Urolithin M6 is released by Gordonibacter spp. and further metabolized by Enterocloster spp
Members of the Gordonibacter genus convert EA to intermediate urolithins uroM5, uroM6, and uroC (Fig. 1a). To determine whether 10-hydroxy urolithins (uroM5 and uroM6) were released by Gordonibacter spp. during EA metabolism, we incubated Gordonibacter pamelaeae, Gordonibacter massiliensis, Gordonibacter sp. 28C, and Gordonibacter urolithinfaciens with EA and monitored urolithin production in both whole cultures and supernatants by liquid chromatography-mass spectrometry (LC-MS). The products of EA metabolism were detected in cell-free supernatants after 2 days (Fig. 1b), albeit at lower concentrations than those observed in whole cultures (Fig. 1c). These data demonstrate that urolithin intermediates derived from EA metabolism are released by Gordonibacter spp. and become available to other urolithin-metabolizing bacteria.
To support this hypothesis, we performed co-culture experiments by incubating combinations of G. urolithinfaciens and Enterocloster spp. (E. bolteae or E. asparagiformis) with EA or uroM6. Based on the EA decarboxylase and urolithin dehydroxylase activities (4-, 9-, and 10-positions) of these 2-member communities, we hypothesized that uroA would be detected in co-cultures. As expected, G. urolithinfaciens alone metabolized EA to uroM5, uroM6, and uroC (Fig. 1d); however, uroM6 was not metabolized to uroC when supplied as the initial substrate (Fig. 1e). In co-cultures of G. urolithinfaciens and E. bolteae (9-position dehydroxylation only) treated with EA, we could detect the intermediate urolithin M7 (3,8,10-trihydroxy-urolithin, uroM7) and the terminal metabolite uroA (Fig. 1d). Direct supplementation with uroM6 yielded uroM7 but no uroA (Fig. 1e), demonstrating that only E. bolteae can dehydroxylate uroM6 under these conditions. In contrast, co-cultures of G. urolithinfaciens and E. asparagiformis (9- and 10-position dehydroxylation) treated with EA yielded uroA as a major product and a trihydroxy-urolithin metabolite (Fig. 1d). Since we did not have a standard for this metabolite, it was tentatively assigned as urolithin G (3,4,8-trihydroxy-urolithin, uroG), likely derived from the 9- and 10-position dehydroxylation of uroM5 released by G. urolithinfaciens. Direct supplementation with uroM6 yielded exclusively uroA (Fig. 1e), highlighting differences between activities of urolithin-metabolizing Enterocloster spp. Collectively, these data show that simple 2-member communities can metabolize EA to different urolithins based on dehydroxylase activity and that urolithin features (hydroxylation status and regiochemistry) dictate whether metabolism will occur in whole cells, likely via substrate-specific transporters or dehydroxylase gene induction by urolithin intermediates.
Based on the observation that urolithin intermediates are released by EA-metabolizing Gordonibacter spp. and that a subset of Enterocloster spp. can metabolize uroM6 to uroA, we sought to identify the genes and enzymes responsible for 10-position urolithin dehydroxylation, as previously done for the 9-position. We reasoned that either the ucd-encoded UcdO (substrate-binding oxidoreductase) subunit of 9- and 10-position-metabolizing species like E. asparagiformis was promiscuous, acting at both the 9- and 10-positions of uroM6, or that a separate dehydroxylase was metabolizing the 10-position in these bacteria. Multiple sequence alignment of UcdO proteins from Enterocloster spp. (E. asparagiformis, E. citroniae, and E. pacaense) revealed that these sequences were >96.32% identical to the previously characterized E. bolteae UcdO (similarities >98.73%, Supplementary Fig. 1). Additionally, predicted urolithin binding site residues W345, Y375, F458, F464, Y536, Y624, and Y632 were conserved among all species (Supplementary Fig. 1), suggesting that substrate specificity should be identical in these bacteria. To validate that the ucd from E. asparagiformis encoded a 9-position-specific dehydroxylase, as observed in E. bolteae, we cloned and heterologously expressed the E. asparagiformis ucd operon in Rhodococcus erythropolis using a thiostrepton-inducible plasmid (pTipQC2-Ea_ucdCFO, Fig. 2a). In crude lysates, we could reliably detect the expression of UcdO in the insoluble fraction (Fig. 2b). Incubation of urolithins (uroM6, uroM7, uroC, or uroA) with ucd-expressing crude lysates confirmed that only 9-position hydroxyl groups, present on uroM6 and uroC, were dehydroxylated to uroM7 and uroA, respectively (Fig. 2c, d). We did not observe any 10-position dehydroxylation in uroM6 or uroM7-treated lysates, demonstrating that the E. asparagiformis ucd operon specifically dehydroxylates 9-hydroxy urolithins. These results suggest that a different, unidentified operon is responsible for 10-position dehydroxylation in some Enterocloster spp.
