA (Fig. 2b). Production of MCBAs was particularly prevalent among Lachnospiraceae species, with all but one capable of performing this transformation. Members of all but one microbial class were able to conjugate bile acids, Akkermansia muciniphila (Verrucomicrobiae) being the exception. The most robust conjugators were Lactiplantibacillus plantarum (Bacilli), Ruminococcus gnavus (Clostridia), Enterococcus faecalis (Bacilli), and Bifidobacterium bifidum (Actinomycetia), indicating there was little association between phylogenetic lineage and the capacity for MCBA production.
Hierarchical clustering was used to group the bacterial conjugation profiles into five clusters based on Bray-Curtis dissimilarity (Fig. 2b, Fig. S3). Strains in cluster 1 conjugated a diverse complement of amino acids comprised of those with large and hydrophobic, charged, or small hydrophilic amino acids; strains in cluster 2 favored the incorporation of glycine and alanine; cluster 3 primarily conjugated small, hydrophilic amino acids; cluster 4 included extensive linkage to lysine; and aspartate conjugation predominated in cluster 5 (Fig. 2b). L. plantarum fits into the first group (cluster 1), being the most robust BA conjugator of the strains screened and produced 17 of 18 observed MCBAs a high total MCBA production (Fig. 2b). Although the clusters revealed little phylogenetic correlation, clusters 3 and 5 were primarily associated with members of the Lachnospiraceae (Fig. 2a).
We then mined the genomes of all 29 strains for the presence of bsh/t to investigate the relationship between translated protein sequences and MCBA profiles (Fig. 2c, Table S3). Two strains, Clostridium sporogenes and Lacrimispora aerotolerans, possess annotated bsh/t but did not produce MCBAs. Clostridium scindens was the only strain capable of producing MCBAs while lacking any annotated bsh/t. The 18 of 19 MCBA producers, not including C. scindens, had at least one bsh/t annotated or predicted in their genome with some containing at least 3, such as E. bolteae. The resulting BSH/T phylogenetic tree topology (Fig. 2c) showed limited correlation to the five MCBA profile clusters (Fig. 2b). There were three main lineages of BSH/T; group I contains a set of diverse and robust MCBA producers that have similar BSH/T sequences, group II associates primarily with MCBA clusters 3 and 5, and group III shows significant sequence diversion from the other two groups and little association with the MCBA profiles (Fig. 2b-c). The last group may represent sequences with a high degree of similarity to enzymes within the N-terminal nucleophile (Ntn)-hydrolase superfamily; thus, these BSH/T homologs may have other functions. E. bolteae and E. clostridioformis contain BSH/T sequences from all three groups, yet the former produced a diverse MCBA profile while the latter products were dominated by Gly-CA. Analysis of the BSH/T amino acid sequence alignment revealed amino acid substitutions that were potentially responsible for the evolutionary divergence, particularly at position 82. Asn82 (Fig. 2d-e, C. perfringens BSH/T as reference15) has been reported as being highly conserved in BSH/T sequences in previous studies16. However, we show that Asn82 is a Tyr82 in several members of the Lachnospiraceae, the bulk of which represent BSH/T group II sequences with a less diverse amino acid conjugation profile. Notably, this residue lies in the active site of the BSH/T crystal structure from C. perfringens (Fig. 2e), adjacent to the carboxylate of co-crystalized deoxycholic acid (DCA) and directly interacting with co-crystalized taurine. This difference at residue 82 has also been a marker for differentiating BSH/T and penicillin-V-acylase (PVA) sequences, with Asn82 corresponding to BSH/T sequences and Tyr82 corresponding to PVA sequences17. We therefore propose that BSH/T sequence variation at the active site dictates whether the enzyme is capable of amino acid conjugation.