Previous work from our lab has shown that dehydroxylation of urolithins by Enterocloster spp. is an inducible process triggered by substrate urolithins. Thus, we hypothesized that, upon exposure to uroM6 (Fig. 1a), Enterocloster spp. would express Ucd and urolithin 10-position (X) dehydroxylase (Uxd) proteins. Therefore, we individually incubated E. asparagiformis, E. citroniae, and E. pacaense with uroM6 (or DMSO as a vehicle control), pooled isolates by treatment group, and performed reference-based proteomics on pooled lysates (Fig. 3a). To ensure that target enzymes were being expressed, we collected samples that were actively metabolizing uroM6. In all 3 tested isolates, uroA could be detected after a 4 h incubation with uroM6 (Fig. 3b). Filtering proteins by presence/absence revealed that 44 individual proteins were overrepresented in the uroM6 group (Fig. 3c). Among these, the most abundant proteins induced by uroM6 mapped to the ucd operon-encoded proteins UcdC and UcdO (Fig. 3d). However, we detected distinct xanthine dehydrogenase family proteins and FAD binding domain-containing proteins, which we annotated as putative Uxd proteins (Fig. 3d). The predicted FAD coenzyme-binding subunit (UxdC) was detected in all 3 species, while the substrate-binding oxidoreductase subunit (UxdO) was only detected in E. pacaense. Additionally, the molybdopterin cytosine dinucleotide (MCD) cofactor biosynthesis protein MogA (MOSC domain-containing protein) and the MCD cofactor chaperone protein XdhC (XdhC/CoxI family protein) were expressed, though to lesser extents than Ucd and Uxd proteins (Fig. 3d).
We next sought to validate the function of Uxd proteins through heterologous expression and noticed that, like Ucd, Uxd protein coding sequences were arranged in an operon with the following arrangement: uxdFOC. Therefore, we cloned the E. asparagiformis uxd operon into a shuttle plasmid to heterologously express UxdF, UxdO, and UxdC proteins in R. erythropolis (pTipQC2-Ea_uxdFOC). Unfortunately, electroporation of R. erythropolis with pTipQC2-Ea_uxdFOC did not yield any transformants capable of producing all three Uxd proteins, despite multiple attempts at transformation. We therefore searched the literature for alternative heterologous expression hosts that were anaerobic, genetically tractable, and ideally capable of xanthine dehydrogenase family protein expression. Recent studies demonstrated that Clostridium sporogenes, found in anaerobic soil and gut environments, is an appropriate host for the heterologous expression of xanthine dehydrogenase family proteins like 2,8-dioxopurine dehydrogenase (DOPDH). C. sporogenes encodes the molybdenum cofactor cytidylyltransferase MocA (WP_003494683.1) and the MCD cofactor chaperone XdhC (WP_033060855.1). These proteins, among others, participate in MCD biosynthesis and xanthine dehydrogenase family protein maturation. In addition, the RiboCas series of plasmids, which utilize theophylline-inducible riboswitches to control Streptococcus pyogenes Cas9 expression, has been used to delete the spoIIE gene in multiple Clostridium spp., including C. sporogenes. Based on the genotype and genetic tractability of C. sporogenes, we hypothesized that it would be an appropriate host for heterologous expression of Uxd proteins.
We therefore cloned the E. asparagiformis uxd operon and replaced the Cas9-encoding region of pRG-Cas2 to produce the pRG-Ea_uxdFOC shuttle plasmid (Fig. 3e). This plasmid was transformed into C. sporogenes via conjugation and resulting transformants were screened for protein expression and 10-position urolithin dehydroxylase activity. There were no obvious differences in protein expression between the wild type C. sporogenes and 10 randomly picked transformants (Supplementary Fig. 2). However, all 10 transformants were able to convert uroM6 to uroC (Fig. 3f). We further validated these findings by incubating uroM7, which lacks a catechol, with wild type C. sporogenes, an enhanced GFP (eGFP)-expressing control transformant, or a Uxd-expressing transformant. No background urolithins were detected in DMSO-treated media or cultures (Fig. 3g). In uroM7-treated cultures, uroA was detected solely in the Uxd-expressing C. sporogenes transformant, demonstrating the 10-position urolithin dehydroxylase activity conferred by the uxd operon (Fig. 3g). Collectively, these data demonstrate that a subset of Enterocloster spp. express distinct enzymes (Ucd and Uxd), along with their accessory proteins, that regioselectively dehydroxylate uroM6, uroM7, and uroC.
Since ucd and uxd operons appeared to be restricted to a subset of Enterocloster spp., we wondered about the prevalence of these operons both between and within Enterocloster spp. Therefore, we downloaded and queried 329 Enterocloster spp. genomes from the National Center for Biotechnology Information (NCBI) genomes database for the presence of ucd and uxd at the nucleotide level. As anticipated, the ucd operon was detected in E. asparagiformis, E. bolteae, E. citroniae, and E. pacaense genomes, whereas the uxd operon was detected in E. asparagiformis, E. citroniae, and E. pacaense (Fig. 4a). There was no evidence of these operons in other Enterocloster spp. genomes, including metagenome-assembled genomes (MAGs). All genomes of urolithin-metabolizing taxa encoded uxd and/or ucd with the exception of one E. asparagiformis MAG (GCA_040307595.1) which had neither of the operons (Fig. 4a). By performing searches using the basic local alignment search tool (BLAST) for sequences similar to the E. asparagiformis uxd, we observed that E. asparagiformis, E. citroniae, and E. pacaense had the highest sequence identities (id) at the nucleotide level (>77% id and >98% coverage, Supplementary Fig. 3). In addition, these 3 taxa had high sequence id for UxdO (>93% id and 100% coverage, Supplementary Fig. 4) and UxdC (>75% id and >97% coverage, Supplementary Fig. 5) at the protein level. However, we identified one bacterium, named Lacrimispora sp. 210928-DFI.3.58 (GCA_020554865.1) in the NCBI, which had a nucleotide sequence with 66% id to uxd with 92% coverage (Supplementary Fig. 3) and proteins with 77% and 63% id to UxdO and UxdC, respectively (>97% coverage, Supplementary Fig. 3, 4). This taxon is annotated as Enterocloster sp900753815 in the Genome Taxonomy Database (GTDB) but was missed in the previous Enterocloster spp. genome search (Fig. 4a) due to its misclassification in the NCBI genomes database (Methods). All other protein hits from the BLAST searches had identities <50% and <57% compared to UxdO and UxdC, respectively. Altogether, ucd and uxd operons, including some homologs, are highly prevalent within a subset of Enterocloster spp.
Next, we compared proteins in the vicinity of the ucd and uxd operons in urolithin-metabolizing Enterocloster spp. to investigate their genomic contexts. The ucd and uxd operons were in close proximity (<5 kb between ucdO and uxdF) in 10-position metabolizing species (Fig. 4b). In these bacteria, Uxd proteins (UxdF, UxdO, and UxdC) were highly similar (>94, >96, and >83% protein similarity, respectively, Supplementary Fig. 6a, b). To our surprise, E. bolteae, which only metabolizes the 9-position of urolithins, possessed a uxd homolog near its ucd operon (Fig. 4b). In E. bolteae, the uxd operon homolog was more distant from the ucd (10.2 kb between ucdO and uxdF), and its Uxd proteins were less similar (<66% protein similarity) compared to those from 10-position metabolizing species (Supplementary Fig. 6a, b). Notably, the substrate-binding UxdO protein was >96% similar between E. asparagiformis, E. citroniae, and E. pacaense, but less than 63% similar (<44% protein id) when compared to E. bolteae (Supplementary Fig. 7). These different protein sequences could explain the lack of 10-position dehydroxylase activity observed in E. bolteae.