Amino acid conjugation alters the antimicrobial effects of bile acids
Free BAs are known to exert antimicrobial activity by damaging cell membranes and chromosomal DNA18, a mechanism not limited to bacterial cells and a known action of secondary BAs19–22. Conversely, host conjugated BAs have significantly decreased antimicrobial activity when compared to free BAs18,23. We therefore hypothesized that microbial BA conjugation also modulates bacterial bile acid toxicity. To test this hypothesis, we first determined the impacts of medium supplementation with CA or individual MCBAs on the first organism identified to produce MCBAs, E. bolteae. When comparing the area under the curve (AUC) growth fold changes, E. bolteae had an increase in AUC for each MCBA tested whereas the presence of 1 mM CA was slightly detrimental to growth compared to the control (DMSO used as the vehicle) (Fig 3A). Here, a 1 mM concentration of CA was used, as BA concentration in the cecum and colon can reach upwards of 1 mM in the human gut1. However, when comparing the impacts of MCBA administration on additional species both aerobically and anaerobically, antimicrobial efficacy varied. The most marked reductions in growth, as determined by fold change in AUC, were seen for C. hylemonae, Blautia coccoides, Peptostreptococcus anaerobius, Lacrimispora aerotolerans, and Lacrimispora indolis, of which only L. indolis was found to produce MCBAs (Fig 3A). This detriment was particularly strong in the cases of P. anaerobius and L. aerotolerans, where growth was significantly reduced if not completely inhibited (Fig. 3b). Interestingly, E. bolteae grown in the presence of any MCBA grew to a greater 24-h AUC compared to growth in CA (Fig 3A-B).
MCBA gavage changes the composition of murine fecal and cecal microbiomes
After observing the changes in antimicrobial efficacy from amino acid conjugation of bile acids in vitro, we examined how these findings translated in vivo. Wild-type C57BL/6 mice were gavaged with 100 mg kg-1 of Phe-CA, Ser-CA, TCA, or vehicle for 13 days. Fecal samples were collected every third day throughout the gavage period and cecum samples were collected on day 14. Fecal and cecum samples were subjected to V4 region 16S rRNA gene microbiome analysis. Shifts in microbial community profiles were stronger in female mice compared to male mice and so subsequent analysis was focused on the female animals (n=5 per group). Significant differences in cecal microbiome communities were observed between the groups (PERMANOVA; F = 9.2081, p<0.001). Phe-CA- and Ser-CA-treated communities were distinct from both TCA-treated and vehicle groups as well as from each other (Fig. 3c), although differences in Phe-CA and Ser-CA gavage alone were less significant (PERMANOVA; F = 1.8692, p = 0.033). Microbiome differences were also seen in the fecal samples, notably at day 13, where differences in community structure were significantly different between gavage groups over time (Fig. 3d, PERMANOVA; F = 7.0358, p<0.001).
Random forest classification was used to determine the effects of BA gavage on cecal and fecal communities (Fig. 3e, g). Of the 30 most impactful amplicon sequence variants (ASVs) on model accuracy, 18 were present in both cecal (Table S4, Table S5) and fecal (Table S6, Table S7) classifications with four being present in the top 15 ASVs (Fig. 3e-h, highlighted in blue). Phe-CA gavage resulted in an increased abundance of cecal Enterococcus (Fig. 3f_viii), which have been shown to produce MCBAs, in addition to a member of the genus Muribaculaceae (Fig. 3h_xv), both of Ser-CA gavage resulted in an increased abundance of an uncultured Muribaculaceae species. TCA and vehicle gavage resulted in increases in a Faecalibacterium species that was absent in mice gavaged with either Ser-CA or Phe-CA. This trend was present in both cecum (Fig. 3f_iv) and fecal (Fig. 3h_xii) samples. Both Ser-CA and Phe-CA gavage resulted in an increased fecal abundance of Dubosiella newyorkensis (Fig. 3h_xiv), a species first isolated in 2017 with little currently known about its role in the murine gut microbiome24. The ratio of Firmicutes to Bacteroidota (F/B ratio), formerly Bacteroidetes, has been of recent interest for its use as a marker for gut health. Previous reports have shown that a higher abundance of Firmicutes in feces has been associated with obesity25,26 while, conversely, an increase in the abundance of Bacteroidota is associated with inflammatory bowel disease26. Female mice gavaged with Phe-CA had a significant increase in F/B ratio compared to vehicle controls for both cecum (Fig. 3i, j) and fecal (Fig. 3k,l) samples following gavage. This increase was not significant in fecal samples of mice gavaged with TCA (p = 0.222) or Ser-CA (p = 0.056), though a trend is apparent. Collectively, these data indicate that MCBAs produced by BSH/T can alter the gut microbiome differently than host-conjugated TCA and a mock control.