We next compared domain annotations between ucd and uxd-encoded proteins of E. asparagiformis, which serves as our reference in this study. Iron-sulfur cluster-binding proteins (UcdF and UxdF) were similar in length, and both possessed two 2Fe-2S cluster binding domains, characteristic of prokaryotic xanthine dehydrogenase family proteins, which use these proteins to shuttle electrons between FAD and the MCD cofactor (Fig. 4c). The MCD- and substrate-binding oxidoreductase subunits (UcdO and UxdO) both had functional domains typical of xanthine dehydrogenase family proteins: an XDH a/b hammerhead and molybdenum cofactor binding domains (Fig. 4c). The FAD-binding subunits (UcdC and UxdC) were more dissimilar, mostly due to the additional length of UxdC, which carried a 4Fe-4S cluster binding domain and an FAD/NAD(P) binding domain in addition to the FAD binding domain common to UcdC and UxdC (Fig. 4c). AlphaFold3 modeling of Ucd and Uxd complexes produced similar overall structures resembling other xanthine dehydrogenase family proteins (Fig. 4d). By superposing these predicted structures onto the crystal structure of the Thauera aromatica 4-hydroxybenzoyl-CoA reductase complex (dehydroxylase), we could place the MCD and 2Fe-2S clusters into UcdO and UxdO, and UcdF and UxdF, respectively. These ligands form an electron transport chain, likely enabling the dehydroxylation of substrates (e.g. urolithins) to their reduced products (Fig. 4e). In sum, ucd and uxd operons, which co-localize in urolithin-metabolizing Enterocloster spp. genomes, encode xanthine dehydrogenase proteins with similar predicted folds.
We next wondered whether uxd operon substrates (10-hydroxy urolithins) induce the expression of uxd (Fig. 5a). Therefore, E. asparagiformis, grown to the mid-exponential phase, was treated with DMSO, uroM6, uroM7, uroC, or uroA (structures in Fig. 1a), and total RNA was extracted for RT-qPCR analysis focusing on ucdO and uxdO gene expression. Curiously, uxd expression followed the same pattern as ucd (Fig. 5b); that is, only 9-hydroxy urolithins (uroM6 and uroC) induced uxd expression. UroM7, which bears a hydroxyl group at the 10-position, and uroA, which is dehydroxylated at both the 9- and 10-positions, did not induce the expression of either operon (Fig. 5b). When we quantified urolithin concentrations in matched cultures, where total RNA and urolithins were isolated from the same samples, we observed dehydroxylation of uroM6 at both the 9- and 10-positions and dehydroxylation of uroC at the 9-position; however, uroM7 was unchanged (Fig. 5c). To validate that 9-hydroxy urolithins are required for 10-position dehydroxylation, E. asparagiformis cultures were induced with 9-hydroxy urolithins lacking a 10-hydroxyl group: urolithin D (3,4,8,9-tetrahydroxy-urolithin, uroD), uroC, or isouroA (Fig. 5d). After a 2 h incubation, 9-position dehydroxylation was observed for uroD- and uroC-treated cultures, yielding uroG and uroA, respectively; however, no dehydroxylation was observed for isouroA at this timepoint since uroB was not detected in culture supernatants (Fig. 5e). Following this 2 h induction, E. asparagiformis cells were washed and resuspended in phosphate buffered saline (PBS) to minimize protein synthesis, then re-treated with urolithins to assess whether the 10-hydroxy urolithins could be dehydroxylated. In DMSO-induced resting cell suspensions, partial 9- and 10-position dehydroxylation was observed for uroM6 and uroC re-treated samples, suggesting basal activity or inducibility in these cell suspensions. Nevertheless, uroM7 remained unchanged (Fig. 5f), as previously observed in exponential phase cultures (Fig. 5c). In 9-hydroxy urolithin-induced resting cell suspensions, significant uroA production was observed for nearly all substrates (compared to DMSO-induced cultures). As anticipated, 10-position dehydroxylation of uroM7 to uroA was only observed in cell suspensions induced with 9-hydroxy urolithins (Fig. 5f), demonstrating the co-regulation of ucd and uxd operons in E. asparagiformis.